docs(dedup): correct the CodeQL claim on the tenant-namespace comment

The comment added in #11649 said keeping the API key id out of the digest "avoids the false-positive CodeQL js/insufficient-password-hash on a cache/dedup key". It does not. CodeQL raised alert #874 on the `createHash` in this very function on the first scan after that merge: once an API-key-derived value reaches the file at all, the query flags the sibling digest regardless of what goes into it. The plaintext prefix is still the right call — a namespace you can read off the key beats one you cannot when debugging a dedup collision, and it matches `semanticCache.generateSignature` (#3740), whose comment makes the same incorrect claim. But a rationale that is measurably false is worse than no rationale: the next person edits here, sees the alert the comment promised would not happen, and doubts the design instead of the comment. Replaced with what actually holds, plus the two things worth knowing: the alert is dismissed per HR#14 and will come back on any edit here, and it must not be "fixed" with a KDF — that would break the determinism dedup depends on. Comment-only. No behaviour change; request-dedup tests 15/15.
2026-09-07 23:42:10 +03:00 · 2026-08-26 19:15:15 -03:00
5664 changed files with 117040 additions and 299799 deletions
--- a/.env.example
+++ b/.env.example
@@ -45,21 +45,10 @@ INITIAL_PASSWORD=CHANGEME
 # executor's on-disk thread-sticky session cache. Leave unset to rely on DATA_DIR.
 # OMNIROUTE_DATA_DIR=/var/lib/omniroute

-# Directory the runtime plugin scanner reads, overriding the home-derived default (#11827).
-# Used by: src/lib/plugins/scanner.ts — getDefaultPluginDir(); it is also the root the
-# plugin manager installs into. Set it in Docker/K8s to point straight at the bind-mounted
-# plugin tree, instead of moving HOME (which changes every other HOME-relative behaviour)
-# just to relocate the scan path. Unset = <HOME>/.omniroute/plugins, and
-# /tmp/.omniroute/plugins when the process exports no home at all.
-# Distinct from the CLI-only variable in section 9 that points the omniroute-cmd-* command
-# loader (bin/cli/plugins.mjs) at a package tree — this one drives the server-side scanner.
-# OMNIROUTE_PLUGINS_DIR=/opt/omniroute/plugins
-
-# Escape hatch for the test/eval DATA_DIR guard (#10428). A test or node eval/print
-# probe (-e/--eval/-p/--print, including --eval=/--print=) that never chose a DATA_DIR
-# is redirected to a throwaway temp dir so it cannot open the operator's real database.
-# Set to 1 only for a deliberate run against the real DATA_DIR — never for CI.
-# Used by: src/lib/dataPaths.ts
+# Escape hatch for the test-context DATA_DIR guard (#10428). A test run that never
+# chose a DATA_DIR is redirected to a throwaway temp dir so it cannot open the
+# operator's real database. Set to 1 only for a deliberate run against the real
+# DATA_DIR — never for CI. Used by: src/lib/dataPaths.ts
 # OMNIROUTE_ALLOW_DEFAULT_DATA_DIR=1

 # Build provenance (#10427). OMNIROUTE_BUILD_SHA lets a container inject the artifact's git
@@ -82,11 +71,6 @@ INITIAL_PASSWORD=CHANGEME
 # Never set this for the running server. Used by: src/lib/buildPhase.ts, src/lib/db/core.ts
 # OMNIROUTE_BUILDING=1

-# Skip the optional native-dependency prebuild check for exotic vendored trees.
-# This does not make a missing dependency buildable. Used by: scripts/check/check-native-deps.mjs
-# Default: 0  |  Set to 1 only when native dependencies are supplied out of band.
-# OMNIROUTE_SKIP_NATIVE_DEP_CHECK=0
-
 # Encryption key for SQLite database encryption at rest.
 # Used by: src/lib/db/encryption.ts — encrypts the entire SQLite database.
 # Generate: openssl rand -hex 32  |  Leave empty to disable DB encryption.
@@ -97,11 +81,9 @@ STORAGE_ENCRYPTION_KEY=
 # Default: v1  |  Increment when rotating STORAGE_ENCRYPTION_KEY.
 STORAGE_ENCRYPTION_KEY_VERSION=v1

-# Routine/pre-write SQLite backups.
-# Used by: src/lib/db/backup.ts. Set true only when those backups are managed externally.
-# This never disables the migration runner's mandatory, content-addressed safety snapshot
-# or its mass-migration guard for an existing persistent database.
-# Default: false (routine backups enabled).
+# Automatic SQLite backup on startup.
+# Used by: src/lib/db/backup.ts — creates a timestamped backup before migrations.
+# Default: false (backups enabled)  |  Set true to skip backup on every restart.
 DISABLE_SQLITE_AUTO_BACKUP=false

 # ── Redis (Rate Limiting) ──
@@ -243,7 +225,7 @@ PORT=20128
 # Used by: src/app/api/v1/relay/chat/completions/route.ts
 # RELAY_IP_PER_MINUTE=30

-# Bundler selection for `npm run dev` and `npm run build`. Set to 0 to fall back to webpack.
+# Bundler selection for `npm run dev`. Set to 0 to fall back to webpack.
 # Default is 1 (Turbopack). PR #4092 had forced webpack because earlier
 # Turbopack 16.2.x panicked on the OmniRoute module graph with "internal error:
 # entered unreachable code: there must be a path to a root"
@@ -253,9 +235,8 @@ PORT=20128
 # /api/v1/models, /api/mcp) and repeated HMR rebuilds: zero panics. Turbopack
 # also keeps dev memory far lower on the edit→rebuild loop (HMR rebuild RSS stays
 # ~flat vs webpack's monotonic growth), which mitigates the dev-server OOM on
-# this 60+ route app. The production build (scripts/build/build-next-isolated.mjs)
-# reads the same flag: Turbopack by default, 0 builds with webpack (`npm run
-# build:contributor` sets it for you).
+# this 60+ route app. The production build still uses webpack (build pipeline is
+# unaffected by this dev-only flag).
 OMNIROUTE_USE_TURBOPACK=1

 # Disable systemd sd_notify (Type=notify / WatchdogSec=) even when running
@@ -418,9 +399,7 @@ ALLOW_API_KEY_REVEAL=false
 # provider dispatch. Heavyweight capacity is reserved before parsing; excess work
 # receives 503 + Retry-After instead of overlapping until the process OOMs.
 # Used by: src/shared/middleware/chatBodyAdmission.ts
-# Actual bodies at or above this size take the heavyweight lease (BYTE path,
-# including POST /v1/responses) and use the same #10437 healthy-headroom escape
-# as structure-heavy. Default 262144 (256 KB).
+# Actual bodies at or above this size require a heavyweight lease. Default 262144 (256 KB).
 # OMNIROUTE_CHAT_LARGE_BODY_BYTES=262144
 # Actual-byte hard cap enforced during bounded ingestion. Default 52428800 (50 MB).
 # OMNIROUTE_CHAT_HARD_MAX_BODY_BYTES=52428800
@@ -428,11 +407,6 @@ ALLOW_API_KEY_REVEAL=false
 # left unset, heavyweight admission is gated by OMNIROUTE_CHAT_MAX_INFLIGHT_BYTES below
 # instead (an auto-derived byte budget), fixing coding-agent fan-out (multiple
 # subagents/CLIs) collapsing to an effective concurrency of ~1 and 503ing.
-# Two overlapping ~750k-token /v1/responses abort ~12 Gi heaps (#7849) — a
-# memory-budget warning, not a hard product max of 2. A healthy heap may admit
-# more via HEALTHY_HEADROOM. Tens of long SSE clients (40-50) is heap +
-# OMNIROUTE_CHAT_MAX_INFLIGHT_BYTES / #10110. Multiply heaps with N independent
-# DATA_DIRs (#11024); never replicas>1 on one SQLite.
 # OMNIROUTE_CHAT_MAX_HEAVY_IN_FLIGHT=1
 # Override for the auto-derived ingest byte budget (#503-fanout). Default: 25% of the
 # process's effective memory ceiling (V8 heap limit, or the tighter cgroup/container
@@ -440,15 +414,13 @@ ALLOW_API_KEY_REVEAL=false
 # 2 GiB; explicit overrides are clamped to the same safe range. Read
 # chatAdmission.maxInflightBytes/budgetSource at /api/monitoring/health before overriding.
 # OMNIROUTE_CHAT_MAX_INFLIGHT_BYTES=134217728
-# Heap-pressure shed ratio (heapUsed/heap_size_limit) for BYTE and STRUCTURE
-# heavyweight admission (#10183, #10268, #10437): a concurrent heavyweight request
-# past OMNIROUTE_CHAT_MAX_HEAVY_IN_FLIGHT is only shed with a retryable 503 when the
-# heap is ALSO under this much pressure — on a healthy heap it is admitted via
-# healthy-headroom instead. Range (0, 1]. Default 0.75.
+# Heap-pressure shed ratio (heapUsed/heap_size_limit) for the structural admission gate
+# (#10183, #10268): a second concurrent heavyweight request past OMNIROUTE_CHAT_MAX_HEAVY_IN_FLIGHT
+# is only shed with a retryable 503 when the heap is ALSO under this much pressure — on a
+# healthy heap it is admitted instead. Range (0, 1]. Default 0.75.
 # OMNIROUTE_CHAT_ADMISSION_HEAP_SHED_RATIO=0.75
 # Bounded extra capacity for the healthy-heap fast path above OMNIROUTE_CHAT_MAX_HEAVY_IN_FLIGHT
-# (#10437, BYTE + STRUCTURE, including bodies >= OMNIROUTE_CHAT_LARGE_BODY_BYTES):
-# once this many concurrent leases are active through the healthy-heap bypass,
+# (#10437): once this many concurrent leases are active through the healthy-heap bypass,
 # further busy requests fall through to the same bounded-wait/shed path used under real heap
 # pressure. 0 disables the bypass entirely. Default 1.
 # OMNIROUTE_CHAT_ADMISSION_HEALTHY_HEADROOM=1
@@ -640,9 +612,11 @@ CLOUD_URL=
 # Default: http://localhost:20128
 NEXT_PUBLIC_BASE_URL=http://localhost:20128

-# Highest-priority OmniRoute public origin override, also used by non-dashboard
-# public-origin validation. Set it when external clients reach OmniRoute through
-# a stable LAN, tunnel, or public origin that differs from its internal URL.
+# Browser-facing OmniRoute origin for generated assets in API responses.
+# Highest-priority public origin override; also used by non-dashboard public-origin validation.
+# Used by: chatgpt-web image generation cache URLs (/v1/chatgpt-web/image/<id>).
+# Set this when OpenWebUI or another relay reaches OmniRoute by an internal URL
+# but the user's browser must fetch images from a LAN, tunnel, or public origin.
 # Do not include /v1; if included accidentally it will be normalized away.
 # OMNIROUTE_PUBLIC_BASE_URL=http://192.168.0.15:20128

@@ -655,6 +629,28 @@ NEXT_PUBLIC_BASE_URL=http://localhost:20128
 # Used by: open-sse/config/providerPluginManifestUrl.ts. Defaults to http.
 # OMNIROUTE_PUBLIC_PROTOCOL=http

+# Max wait time for an async chatgpt-web image to land via the celsius
+# WebSocket, in milliseconds. Default 180000 (3 minutes). Increase during
+# upstream queue-deep windows ("Lots of people are creating images right now").
+# OMNIROUTE_CGPT_WEB_IMAGE_TIMEOUT_MS=180000
+
+# Total in-memory byte budget for the chatgpt-web image cache (used to serve
+# /v1/chatgpt-web/image/<id>), in megabytes. Default 256. Lower this if you
+# run OmniRoute on a memory-constrained host; raise it if image generation
+# is heavy and clients are racing the 30-minute TTL.
+# OMNIROUTE_CGPT_WEB_IMAGE_CACHE_MAX_MB=256
+
+# Overall wait budget for a chatgpt-web GPT-5.5 Pro background-poll handoff,
+# in milliseconds. Default 1200000 (20 minutes). Pro reasoning runs are slow
+# and complete out-of-band, so OmniRoute polls until the answer lands or this
+# budget elapses. Raise it if Pro requests time out before finishing.
+# OMNIROUTE_CGPT_WEB_PRO_TIMEOUT_MS=1200000
+
+# Interval between chatgpt-web GPT-5.5 Pro background-poll attempts, in
+# milliseconds. Default 4000 (4 seconds). Lower for snappier completion at the
+# cost of more upstream polling; raise to reduce request volume.
+# OMNIROUTE_CGPT_WEB_PRO_POLL_INTERVAL_MS=4000
+
 # Public cloud URL — client-side mirror of CLOUD_URL.
 NEXT_PUBLIC_CLOUD_URL=

@@ -686,11 +682,21 @@ NEXT_PUBLIC_CLOUD_URL=
 # open-sse/services/usage.ts.
 #OMNIROUTE_CROF_USAGE_URL=https://crof.ai/usage_api/
 #OMNIROUTE_CODEWHISPERER_BASE_URL=https://codewhisperer.us-east-1.amazonaws.com
-# Official OpenCode Go usage endpoint, authenticated with the connection API key.
-# Override only for relays or test fixtures.
-#OMNIROUTE_OPENCODE_QUOTA_URL=https://opencode.ai/zen/go/v1/usage
+#OMNIROUTE_OPENCODE_QUOTA_URL=https://opencode.ai/zen/go/v1/quota
+# OpenCode Go has no public quota API — this has no default and stays
+# unset unless you explicitly opt in to a self-hosted/mirrored endpoint:
+#OMNIROUTE_OPENCODE_GO_QUOTA_URL=
+#OMNIROUTE_OPENCODE_GO_DASHBOARD_URL=https://opencode.ai/workspace
 #OMNIROUTE_OLLAMA_CLOUD_USAGE_URL=https://ollama.com/settings

+# OpenCode Go dashboard quota scraping. Prefer configuring these per connection
+# in Dashboard → Providers → OpenCode Go. Env vars are useful for headless
+# deployments or shared server defaults. The cookie is sensitive.
+#OPENCODE_GO_WORKSPACE_ID=wrk_...
+#OMNIROUTE_OPENCODE_GO_WORKSPACE_ID=wrk_...
+#OPENCODE_GO_AUTH_COOKIE=auth=...
+#OMNIROUTE_OPENCODE_GO_AUTH_COOKIE=auth=...
+
 # OpenCode Go/Zen VPS egress (#5997): on a datacenter VPS, Cloudflare in front of
 # opencode.ai/zen/go 403s chat requests that lack OpenCode CLI identity headers.
 # When your clients don't already send them, set this to synthesize the CLI headers
@@ -825,14 +831,9 @@ NEXT_PUBLIC_ENABLE_SOCKS5_PROXY=true
 # CLI_CRUSH_BIN=crush
 # CLI_OMP_BIN=omp
 # CLI_LETTA_BIN=letta
-# CLI_PRIME_AGENT_BIN=prime-agent
 # Windsurf has no default binary — set this to enable binary detection for it.
 # CLI_WINDSURF_BIN=windsurf
 # CLI_AUGGIE_BIN=auggie
-# CLI_5DIVE_BIN=5dive
-# 5dive keeps root-owned auth profiles under a system state dir (its own STATE_DIR,
-# default /var/lib/5dive); override here when it lives elsewhere.
-# CLI_5DIVE_STATE_DIR=/var/lib/5dive
 # AUGGIE_BIN=auggie

 # ── ZCode (Z.ai GLM coding-plan CLI) local provider ──
@@ -876,17 +877,6 @@ NEXT_PUBLIC_ENABLE_SOCKS5_PROXY=true
 # Legacy alias for OMNIROUTE_API_KEY.
 # ROUTER_API_KEY=

-# Days of A2A task history to keep before the daily purge deletes a row.
-# Used by: src/lib/a2a/taskManager.ts (historyRetentionDays). Unset, non-numeric,
-# or <= 0 falls back to the default.
-# OMNIROUTE_A2A_HISTORY_RETENTION_DAYS=30
-
-# Kill-switch for the A2A memory-hits observability feature (Orchestration Canvas
-# Fase 2). Set to "0" to skip the memory recall lookup entirely; any other value
-# (including unset) keeps it enabled.
-# Used by: src/lib/a2a/taskExecution.ts (collectMemoryHits).
-# OMNIROUTE_A2A_MEMORY_HITS=1
-
 # Enable the offline/local Issue Agent recorded-triage endpoint.
 # Used by: src/app/api/issue-agent/runs/route.ts. Default: disabled.
 # OMNIROUTE_ISSUE_AGENT_ENABLED=false
@@ -1003,11 +993,6 @@ PROVIDER_LIMITS_SYNC_SPACING_MS=1500
 # Used by: src/lib/jobs/budgetResetJob.ts.  Floor: 10000.
 #OMNIROUTE_BUDGET_RESET_JOB_INTERVAL_MS=600000

-# Cron expression for the call-log export job (destinations configured in the
-# dashboard under Integrations > Log export).  Default: hourly, on the hour.
-# Used by: src/lib/jobs/logExportJob.ts.  Timezone: UTC.
-#OMNIROUTE_LOG_EXPORT_CRON=0 * * * *
-
 # Emergency budget-exhaustion fallback (set false or 0 to disable the reroute to
 # nvidia/openai/gpt-oss-120b when a request fails with a 402 budget error).
 # Used by: open-sse/services/emergencyFallback.ts.  Default: enabled.
@@ -1052,13 +1037,6 @@ PROVIDER_LIMITS_SYNC_SPACING_MS=1500
 # to disable the check. Used by: src/lib/db/migrationRunner.ts.  Default: 50.
 #OMNIROUTE_MAX_PENDING_MIGRATIONS=50

-# Working directory for the check:install-upgrade release gate. It builds two ~3 GB
-# install trees plus a ~275 MB tarball, so it needs roughly 12 GB — more than the
-# 12 GB RAM-backed tmpfs that /tmp is on the self-hosted runner, where it exhausted
-# the tmpfs and npm silently truncated the package. Defaults to <repo>/.install-upgrade
-# on real disk. Used by: scripts/check/check-install-upgrade.mjs.  Default: <repo>/.install-upgrade.
-#OMNIROUTE_INSTALL_UPGRADE_WORKDIR=/var/tmp/omniroute-install-upgrade
-
 # Trust user-managed RTK project filter rules without strict signature checks.
 # Used by: open-sse/services/compression/engines/rtk/filterLoader.ts.  Default: 0.
 #OMNIROUTE_RTK_TRUST_PROJECT_FILTERS=0
@@ -1175,6 +1153,11 @@ CODEX_OAUTH_CLIENT_ID=app_EMoamEEZ73f0CkXaXp7hrann
 # Trae OAuth token override. Used by: open-sse/executors/trae.ts.
 # TRAE_TOKEN=

+# ── The Old LLM (theoldllm) ──
+# Playwright navigation timeout (ms) for the browser-backed token capture.
+# Used by: open-sse/executors/theoldllm.ts. Default: 30000 (30s).
+# THEOLDLLM_NAV_TIMEOUT_MS=30000
+
 # ── Gemini / Antigravity (Google-based) ──
 # These providers ship public OAuth client_id/secret values embedded in their
 # public CLIs. Defaults are baked into the code via
@@ -1273,6 +1256,17 @@ GITHUB_OAUTH_CLIENT_ID=Iv1.b507a08c87ecfe98
 # VISION_BRIDGE_BASE_URL=
 # VISION_BRIDGE_API_KEY=

+# ── Raycast Pro (local auto-import) ──
+# Raycast Pro AI is a reverse-engineered, unofficial API — local/personal use
+# only (no OAuth client_id/secret; token is captured via macOS Auto-Import
+# from the Keychain + local Raycast SQLite DB, or pasted manually). These
+# vars are optional manual overrides used by open-sse/services/raycast.ts
+# and the direct-probe benchmark script scripts/raycast/usage-benchmark.mjs.
+# RAYCAST_BEARER_TOKEN=
+# RAYCAST_DEVICE_ID=
+# RAYCAST_AID=
+# RAYCAST_SIG_SECRET=
+
 # ─────────────────────────────────────────────────────────────────────────────
 # ⚠️  GOOGLE OAUTH (Antigravity) & OTHER PROVIDERS — REMOTE SERVERS
 # ─────────────────────────────────────────────────────────────────────────────
@@ -1302,7 +1296,7 @@ GITHUB_OAUTH_CLIENT_ID=Iv1.b507a08c87ecfe98
 # Used by: open-sse/executors/base.ts — buildHeaders() dynamic lookup.
 # Update these when providers release new CLI versions to avoid blocks.

-CLAUDE_USER_AGENT="claude-cli/2.1.258 (external, cli)"
+CLAUDE_USER_AGENT="claude-cli/2.1.219 (external, cli)"

 # Disable the deterministic tool-name cloak applied on both Anthropic-bound paths
 # (executors/base.ts native OAuth + executors/cliproxyapi.ts CLIProxyAPI) —
@@ -1332,16 +1326,6 @@ CURSOR_USER_AGENT="Cursor/3.4"
 # Override Codex client version sent in headers independently of the
 # CODEX_USER_AGENT string. Used by: open-sse/config/codexClient.ts.
 # CODEX_CLIENT_VERSION=0.144.1
-#
-# Override the advertised Claude Code client version independently of
-# CLAUDE_USER_AGENT. Anthropic gates some models (Fable 5.1) on this
-# value; a UA-only override is not enough (#12417). Used by:
-# src/shared/constants/claudeCodeClient.ts.
-# CLAUDE_CODE_CLIENT_VERSION=2.1.259
-#
-# Override the advertised GitHub Copilot CLI version independently of
-# GITHUB_USER_AGENT. Used by: open-sse/config/providerHeaderProfiles.ts.
-# GITHUB_COPILOT_CLI_VERSION=1.0.82

 # Kill-switch to strip non-standard `codex.*` SSE events (e.g. codex.rate_limits)
 # from the Codex Responses stream. These frames break the OpenAI SDK's
@@ -1499,15 +1483,28 @@ CURSOR_USER_AGENT="Cursor/3.4"
 # FIRECRAWL_BASE_URL=https://api.firecrawl.dev
 # FIRECRAWL_TIMEOUT_MS=30000            # Per-request timeout (default: 30000 = 30s)

-# ── Claude TLS transport (Chromium-fingerprinted client) ──
-# Used by: open-sse/services/claudeTlsClient.ts — native wreq-js request timeout
-# plus the absolute JS hard-deadline grace when the native request is wedged.
+# ── ChatGPT TLS sidecar (Firefox-fingerprinted client) ──
+# Used by: open-sse/services/chatgptTlsClient.ts — wire-level timeout for
+# the bogdanfinn/tls-client koffi binding and the JS-side grace window
+# layered on top of it when the native library is wedged.
+# OMNIROUTE_CHATGPT_TLS_TIMEOUT_MS=60000
+# OMNIROUTE_CHATGPT_TLS_GRACE_MS=10000
+# Max wait for the FIRST streamed byte from the ChatGPT TLS sidecar before the
+# request is aborted as a dead stream, in milliseconds. Default 30000 (30s).
+# Raise it if upstream cold-starts routinely exceed the window.
+# OMNIROUTE_CHATGPT_STREAM_FIRST_BYTE_TIMEOUT_MS=30000
+
+# ── Claude TLS sidecar (Chromium-fingerprinted client) ──
+# Used by: open-sse/services/claudeTlsClient.ts — wire-level timeout for
+# the bogdanfinn/tls-client koffi binding and the JS-side grace window
+# layered on top of it when the native library is wedged.
 # OMNIROUTE_CLAUDE_TLS_TIMEOUT_MS=60000
 # OMNIROUTE_CLAUDE_TLS_GRACE_MS=10000

-# ── Perplexity TLS transport (Firefox-fingerprinted client) ──
-# Used by: open-sse/services/perplexityTlsClient.ts — native wreq-js request
-# timeout plus the absolute JS hard-deadline grace.
+# ── Perplexity TLS sidecar (Firefox-fingerprinted client) ──
+# Used by: open-sse/services/perplexityTlsClient.ts — wire-level timeout for
+# the bogdanfinn/tls-client koffi binding and the JS-side grace window
+# layered on top of it when the native library is wedged.
 # OMNIROUTE_PPLX_TLS_TIMEOUT_MS=30000
 # OMNIROUTE_PPLX_TLS_GRACE_MS=10000

@@ -1519,16 +1516,18 @@ CURSOR_USER_AGENT="Cursor/3.4"
 # meta-commentary. Set to 1/true/yes/on to restore the old behavior.
 # OMNIROUTE_PPLX_SEARCH_HINT=0

-# ── Grok web TLS transport (Chrome-fingerprinted client) ──
-# Used by: open-sse/services/grokTlsClient.ts — native wreq-js request timeout
-# plus the absolute JS hard-deadline grace.
+# ── Grok web TLS sidecar (Chrome-fingerprinted client) ──
+# Used by: open-sse/services/grokTlsClient.ts — wire-level timeout for the
+# bogdanfinn/tls-client koffi binding and the JS-side grace window layered on
+# top of it when the native library is wedged.
 # OMNIROUTE_GROK_TLS_TIMEOUT_MS=60000
 # OMNIROUTE_GROK_TLS_GRACE_MS=10000

-# ── Notion web TLS transport (Chrome-fingerprinted client) ──
-# Used by: open-sse/services/notionTlsClient.ts — native wreq-js request timeout
-# plus the absolute JS hard-deadline grace. The notion-web executor raises the
-# native timeout per request to 180000 for long generations.
+# ── Notion web TLS sidecar (Chrome-fingerprinted client) ──
+# Used by: open-sse/services/notionTlsClient.ts — wire-level timeout for the
+# bogdanfinn/tls-client koffi binding and the JS-side grace window layered on
+# top of it when the native library is wedged. The notion-web executor raises
+# the wire timeout per-request to 180000 for long generations.
 # OMNIROUTE_NOTION_TLS_TIMEOUT_MS=30000
 # OMNIROUTE_NOTION_TLS_GRACE_MS=10000

@@ -1546,15 +1545,6 @@ CURSOR_USER_AGENT="Cursor/3.4"
 # request into the browser-backed path.
 # OMNIROUTE_BROWSER_POOL=on
 # WEB_COOKIE_USE_BROWSER=0
-# Obscura (https://github.com/h4ckf0r0day/obscura) is the primary headless
-# engine: a lightweight CDP server the pool and cloudflare-playground connect
-# to before falling back to Chromium. Unset OBSCURA_BIN to auto-detect from
-# PATH; set OBSCURA_CDP_ENDPOINT to reuse an already-running Obscura instead
-# of spawning one; set OBSCURA_PORT to pin the spawned serve port.
-# Used by: open-sse/services/obscura.ts
-# OBSCURA_BIN=
-# OBSCURA_CDP_ENDPOINT=
-# OBSCURA_PORT=

 # ── Kimi Web (international kimi.ai Connect-RPC) ──
 # Used by: open-sse/executors/kimi-web.ts. Override the base/chat URLs only if
@@ -1798,7 +1788,6 @@ APP_LOG_TO_FILE=true

 # Custom directory for CLI plugin discovery (omniroute-cmd-* packages).
 # Default: ~/.omniroute/plugins/  Override in dev/CI to point at a local plugin tree.
-# CLI-only: the server-side plugin scanner is pointed by OMNIROUTE_PLUGINS_DIR (section 2).
 # OMNIROUTE_PLUGIN_PATH=

 # ── Prompt cache (system prompt deduplication) ──
@@ -1863,7 +1852,7 @@ APP_LOG_TO_FILE=true
 # short alias prefix and the canonical provider prefix for each model (cc/claude-sonnet-4-6
 # AND claude/claude-sonnet-4-6) so client configs that hardcoded either form keep working —
 # which roughly doubles the catalog. "alias" emits one id per model; "canonical" emits only
-# the full provider-id prefix (providers whose alias is already canonical keep their one id).
+# the full provider-id prefix (and drops providers whose alias is already canonical).
 # A client can override per request with GET /v1/models?prefix=alias instead.
 # Also configurable from Dashboard > Settings > Feature Flags.
 # Used by: src/shared/constants/featureFlagDefinitions.ts, src/app/api/v1/models/catalog.ts
@@ -1933,11 +1922,6 @@ APP_LOG_TO_FILE=true
 # Default: true
 # MODEL_CATALOG_INCLUDE_NAMES=true

-# Cold-path wait bound for a coalesced GET /v1/models catalog rebuild (#12627).
-# Used by: src/app/api/v1/models/catalogCache.ts
-# Default: 8000 (8 seconds). On timeout, a last-good 200 is served when available.
-# CATALOG_BUILD_TIMEOUT_MS=8000
-
 # ── NanoBanana (Image Generation) ──
 # Polling config for async image generation jobs.
 # Used by: open-sse/handlers/imageGeneration.ts
@@ -1957,6 +1941,12 @@ APP_LOG_TO_FILE=true
 # Base backoff after a transient 408 response (ms); five attempts maximum.
 # ADOBE_FIREFLY_SUBMIT_BASE_DELAY_MS=8000

+# ── Microsoft Designer Web (Image Generation) ──
+# Polling config for the microsoft-designer-web submit-then-poll image job.
+# Used by: open-sse/handlers/imageGeneration/providers/designerWeb.ts
+# DESIGNER_WEB_POLL_TIMEOUT_MS=60000   # Max wait for job completion (default: 60s)
+# DESIGNER_WEB_POLL_INTERVAL_MS=2000   # Poll frequency (default: 2s)
+
 # ── Adobe Firefly (Image Upscale) ──
 # Base delay (ms) for the submit-retry exponential backoff when Adobe Firefly's
 # upscale job submission is rate-limited. Used by:
@@ -2072,8 +2062,6 @@ APP_LOG_TO_FILE=true
 # CLIPROXYAPI_HOST=127.0.0.1
 # CLIPROXYAPI_PORT=5544
 # CLIPROXYAPI_CONFIG_DIR=~/.cli-proxy-api
-# Data-plane key fallback; the cliproxyapi_api_key setting takes precedence.
-# CLIPROXYAPI_API_KEY=
 # Management key for an externally managed instance. Embedded instances use
 # OmniRoute's encrypted service key.
 # CLIPROXYAPI_MANAGEMENT_KEY=
@@ -2177,12 +2165,6 @@ APP_LOG_TO_FILE=true
 # Used by: open-sse/services/rateLimitManager.ts
 # RATE_LIMIT_MAX_WAIT_MS=15000

-# Limiter-managed execution backstop (Bottleneck `expiration`): bounds a job's
-# post-dispatch execution, never queue wait. Must stay ABOVE upstream
-# fetch-start timeouts on non-incremental gateways. Default: 600000 (10 min)
-# Used by: open-sse/services/rateLimitManager.ts
-# RATE_LIMIT_EXECUTION_MAX_WAIT_MS=600000
-
 # Rate limit queue admission cap: reject with 429 queue_full once this many requests
 # are already queued (0 = disabled/unbounded, the default). Used by: open-sse/services/rateLimitManager.ts
 # RATE_LIMIT_MAX_QUEUE_DEPTH=0
@@ -2306,16 +2288,6 @@ APP_LOG_TO_FILE=true
 # Cursor image-generation wall clock (ms). Default: 210000.
 # CURSOR_IMG_TIMEOUT_MS=210000

-# UC (uncensored.com) image-generation result-poll cadence + wall clock (ms).
-# Used by: open-sse/handlers/imageGeneration/providers/ucImage.ts. Defaults: 2000 / 60000.
-# UC_IMAGE_POLL_INTERVAL_MS=2000
-# UC_IMAGE_POLL_TIMEOUT_MS=60000
-
-# UC (uncensored.com) video-generation result-poll cadence + wall clock (ms).
-# Used by: open-sse/handlers/videoGeneration/providers/ucVideo.ts. Defaults: 3000 / 300000.
-# UC_VIDEO_POLL_INTERVAL_MS=3000
-# UC_VIDEO_POLL_TIMEOUT_MS=300000
-
 # Shared-seat concurrency gate for Cursor image jobs. Default: 2.
 # CURSOR_IMG_MAX_CONCURRENT=2

@@ -2447,9 +2419,6 @@ APP_LOG_TO_FILE=true
 # Bundled Codeium/language-server extension_version, distinct from Desktop.
 # Must use x.y.z format; invalid/unset values use the bundled default 1.48.2.
 # DEVIN_DESKTOP_EXTENSION_VERSION=1.48.2
-# Optional override for the Codeium seat-management API used by Devin CLI quota.
-# Used by: open-sse/services/usage/devinCli.ts. Default: https://server.codeium.com
-# DEVIN_SEAT_API_URL=https://server.codeium.com

 # ── Command Code (custom CLI) callback ──
 # Local port used for OAuth-style callbacks from the Command Code CLI helper.
@@ -2519,6 +2488,14 @@ APP_LOG_TO_FILE=true
 # When enabled, the node authenticates with the API key stored on its connection.
 # AUDIO_REMOTE_PROVIDER_NODES=false

+# ── 1Proxy egress pool ──
+# Used by: src/lib/oneproxySync.ts — fetches proxy nodes from the OmniRoute
+# CrofAI 1Proxy service. Disable, override URL, or tune the import quality.
+# ONEPROXY_ENABLED=true
+# ONEPROXY_API_URL=https://1proxy-api.aitradepulse.com
+# ONEPROXY_MAX_PROXIES=500
+# ONEPROXY_MIN_QUALITY_THRESHOLD=50
+
 # ── Free Proxy Pool (auto-sync scheduler) ──
 # Background refresh of the free-proxy pool. Opt-in, OFF by default (parallels
 # Hard Rule #20's default-off posture for data-mutating background features).
@@ -2589,6 +2566,11 @@ APP_LOG_TO_FILE=true
 # Used by: src/lib/jobs/backupScheduleJob.ts
 # OMNIROUTE_BACKUP_SCHEDULE_JOB_INTERVAL_MS=30000

+# ── TLS sidecar override ──
+# Used by: open-sse/services/chatgptTlsClient.ts tests. Production deployments
+# should leave this unset; the sidecar is auto-managed.
+# OMNIROUTE_TLS_PROXY_URL=
+
 # ── Skills sandbox (experimental) ──
 # Used by: src/lib/skills/builtins.ts. All values support comma lists where
 # noted in the source.
@@ -2937,14 +2919,6 @@ QUOTA_STORE_DRIVER=sqlite
 # PROMPTQL_TOKEN_REFRESH_URL=https://auth.pro.ql.app/ddn/project/token
 # PROMPTQL_POLL_TIMEOUT_MS=180000

-# ─────────────────────────────────────────────────────────────────────────────
-# Kilo Code usage quotas (src/shared/constants/providers/kilocode.ts)
-# Personal USD balance and Kilo Pass usage lookup. Optional — the default
-# points at the public Kilo API; override only for a relay/test fixture.
-# Authentication uses the connection's existing OAuth access token.
-# Used by: open-sse/services/usage/kilocode.ts
-# ─────────────────────────────────────────────────────────────────────────────
-# KILO_API_URL=https://api.kilo.ai
 # ─────────────────────────────────────────────────────────────────────────────
 # HyperAgent web provider (Unofficial/Experimental — src/shared/constants/providers/web-cookie.ts)
 # Reverse-engineered session bridge for hyperagent.com. Optional — defaults
@@ -2964,12 +2938,7 @@ QUOTA_STORE_DRIVER=sqlite
 # CHATGPT_WEB_CODEX_CDP_URL=http://chatgpt-web-codex-browser:9223
 # CHATGPT_WEB_CODEX_TUNNEL_ID=tunnel_0123456789abcdef0123456789abcdef
 # CHATGPT_WEB_CODEX_RUNTIME_KEY=
-# CHATGPT_WEB_CODEX_CONNECTOR_NAME=OmniRoute Codex v2
-# CODEX_CHATGPT_WEB_HOME=/var/lib/omniroute/chatgpt-web-codex
-# CODEX_CHATGPT_WEB_BROWSER_DIAGNOSTICS=0
-# CODEX_CHATGPT_WEB_LAUNCHER=/absolute/path/to/codex-chatgpt-web
-# CODEX_CHATGPT_WEB_BUN=/absolute/path/to/bun
-# CODEX_WEB_GPT_BUN=/absolute/path/to/bun
+# CHATGPT_WEB_CODEX_CONNECTOR_NAME=OmniRoute Codex

 # ─────────────────────────────────────────────────────────────────────────────
 # Browser-login VNC sessions (optional — src/lib/vncSession/manifest.ts)
@@ -3104,11 +3073,3 @@ QUOTA_STORE_DRIVER=sqlite
 # corpus-aware retrieval. Higher values keep more index entries hot.
 # Used by: src/lib/localCorpus/configured.ts
 # OMNIROUTE_CORPUS_CACHE_SIZE=5
-
-# Service-worker cache-busting id for the PWA shell (#11779). NEXT_PUBLIC_SW_BUILD_ID is
-# derived at build time from OMNIROUTE_SW_BUILD_ID, then SOURCE_VERSION (set by some PaaS
-# builders), then the git SHA — override only when the build cannot see git. Used by:
-# next.config.mjs, scripts/build/assembleStandalone.mjs, src/shared/components/PwaRegister.tsx.
-#OMNIROUTE_SW_BUILD_ID=2026-08-28T12-00-00
-#SOURCE_VERSION=abcdef0123456789
-#NEXT_PUBLIC_SW_BUILD_ID=abcdef0123456789
--- a/.github/actions/npm-ci-retry/action.yml
+++ b/.github/actions/npm-ci-retry/action.yml
@@ -1,45 +1,9 @@
 name: npm ci with retry
-description: >-
-  Install dependencies. Restores node_modules from the Actions cache when the exact
-  lockfile / runner / Node version / postinstall inputs match; otherwise runs npm ci
-  with retries for transient registry/network failures and saves the tree for the
-  next run.
-inputs:
-  cache:
-    description: Set to "false" to skip the node_modules cache and always run npm ci.
-    required: false
-    default: "true"
+description: Run npm ci with retries for transient registry/network failures.
 runs:
  using: composite
  steps:
-    - name: Resolve Node version for the cache key
-      id: node
-      shell: bash
-      run: echo "version=$(node --version)" >> "$GITHUB_OUTPUT"
-
-    # #8084 D3 (plan 3.8.51 task 5): every job used to pay ~80-90 s of `npm ci` even
-    # with setup-node's npm tarball cache warm — 36 jobs per ci.yml run, ~55 min of
-    # runner time per run just installing. A node_modules cache keyed on EVERYTHING
-    # that shapes the tree lets a hit skip the install entirely.
-    #
-    # No restore-keys on purpose (same rule as the ESLint cache, #11600): a partial
-    # tree from another lockfile / Node / postinstall script is exactly the kind of
-    # silent drift a lockfile-pinned CI must never inherit. Exact key or a full npm ci.
-    #
-    # postinstall (scripts/build/postinstall.mjs + helpers) only mutates node_modules
-    # on a plain install — its dist/ branch is gated on dist/ existing, which never
-    # holds at install time in CI — so the cached tree already carries its effects.
-    - name: Restore node_modules
-      id: node-modules
-      if: inputs.cache == 'true'
-      uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
-      with:
-        path: node_modules
-        key: node-modules-${{ runner.os }}-${{ runner.arch }}-${{ steps.node.outputs.version }}-${{ hashFiles('package-lock.json', '.npmrc', 'scripts/build/postinstall.mjs', 'scripts/build/postinstallSupport.mjs', 'scripts/build/colocateOptionals.mjs', 'scripts/build/wreqJsNative.mjs', 'scripts/build/fixPlaywrightAndroid.mjs', 'scripts/build/native-binary-compat.mjs') }}
-
-    - name: npm ci (with retry)
-      if: steps.node-modules.outputs.cache-hit != 'true'
-      shell: bash
+    - shell: bash
      run: |
        set -euo pipefail

@@ -51,8 +15,7 @@ runs:
            echo "npm ci attempt $attempt/$max_attempts after transient failure"
          fi

-          # --no-audit: `audit:deps` is its own gate; the inline audit only adds latency.
-          if npm ci --no-audit --no-fund; then
+          if npm ci; then
            exit 0
          fi

@@ -64,8 +27,3 @@ runs:
          sleep "$delay_seconds"
          delay_seconds=$((delay_seconds * 2))
        done
-
-    - name: node_modules restored from cache
-      if: steps.node-modules.outputs.cache-hit == 'true'
-      shell: bash
-      run: echo "node_modules restored from cache (key hit) — npm ci skipped"
--- a/.github/dependabot.yml
+++ b/.github/dependabot.yml
@@ -58,22 +58,6 @@ updates:
      # on the VPS — so keep auto-bumps frozen (no update-types = ignore every version).
      # Migrate it intentionally, not via dependabot (#4050).
      - dependency-name: "@huggingface/transformers"
-      # onnxruntime-node is the OTHER HALF of the @huggingface/transformers pair frozen
-      # above: the hoisted copy must equal the exact version transformers pins, or npm
-      # nests a second ABI-incompatible native copy (contract test
-      # tests/unit/onnxruntime-single-copy.test.ts, pair established in #9962). A solo
-      # bump can never be correct — it only ever moves together with transformers, in
-      # the same deliberate migration PR. Freezing it keeps the production group PRs
-      # (e.g. #12219) from being born red on the pair contract.
-      - dependency-name: "onnxruntime-node"
-      # eslint-plugin-react-hooks is pinned to 7.0.1 by a contract test
-      # (tests/unit/eslint-react-hooks-version-pinned.test.ts) until the 7.1.1 rule set
-      # is adopted deliberately — that adoption needs a full cold lint run and its own
-      # PR (the #12146 react-hooks migration finished on 2026-09-01, so the path is
-      # open; the bump still must not ride a dependabot group, where it reds the
-      # development group PRs, e.g. #12220). Remove this ignore in the adoption PR
-      # together with the pin test.
-      - dependency-name: "eslint-plugin-react-hooks"

  - package-ecosystem: "github-actions"
    directory: "/"
--- a/.github/workflows/api-route-typecheck.yml
+++ b/.github/workflows/api-route-typecheck.yml
@@ -1,36 +0,0 @@
-name: API Route Typecheck
-
-on:
-  pull_request:
-    branches:
-      - main
-      - "release/**"
-    types: [opened, synchronize, reopened, ready_for_review]
-  push:
-    branches: [main]
-  workflow_dispatch:
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
-
-permissions:
-  contents: read
-
-jobs:
-  api-typecheck:
-    name: API Route Typecheck
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
-        with:
-          persist-credentials: false
-      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
-        with:
-          node-version: "24"
-          cache: npm
-      - uses: ./.github/actions/npm-ci-retry
-      - name: Reject new API-route TypeScript diagnostics
-        run: node scripts/check/check-api-typecheck.mjs
-      - name: API typecheck gate unit tests
-        run: node --import tsx/esm --test tests/unit/build/check-api-typecheck.test.ts
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -1,16 +1,9 @@
 name: Build App

-# Manual-only since #11946. The hosted 7 GB runner can no longer build this tree — 19 of
-# the last 30 runs died with "The runner has received a shutdown signal" (VM out of
-# memory) ~8 min into `next build`, release/v3.8.51 itself included, even with the 10 GB
-# swapfile below. Triggered on `push: branches: ["**"]` it painted every branch and every
-# PR red while producing an artefact nothing downloads. The bundle is validated where a
-# build actually fits:
-#   - main:       ci.yml `Build` (self-hosted omni-build pool) on every merge
-#   - release/**: nightly-release-green.yml (same pool, continuous)
-# Dispatch this workflow by hand when a hosted build artefact is genuinely needed.
 on:
  workflow_dispatch:
+  push:
+    branches: ["**"]

 permissions:
  contents: read
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -93,7 +93,6 @@ jobs:
    steps:
      - uses: actions/checkout@v7
        with:
-          fetch-depth: 0
          persist-credentials: false
      - uses: actions/setup-node@v7
        with:
@@ -109,11 +108,8 @@ jobs:
            .eslintcache
            .eslintcache-complexity
          key: eslint-${{ runner.os }}-${{ hashFiles('eslint.config.mjs', 'eslint.complexity-ratchets.config.mjs', 'config/quality/eslint-suppressions.json', 'package-lock.json') }}
-          # No restore-keys fallback on purpose (#11600, P-II.1 of the v3.8.50 postmortem): a
-          # cache built under a different suppressions file / lint config / lockfile reports
-          # stale per-file verdicts, which is exactly how 215 pre-existing errors stayed
-          # invisible for a whole cycle. Exact key or a cold full lint (~13 min) — never a
-          # partial cache from another configuration.
+          restore-keys: |
+            eslint-${{ runner.os }}-
      # Single ESLint inventory (JSON) — quality-gate reuses the artifact instead of
      # a second cold full-tree pass for eslintWarnings ratchet counts.
      - name: ESLint (JSON report)
@@ -131,7 +127,6 @@ jobs:
      - run: npm run check:any-budget:t11
      - run: npm run check:provider-consistency
      - run: npm run check:model-lifecycle
-      - run: npm run check:provider-asset-provenance
      - run: npm run check:fetch-targets
      - run: npm run check:deps
      - run: npm run check:file-size
@@ -212,11 +207,8 @@ jobs:
            .eslintcache
            .eslintcache-complexity
          key: eslint-${{ runner.os }}-${{ hashFiles('eslint.config.mjs', 'eslint.complexity-ratchets.config.mjs', 'config/quality/eslint-suppressions.json', 'package-lock.json') }}
-          # No restore-keys fallback on purpose (#11600, P-II.1 of the v3.8.50 postmortem): a
-          # cache built under a different suppressions file / lint config / lockfile reports
-          # stale per-file verdicts, which is exactly how 215 pre-existing errors stayed
-          # invisible for a whole cycle. Exact key or a cold full lint (~13 min) — never a
-          # partial cache from another configuration.
+          restore-keys: |
+            eslint-${{ runner.os }}-
      # Coverage mergeada (coverage-summary.json) p/ o ratchet de cobertura.
      # continue-on-error: o artifact pode não existir se a job test-coverage foi
      # SKIPPED (shard flaky). Nesse caso collect-metrics pula coverage.* (ausente sem
@@ -502,11 +494,6 @@ jobs:
          cache: npm
      - uses: ./.github/actions/npm-ci-retry
      - run: node scripts/i18n/check-ui-keys-coverage.mjs --threshold=65
-      # Real-translation ratchet: a leaf copied verbatim from en.json passes key
-      # parity above but is still English to the user (es shipped 55% English).
-      # Advisory in PR-0; flipped to blocking once the backlog is retranslated (PR-4).
-      - name: i18n real-translation ratio (advisory)
-        run: node scripts/i18n/check-translation-ratio.mjs --warn
      # #8463: a rewritten English value used to leave its 39 translations behind
      # silently (googleOAuthWarning shipped wrong copy in 39 locales for months).
      # Key parity above cannot see it — a stale translation counts as covered.
@@ -620,24 +607,13 @@ jobs:
    # Dynamic runner: when the release captain flips the USE_VPS_RUNNER repo var to
    # 'true' (scripts/vps/release-runner-up.sh does it after the self-hosted VM is
    # online), the heavy jobs run on the dedicated 32-core VPS runners (label
-    # omni-build) instead of queueing on the 20-concurrent-job hosted pool.
+    # omni-release) instead of queueing on the 20-concurrent-job hosted pool.
    # Safety: fork PRs NEVER reach the self-hosted runner — the expression falls
    # back to ubuntu-latest unless the PR head repo is this repository (push /
    # dispatch events are own-origin by definition). Any failure path (VM down,
    # var unset/false) also falls back to ubuntu-latest.
-    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && (github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository)) && fromJSON('["self-hosted","omni-build"]') || 'ubuntu-latest' }}
+    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && (github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository)) && fromJSON('["self-hosted","omni-release"]') || 'ubuntu-latest' }}
    needs: changes
-    # The .113 pool runs ONE next-build with room to spare and two at the edge: the
-    # box has 31 GB and a single next-build peaks at 14–16 GB RSS. On 2026-08-28
-    # 13:50Z the kernel OOM-killed main's build while a PR build ran beside it
-    # (five Build jobs had been queued by a burst of PRs). Two lanes: main keeps
-    # its own so a release is never queued behind PR traffic; PR builds serialize
-    # among themselves. docker-publish.yml's amd64 leg joins `heavy-build-main`
-    # so a :next image build waits beside this artefact instead of becoming the
-    # third heavy (#11976). GitHub keeps one running + one pending per group.
-    concurrency:
-      group: heavy-build-${{ github.ref == 'refs/heads/main' && 'main' || 'pr' }}
-      cancel-in-progress: false
    if: ${{ github.event_name != 'pull_request' || (needs.changes.outputs.code == 'true' && github.event.pull_request.draft == false) }}
    steps:
      - uses: actions/checkout@v7
@@ -671,14 +647,14 @@ jobs:
        # Keep standalone/node_modules intact: package/electron jobs consume the
        # Next-traced standalone tree and must not replace it with root node_modules.
        run: |
-          tar -czf "$RUNNER_TEMP/e2e-build.tar.gz" \
+          tar -czf /tmp/e2e-build.tar.gz \
            --exclude='.build/next/cache' \
            .build/next
      - name: Upload Next.js build for downstream jobs
        uses: actions/upload-artifact@v7
        with:
          name: next-build
-          path: ${{ runner.temp }}/e2e-build.tar.gz
+          path: /tmp/e2e-build.tar.gz
          retention-days: 1

  package-artifact:
@@ -701,63 +677,16 @@ jobs:
        uses: actions/download-artifact@v8
        with:
          name: next-build
-          # Workspace-relative on purpose: the matrix below includes windows-latest, whose
-          # default shell is pwsh, where $RUNNER_TEMP is empty (it is $env:RUNNER_TEMP) —
-          # #11896's first cut broke the Electron smoke on exactly that. A relative path
-          # works in bash and pwsh alike; hosted workspaces are ephemeral.
-          path: next-build-artifact
+          path: /tmp/
      - name: Extract Next.js build artifact
        run: |
-          tar -xzf next-build-artifact/e2e-build.tar.gz
+          tar -xzf /tmp/e2e-build.tar.gz
      # build:cli consumes the downloaded .build/next standalone artifact and assembles dist/;
      # it only rebuilds if the downloaded standalone artifact is missing.
      - run: npm run build:cli
      - name: Assert dist/server.js exists
        run: test -f dist/server.js || (echo "dist/server.js missing — build:cli did not assemble correctly" && exit 1)
-      # `build:cli` monta dist/ mas NAO grava dist/BUILD_SHA — so `build:release` faz
-      # isso, chamando write-build-sha.mjs. O guard de proveniencia do #10427, dentro
-      # de check:pack-artifact, rejeita um artefato sem SHA (e rejeita mesmo com
-      # OMNIROUTE_ALLOW_CANARY_BUILD=1: o que nao da para identificar nao da para
-      # vouchear). Sem este passo o par build+validate deste job e estruturalmente
-      # incompativel e falha 100% das vezes.
-      - name: Stamp dist/BUILD_SHA for the provenance guard (#10427)
-        # O SHA TEM de vir do head da PR, nao de `git rev-parse HEAD`. Este workflow
-        # roda em `pull_request`, entao o checkout e o MERGE COMMIT efemero que o
-        # GitHub cria — um commit que nao existe em branch nenhuma e portanto nunca e
-        # ancestral da release. O guard de proveniencia (#10427) rejeita exatamente
-        # isso, e com razao: um artefato carimbado com o merge commit nao pode ser
-        # rastreado ate codigo que passou pelos gates.
-        env:
-          OMNIROUTE_BUILD_SHA: ${{ github.event.pull_request.head.sha || github.sha }}
-        run: |
-          export OMNIROUTE_BUILD_SHA="${OMNIROUTE_BUILD_SHA:0:7}"
-          node scripts/build/write-build-sha.mjs
-      # O guard de proveniencia checa ancestralidade contra `origin/main` por padrao.
-      # Esse e o ref certo na PUBLICACAO (npm-publish.yml roda em main), mas em
-      # `pull_request` e estruturalmente impossivel: enquanto a PR esta aberta o head
-      # dela NUNCA e ancestral de main — e o checkout raso nem traz `origin/main` para
-      # o grafo local, entao a sonda responde `false` de qualquer jeito. Resultado: o
-      # gate falhava 100% das vezes em PR. Pre-merge o unico invariante checavel e "o
-      # stamp corresponde a branch sob teste", entao apontamos o ref para o head da PR.
-      # Usamos `refs/pull/<N>/head` e nao `head.ref` porque aquele existe no PROPRIO
-      # origin mesmo quando a PR vem de um fork; `head.ref` so existe no repo do autor.
-      - name: Resolve the provenance ref for the pack gate (#10427)
-        id: provenance-ref
-        env:
-          PR_NUMBER: ${{ github.event.pull_request.number }}
-        run: |
-          if [ -n "$PR_NUMBER" ]; then
-            git fetch --no-tags --depth=50 origin \
-              "+refs/pull/$PR_NUMBER/head:refs/remotes/origin/pr-head"
-            echo "ref=origin/pr-head" >> "$GITHUB_OUTPUT"
-          else
-            git fetch --no-tags --depth=50 origin \
-              "+refs/heads/$GITHUB_REF_NAME:refs/remotes/origin/$GITHUB_REF_NAME"
-            echo "ref=origin/$GITHUB_REF_NAME" >> "$GITHUB_OUTPUT"
-          fi
      - run: npm run check:pack-artifact
-        env:
-          OMNIROUTE_RELEASE_REF: ${{ steps.provenance-ref.outputs.ref }}
      # WS1.2 (#7065 class): pack the real tarball, install it into a clean prefix and
      # BOOT it to a healthy /api/monitoring/health — the gate that structure checks
      # cannot provide (3 releases shipped boot-crashing tarballs with green lists).
@@ -796,14 +725,10 @@ jobs:
        uses: actions/download-artifact@v8
        with:
          name: next-build
-          # Workspace-relative on purpose: the matrix below includes windows-latest, whose
-          # default shell is pwsh, where $RUNNER_TEMP is empty (it is $env:RUNNER_TEMP) —
-          # #11896's first cut broke the Electron smoke on exactly that. A relative path
-          # works in bash and pwsh alike; hosted workspaces are ephemeral.
-          path: next-build-artifact
+          path: /tmp/
      - name: Extract Next.js build artifact
        run: |
-          tar -xzf next-build-artifact/e2e-build.tar.gz
+          tar -xzf /tmp/e2e-build.tar.gz
      - name: Install Electron dependencies
        working-directory: electron
        run: npm install --no-audit --no-fund
@@ -868,23 +793,9 @@ jobs:
      # D3 (plano mestre): a coverage é coletada NESTE mesmo run (c8/NODE_V8_COVERAGE propaga
      # aos filhos através do npm) — elimina a matrix Coverage Shard ×8, que re-executava a
      # suíte inteira só para medir o gate. Padrão usado pelo CI do próprio nodejs/node.
-      # Heap: os shards rodam sob instrumentacao de cobertura do V8, que retem muito
-      # mais memoria que a suite crua. Com o teto antigo de 4096 MB os shards passaram
-      # a abortar com SIGABRT (exit 134, "Ineffective mark-compacts near heap limit")
-      # ao redor de 4086 MB conforme o catalogo de providers cresceu no ciclo v3.8.50 —
-      # todos os testes passavam e o processo morria no fim, o que le como falha de
-      # teste sem ser. O teto vive em `test:unit:ci:shard` (package.json) e agora
-      # acompanha os 8192 MB ja usados pelas variantes nao-shardadas; os runners
-      # GitHub-hosted tem 16 GB.
      - name: Unit tests (shard ${{ matrix.shard }}/8) with V8 coverage
        env:
          TEST_SHARD: ${{ matrix.shard }}/8
-          # NODE_OPTIONS (nao so o flag em test:unit:ci:shard) porque quem estoura o
-          # heap e o processo `c8` que embrulha a suite — ele agrega ~577 MB de JSON
-          # de cobertura bruta. Subir o teto so no node filho deixa o pai no default
-          # do V8 (~4 GB) e o OOM continua igual, em ~4083 MB. Mesmo padrao ja usado
-          # pelo job de merge de cobertura mais abaixo.
-          NODE_OPTIONS: --max-old-space-size=8192
        run: |
          rm -rf coverage-shard coverage-shard-report
          npx c8 \
@@ -990,11 +901,7 @@ jobs:
    # 10min was sized before #7114 added the lcov reporter (Codecov/Sonar need it);
    # merging 8 shard JSONs + text+json+lcov now takes ~10-12min — three consecutive
    # release-tip runs died at exactly 10m as job-timeout "cancelled" (2026-07-15/16).
-    # 30, not 20 (2026-08-29): the informational Codecov upload below hung for the rest of
-    # the budget on two consecutive main runs (33207760653, 33215115341); the job ended
-    # `cancelled` and dragged the whole run's conclusion to `cancelled` although every
-    # blocking job was green. The upload step now has its own ceiling; this is headroom.
-    timeout-minutes: 30
+    timeout-minutes: 20
    needs: test-unit
    if: ${{ !cancelled() && needs.test-unit.result == 'success' && !contains(github.event.pull_request.labels.*.name, 'hotfix') }}
    env:
@@ -1073,10 +980,6 @@ jobs:
      # (if-no-files-found: warn) — Sonar consumes the same file.
      - name: Upload coverage to Codecov (informational)
        if: always()
-        # Informational means informational: its own ceiling and continue-on-error, so a
-        # stalled upload can neither eat the job's budget nor turn a green job cancelled.
-        timeout-minutes: 5
-        continue-on-error: true
        uses: codecov/codecov-action@fb8b3582c8e4def4969c97caa2f19720cb33a72f # v7.0.0
        with:
          files: coverage/lcov.info
@@ -1271,14 +1174,10 @@ jobs:
        uses: actions/download-artifact@v8
        with:
          name: next-build
-          # Workspace-relative on purpose: the matrix below includes windows-latest, whose
-          # default shell is pwsh, where $RUNNER_TEMP is empty (it is $env:RUNNER_TEMP) —
-          # #11896's first cut broke the Electron smoke on exactly that. A relative path
-          # works in bash and pwsh alike; hosted workspaces are ephemeral.
-          path: next-build-artifact
+          path: /tmp/
      - name: Extract Next.js build artifact
        run: |
-          tar -xzf next-build-artifact/e2e-build.tar.gz
+          tar -xzf /tmp/e2e-build.tar.gz
      # WS4.1: duration-balanced shards (LPT over config/quality/e2e-timings.json).
      # Measured skew of plain --shard was 14× (24m47s vs 1m47s) — E2E was the CI
      # critical path. The balancer self-verifies completeness and exits non-zero on
--- a/.github/workflows/codeql.yml
+++ b/.github/workflows/codeql.yml
@@ -22,10 +22,10 @@ jobs:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
-      - uses: github/codeql-action/init@cdf488f595d80d6e07e03d4674febd5ab45fa938 # v4.37.9
+      - uses: github/codeql-action/init@ff2f1c621b7f889edc0d3c761ac2e6a3f8cdb0dd # v4.37.7
        with:
          languages: javascript-typescript
          queries: security-extended
-      - uses: github/codeql-action/analyze@cdf488f595d80d6e07e03d4674febd5ab45fa938 # v4.37.9
+      - uses: github/codeql-action/analyze@ff2f1c621b7f889edc0d3c761ac2e6a3f8cdb0dd # v4.37.7
        with:
          category: "/language:javascript-typescript"
--- a/.github/workflows/dast-smoke.yml
+++ b/.github/workflows/dast-smoke.yml
@@ -1,15 +1,7 @@
 name: DAST smoke (PR)
-# PRs into main only since #11946. The job's "Build CLI bundle" step is a backend-only
-# `next build`; on the hosted 7 GB runner it fits main's tree (~5.5 min) but dies on
-# release/v3.8.51 (VM shutdown ~7 min in, before the server even starts), and because the
-# job is continue-on-error the result was a permanently red advisory check on every
-# release PR — noise, not signal. DAST coverage for release/** lives on the nightly rail
-# (nightly-schemathesis.yml, nightly-llm-security.yml); dispatch this workflow by hand
-# to smoke a release branch on demand.
 on:
-  workflow_dispatch:
  pull_request:
-    branches: ["main"]
+    branches: ["main", "release/**"]
    # Runner-cost guard (#8084): the CLI-bundle build alone is 6-11min; a docs-only PR
    # cannot change DAST behavior, so skip the whole workflow for pure docs/markdown
    # changes. Any code path in the diff still runs the full smoke.
--- a/.github/workflows/docker-publish.yml
+++ b/.github/workflows/docker-publish.yml
@@ -26,14 +26,6 @@ on:
        type: boolean
        default: false

-# One publish per ref. A merge storm used to fan out 8 concurrent hosted builds,
-# every one OOM-killing `npm run build` inside BuildKit (#11976). The :next
-# channel only needs the newest SHA; cancel-in-progress is the same pattern as
-# quality.yml / nightly-release-green.
-concurrency:
-  group: docker-publish-${{ github.ref }}
-  cancel-in-progress: true
-
 # Least-privilege default: read-only at the top level; the build and merge jobs that
 # push to GHCR grant packages: write themselves (Scorecard TokenPermissions).
 permissions:
@@ -126,23 +118,7 @@ jobs:
    name: Build Docker (${{ matrix.platform }})
    needs: prepare
    if: needs.prepare.outputs.skip != 'true'
-    # amd64: the .113 omni-build pool (31 GB / 32 cores, ONE listener since
-    # #12048). Hosted ubuntu-24.04 is ~7 GB and dies ResourceExhausted (#11976).
-    # Falls back to hosted when USE_VPS_RUNNER is off. arm64: no ARM box — stay
-    # on GitHub's ubuntu-24.04-arm.
-    # Webpack on BOTH arches: Turbopack on omniroute-113-6 hit
-    # TurbopackInternalError "there must be a path to a root" after 26 min
-    # (run 33253576569). The same tree's arm64 webpack build on hosted ARM
-    # succeeded (run 33264823398). Dockerfile already documents webpack as the
-    # Docker escape hatch (OMNIROUTE_USE_TURBOPACK=0).
-    runs-on: ${{ matrix.arch == 'amd64' && (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-build"]') || 'ubuntu-24.04') || 'ubuntu-24.04-arm' }}
-    # Share the 1-slot omni-build ceiling (#12048) with ci.yml `Build` /
-    # npm-publish. Same group as main's Build so a :next publish waits beside
-    # the artefact instead of sitting next to it. arm64 is hosted — its own
-    # group, cancelled by the workflow-level concurrency.
-    concurrency:
-      group: ${{ matrix.arch == 'amd64' && 'heavy-build-main' || format('docker-publish-arm-{0}', github.ref) }}
-      cancel-in-progress: ${{ matrix.arch != 'amd64' }}
+    runs-on: ${{ matrix.runner }}
    permissions:
      contents: read
      packages: write
@@ -151,8 +127,10 @@ jobs:
      matrix:
        include:
          - platform: linux/amd64
+            runner: ubuntu-24.04
            arch: amd64
          - platform: linux/arm64
+            runner: ubuntu-24.04-arm
            arch: arm64
    env:
      IMAGE_NAME: diegosouzapw/omniroute
@@ -165,9 +143,6 @@ jobs:
          ref: ${{ github.event_name == 'workflow_dispatch' && format('refs/tags/v{0}', inputs.version) || '' }}
          fetch-depth: 0

-      - name: Assert Docker Engine
-        run: docker info
-
      - name: Set up Docker Buildx
        uses: docker/setup-buildx-action@v4

@@ -191,14 +166,12 @@ jobs:
          context: .
          target: runner-base
          platforms: ${{ matrix.platform }}
-          build-args: |
-            OMNIROUTE_USE_TURBOPACK=0
          outputs: type=image,push-by-digest=true,name-canonical=true,push=true
          tags: |
            ${{ env.IMAGE_NAME }}
            ${{ env.GHCR_IMAGE_NAME }}
          cache-from: type=gha,scope=docker-${{ matrix.arch }}
-          cache-to: type=gha,scope=docker-${{ matrix.arch }},mode=max,ignore-error=true
+          cache-to: type=gha,scope=docker-${{ matrix.arch }},mode=max
          no-cache: false
        env:
          DOCKER_BUILDKIT_INLINE_CACHE: 1
@@ -210,68 +183,48 @@ jobs:
          context: .
          target: runner-web
          platforms: ${{ matrix.platform }}
-          build-args: |
-            OMNIROUTE_USE_TURBOPACK=0
          outputs: type=image,push-by-digest=true,name-canonical=true,push=true
          tags: |
            ${{ env.IMAGE_NAME }}
            ${{ env.GHCR_IMAGE_NAME }}
          cache-from: type=gha,scope=docker-web-${{ matrix.arch }}
-          cache-to: type=gha,scope=docker-web-${{ matrix.arch }},mode=max,ignore-error=true
+          cache-to: type=gha,scope=docker-web-${{ matrix.arch }},mode=max
          no-cache: false
        env:
          DOCKER_BUILDKIT_INLINE_CACHE: 1

      - name: Build and push BUN base platform image by digest
        id: build-bun-base
-        # Bun is a best-effort compatibility target, not a supported runtime
-        # (AGENTS.md -> Environment). Its `bun run build` has been OOM-killing on
-        # both arches; letting that sink the whole publish means the SUPPORTED
-        # runner-base / runner-web images never reach the registry either. The
-        # image is still built and pushed whenever it succeeds — only its power to
-        # block the release is removed.
-        continue-on-error: true
        uses: docker/build-push-action@53b7df96c91f9c12dcc8a07bcb9ccacbed38856a # v7
        with:
          context: .
          file: Dockerfile.bun
          target: runner-base
          platforms: ${{ matrix.platform }}
-          build-args: |
-            OMNIROUTE_USE_TURBOPACK=0
          outputs: type=image,push-by-digest=true,name-canonical=true,push=true
          tags: |
            ${{ env.IMAGE_NAME }}
            ${{ env.GHCR_IMAGE_NAME }}
          cache-from: type=gha,scope=docker-bun-base-${{ matrix.arch }}
-          cache-to: type=gha,scope=docker-bun-base-${{ matrix.arch }},mode=max,ignore-error=true
+          cache-to: type=gha,scope=docker-bun-base-${{ matrix.arch }},mode=max
          no-cache: false
        env:
          DOCKER_BUILDKIT_INLINE_CACHE: 1

      - name: Build and push BUN web platform image by digest
        id: build-bun-web
-        # Bun is a best-effort compatibility target, not a supported runtime
-        # (AGENTS.md -> Environment). Its `bun run build` has been OOM-killing on
-        # both arches; letting that sink the whole publish means the SUPPORTED
-        # runner-base / runner-web images never reach the registry either. The
-        # image is still built and pushed whenever it succeeds — only its power to
-        # block the release is removed.
-        continue-on-error: true
        uses: docker/build-push-action@53b7df96c91f9c12dcc8a07bcb9ccacbed38856a # v7
        with:
          context: .
          file: Dockerfile.bun
          target: runner-web
          platforms: ${{ matrix.platform }}
-          build-args: |
-            OMNIROUTE_USE_TURBOPACK=0
          outputs: type=image,push-by-digest=true,name-canonical=true,push=true
          tags: |
            ${{ env.IMAGE_NAME }}
            ${{ env.GHCR_IMAGE_NAME }}
          cache-from: type=gha,scope=docker-bun-web-${{ matrix.arch }}
-          cache-to: type=gha,scope=docker-bun-web-${{ matrix.arch }},mode=max,ignore-error=true
+          cache-to: type=gha,scope=docker-bun-web-${{ matrix.arch }},mode=max
          no-cache: false
        env:
          DOCKER_BUILDKIT_INLINE_CACHE: 1
@@ -287,15 +240,8 @@ jobs:
          mkdir -p /tmp/digests/base /tmp/digests/web /tmp/digests/bun-base /tmp/digests/bun-web
          touch "/tmp/digests/base/${DIGEST_BASE#sha256:}"
          touch "/tmp/digests/web/${DIGEST_WEB#sha256:}"
-          # Empty when the (non-blocking) bun build produced no image. `if` blocks,
-          # not `[ -n ] && touch`: under `set -e` a failing AND-list aborts the step,
-          # which is precisely the case being handled here.
-          if [ -n "$DIGEST_BUN_BASE" ]; then
-            touch "/tmp/digests/bun-base/${DIGEST_BUN_BASE#sha256:}"
-          fi
-          if [ -n "$DIGEST_BUN_WEB" ]; then
-            touch "/tmp/digests/bun-web/${DIGEST_BUN_WEB#sha256:}"
-          fi
+          touch "/tmp/digests/bun-base/${DIGEST_BUN_BASE#sha256:}"
+          touch "/tmp/digests/bun-web/${DIGEST_BUN_WEB#sha256:}"

      - name: Upload base digests
        uses: actions/upload-artifact@v7
@@ -318,11 +264,7 @@ jobs:
        with:
          name: digests-bun-base-${{ matrix.arch }}
          path: /tmp/digests/bun-base/*
-          # `ignore`, not `error`: the bun build is non-blocking, so an absent
-          # digest is the expected outcome of a failed/skipped bun image — the
-          # manifest step already treats these tags as optional. Leaving `error`
-          # here just relocates the blocker from the manifest to the upload.
-          if-no-files-found: ignore
+          if-no-files-found: error
          retention-days: 1

      - name: Upload bun-web digests
@@ -330,11 +272,7 @@ jobs:
        with:
          name: digests-bun-web-${{ matrix.arch }}
          path: /tmp/digests/bun-web/*
-          # `ignore`, not `error`: the bun build is non-blocking, so an absent
-          # digest is the expected outcome of a failed/skipped bun image — the
-          # manifest step already treats these tags as optional. Leaving `error`
-          # here just relocates the blocker from the manifest to the upload.
-          if-no-files-found: ignore
+          if-no-files-found: error
          retention-days: 1

  merge:
@@ -392,9 +330,6 @@ jobs:
          merge-multiple: true

      - name: Download bun-base digests
-        # Non-blocking: the bun image is best-effort, so its artifact may not
-        # exist at all. The manifest step treats these tags as optional.
-        continue-on-error: true
        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8
        with:
          pattern: digests-bun-base-*
@@ -402,9 +337,6 @@ jobs:
          merge-multiple: true

      - name: Download bun-web digests
-        # Non-blocking: the bun image is best-effort, so its artifact may not
-        # exist at all. The manifest step treats these tags as optional.
-        continue-on-error: true
        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8
        with:
          pattern: digests-bun-web-*
@@ -416,7 +348,7 @@ jobs:
          set -euo pipefail

          create_manifest() {
-            local image="$1" suffix="$2" dir="$3" optional="${4:-}"
+            local image="$1" suffix="$2" dir="$3"
            local tags=(-t "${image}:${VERSION}${suffix}")
            if [ "$PROMOTE_LATEST" = "true" ]; then
              tags+=(-t "${image}:latest${suffix}")
@@ -426,10 +358,6 @@ jobs:
              refs+=("${image}@sha256:$(basename "$digest_file")")
            done < <(find "$dir" -type f | sort)
            if [ "${#refs[@]}" -eq 0 ]; then
-              if [ -n "$optional" ]; then
-                echo "::warning::No image digests in $dir — skipping optional tag ${image}:${VERSION}${suffix}" >&2
-                return 0
-              fi
              echo "No image digests in $dir" >&2
              exit 1
            fi
@@ -438,15 +366,15 @@ jobs:

          create_manifest "${IMAGE_NAME}" "" /tmp/digests/base
          create_manifest "${IMAGE_NAME}" "-web" /tmp/digests/web
-          create_manifest "${IMAGE_NAME}" "-bun" /tmp/digests/bun-base optional
-          create_manifest "${IMAGE_NAME}" "-web-bun" /tmp/digests/bun-web optional
+          create_manifest "${IMAGE_NAME}" "-bun" /tmp/digests/bun-base
+          create_manifest "${IMAGE_NAME}" "-web-bun" /tmp/digests/bun-web

      - name: Create GHCR manifest
        run: |
          set -euo pipefail

          create_manifest() {
-            local image="$1" suffix="$2" dir="$3" optional="${4:-}"
+            local image="$1" suffix="$2" dir="$3"
            local tags=(-t "${image}:${VERSION}${suffix}")
            if [ "$PROMOTE_LATEST" = "true" ]; then
              tags+=(-t "${image}:latest${suffix}")
@@ -456,10 +384,6 @@ jobs:
              refs+=("${image}@sha256:$(basename "$digest_file")")
            done < <(find "$dir" -type f | sort)
            if [ "${#refs[@]}" -eq 0 ]; then
-              if [ -n "$optional" ]; then
-                echo "::warning::No image digests in $dir — skipping optional tag ${image}:${VERSION}${suffix}" >&2
-                return 0
-              fi
              echo "No image digests in $dir" >&2
              exit 1
            fi
@@ -468,8 +392,8 @@ jobs:

          create_manifest "${GHCR_IMAGE_NAME}" "" /tmp/digests/base
          create_manifest "${GHCR_IMAGE_NAME}" "-web" /tmp/digests/web
-          create_manifest "${GHCR_IMAGE_NAME}" "-bun" /tmp/digests/bun-base optional
-          create_manifest "${GHCR_IMAGE_NAME}" "-web-bun" /tmp/digests/bun-web optional
+          create_manifest "${GHCR_IMAGE_NAME}" "-bun" /tmp/digests/bun-base
+          create_manifest "${GHCR_IMAGE_NAME}" "-web-bun" /tmp/digests/bun-web

      - name: Inspect image
        if: needs.prepare.outputs.version != 'main'
@@ -528,14 +452,11 @@ jobs:
          severity: CRITICAL
          ignore-unfixed: true
          exit-code: "1"
-          # Explicit: the advisory scan above already points at it, and the blocking
-          # gate must honour the same accepted-risk list (#12084).
-          trivyignores: .trivyignore

      - name: Upload Trivy SARIF to Security tab
        if: needs.prepare.outputs.version != 'main'
        continue-on-error: true
-        uses: github/codeql-action/upload-sarif@v4.37.9
+        uses: github/codeql-action/upload-sarif@v4.37.7
        with:
          sarif_file: trivy-results.sarif
          category: trivy-image
--- a/.github/workflows/electron-release.yml
+++ b/.github/workflows/electron-release.yml
@@ -4,21 +4,12 @@ on:
  push:
    tags:
      - "v*"
-  # A dispatch builds the ref it is dispatched ON (`gh workflow run … --ref v3.8.50` rebuilds
-  # that tag; `--ref main` builds the repaired line). The ref is deliberately NOT an input:
-  # CodeQL flags an input-controlled checkout next to the npm cache on the default branch as
-  # cache poisoning (actions/cache-poisoning/poisonable-step), and `github.ref` is trusted.
  workflow_dispatch:
    inputs:
      version:
        description: "Release version (e.g., v1.6.8)"
        required: true
        type: string
-      publish_npm:
-        description: "Also run the npm publish leg (turn off when re-attaching desktop assets to a release whose npm package already shipped)"
-        required: false
-        default: true
-        type: boolean

 # Least-privilege default: read-only at the top level; each job grants the writes it
 # needs (build/release upload assets, publish-npm forwards npm provenance / packages
@@ -85,9 +76,6 @@ jobs:
      - uses: actions/checkout@v7
        with:
          persist-credentials: false
-          # workflow_dispatch: build the tag being (re)built, not the dispatching branch. On a
-          # tag push this resolves to the same commit.
-          ref: ${{ needs.validate.outputs.version }}
      - name: Setup Node
        uses: actions/setup-node@v7
        with:
@@ -173,9 +161,6 @@ jobs:
      - uses: actions/checkout@v7
        with:
          persist-credentials: false
-          # workflow_dispatch: build the tag being (re)built, not the dispatching branch. On a
-          # tag push this resolves to the same commit.
-          ref: ${{ needs.validate.outputs.version }}
      - name: Setup Node
        uses: actions/setup-node@v7
        with:
@@ -187,22 +172,6 @@ jobs:
        env:
          NPM_CONFIG_LEGACY_PEER_DEPS: true

-      # The Linux leg produces x64 + arm64 installers from one x64 runner. npm
-      # deliberately installs only host-compatible optional dependencies, so
-      # hydrateNativeDeps cannot source the arm64 fork unless we fetch the exact
-      # package pinned in package-lock before either build path runs.
-      - name: Install Linux arm64 wreq binding for cross-package
-        if: matrix.platform == 'linux'
-        shell: bash
-        run: |
-          npm install --no-save --ignore-scripts --force --legacy-peer-deps \
-            @wreq-js/binding-linux-arm64-gnu@3.2.0
-          git diff --exit-code -- package.json package-lock.json
-          mkdir -p "$RUNNER_TEMP/omniroute-wreq-verify"
-          DATA_DIR="$RUNNER_TEMP/omniroute-wreq-verify" node --import tsx/esm --test \
-            --test-name-pattern='wreq-js 3.2 manifest pins all nine' \
-            tests/unit/wreq-native-manifest.test.ts
-
      - name: Sanitize Windows home directory
        if: runner.os == 'Windows'
        shell: bash
@@ -251,9 +220,9 @@ jobs:
        #          targets, and no unlisted files) byte-for-byte.
        # hydrate: the bundle was built on ubuntu, so install-machine-forked native
        #          optionals (@img/sharp-*, @img/sharp-libvips-*, @ngrok/ngrok-*,
-        #          @wreq-js/binding-*, fsevents) carry linux forks. Replace them with the forks this
+        #          fsevents) carry linux forks. Replace them with the forks this
        #          leg's own `npm ci` resolved, then assert every bundled native
-        #          (better-sqlite3 prebuilds, wreq-js, onnxruntime)
+        #          (koffi triplets, better-sqlite3 prebuilds, wreq-js, onnxruntime)
        #          can service this leg's platform/arch before packaging starts.
        run: |
          node scripts/build/standaloneBundle.mjs restore --archive web-bundle.tar.gz
@@ -378,8 +347,6 @@ jobs:
        with:
          persist-credentials: false
          fetch-depth: 0
-          # Source archives + SBOM come from the tag being released, not the dispatching branch.
-          ref: ${{ needs.validate.outputs.version }}

      # `merge-multiple` is deliberately OFF. It resolves same-name collisions by ARRIVAL
      # ORDER, and the two macOS jobs each emit their own `latest-mac.yml` listing only their
@@ -437,14 +404,7 @@ jobs:
          tag_name: ${{ needs.validate.outputs.version }}
          draft: false
          prerelease: false
-          # NEVER. Phase 3 of the release flow creates the GitHub Release with the curated
-          # notes seconds after pushing the tag, so by the time this step runs (1-2 h of
-          # builds later) the body already exists — and `true` APPENDS GitHub's
-          # auto-generated "What's Changed" block to it (v3.8.48 shipped that way; the
-          # v3.8.50 re-attach dispatch added +1,416 chars to a 121 KB body, run
-          # 33238093090). A curated body sits ~3 KB under the 125,000-char cap, so the
-          # append can also turn this step RED and leave the release with no assets.
-          generate_release_notes: false
+          generate_release_notes: true
          fail_on_unmatched_files: false
          files: |
            release-assets/*.dmg
@@ -502,20 +462,11 @@ jobs:
  publish-npm:
    name: Publish to npm
    needs: [validate, release]
-    # A re-dispatch that only re-attaches desktop assets must not publish the npm package again.
-    if: ${{ github.event_name != 'workflow_dispatch' || inputs.publish_npm }}
    permissions:
      # Must be `write`, not `read`: this job calls the reusable npm-publish.yml whose
      # `publish` job needs `contents: write` (gh release upload — attach the SBOM, #3874).
      # A reusable workflow's job cannot request more permission than the caller grants,
      # so a `read` here makes GitHub reject the run at startup (startup_failure).
-      #
-      # `actions: read` for the same reason: the called `publish` job downloads the next-build
-      # artefact and requests it. v3.8.50 (run 33005490476) died at startup with "The nested
-      # job 'publish' is requesting 'actions: read', but is only allowed 'actions: none'" — and
-      # because `release` lives in this same workflow, the tag shipped with ZERO assets. Keep
-      # this block a superset of every job's permissions in npm-publish.yml.
-      actions: read
      contents: write
      id-token: write # npm provenance (forwarded to the reusable workflow)
      packages: write # publish to npm.pkg.github.com
--- a/.github/workflows/nightly-llm-security.yml
+++ b/.github/workflows/nightly-llm-security.yml
@@ -10,10 +10,7 @@ permissions:
 jobs:
  promptfoo-guard:
    name: promptfoo — injection guard (block mode, no secret)
-    # #11965: this job runs a backend-only `next build`; the hosted 7 GB runner cannot build
-    # release/v3.8.51 (VM shutdown ~7 min in), so it targets the box's light pool (`omni-light`:
-    # two listeners, jobs ≤ ~6 GB). Falls back to hosted when USE_VPS_RUNNER is off.
-    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-light"]')) || 'ubuntu-latest' }}
+    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v7
        with:
@@ -49,10 +46,7 @@ jobs:

  garak:
    name: garak probes (skip without provider secret)
-    # #11965: this job runs a backend-only `next build`; the hosted 7 GB runner cannot build
-    # release/v3.8.51 (VM shutdown ~7 min in), so it targets the box's light pool (`omni-light`:
-    # two listeners, jobs ≤ ~6 GB). Falls back to hosted when USE_VPS_RUNNER is off.
-    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-light"]')) || 'ubuntu-latest' }}
+    runs-on: ubuntu-latest
    # NOTE: the `secrets` context is NOT available in a job-level `if:` — referencing
    # it there makes GitHub reject the file on push (startup_failure on every push).
    # Map the secret into a job-level env and gate each step on a presence check, so
--- a/.github/workflows/nightly-release-green.yml
+++ b/.github/workflows/nightly-release-green.yml
@@ -68,7 +68,7 @@ jobs:
    # this runs on the dedicated VPS runner — clean env (no operator OMNIROUTE_API_KEY,
    # no local noauth CLIs => zero machine-specific false positives) and no contention.
    # Nightly cron normally finds the var false (VM off) and falls back to hosted.
-    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-build"]')) || 'ubuntu-latest' }}
+    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-release"]')) || 'ubuntu-latest' }}
    env:
      JWT_SECRET: ci-nightly-secret-with-sufficient-length-for-validation
      API_KEY_SECRET: ci-nightly-api-key-secret-long
@@ -196,26 +196,6 @@ jobs:
            gh issue create --repo "$GITHUB_REPOSITORY" --title "$TITLE" --label base-red --body-file issue-body.md
          fi

-      - name: Close tracking issue when the branch is green again
-        if: steps.validate.outputs.exit == '0'
-        env:
-          GH_TOKEN: ${{ github.token }}
-          TARGET: ${{ steps.branch.outputs.target }}
-          RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
-        run: |
-          set -euo pipefail
-          # The open/update step above is the UPWARD half of the loop; without this
-          # step a stale "not green" issue outlives the fix and every base-green check
-          # (`AGENTS.md` → "Base-green check") keeps stamping new PRs as base-red inherited.
-          TITLE="🔴 Release branch not green: ${TARGET}"
-          EXISTING=$(gh issue list --repo "$GITHUB_REPOSITORY" --state open \
-            --search "in:title $TITLE" --json number --jq '.[0].number' 2>/dev/null || echo "")
-          if [ -n "$EXISTING" ]; then
-            gh issue close "$EXISTING" --repo "$GITHUB_REPOSITORY" --reason completed \
-              --comment "✅ \`${TARGET}\` is release-green again at \`${GITHUB_SHA:0:9}\` — ${RUN_URL}. Auto-closed by Release-Green (continuous)."
-            echo "Closed issue #$EXISTING"
-          fi
-
      - name: Upload report artifact
        if: always()
        uses: actions/upload-artifact@v7
@@ -237,7 +217,7 @@ jobs:
    # On a push, only run for a push to main — a push to release/* is handled by
    # release-green above. Schedule/dispatch always run (they also sweep main).
    if: ${{ github.event_name != 'push' || github.ref_name == 'main' }}
-    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-build"]')) || 'ubuntu-latest' }}
+    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-release"]')) || 'ubuntu-latest' }}
    env:
      JWT_SECRET: ci-nightly-secret-with-sufficient-length-for-validation
      API_KEY_SECRET: ci-nightly-api-key-secret-long
@@ -314,25 +294,6 @@ jobs:
            gh issue create --repo "$GITHUB_REPOSITORY" --title "$TITLE" --label base-red --body-file issue-body.md
          fi

-      - name: Close tracking issue when the branch is green again
-        if: steps.validate.outputs.exit == '0'
-        env:
-          GH_TOKEN: ${{ github.token }}
-          RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
-        run: |
-          set -euo pipefail
-          # The open/update step above is the UPWARD half of the loop; without this
-          # step a stale "not green" issue outlives the fix and every base-green check
-          # (`AGENTS.md` → "Base-green check") keeps stamping new PRs as base-red inherited.
-          TITLE="🔴 main branch not green"
-          EXISTING=$(gh issue list --repo "$GITHUB_REPOSITORY" --state open \
-            --search "in:title $TITLE" --json number --jq '.[0].number' 2>/dev/null || echo "")
-          if [ -n "$EXISTING" ]; then
-            gh issue close "$EXISTING" --repo "$GITHUB_REPOSITORY" --reason completed \
-              --comment "✅ \`main\` is main-green again at \`${GITHUB_SHA:0:9}\` — ${RUN_URL}. Auto-closed by Release-Green (continuous)."
-            echo "Closed issue #$EXISTING"
-          fi
-
      - name: Upload report artifact
        if: always()
        uses: actions/upload-artifact@v7
@@ -423,13 +384,6 @@ jobs:
          # on `improvements`, complexity-ratchets only when `.improved`), and both exit
          # non-zero while the branch is over baseline — which is exactly when there is
          # nothing to bank. Their exit code is not the signal; the verifier below is.
-          # Velocity phase (quality-baseline.json `_policy`, relax-baselines.mjs): the caps
-          # were raised on purpose, so banking the measured shrink would silently undo the
-          # 20% headroom every night. Pause the downward ratchet until the phase closes.
-          if node -e 'process.exit(require("./config/quality/quality-baseline.json")._policy?.phase === "velocity" ? 0 : 1)'; then
-            echo "Velocity phase active — ratchet banking paused (see docs/architecture/QUALITY_GATES.md → Velocity phase)."
-            exit 0
-          fi
          set +e
          node scripts/check/check-file-size.mjs --update
          node scripts/check/check-complexity-ratchets.mjs --update
@@ -491,72 +445,3 @@ jobs:
            gh pr create --repo "$GITHUB_REPOSITORY" --base "$TARGET" --head "$BANK_BRANCH" \
              --title "chore(quality): bank ratchet shrinks (${TARGET})" --body-file pr-body.md
          fi
-
-  # ── Baseline headroom (velocity phase, 2026-08-30 → v4.0) ──────────────────────
-  # The ratchets only speak when a baseline is crossed. With every baseline loosened by
-  # 20% (scripts/quality/relax-baselines.mjs) the question is how fast the budget is
-  # being consumed — this job measures each gate the way CI does and posts the headroom
-  # table to one living issue, so a budget that fills in a week is visible before the
-  # first red PR. Advisory: never fails the workflow.
-  baseline-headroom:
-    name: Baseline headroom
-    if: ${{ github.event_name != 'push' }}
-    timeout-minutes: 60
-    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-light"]')) || 'ubuntu-latest' }}
-    permissions:
-      contents: read
-      issues: write
-    steps:
-      - uses: actions/checkout@v7
-        with:
-          fetch-depth: 1
-          persist-credentials: false
-
-      - uses: actions/setup-node@v7
-        with:
-          node-version: "24"
-          cache: npm
-
-      - uses: ./.github/actions/npm-ci-retry
-
-      - name: Measure headroom on ${{ github.ref_name }}
-        run: |
-          set -euo pipefail
-          node scripts/quality/baseline-headroom.mjs \
-            --json reports/quality/headroom.json --md reports/quality/headroom.md
-          cat reports/quality/headroom.md >> "$GITHUB_STEP_SUMMARY"
-
-      - name: Upload headroom report
-        if: always()
-        uses: actions/upload-artifact@v7
-        with:
-          name: baseline-headroom-${{ github.run_id }}
-          path: reports/quality/headroom.*
-          retention-days: 90
-
-      - name: Post to the living issue
-        env:
-          GH_TOKEN: ${{ github.token }}
-          RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
-        run: |
-          set -euo pipefail
-          TITLE="📈 Baseline headroom (velocity phase)"
-          BAD=$(node -e 'const r=require("./reports/quality/headroom.json").rows;console.log(r.filter(x=>x.status==="critical"||x.status==="warn").length)')
-          {
-            echo "Branch: \`${GITHUB_REF_NAME}\` · run: ${RUN_URL}"
-            echo ""
-            cat reports/quality/headroom.md
-          } > headroom-comment.md
-          EXISTING=$(gh issue list --repo "$GITHUB_REPOSITORY" --state open \
-            --search "in:title $TITLE" --json number --jq '.[0].number' 2>/dev/null || echo "")
-          if [ -z "$EXISTING" ]; then
-            EXISTING=$(gh issue create --repo "$GITHUB_REPOSITORY" --title "$TITLE" --label quality-gate-finding \
-              --body "Living tracker for the velocity-phase baseline budget (docs/architecture/QUALITY_GATES.md → Velocity phase). One comment per nightly run; the newest comment is the current state." \
-              | grep -oE '[0-9]+$')
-          fi
-          gh issue comment "$EXISTING" --repo "$GITHUB_REPOSITORY" --body-file headroom-comment.md
-          if [ "$BAD" != "0" ]; then
-            gh issue edit "$EXISTING" --repo "$GITHUB_REPOSITORY" --add-label "headroom-alert" 2>/dev/null || true
-          else
-            gh issue edit "$EXISTING" --repo "$GITHUB_REPOSITORY" --remove-label "headroom-alert" 2>/dev/null || true
-          fi
--- a/.github/workflows/nightly-resilience.yml
+++ b/.github/workflows/nightly-resilience.yml
@@ -78,10 +78,7 @@ jobs:

  a11y:
    name: A11y axe (nightly, freeze-and-alert)
-    # #11965: this job runs a backend-only `next build`; the hosted 7 GB runner cannot build
-    # release/v3.8.51 (VM shutdown ~7 min in), so it targets the box's light pool (`omni-light`:
-    # two listeners, jobs ≤ ~6 GB). Falls back to hosted when USE_VPS_RUNNER is off.
-    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-light"]')) || 'ubuntu-latest' }}
+    runs-on: ubuntu-latest
    # The Playwright webServer (`start` mode) builds Next via build-next-isolated.mjs and
    # boots the standalone server itself (waits on /api/monitoring/health, 15min webServer
    # timeout). Unlike the per-PR test-e2e job, this nightly job has no pre-built artifact,
--- a/.github/workflows/nightly-schemathesis.yml
+++ b/.github/workflows/nightly-schemathesis.yml
@@ -10,10 +10,7 @@ permissions:
 jobs:
  schemathesis:
    name: Schemathesis — OpenAPI contract fuzz (advisory)
-    # #11965: this job runs a backend-only `next build`; the hosted 7 GB runner cannot build
-    # release/v3.8.51 (VM shutdown ~7 min in), so it targets the box's light pool (`omni-light`:
-    # two listeners, jobs ≤ ~6 GB). Falls back to hosted when USE_VPS_RUNNER is off.
-    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-light"]')) || 'ubuntu-latest' }}
+    runs-on: ubuntu-latest
    timeout-minutes: 30
    steps:
      - uses: actions/checkout@v7
--- a/.github/workflows/npm-publish.yml
+++ b/.github/workflows/npm-publish.yml
@@ -23,12 +23,11 @@ on:
          - next
          - historic
      publish_mode:
-        description: "auto = publish through npm Trusted Publishing (OIDC, no token, no 2FA prompt — the default); staged = npm stage publish (owner approves with 2FA); direct = legacy token publish (emergency fallback only)"
+        description: "staged = npm stage publish (owner approves with 2FA after the staged boot-verify); direct = legacy immediate publish (emergency fallback only)"
        required: false
-        default: "auto"
+        default: "staged"
        type: choice
        options:
-          - auto
          - staged
          - direct
  workflow_call:
@@ -63,15 +62,11 @@ jobs:
    # mid-"Creating an optimized production build" while v3.8.48 had still fit in 16min.
    # This job never runs on `pull_request`, so the fork-safety clause is always true here;
    # it is kept verbatim so the expression stays greppable against ci.yml.
-    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && (github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository)) && fromJSON('["self-hosted","omni-build"]') || 'ubuntu-latest' }}
-    outputs:
-      version: ${{ steps.resolve.outputs.version }}
-      tag: ${{ steps.resolve.outputs.tag }}
-      skip: ${{ steps.resolve.outputs.skip }}
+    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && (github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository)) && fromJSON('["self-hosted","omni-release"]') || 'ubuntu-latest' }}
    permissions:
      actions: read # find + download the CI run's next-build artifact for this SHA
      contents: write # gh release upload (attach SBOM to the GitHub Release)
-      id-token: write # npm provenance (GitHub Packages step)
+      id-token: write # npm provenance
      packages: write # publish to npm.pkg.github.com
    steps:
      - name: Checkout
@@ -205,11 +200,8 @@ jobs:
            exit 0
          fi
          RUN=""
-          # $RUNNER_TEMP, never /tmp: on the .113 pool /tmp is a 12 GB tmpfs (RAM). Parking
-          # this 1.3 GB artefact there took 27–32 min of the 76-min publish job — the
-          # same bytes upload from disk in 2 min. RUNNER_TEMP is per-runner and on disk.
          for candidate in $CANDIDATES; do
-            if gh run download "$candidate" --repo "$REPO" --name next-build --dir "$RUNNER_TEMP/next-build" 2>/dev/null; then
+            if gh run download "$candidate" --repo "$REPO" --name next-build --dir /tmp/next-build 2>/dev/null; then
              RUN="$candidate"
              break
            fi
@@ -219,8 +211,8 @@ jobs:
            echo "::notice::none of the candidate runs still carries next-build (1-day retention) — falling back to a full build"
            exit 0
          fi
-          tar -xzf "$RUNNER_TEMP/next-build/e2e-build.tar.gz" -C .
-          rm -rf "$RUNNER_TEMP/next-build"
+          tar -xzf /tmp/next-build/e2e-build.tar.gz -C .
+          rm -rf /tmp/next-build
          if [ -f .build/next/standalone/server.js ]; then
            echo "✅ standalone tree restored from CI run $RUN — build:cli will skip next build"
          else
@@ -234,28 +226,6 @@ jobs:
          JWT_SECRET: ci-build-secret-with-sufficient-length-for-validation
        run: npm run build:cli

-      # `build:cli` assembles dist/ but does NOT write dist/BUILD_SHA — only
-      # `build:release` does, by calling write-build-sha.mjs. The #10427 provenance
-      # guard inside check:pack-artifact rejects an artifact with no SHA (and rejects
-      # it even under OMNIROUTE_ALLOW_CANARY_BUILD=1: what cannot be identified cannot
-      # be vouched for). Without this step the build+validate pair in this job is
-      # structurally incompatible and fails 100% of the time — the same gap that was
-      # fixed in ci.yml's Package Artifact job.
-      - name: Stamp dist/BUILD_SHA for the provenance guard (#10427)
-        if: steps.resolve.outputs.skip != 'true'
-        env:
-          OMNIROUTE_BUILD_SHA: ${{ github.sha }}
-        run: |
-          export OMNIROUTE_BUILD_SHA="${OMNIROUTE_BUILD_SHA:0:7}"
-          node scripts/build/write-build-sha.mjs
-
-      # The guard checks ancestry against origin/main by default, which is correct
-      # here (a release tag is cut from main), but the ref has to exist locally for
-      # `git merge-base` to resolve it.
-      - name: Fetch main for the provenance probe
-        if: steps.resolve.outputs.skip != 'true'
-        run: git fetch --no-tags --depth=50 origin +refs/heads/main:refs/remotes/origin/main
-
      - name: Validate npm package artifact
        if: steps.resolve.outputs.skip != 'true'
        run: npm run check:pack-artifact
@@ -273,20 +243,11 @@ jobs:
          if-no-files-found: error

      - name: Attach SBOM to GitHub Release
-        # Not only on the `release` event: the v3.8.50 package shipped through a
-        # workflow_dispatch (staged publish, 11 attempts) and this step was skipped, so the
-        # GitHub Release carried no SBOM until it was attached by hand from the run's
-        # `sbom-npm` artifact. Attach whenever a release for the published tag exists.
-        if: steps.resolve.outputs.skip != 'true' && (github.event_name == 'release' || github.event_name == 'workflow_dispatch')
+        if: steps.resolve.outputs.skip != 'true' && github.event_name == 'release'
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          TAG: ${{ github.event_name == 'release' && github.ref_name || format('v{0}', inputs.version) }}
-        run: |
-          if ! gh release view "$TAG" --repo "$GITHUB_REPOSITORY" >/dev/null 2>&1; then
-            echo "::notice::no GitHub Release for $TAG yet — SBOM stays on the sbom-npm workflow artifact"
-            exit 0
-          fi
-          gh release upload "$TAG" sbom-npm.cdx.json --repo "$GITHUB_REPOSITORY" --clobber
+          TAG: ${{ github.ref_name }}
+        run: gh release upload "$TAG" sbom-npm.cdx.json --clobber

      # WS1.2/WS1.3 (#7065 class): the artifact that is about to be published must
      # BOOT. build:cli already assembled dist/ above; this packs+installs+boots the
@@ -304,12 +265,7 @@ jobs:
      # a staged package that is never approved simply expires, with no `npm deprecate` needed.
      - name: Prove clean-install AND upgrade-over-previous both boot
        if: steps.resolve.outputs.skip != 'true'
-        # 60, not 30. This gate was added in #8953 and the 2026-08-27 v3.8.50 publish
-        # was the FIRST run to ever reach it — every earlier attempt died upstream, so
-        # its budget had never been measured against a real run. It then blew the limit
-        # on its debut: `npm pack` alone took 24m37s, leaving 5 minutes for two installs
-        # and two boots. 30 was a guess; 60 is sized to the one measurement we have.
-        timeout-minutes: 60
+        timeout-minutes: 30
        run: npm run check:install-upgrade

      # WS1.3 (D2, v3.8.49 plan): STAGED publishing by default — `npm stage publish`
@@ -331,133 +287,17 @@ jobs:
          fi
          npm --version

-      # The registry upload itself moved to the `stage-npm` job below: npm REFUSES
-      # `--provenance` from a self-hosted runner (422 "Unsupported GitHub Actions
-      # runner environment"), and the heavy verification above cannot move to a
-      # hosted one (16 GB is not enough for build:cli's next-build fallback — see
-      # this job's runs-on comment). So this job proves the bytes and hands them
-      # over; a tiny hosted job does the upload.
-      - name: Pack the verified tarball for the upload job
-        if: steps.resolve.outputs.skip != 'true'
-        env:
-          VERSION: ${{ steps.resolve.outputs.version }}
-        run: |
-          set -euo pipefail
-          # --ignore-scripts: prepublishOnly would re-run build:cli-api && build:cli,
-          # rebuilding bytes this job has already built, validated and boot-smoked.
-          npm pack --ignore-scripts
-          TARBALL="omniroute-${VERSION}.tgz"
-          test -f "$TARBALL" || { echo "expected $TARBALL to exist after npm pack" >&2; ls -la ./*.tgz || true; exit 1; }
-          echo "packed $TARBALL ($(du -h "$TARBALL" | cut -f1))"
-
-      - name: Hand the tarball to the hosted publish job
-        if: steps.resolve.outputs.skip != 'true'
-        uses: actions/upload-artifact@v7
-        with:
-          name: npm-tarball
-          path: omniroute-${{ steps.resolve.outputs.version }}.tgz
-          retention-days: 1
-          if-no-files-found: error
-
-      - name: Publish to GitHub Packages
-        if: steps.resolve.outputs.skip != 'true'
+      - name: Publish to npm (staged — owner approves with 2FA)
+        if: steps.resolve.outputs.skip != 'true' && (github.event_name != 'workflow_dispatch' || inputs.publish_mode != 'direct')
        env:
          VERSION: ${{ steps.resolve.outputs.version }}
          TAG: ${{ steps.resolve.outputs.tag }}
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          set -euo pipefail
-          echo "Configuring for GitHub Packages..."
-          echo "//npm.pkg.github.com/:_authToken=${GITHUB_TOKEN}" > .npmrc
-          npm pkg set name="@diegosouzapw/omniroute"
-          npm publish --registry=https://npm.pkg.github.com --tag "$TAG" \
-            || echo "⚠️ omniroute@${VERSION} might already be published on GitHub Packages."
-          echo "✅ Action finished for GitHub Packages"
-
-  # npm REFUSES `--provenance` from a self-hosted runner:
-  #   422 Unprocessable Entity - Error verifying sigstore provenance bundle:
-  #   Unsupported GitHub Actions runner environment: "self-hosted".
-  #   Only "github-hosted" runners are supported when publishing with provenance.
-  # v3.8.49 published fine because it predates USE_VPS_RUNNER being turned on
-  # (2026-08-02); v3.8.50 was the first release after it, so this had been latent
-  # for four weeks. Dropping --provenance was not an option: 3.8.49 carries a
-  # SLSA attestation and 3.8.50 must not regress that.
-  # The `publish` job cannot simply move to a hosted runner either — 16 GB is not
-  # enough for build:cli's next-build fallback. So it keeps proving the bytes and
-  # this job, which needs no memory at all, performs the upload.
-  stage-npm:
-    needs: publish
-    if: needs.publish.outputs.skip != 'true'
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-      id-token: write # npm provenance — the whole reason this job is separate
-    steps:
-      - name: Download the tarball the publish job proved
-        uses: actions/download-artifact@v8
-        with:
-          name: npm-tarball
-          path: .
-
-      - name: Setup Node.js
-        uses: actions/setup-node@v7
-        with:
-          node-version: ${{ env.NPM_PUBLISH_NODE_VERSION }}
-          registry-url: https://registry.npmjs.org
-
-      - name: Ensure npm supports staged publishing
-        run: |
-          set -euo pipefail
-          CUR=$(npm --version)
-          if ! node -e "const [a,b]='$(npm --version)'.split('.').map(Number); process.exit(a>11||(a===11&&b>=15)?0:1)"; then
-            # Pinned exact version (supply-chain: never float @latest in a publish
-            # job); bump deliberately when a newer npm is required.
-            echo "npm $CUR < 11.15 — installing pinned npm 11.15.0 for staged publishing"
-            npm install -g --ignore-scripts npm@11.15.0
-          fi
-          npm --version
-
-      # Trusted Publishing (OIDC): npm mints a short-lived credential for THIS run from
-      # GitHub's id-token — no NPM_TOKEN secret, no 2FA prompt, provenance included, and
-      # it is the bypass npm sanctions now that tokens which skip 2FA are being retired
-      # (gh.io/npm-gat-bypass2fa-deprecation). Requires the package's Trusted Publisher to
-      # be configured on npmjs.com (owner: diegosouzapw/OmniRoute, workflow
-      # npm-publish.yml) and a github-hosted runner — which is why this job exists.
-      # Without that configuration `npm publish` fails with ENEEDAUTH: re-dispatch with
-      # publish_mode=staged or direct. Automatic publishing was the flow up to v3.8.48;
-      # v3.8.49 moved to staged (WS1.3) to keep a leaked token from publishing alone —
-      # OIDC gives the same guarantee without the manual approve.
-      - name: Publish to npm (Trusted Publishing / OIDC — automatic)
-        if: github.event_name != 'workflow_dispatch' || inputs.publish_mode == 'auto'
-        env:
-          VERSION: ${{ needs.publish.outputs.version }}
-          TAG: ${{ needs.publish.outputs.tag }}
-        run: |
-          set -euo pipefail
-          TARBALL="omniroute-${VERSION}.tgz"
-          test -f "$TARBALL" || { echo "tarball $TARBALL did not arrive from the publish job" >&2; ls -la; exit 1; }
-          # Deliberately NO NODE_AUTH_TOKEN in this step: npm >= 11.5 detects the GitHub
-          # OIDC token itself. Always pass --tag explicitly (defense in depth: an older
-          # VERSION can never claim `@latest`).
-          npm publish "$TARBALL" --provenance --access public --tag "$TAG" --ignore-scripts
-          echo "✅ Published omniroute@$VERSION (dist-tag=$TAG) via Trusted Publishing"
-
-      - name: Publish to npm (staged — owner approves with 2FA)
-        # Only on an explicit request now: Trusted Publishing below is the default.
-        if: github.event_name == 'workflow_dispatch' && inputs.publish_mode == 'staged'
-        env:
-          VERSION: ${{ needs.publish.outputs.version }}
-          TAG: ${{ needs.publish.outputs.tag }}
          NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}
        run: |
          set -euo pipefail
-          TARBALL="omniroute-${VERSION}.tgz"
-          test -f "$TARBALL" || { echo "tarball $TARBALL did not arrive from the publish job" >&2; ls -la; exit 1; }
          # Always pass --tag explicitly. Defense in depth: even if VERSION is
          # accidentally an older release, the historic tag will NOT claim `@latest`.
-          # --ignore-scripts: publishing a built tarball must never re-run
-          # prepublishOnly (build:cli-api && build:cli) on this small runner.
-          npm stage publish "$TARBALL" --provenance --access public --tag "$TAG" --ignore-scripts
+          npm stage publish --provenance --access public --tag "$TAG"
          {
            echo "## 📦 omniroute@$VERSION STAGED (not yet installable)"
            echo ""
@@ -473,18 +313,31 @@ jobs:
          echo "✅ Staged omniroute@$VERSION (dist-tag=$TAG) — awaiting owner 'npm stage approve'"

      - name: Publish to npm (DIRECT — emergency fallback)
-        if: github.event_name == 'workflow_dispatch' && inputs.publish_mode == 'direct'
+        if: steps.resolve.outputs.skip != 'true' && github.event_name == 'workflow_dispatch' && inputs.publish_mode == 'direct'
        env:
-          VERSION: ${{ needs.publish.outputs.version }}
-          TAG: ${{ needs.publish.outputs.tag }}
+          VERSION: ${{ steps.resolve.outputs.version }}
+          TAG: ${{ steps.resolve.outputs.tag }}
          NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}
        run: |
          set -euo pipefail
-          TARBALL="omniroute-${VERSION}.tgz"
-          test -f "$TARBALL" || { echo "tarball $TARBALL did not arrive from the publish job" >&2; ls -la; exit 1; }
-          npm publish "$TARBALL" --provenance --access public --tag "$TAG" --ignore-scripts
+          npm publish --provenance --access public --tag "$TAG"
          echo "✅ Published omniroute@$VERSION (dist-tag=$TAG) [DIRECT mode]"

+      - name: Publish to GitHub Packages
+        if: steps.resolve.outputs.skip != 'true'
+        env:
+          VERSION: ${{ steps.resolve.outputs.version }}
+          TAG: ${{ steps.resolve.outputs.tag }}
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          set -euo pipefail
+          echo "Configuring for GitHub Packages..."
+          echo "//npm.pkg.github.com/:_authToken=${GITHUB_TOKEN}" > .npmrc
+          npm pkg set name="@diegosouzapw/omniroute"
+          npm publish --registry=https://npm.pkg.github.com --tag "$TAG" \
+            || echo "⚠️ omniroute@${VERSION} might already be published on GitHub Packages."
+          echo "✅ Action finished for GitHub Packages"
+
  publish-opencode-plugin:
    runs-on: ubuntu-latest
    permissions:
@@ -573,92 +426,3 @@ jobs:
          fi
          npm publish --provenance --access public --ignore-scripts
          echo "✅ Published ${PKG_NAME}@${PKG_VERSION}"
-
-  publish-opencode-plugin-v2:
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-      id-token: write # npm provenance
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v7
-        with:
-          persist-credentials: false
-          fetch-depth: 0
-          # Full history needed for auto-bump: git diff against previous release tag
-
-      - name: Setup Node.js
-        uses: actions/setup-node@v7
-        with:
-          node-version: ${{ env.NPM_PUBLISH_NODE_VERSION }}
-          registry-url: https://registry.npmjs.org
-
-      - name: Auto-bump plugin-v2 version if plugin-v2 changed since last release
-        id: bump
-        working-directory: "@omniroute/opencode-plugin-v2"
-        env:
-          CURRENT_TAG: ${{ github.ref_name }}
-        run: |
-          set -euo pipefail
-
-          PKG_VERSION=$(node -p "require('./package.json').version")
-          PKG_NAME=$(node -p "require('./package.json').name")
-
-          # 1) Skip if current version is not yet published (no bump needed)
-          PUBLISHED="$(npm view "${PKG_NAME}@${PKG_VERSION}" version 2>/dev/null || true)"
-          if [ "$PUBLISHED" != "$PKG_VERSION" ]; then
-            echo "✅ ${PKG_NAME}@${PKG_VERSION} is new — no bump needed."
-            echo "bumped=false" >> "$GITHUB_OUTPUT"
-            exit 0
-          fi
-
-          # 2) Find the previous release tag (exclude the current one)
-          PREV_TAG=$(git tag -l 'v*' --sort=-version:refname \
-            | grep -v "^${CURRENT_TAG}$" | head -1 || echo "")
-          if [ -z "$PREV_TAG" ]; then
-            echo "No previous tag to compare — skipping bump."
-            echo "bumped=false" >> "$GITHUB_OUTPUT"
-            exit 0
-          fi
-
-          # 3) Check if plugin-v2 dir actually changed since that tag
-          if git diff --quiet "$PREV_TAG" -- "@omniroute/opencode-plugin-v2/"; then
-            echo "⏭️ No plugin-v2 changes since $PREV_TAG — nothing to publish."
-            echo "bumped=false" >> "$GITHUB_OUTPUT"
-            exit 0
-          fi
-
-          # 4) Auto-bump patch version
-          npm version patch --no-git-tag-version --allow-same-version
-          NEW_VERSION=$(node -p "require('./package.json').version")
-          echo "bumped=true" >> "$GITHUB_OUTPUT"
-          echo "📦 Auto-bumped ${PKG_NAME} from ${PKG_VERSION} to ${NEW_VERSION}"
-
-      - name: Install plugin-v2 dependencies
-        working-directory: "@omniroute/opencode-plugin-v2"
-        run: npm install --no-audit --no-fund
-
-      - name: Build plugin-v2
-        working-directory: "@omniroute/opencode-plugin-v2"
-        run: npm run clean && npm run build
-
-      - name: Test plugin-v2
-        working-directory: "@omniroute/opencode-plugin-v2"
-        run: npm test
-
-      - name: Publish @omniroute/opencode-plugin-v2 to npm
-        working-directory: "@omniroute/opencode-plugin-v2"
-        env:
-          NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}
-        run: |
-          set -euo pipefail
-          PKG_VERSION=$(node -p "require('./package.json').version")
-          PKG_NAME=$(node -p "require('./package.json').name")
-          # Same hardened skip-check as the main job (no --silent flag).
-          PUBLISHED="$(npm view "${PKG_NAME}@${PKG_VERSION}" version 2>/dev/null || true)"
-          if [ "$PUBLISHED" = "$PKG_VERSION" ]; then
-            echo "⚠️ ${PKG_NAME}@${PKG_VERSION} is already published on npm — skipping."
-            exit 0
-          fi
-          npm publish --provenance --access public --ignore-scripts
-          echo "✅ Published ${PKG_NAME}@${PKG_VERSION}"
--- a/.github/workflows/opencode-plugin-ci.yml
+++ b/.github/workflows/opencode-plugin-ci.yml
@@ -2,15 +2,13 @@ name: opencode-plugin CI

 on:
  push:
-    branches: [main, "release/**"]
+    branches: [main, release/v3.8.2]
    paths:
      - "@omniroute/opencode-plugin/**"
-      - "@omniroute/opencode-plugin-v2/**"
  pull_request:
-    branches: [main, "release/**"]
+    branches: [main, release/v3.8.2]
    paths:
      - "@omniroute/opencode-plugin/**"
-      - "@omniroute/opencode-plugin-v2/**"
    types: [opened, synchronize, reopened, ready_for_review]
  workflow_dispatch:

@@ -46,33 +44,10 @@ jobs:
      - run: npm run build
      - run: npm test

-  test-v2:
-    name: Test v2 (Node ${{ matrix.node }})
-    runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-      matrix:
-        node: ["22", "24"]
-    defaults:
-      run:
-        working-directory: "@omniroute/opencode-plugin-v2"
-    steps:
-      - uses: actions/checkout@v7
-        with:
-          persist-credentials: false
-      - uses: actions/setup-node@v7
-        with:
-          node-version: ${{ matrix.node }}
-          cache: npm
-          cache-dependency-path: "@omniroute/opencode-plugin-v2/package-lock.json"
-      - run: npm ci --no-audit --no-fund
-      - run: npm run build
-      - run: npm test
-
  build:
    name: Build
    runs-on: ubuntu-latest
-    needs: [test, test-v2]
+    needs: test
    steps:
      - uses: actions/checkout@v7
        with:
@@ -80,11 +55,12 @@ jobs:
      - uses: actions/setup-node@v7
        with:
          node-version: "22"
-      - name: Build plugin-v2 artifact
-        working-directory: "@omniroute/opencode-plugin-v2"
-        run: npm ci --no-audit --no-fund && npm run build
+          cache: npm
+          cache-dependency-path: "@omniroute/opencode-plugin/package-lock.json"
+      - run: npm install --no-audit --no-fund
+      - run: npm run build
      - uses: actions/upload-artifact@v7
        with:
-          name: opencode-plugin-v2-dist
-          path: "@omniroute/opencode-plugin-v2/dist"
+          name: opencode-plugin-dist
+          path: "@omniroute/opencode-plugin/dist"
          retention-days: 7
--- a/.github/workflows/quality.yml
+++ b/.github/workflows/quality.yml
@@ -61,25 +61,13 @@ jobs:
    name: Build (advisory)
    needs: changes
    # FORK PRs ONLY. build.yml's `Fast Production Build` triggers on `push: branches: ["**"]`
-    # (#11946, 2026-08-29: build.yml is now workflow_dispatch-only — the hosted runner cannot
-    # build this tree in any profile, 8/8 recent fork PRs included — so own-origin PRs rely on
-    # ci.yml `Build` after merge to main and on nightly-release-green for release/**.)
    # and runs `build:release` — a superset of this job — so for an own-origin branch this job
    # was building the same tree twice. A fork contributor pushes to THEIR repo, so that push
    # never fires here, and this is the only pre-merge build signal they get. Measured
    # 2026-08-14: 72 of the last 100 PRs into release/** came from forks, so the fork case is
    # the majority of the traffic, not the exception — this job earns its place, it just should
    # not duplicate build.yml for the own-origin 28%.
-    # Disabled 2026-08-29 (#11976 follow-up). `continue-on-error: true` still
-    # reports a GitHub check FAILURE, so every fork PR into release/** was born
-    # with a red "Build (advisory)" even when every required gate was green
-    # (sweep-reds, 41 PRs). Hosted ubuntu-latest cannot finish `npm run build`
-    # on this tree — VM shutdown ~6 min in, same class as build.yml going
-    # workflow_dispatch-only in #11962. Pre-merge build signal for release/**
-    # is nightly-release-green (omni-build); for main it is ci.yml `Build`.
-    # Restore this job when a runner that actually fits the tree is wired here.
-    # Bare `false` (not `${{ false }}`) — zizmor obfuscation flags the expression form.
-    if: false
+    if: ${{ github.event_name != 'pull_request' || ((github.event.pull_request.draft == false || startsWith(github.head_ref, 'mergify/merge-queue/')) && needs.changes.outputs.code == 'true' && github.event.pull_request.head.repo.full_name != github.repository) }}
    # PINNED to hosted — this was the last job in THIS workflow still on the USE_VPS_RUNNER
    # switch (ci.yml's Build, nightly-release-green and npm-publish keep it, so the variable
    # stays meaningful), and with USE_VPS_RUNNER=true it produced NO signal at all here.
@@ -201,11 +189,8 @@ jobs:
            .eslintcache
            .eslintcache-complexity
          key: eslint-${{ runner.os }}-${{ hashFiles('eslint.config.mjs', 'eslint.complexity-ratchets.config.mjs', 'config/quality/eslint-suppressions.json', 'package-lock.json') }}
-          # No restore-keys fallback on purpose (#11600, P-II.1 of the v3.8.50 postmortem): a
-          # cache built under a different suppressions file / lint config / lockfile reports
-          # stale per-file verdicts, which is exactly how 215 pre-existing errors stayed
-          # invisible for a whole cycle. Exact key or a cold full lint (~13 min) — never a
-          # partial cache from another configuration.
+          restore-keys: |
+            eslint-${{ runner.os }}-
      # Security scanners — same hardened install as ci.yml quality-extended
      # (gh release download = authenticated, 5000 req/hr; curl to api.github.com
      # is rate-limited to 60/hr and silently no-ops when throttled). The blocking
@@ -284,7 +269,7 @@ jobs:
        run: |
          set -uo pipefail
          gates=(
-            provider-consistency provider-asset-provenance fetch-targets deps file-size error-helper
+            provider-consistency fetch-targets deps file-size error-helper
            migration-numbering public-creds db-rules known-symbols
            route-guard-membership test-discovery test-runner-api
            mutation-test-coverage any-budget:t11 build-scope pack-policy
@@ -303,12 +288,7 @@ jobs:
            # #8522: file-size is base-relative on PR events (compare against
            # max(frozen, base)) so inherited drift doesn't red an innocent PR;
            # workflow_dispatch (no PR base) falls back to absolute comparison.
-            # New-code mode (Clean-as-You-Code, 2026-08-30): complexity-ratchets and
-            # dead-code compare the PR's files against the merge-base and block only on
-            # what the PR added; the global totals are advisory on PRs and re-frozen at
-            # release. See scripts/check/newCodeMode.mjs.
-            case "$g" in file-size|complexity-ratchets|dead-code) NEW_CODE=1 ;; *) NEW_CODE= ;; esac
-            if [ -n "$NEW_CODE" ] && [ -n "${PR_BASE_SHA:-}" ]; then
+            if [ "$g" = "file-size" ] && [ -n "${PR_BASE_SHA:-}" ]; then
              npm run "check:$g" -- --base-ref "$PR_BASE_SHA" || failed+=("$g")
            else
              npm run "check:$g" || failed+=("$g")
@@ -480,12 +460,6 @@ jobs:
    # cache restore saturating the link), while the tests themselves tied, 2m54 vs 2m31. So
    # self-hosted is strictly worse here and there is nothing to configure.
    runs-on: ubuntu-latest
-    # A shard finishes in ~10 min. Without a ceiling a hung test process holds the PR for
-    # GitHub's 6 h default: on 2026-08-28 shard 1/4 sat 64 min without a line of output
-    # (twice, same spot — a timing race, gone on the third run) while the other three
-    # shards were long green. 30 min = 3x the normal wall-clock; a shard that needs more
-    # is a hang, not a slow run, and a fast red with a re-run beats a silent 6 h hold.
-    timeout-minutes: 30
    strategy:
      fail-fast: false
      matrix:
@@ -526,10 +500,7 @@ jobs:
    name: No new ESLint warnings
    needs: changes
    if: ${{ github.event_name != 'pull_request' || ((github.event.pull_request.draft == false || startsWith(github.head_ref, 'mergify/merge-queue/')) && needs.changes.outputs.code == 'true') }}
-    runs-on: ${{ (vars.USE_VPS_RUNNER == 'true' && fromJSON('["self-hosted","omni-light"]')) || 'ubuntu-latest' }}
-    # 2026-08-30: a cold full lint with the eslint-plugin-react-hooks 7 compiler rules is
-    # killed on the 7 GB hosted runner without a message (status null → exit 1, the
-    # JSON never written); the box lints it in ~12 min with the heap below.
+    runs-on: ubuntu-latest
    continue-on-error: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.repo.fork == true }}
    # G0 (trilho .50): security-events:read lets the CodeQL ratchet below read open
    # code-scanning alerts via `gh api .../code-scanning/alerts` (same as ci.yml's
@@ -553,16 +524,11 @@ jobs:
            .eslintcache
            .eslintcache-complexity
          key: eslint-${{ runner.os }}-${{ hashFiles('eslint.config.mjs', 'eslint.complexity-ratchets.config.mjs', 'config/quality/eslint-suppressions.json', 'package-lock.json') }}
-          # No restore-keys fallback on purpose (#11600, P-II.1 of the v3.8.50 postmortem): a
-          # cache built under a different suppressions file / lint config / lockfile reports
-          # stale per-file verdicts, which is exactly how 215 pre-existing errors stayed
-          # invisible for a whole cycle. Exact key or a cold full lint (~13 min) — never a
-          # partial cache from another configuration.
+          restore-keys: |
+            eslint-${{ runner.os }}-
      - name: ESLint (baseline congelado — warning novo = vermelho)
        # lint:json writes the report; --max-warnings 0 keeps no-new-warnings policy.
        run: npm run lint:json -- --max-warnings 0
-        env:
-          NODE_OPTIONS: --max-old-space-size=8192
      # ── G0 (trilho .50): motor de ratchet também no trilho B ─────────────────────
      # This job just wrote .artifacts/eslint-results.json — collect-metrics prefers
      # that file, so the ratchet engine lands here at ZERO extra ESLint cost (one
--- a/.github/workflows/scorecard.yml
+++ b/.github/workflows/scorecard.yml
@@ -4,15 +4,12 @@ on:
  schedule:
    - cron: "27 7 * * 1"
  push:
-    # Scorecard only accepts the DEFAULT branch — here the active release/vX.Y.Z,
-    # not `main`. The job below guards on it so a push to any other branch skips.
-    branches: ["main", "release/**"]
+    branches: ["main"]

 permissions: read-all

 jobs:
  analysis:
-    if: ${{ github.event_name != 'push' || github.ref == format('refs/heads/{0}', github.event.repository.default_branch) }}
    name: Scorecard analysis
    runs-on: ubuntu-latest
    permissions:
--- a/.gitignore
+++ b/.gitignore
@@ -293,6 +293,3 @@ docker-compose.yml.bak
 # Ad-hoc test sandboxes (never tracked — may contain local DBs)
 /.sandbox/
 .aider*
-
-# check:install-upgrade work trees (~12 GB, disposable)
-/.install-upgrade/
--- a/.husky/pre-commit
+++ b/.husky/pre-commit
@@ -7,7 +7,6 @@ fi

 # Cheap, deterministic local gates (re-enabled). Slower checks (i18n drift,
 # openapi coverage/security-tiers, env-doc sync) run in CI to keep commits fast.
-sh scripts/check/check-git-identity.sh
 npx lint-staged
 node scripts/check/check-docs-sync.mjs
 npm run check:any-budget:t11
--- a/.i18n-state.json
+++ b/.i18n-state.json
--- a/.mailmap
+++ b/.mailmap
@@ -1,40 +0,0 @@
-# .mailmap — canonical author identities for git log/shortlog/blame.
-#
-# Why this file exists: between 2026-08-13 and 2026-08-26 this checkout carried a
-# `git config --local` whose user.name was one contributor's ("Xiangzhe" / @xz-dev)
-# and whose user.email was ANOTHER contributor's (@backryun). Every commit produced
-# on this machine in that window was therefore signed with @backryun's address —
-# 237 commits, all in the -0300 timezone, while @backryun's own work commits from
-# +0900 and continued normally throughout. The local override was removed on
-# 2026-08-26; this file repairs the RECORD without rewriting published history
-# (those commits live on release/v3.8.50 and release/v3.8.51, which other sessions
-# and open PRs build on — a rewrite would force-push both and orphan the v3.8.50 tag).
-#
-# Format: Canonical Name <canonical@email> Commit Name <commit@email>
-
-# --- Maintainer: several addresses used over the project's life ---
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> <diegosouza.pw@gmail.com>
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> <diegosouza.pw@outlook.com>
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> <diegosouzapw@users.noreply.github.com>
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> Diego Rodrigues de Sa e Souza <8016841+diegosouzapw@users.noreply.github.com>
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> Diego Souza <8016841+diegosouzapw@users.noreply.github.com>
-
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> <diego.souza.pw@gmail.com>
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> <souzamiriamrodrigues790@gmail.com>
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> <diego.souza@cdwasolutions.com.br>
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> <diegosouzapw@devbox.local>
-
-# --- The misattribution window: name Xiangzhe + @backryun's email, from -0300.
-# These are maintainer/session commits, NOT @backryun's contributions.
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> Xiangzhe <bakryun0718@proton.me>
-diegosouzapw <8016841+diegosouzapw@users.noreply.github.com> Xiangzhe <diegosouza.pw@gmail.com>
-
-# --- Xiangzhe (@xz-dev) — a distinct contributor; keep their own work intact ---
-Xiangzhe <32761048+xz-dev@users.noreply.github.com> <xz-dev@users.noreply.github.com>
-Xiangzhe <32761048+xz-dev@users.noreply.github.com> <xiangzhedev@gmail.com>
-
-# --- @backryun's own alternate addresses (their real work, kept intact) ---
-backryun <24198422+backryun@users.noreply.github.com> <bakryun0718@proton.me>
-backryun <24198422+backryun@users.noreply.github.com> <backryun@daonlab.local>
-backryun <24198422+backryun@users.noreply.github.com> <busan011@ormbiz.co.kr>
-backryun <24198422+backryun@users.noreply.github.com> <backryun@users.noreply.github.com>
--- a/.trivyignore
+++ b/.trivyignore
@@ -18,3 +18,5 @@
 #
 # Keep this list SHORT and reviewed every release. Prefer fixing (rebuild on a
 # patched base / bump the dep) over suppressing. Stale entries are debt.
+#
+# (No accepted-risk suppressions at present — ignore-unfixed covers the noise.)
--- a/@omniroute/opencode-plugin-v2/.gitignore
+++ b/@omniroute/opencode-plugin-v2/.gitignore
@@ -1,4 +0,0 @@
-node_modules
-dist
-*.log
-.DS_Store
--- a/@omniroute/opencode-plugin-v2/LICENSE
+++ b/@omniroute/opencode-plugin-v2/LICENSE
@@ -1,21 +0,0 @@
-MIT License
-
-Copyright (c) 2026 OmniRoute contributors
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
--- a/@omniroute/opencode-plugin-v2/README.md
+++ b/@omniroute/opencode-plugin-v2/README.md
@@ -1,107 +0,0 @@
-# @omniroute/opencode-plugin-v2
-
-OpenCode v2 plugin (`define({ id, setup })`, Promise API) that publishes the live OmniRoute catalog — models from `/v1/models`, combos from `/api/combos` (least-common-denominator join), auto-combos from `/api/combos/auto`, enrichment (names + pricing), and usable-provider filtering — into the v2 `catalog.transform`, with `key` + `env` auth via `integration.transform`.
-
-Companion to `@omniroute/opencode-plugin` (OpenCode v1, same repo). The two packages are independent: this one carries its own catalog-mapping logic and the v1 plugin is left untouched.
-
-## Install
-
-```sh
-npm install @omniroute/opencode-plugin-v2
-```
-
-`opencode.json`:
-
-```json
-{
-  "plugins": [
-    {
-      "package": "@omniroute/opencode-plugin-v2",
-      "options": {
-        "providerId": "omniroute",
-        "baseURL": "http://localhost:20128"
-      }
-    }
-  ]
-}
-```
-
-## Credentials
-
-The plugin needs a gateway key to read the catalog, and looks for one in this
-order:
-
-1. **The credential you connected in OpenCode.** The plugin registers an
-   integration, so `opencode auth` (or the Connect action in the model picker)
-   can store a key for it. Nothing is written to `opencode.json` — this is the
-   recommended route.
-2. **`apiKey` in the plugin options**, when you want a per-project override.
-   Remember that this puts the key in a config file you may be committing.
-3. **`OMNIROUTE_API_KEY` in the environment.**
-
-If none of the three yields a key, the catalog is empty and the plugin says so
-once at startup rather than leaving you with a silent empty model list.
-
-### The management token is a different key
-
-Combos, provider health and enrichment (display names, pricing, free-tier
-budgets) come from the gateway's `/api/*` endpoints, which most deployments
-gate behind a **management** token rather than the inference key. Set it
-explicitly:
-
-```json
-"options": {
-  "baseURL": "http://localhost:20128",
-  "managementReadToken": "<management read token>"
-}
-```
-
-Left unset, `managementReadToken` falls back to `apiKey` for backwards
-compatibility. When a gateway rejects that fallback, the catalog still
-publishes — but with raw model ids instead of display names, no canonical
-alias dedupe, no pricing and no combos. The plugin warns once per endpoint
-when this happens, naming the endpoint and the consequence, so the degraded
-catalog is never a mystery.
-
-## Options
-
-| Key                              | Default                                        | Notes                                                                                                                 |
-| -------------------------------- | ---------------------------------------------- | --------------------------------------------------------------------------------------------------------------------- |
-| `providerId`                     | `"omniroute"`                                  | Provider id and integration id; models publish under `<providerId>/…`                                                 |
-| `baseURL`                        | required                                       | OmniRoute gateway root (no `/v1` suffix needed)                                                                       |
-| `apiKey`                         | connected credential, then `OMNIROUTE_API_KEY` | Chat key for `/v1/*` — see [Credentials](#credentials)                                                                |
-| `managementReadToken`            | falls back to `apiKey`                         | Management key for `/api/*` (combos, providers, enrichment) — usually **not** the same key                            |
-| `displayName`                    | `"OmniRoute"`                                  | Provider display name                                                                                                 |
-| `timeoutMs`                      | `10000`                                        | Per-endpoint fetch timeout (auto-combos use 5s)                                                                       |
-| `modelCacheTtlMs`                | `300000`                                       | Catalog cache TTL; disk snapshot warms cold starts                                                                    |
-| `timeouts`                       | per-endpoint override                          | `{ models, combos, autoCombos, enrichment }` in ms; falls back to `timeoutMs`                                         |
-| `enrichment`                     | `true`                                         | Fetch names + pricing (`/api/pricing*`, `/api/free-tier/summary`)                                                     |
-| `providerTag`                    | `true`                                         | Prefix a display name with the upstream provider it routes to                                                         |
-| `geminiSanitization`             | `true`                                         | Strip `$schema`/`additionalProperties` from tool schemas sent to Gemini models (`$ref` tools are forwarded untouched) |
-| `usableOnly`                     | `false`                                        | Filter to healthy provisioned providers (`/api/providers`)                                                            |
-| `visibleModels` / `hiddenModels` | `[]`                                           | Exact-or-suffix allowlists, deny wins                                                                                 |
-| `apiFormat.allowAnthropic`       | `false`                                        | Route allowlisted ids to the Anthropic API block                                                                      |
-| `apiFormat.anthropicModels`      | `[]`                                           | Full model ids routed to Anthropic                                                                                    |
-| `apiFormat.anthropicPrefixes`    | v1 defaults                                    | Deprecated, warns once — prefer `anthropicModels`                                                                     |
-| `logLevel` / `startupDebug`      | `warn` / `false`                               | Logger verbosity                                                                                                      |
-
-## Tool calling on Gemini models
-
-Gemini answers `400 INVALID_ARGUMENT` — for the whole request, not just the
-offending tool — when a tool declaration carries `$schema` or
-`additionalProperties`. Anything that emits standard JSON Schema therefore
-breaks tool calling as soon as the chain routes to Gemini.
-
-The plugin strips those keywords from tool schemas bound for a Gemini model of
-this provider, and leaves every other request untouched. A tool carrying a
-`$ref` is forwarded untouched instead of stripped: removing the reference
-would widen the schema to "accept anything". Set
-`"geminiSanitization": false` to turn it off.
-
-## Migrating from the v1 plugin
-
-The v2 plugin publishes provider id `X` bare. The v1 plugin published `opencode-X` (native-adapter gate). Sessions pinned to `opencode-X/...` must re-select the model under `X/...`.
-
-## License
-
-MIT
--- a/@omniroute/opencode-plugin-v2/RELEASE.md
+++ b/@omniroute/opencode-plugin-v2/RELEASE.md
@@ -1,9 +0,0 @@
-# Release process — `@omniroute/opencode-plugin-v2`
-
-## Publishing
-
-One package, no ordering: bump `@omniroute/opencode-plugin-v2` (`npm version patch`) and publish it. The plugin carries its own copy of the mapping logic, so a release never has to be coordinated with another package.
-
-## Migration note (`opencode-X` → `X`)
-
-The v1 plugin published provider id `opencode-X` (native-adapter gate). The v2 plugin publishes `X` bare. Sessions pinned to `opencode-X/...` resolve `ModelUnavailableError` — users must re-select the model under `X/...`.
--- a/@omniroute/opencode-plugin-v2/package-lock.json
+++ b/@omniroute/opencode-plugin-v2/package-lock.json
--- a/@omniroute/opencode-plugin-v2/package.json
+++ b/@omniroute/opencode-plugin-v2/package.json
@@ -1,70 +0,0 @@
-{
-  "name": "@omniroute/opencode-plugin-v2",
-  "version": "0.1.0",
-  "description": "OmniRoute OpenCode plugin (v2 Promise API): catalog transform with models, combos, enrichment, and naming.",
-  "type": "module",
-  "main": "./dist/index.js",
-  "types": "./dist/index.d.ts",
-  "exports": {
-    ".": {
-      "types": "./dist/index.d.ts",
-      "import": "./dist/index.js"
-    }
-  },
-  "files": [
-    "dist",
-    "README.md",
-    "LICENSE"
-  ],
-  "scripts": {
-    "build": "tsup",
-    "clean": "rm -rf dist",
-    "test": "node --import tsx/esm --test tests/*.test.ts",
-    "prepublishOnly": "npm run clean && npm run build && npm test"
-  },
-  "dependencies": {
-    "zod": "^4.4.3"
-  },
-  "devDependencies": {
-    "@opencode-ai/plugin": "1.18.29",
-    "@types/node": "^22.19.19",
-    "tsup": "^8.5.1",
-    "tsx": "^4.22.3",
-    "typescript": "^5.9.3"
-  },
-  "engines": {
-    "node": ">=22.22.3"
-  },
-  "license": "MIT",
-  "author": "OmniRoute contributors",
-  "repository": {
-    "type": "git",
-    "url": "https://github.com/diegosouzapw/OmniRoute.git",
-    "directory": "@omniroute/opencode-plugin-v2"
-  },
-  "homepage": "https://github.com/diegosouzapw/OmniRoute/tree/main/%40omniroute/opencode-plugin-v2#readme",
-  "bugs": {
-    "url": "https://github.com/diegosouzapw/OmniRoute/issues"
-  },
-  "keywords": [
-    "omniroute",
-    "opencode",
-    "opencode-plugin",
-    "opencode-v2",
-    "ai-sdk",
-    "openai-compatible",
-    "provider",
-    "catalog",
-    "combos",
-    "gemini"
-  ],
-  "publishConfig": {
-    "access": "public"
-  },
-  "peerDependencies": {
-    "@opencode-ai/plugin": ">=1.18.29 <2"
-  },
-  "overrides": {
-    "esbuild": "^0.28.1"
-  }
-}
--- a/@omniroute/opencode-plugin-v2/src/cache.ts
+++ b/@omniroute/opencode-plugin-v2/src/cache.ts
@@ -1,211 +0,0 @@
-import { createHash } from "node:crypto";
-import { homedir } from "node:os";
-import { mkdir, readFile, unlink, writeFile } from "node:fs/promises";
-import { dirname, join } from "node:path";
-import type {
-  OmniRouteEnrichmentEntry,
-  OmniRouteEnrichmentMap,
-  OmniRouteProviderConnection,
-  OmniRouteRawAutoCombo,
-  OmniRouteRawCombo,
-  OmniRouteRawModelEntry,
-} from "./shared/index.js";
-
-export const DEFAULT_MODEL_CACHE_TTL_MS = 300_000 as const;
-
-/**
- * Breather after a refresh whose models fetch came back empty (gateway down
- * or refusing). Transforms inside the window serve last-known-good without
- * re-firing the fetch suite. Short on purpose: it only guards the
- * pathological case, normal TTL expiry still refetches every window.
- */
-export const UNREACHABLE_COOLDOWN_MS = 15_000 as const;
-
-export interface CatalogSnapshot {
-  models: OmniRouteRawModelEntry[];
-  combos: OmniRouteRawCombo[];
-  autoCombos: OmniRouteRawAutoCombo[];
-  providers?: OmniRouteProviderConnection[];
-  enrichment?: OmniRouteEnrichmentMap;
-  fetchedAt: number;
-}
-
-export const SNAPSHOT_FORMAT_VERSION = 2 as const;
-
-/**
- * A raw snapshot entry is stale when it cannot be mapped to a publishable
- * model: no string `id` (unroutable) or a pre-mapped `api` block without a
- * valid `npm` package (the runner would reject it as `Unsupported package`).
- * Plain `/v1/models` entries carry no `api` block -- it is synthesized at
- * publish time -- so only a present-but-invalid block drops the entry.
- */
-export function isStaleSnapshotModel(entry: unknown): boolean {
-  if (!entry || typeof entry !== "object") return true;
-  const id = (entry as { id?: unknown }).id;
-  if (typeof id !== "string" || id.length === 0) return true;
-  const api = (entry as { api?: unknown }).api;
-  if (api === undefined) return false;
-  if (!api || typeof api !== "object") return true;
-  const npm = (api as { npm?: unknown }).npm;
-  return typeof npm !== "string" || npm.length === 0;
-}
-
-interface DiskSnapshotV2 {
-  v: 2;
-  identityFingerprint: string;
-  models: OmniRouteRawModelEntry[];
-  combos: OmniRouteRawCombo[];
-  autoCombos?: OmniRouteRawAutoCombo[];
-  providers?: OmniRouteProviderConnection[];
-  /**
-   * Display names, provider labels, pricing and free-tier budgets, as
-   * `[key, entry]` pairs (a Map does not survive JSON). Persisted because a
-   * cold start otherwise publishes raw model ids until the first refresh
-   * completes — which is the moment the snapshot exists to cover.
-   */
-  enrichment?: [string, OmniRouteEnrichmentEntry][];
-  writtenAt: number;
-}
-
-/**
- * Ceiling on what one snapshot may occupy on disk. A gateway with thousands of
- * models makes this file grow without bound otherwise; past the cap the
- * enrichment overlay is dropped first (it is rebuilt on the next refresh)
- * rather than losing the catalog itself.
- */
-const MAX_SNAPSHOT_BYTES = 32 * 1024 * 1024;
-
-function trimTrailingSlashes(value: string): string {
-  let i = value.length;
-  while (i > 0 && value.charCodeAt(i - 1) === 0x2f) i -= 1;
-  return i === value.length ? value : value.slice(0, i);
-}
-
-function normalizeBaseURL(baseURL: string): string {
-  try {
-    const parsed = new URL(baseURL);
-    parsed.hash = "";
-    parsed.pathname = trimTrailingSlashes(parsed.pathname) || "/";
-    return parsed.toString();
-  } catch {
-    return trimTrailingSlashes(baseURL);
-  }
-}
-
-export function memoryCacheKey(baseURL: string, credentialId: string): string {
-  return `${baseURL}::${createHash("sha256").update(credentialId).digest("hex")}`;
-}
-
-export function snapshotIdentityFingerprint(
-  baseURL: string,
-  apiKey: string,
-  managementReadToken: string
-): string {
-  return createHash("sha256")
-    .update(JSON.stringify([normalizeBaseURL(baseURL), apiKey, managementReadToken]))
-    .digest("hex");
-}
-
-export function diskSnapshotPath(providerId: string): string {
-  // OPENCODE_DATA_DIR is honoured verbatim when set: whoever controls the
-  // process environment already chooses where the process writes, so
-  // resolving it further would only surprise. The providerId segment stays
-  // bounded by the options schema (letters, digits, '.', '_' and '-'; never
-  // "." or ".."), keeping the file inside <dir>/plugins/.
-  const dir = process.env.OPENCODE_DATA_DIR ?? join(homedir(), ".local", "share", "opencode");
-  return join(dir, "plugins", `omniroute-${providerId}.json`);
-}
-
-export async function readDiskSnapshot(
-  providerId: string,
-  identityFingerprint: string,
-  logger?: { warn: (message: string) => void }
-): Promise<CatalogSnapshot | undefined> {
-  try {
-    const body = await readFile(diskSnapshotPath(providerId), "utf8");
-    const parsed = JSON.parse(body) as Partial<DiskSnapshotV2>;
-    if (
-      !parsed ||
-      typeof parsed.v !== "number" ||
-      parsed.v < SNAPSHOT_FORMAT_VERSION ||
-      typeof parsed.identityFingerprint !== "string" ||
-      parsed.identityFingerprint !== identityFingerprint
-    ) {
-      return undefined;
-    }
-    if (
-      !Array.isArray(parsed.models) ||
-      parsed.models.length === 0 ||
-      !Array.isArray(parsed.combos)
-    ) {
-      return undefined;
-    }
-    const stale = (parsed.models as unknown[]).filter(isStaleSnapshotModel).length;
-    const models = (parsed.models as OmniRouteRawModelEntry[]).filter(
-      (entry) => !isStaleSnapshotModel(entry)
-    );
-    if (stale > 0) {
-      logger?.warn(`[omniroute-v2] dropping ${stale} stale snapshot entries without api block`);
-    }
-    if (models.length === 0) return undefined;
-    return {
-      models,
-      combos: parsed.combos as OmniRouteRawCombo[],
-      autoCombos: Array.isArray(parsed.autoCombos)
-        ? (parsed.autoCombos as OmniRouteRawAutoCombo[])
-        : [],
-      providers: Array.isArray(parsed.providers)
-        ? (parsed.providers as OmniRouteProviderConnection[])
-        : [],
-      // A snapshot written before this field existed, or one whose overlay was
-      // dropped for size, simply starts unenriched and recovers on the first
-      // refresh — the same state as before it was persisted at all.
-      enrichment: Array.isArray(parsed.enrichment)
-        ? new Map(parsed.enrichment as [string, OmniRouteEnrichmentEntry][])
-        : undefined,
-      fetchedAt: typeof parsed.writtenAt === "number" ? parsed.writtenAt : Date.now(),
-    };
-  } catch {
-    return undefined;
-  }
-}
-
-export async function writeDiskSnapshot(
-  providerId: string,
-  snapshot: CatalogSnapshot,
-  identityFingerprint: string
-): Promise<void> {
-  try {
-    if (snapshot.models.length === 0) return;
-    const file = diskSnapshotPath(providerId);
-    await mkdir(dirname(file), { recursive: true, mode: 0o700 });
-    const envelope: DiskSnapshotV2 = {
-      v: 2,
-      identityFingerprint,
-      models: snapshot.models,
-      combos: snapshot.combos,
-      autoCombos: snapshot.autoCombos,
-      providers: snapshot.providers ?? [],
-      enrichment: snapshot.enrichment ? [...snapshot.enrichment.entries()] : undefined,
-      writtenAt: Date.now(),
-    };
-    let payload = JSON.stringify(envelope);
-    if (payload.length > MAX_SNAPSHOT_BYTES && envelope.enrichment !== undefined) {
-      delete envelope.enrichment;
-      payload = JSON.stringify(envelope);
-    }
-    if (payload.length > MAX_SNAPSHOT_BYTES) return;
-    await writeFile(file, payload, { encoding: "utf8", mode: 0o600 });
-  } catch {
-    // Best-effort: callers already hold the in-memory entry.
-  }
-}
-
-export async function clearDiskSnapshot(providerId: string): Promise<boolean> {
-  try {
-    await unlink(diskSnapshotPath(providerId));
-    return true;
-  } catch {
-    return false;
-  }
-}
--- a/@omniroute/opencode-plugin-v2/src/catalog.ts
+++ b/@omniroute/opencode-plugin-v2/src/catalog.ts
@@ -1,798 +0,0 @@
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import { type HostContract, detectHostContract, emitsLegacyFields } from "./compat.js";
-import type { Model as LegacyModelV2 } from "@opencode-ai/sdk/v2";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-import {
-  type ApiFormatV2,
-  type LogLevel,
-  type Logger,
-  type OmniRouteAutoCombosFetcher,
-  type OmniRouteCombosFetcher,
-  type OmniRouteEnrichmentFetcher,
-  type OmniRouteEnrichmentMap,
-  type OmniRouteModelsFetcher,
-  type OmniRouteProviderConnection,
-  type OmniRouteProvidersFetcher,
-  type OmniRouteRawAutoCombo,
-  type OmniRouteRawCombo,
-  type OmniRouteRawModelEntry,
-  applyEnrichment,
-  buildCanonicalToAliasMap,
-  canonicalDedupSet,
-  createLogger,
-  defaultOmniRouteEnrichmentFetcher,
-  defaultOmniRouteProvidersFetcher,
-  ensureV1Suffix,
-  isUsableCombo,
-  isUsableRawModelId,
-  lookupEnrichment,
-  mapAutoComboToModelV2,
-  mapComboToModelV2,
-  mapRawModelToModelV2,
-  usableProviderAliasSet,
-} from "./shared/index.js";
-
-export type ModelsFetcher = OmniRouteModelsFetcher;
-export type CombosFetcher = OmniRouteCombosFetcher;
-export type AutoCombosFetcher = OmniRouteAutoCombosFetcher;
-export type ProvidersFetcher = OmniRouteProvidersFetcher;
-export type EnrichmentFetcher = OmniRouteEnrichmentFetcher;
-
-export interface EndpointTimeouts {
-  models?: number;
-  combos?: number;
-  autoCombos?: number;
-  enrichment?: number;
-}
-
-export interface ResolvedOptions {
-  providerId: string;
-  baseURL: string;
-  apiKey: string;
-  managementReadToken?: string;
-  timeoutMs: number;
-  timeouts?: EndpointTimeouts;
-  logger?: Logger;
-  logLevel?: LogLevel;
-  startupDebug?: boolean;
-  modelCacheTtlMs: number;
-  /** v1 parity: prefix the display name with the upstream provider label. */
-  providerTag?: boolean;
-  displayName?: string;
-  apiFormat?: ApiFormatV2;
-  visibleModels?: string[];
-  hiddenModels?: string[];
-  usableOnly: boolean;
-  enrichment?: OmniRouteEnrichmentMap | boolean;
-  /**
-   * Shared collision-warning dedupe set keyed `cacheKey::comboKey`. When
-   * omitted a fresh per-publish set is used. index.ts passes one setup-wide
-   * set so a repeated publish (stale replay + refresh) warns once per key.
-   */
-  collisionWarned?: Set<string>;
-}
-
-export interface CatalogFetchers {
-  fetcher?: ModelsFetcher;
-  combosFetcher?: CombosFetcher;
-  autoCombosFetcher?: AutoCombosFetcher;
-  providersFetcher?: ProvidersFetcher;
-  enrichmentFetcher?: EnrichmentFetcher;
-  models?: ModelsFetcher;
-  combos?: CombosFetcher;
-  autoCombos?: AutoCombosFetcher;
-  providers?: ProvidersFetcher;
-  enrichment?: EnrichmentFetcher;
-  /**
-   * Called when a gateway source cannot be read. Without it this function
-   * degrades silently — the catalog publishes with raw ids and no combos and
-   * nothing says why, which is the failure the plugin path reports.
-   */
-  onSourceError?: (endpoint: string, reason: string) => void;
-}
-
-// The shared mappers speak the legacy (`Provider.models[id]`) `Model` shape
-// (imported from `@opencode-ai/sdk/v2`, also re-exported by the plugin root
-// as `ModelV2`); the real v2 `CatalogDraft` carries `ModelV2Info` instead.
-// Convert the fields 1:1 at the draft boundary -- NEVER `as unknown as` the
-// whole model.
-//
-// Binary-compat note: the prod binary (beta-17823) reads a top-level
-// `package` field on both Model and Provider structs (`package:a.Package`,
-// gated by `isAISDK = startsWith("aisdk:")`), with a model-to-provider
-// fallback (`package: u.package ?? s.package`). The pinned SDK types
-// (1.18.29) only know the `api` block, so the binary field is published via
-// the typed extensions below (spread/Object.assign, never `any`).
-export const BINARY_AISDK_PREFIX = "aisdk:";
-
-/** Top-level `package` as the legacy contract expects it (`aisdk:<npm>`). */
-export interface BinaryCompatPackage {
-  package: string;
-}
-
-/**
- * The legacy contract keeps on the model/provider itself what the `api` block
- * carries in the pinned types: the aisdk package, the endpoint (as
- * `settings.baseURL`) and the per-request headers. None of these keys collide
- * with a key of `ModelV2Info`/`ProviderV2Info`, so both field sets can be
- * published on the same object.
- */
-export interface BinaryCompatFields extends BinaryCompatPackage {
-  settings: Record<string, unknown>;
-  headers: Record<string, string>;
-}
-
-/** Legacy variants read their options from `settings`, not `headers`/`body`. */
-export type BinaryCompatVariant = ModelV2Info["variants"][number] & {
-  settings: Record<string, unknown>;
-};
-
-export type BinaryCompatModel = ModelV2Info & BinaryCompatFields;
-export type BinaryCompatProvider = ProviderV2Info &
-  BinaryCompatPackage & {
-    settings: Record<string, unknown>;
-  };
-
-export function toBinaryPackage(npm: string): string {
-  return npm.startsWith(BINARY_AISDK_PREFIX) ? npm : `${BINARY_AISDK_PREFIX}${npm}`;
-}
-export function legacyApiToInfoApi(api: LegacyModelV2["api"]): ModelV2Info["api"] {
-  if (!api || typeof api.npm !== "string" || api.npm.length === 0) {
-    throw new Error(
-      "[omniroute-v2] refusing to publish a model without an api block (missing api.npm)"
-    );
-  }
-  return { id: api.id, type: "aisdk", package: api.npm, url: api.url };
-}
-
-function legacyCostToInfoCost(cost: LegacyModelV2["cost"]): ModelV2Info["cost"] {
-  return [{ input: cost.input, output: cost.output, cache: cost.cache }];
-}
-
-function legacyCapabilitiesToInfoCapabilities(
-  caps: LegacyModelV2["capabilities"]
-): ModelV2Info["capabilities"] {
-  const input: string[] = [];
-  if (caps.input.text) input.push("text");
-  if (caps.input.audio) input.push("audio");
-  if (caps.input.image) input.push("image");
-  if (caps.input.video) input.push("video");
-  if (caps.input.pdf) input.push("pdf");
-  const output: string[] = [];
-  if (caps.output.text) output.push("text");
-  if (caps.output.audio) output.push("audio");
-  if (caps.output.image) output.push("image");
-  if (caps.output.video) output.push("video");
-  if (caps.output.pdf) output.push("pdf");
-  return { tools: caps.toolcall, input, output };
-}
-
-function legacyToInfo(providerID: string, modelID: string, m: LegacyModelV2): ModelV2Info {
-  const variants = Object.entries(m.variants ?? {}).map(([id, body]) => ({
-    id,
-    headers: {},
-    body: body as Record<string, unknown>,
-  }));
-  const parsed = Date.parse(m.release_date);
-  return {
-    id: modelID,
-    providerID,
-    ...(m.family !== undefined ? { family: m.family } : {}),
-    name: m.name,
-    api: legacyApiToInfoApi(m.api),
-    capabilities: legacyCapabilitiesToInfoCapabilities(m.capabilities),
-    request: { headers: { ...m.headers }, body: { ...m.options } },
-    variants,
-    time: { released: Number.isNaN(parsed) ? 0 : parsed },
-    cost: legacyCostToInfoCost(m.cost),
-    status: m.status,
-    enabled: true,
-    limit: { ...m.limit },
-  };
-}
-
-export interface PublishCounts {
-  models: number;
-  combos: number;
-  autoCombos: number;
-}
-
-export interface ModelListFilter {
-  exact: Set<string>;
-  suffixes: Set<string>;
-}
-
-export function compileModelListFilter(list?: string[]): ModelListFilter | undefined {
-  if (!list || list.length === 0) return undefined;
-  const exact = new Set<string>();
-  const suffixes = new Set<string>();
-  for (const id of list) {
-    if (id.includes("/")) {
-      exact.add(id);
-    } else {
-      suffixes.add(id);
-    }
-  }
-  if (exact.size === 0 && suffixes.size === 0) return undefined;
-  return { exact, suffixes };
-}
-
-function matchesSuffix(id: string, suffixes: Set<string>): boolean {
-  if (suffixes.size === 0) return false;
-  const slash = id.indexOf("/");
-  const suffix = slash > 0 ? id.slice(slash + 1) : id;
-  return suffixes.has(suffix);
-}
-
-export function passesModelAllowlist(
-  id: string,
-  visible?: ModelListFilter,
-  hidden?: ModelListFilter
-): boolean {
-  if (hidden) {
-    if (hidden.exact.has(id) || matchesSuffix(id, hidden.suffixes)) return false;
-  }
-  if (visible) {
-    if (!visible.exact.has(id) && !matchesSuffix(id, visible.suffixes)) return false;
-  }
-  return true;
-}
-
-export function passesComboAllowlist(combo: OmniRouteRawCombo, visible?: ModelListFilter): boolean {
-  if (!visible) return true;
-  const steps = Array.isArray(combo.models) ? combo.models : [];
-  if (steps.length === 0) return true;
-  let sawResolvableMember = false;
-  for (const step of steps) {
-    if (step?.kind === "combo-ref") continue;
-    const modelId = typeof step?.model === "string" ? step.model : "";
-    if (modelId.length === 0) continue;
-    sawResolvableMember = true;
-    if (visible.exact.has(modelId) || matchesSuffix(modelId, visible.suffixes)) return true;
-  }
-  if (!sawResolvableMember) return true;
-  return false;
-}
-
-/**
- * Project the `api` block onto the legacy top-level fields. Only the `aisdk`
- * variant of `ModelApi`/`ProviderApi` carries a package, so the caller narrows
- * before calling; a `native` api has no legacy equivalent and publishes
- * nothing (the legacy contract has no native models).
- */
-function legacyModelFields(info: ModelV2Info): BinaryCompatFields | undefined {
-  if (info.api.type !== "aisdk") return undefined;
-  const settings: Record<string, unknown> = {
-    ...(info.api.settings ?? {}),
-    ...info.request.body,
-  };
-  if (info.api.url !== undefined) settings.baseURL = info.api.url;
-  return {
-    package: toBinaryPackage(info.api.package),
-    settings,
-    headers: { ...info.request.headers },
-  };
-}
-
-/** `{id, headers, body}` (pinned types) plus `{settings}` (legacy contract). */
-function legacyVariants(variants: ModelV2Info["variants"]): BinaryCompatVariant[] {
-  return variants.map((variant) => ({ ...variant, settings: { ...variant.body } }));
-}
-
-function assignModelFields(
-  target: ModelV2Info,
-  source: LegacyModelV2,
-  contract: HostContract
-): void {
-  const info = legacyToInfo(target.providerID || source.providerID, target.id || source.id, source);
-  target.name = info.name;
-  target.api = info.api;
-  target.capabilities = info.capabilities;
-  target.request = info.request;
-  target.variants = info.variants;
-  target.time = info.time;
-  target.cost = info.cost;
-  target.status = info.status;
-  target.enabled = info.enabled;
-  target.limit = info.limit;
-  if (info.family !== undefined) {
-    target.family = info.family;
-  }
-  if (!emitsLegacyFields(contract)) return;
-  const legacy = legacyModelFields(info);
-  if (legacy !== undefined) {
-    Object.assign(target, legacy);
-    target.variants = legacyVariants(info.variants);
-  }
-}
-
-function assignProviderFields(
-  target: ProviderV2Info,
-  source: { name: string; api: ProviderV2Info["api"]; integrationID: string },
-  contract: HostContract
-): void {
-  target.name = source.name;
-  target.api = source.api;
-  target.integrationID = source.integrationID;
-  if (!emitsLegacyFields(contract)) return;
-  // The legacy contract defaults `Provider.Info.package` to `""` and model
-  // resolution falls back to it (`package: model.package ?? provider.package`),
-  // so the provider carries the same `aisdk:<npm>` value as its models, and
-  // the endpoint as `settings.baseURL`.
-  if (source.api.type !== "aisdk") return;
-  const settings: Record<string, unknown> = { ...(source.api.settings ?? {}) };
-  if (source.api.url !== undefined) settings.baseURL = source.api.url;
-  Object.assign(target, { package: toBinaryPackage(source.api.package), settings });
-}
-
-/** A widened capability flag (`boolean | { field }`) read back as a plain flag. */
-function isCapabilityEnabled(value: boolean | { field: string }): boolean {
-  return value !== false;
-}
-
-/**
- * Combo steps reach us from the gateway with a shape the SDK types do not
- * describe (`kind`, `comboName`, `model` appear per step kind). One reader
- * keeps that single untyped boundary in one place instead of scattering casts.
- */
-function readStepField(step: unknown, key: "kind" | "comboName" | "model"): unknown {
-  return (step as Record<string, unknown> | null | undefined)?.[key];
-}
-
-/**
- * Resolve the display-name + pricing overlay. A caller may hand over a
- * ready-made map (tests, pre-resolved overlays) or turn the fetch off; a
- * failed fetch soft-fails to an empty map so the catalog still publishes,
- * with mapper-default names and zeroed pricing rather than nothing at all.
- */
-async function resolveEnrichmentOverlay(
-  opts: ResolvedOptions,
-  fetchers: CatalogFetchers | undefined,
-  log: Logger
-): Promise<OmniRouteEnrichmentMap> {
-  if (opts.enrichment instanceof Map) return opts.enrichment;
-  if (opts.enrichment === false) return new Map();
-  const fetchEnrichment =
-    fetchers?.enrichmentFetcher ?? fetchers?.enrichment ?? defaultOmniRouteEnrichmentFetcher;
-  try {
-    return await fetchEnrichment(
-      opts.baseURL,
-      opts.managementReadToken ?? opts.apiKey,
-      opts.timeouts?.enrichment ?? opts.timeoutMs,
-      fetchers?.onSourceError
-    );
-  } catch (err) {
-    log.warn(
-      `[omniroute-v2] enrichment fetch failed, continuing without names/pricing: ${err instanceof Error ? err.message : String(err)}`
-    );
-    return new Map();
-  }
-}
-
-/**
- * Resolve the provider aliases worth publishing when `usableOnly` is on.
- * Gated on the flag, so the default configuration issues no request at all.
- * The filter subtracts: a failed or empty connections fetch yields
- * `undefined` and keeps the whole catalog, because only a prefix proven not
- * provisioned may be dropped.
- */
-async function resolveUsableAliases(
-  opts: ResolvedOptions,
-  providersFetcher: OmniRouteProvidersFetcher | undefined,
-  onSourceError: ((endpoint: string, reason: string) => void) | undefined,
-  enrichment: OmniRouteEnrichmentMap,
-  timeoutMs: number,
-  log: Logger
-): Promise<ReturnType<typeof usableProviderAliasSet> | undefined> {
-  if (!opts.usableOnly) return undefined;
-  let rawConnections: OmniRouteProviderConnection[];
-  try {
-    const fetchProviders = providersFetcher ?? defaultOmniRouteProvidersFetcher;
-    rawConnections = await fetchProviders(
-      opts.baseURL,
-      opts.managementReadToken ?? opts.apiKey,
-      timeoutMs,
-      onSourceError
-    );
-  } catch (err) {
-    log.warn(
-      `[omniroute-v2] providers fetch failed, usableOnly filter disabled for this refresh: ${err instanceof Error ? err.message : String(err)}`
-    );
-    rawConnections = [];
-  }
-  return rawConnections.length > 0 ? usableProviderAliasSet(rawConnections, enrichment) : undefined;
-}
-
-/** Everything the combo publishing pass reads, passed as one value. */
-interface PublishContext {
-  draft: CatalogDraft;
-  opts: ResolvedOptions;
-  log: Logger;
-  providerId: string;
-  hostContract: HostContract;
-  enrichment: OmniRouteEnrichmentMap;
-  rawModelById: Map<string, OmniRouteRawModelEntry>;
-  publishedKeys: Set<string>;
-  publishedModelIds: Map<string, string>;
-  visibleFilter: ReturnType<typeof compileModelListFilter>;
-  hiddenFilter: ReturnType<typeof compileModelListFilter>;
-  usable: ReturnType<typeof usableProviderAliasSet> | undefined;
-  canonicalToAlias: ReturnType<typeof buildCanonicalToAliasMap>;
-  combosFetcher: CatalogFetchers["combos"] | undefined;
-  combosTimeout: number;
-  /** Shared with the auto-combos pass: one collision warning per key, per run. */
-  warnedCombos: Set<string>;
-  cacheKey: string;
-}
-
-/**
- * Fetch the gateway's combos and publish them, resolving nested combo-refs to
- * a fixpoint first: a combo whose members are themselves combos only knows its
- * lowest common denominator once those are known. Combos that never resolve
- * are dropped rather than published with a fabricated capability set, and
- * reported once.
- *
- * Returns the number published, or `undefined` when the combos fetch failed —
- * the caller then publishes a models-only catalog instead of an empty one.
- */
-async function publishCombos(ctx: PublishContext): Promise<number | undefined> {
-  const {
-    draft,
-    opts,
-    log,
-    providerId: X,
-    hostContract,
-    enrichment,
-    rawModelById,
-    publishedKeys,
-    publishedModelIds,
-    visibleFilter,
-    hiddenFilter,
-    usable,
-    canonicalToAlias,
-    combosFetcher,
-    combosTimeout,
-    warnedCombos,
-    cacheKey,
-  } = ctx;
-  let rawCombos: OmniRouteRawCombo[];
-  try {
-    rawCombos = combosFetcher
-      ? await combosFetcher(opts.baseURL, opts.managementReadToken ?? opts.apiKey, combosTimeout)
-      : [];
-  } catch (err) {
-    log.warn(
-      `[omniroute-v2] combos fetch failed, falling back to models-only catalog: ${err instanceof Error ? err.message : String(err)}`
-    );
-    return undefined;
-  }
-
-  let comboCount = 0;
-  // Ported from v1 (fixpoint 8 passes + warn once per (cacheKey, comboKey)
-  // + intentional-dedup exception). Nested combo-refs resolve against the
-  // friendly combo name; unresolvable combos are dropped (never published
-  // with a fabricated empty LCD) and reported once.
-  const MAX_COMBO_PASSES = 8;
-  const pending = rawCombos.filter((combo) => {
-    if (!combo || !combo.id) return false;
-    if (combo.isHidden === true) return false;
-    if (usable && !isUsableCombo(combo, usable)) return false;
-    if (visibleFilter && !passesComboAllowlist(combo, visibleFilter)) return false;
-    // Deny wins for combos too: a user who hides an id expects it gone from
-    // the picker whether it is a model or a combo built on it.
-    if (hiddenFilter && passesComboAllowlist(combo, hiddenFilter)) return false;
-    return true;
-  });
-  const resolvedByName = new Map<string, LegacyModelV2>();
-  let unresolved: typeof pending = [];
-
-  for (let pass = 0; pass < MAX_COMBO_PASSES && pending.length > 0; pass++) {
-    const stillPending: typeof pending = [];
-    for (const combo of pending) {
-      const memberSteps = Array.isArray(combo.models) ? combo.models : [];
-      const memberEntries: OmniRouteRawModelEntry[] = [];
-      let deferred = false;
-      for (const step of memberSteps) {
-        const kind = readStepField(step, "kind");
-        if (kind === "combo-ref") {
-          const comboName = readStepField(step, "comboName");
-          if (typeof comboName !== "string" || comboName.length === 0) continue;
-          const nested = resolvedByName.get(comboName);
-          if (!nested) {
-            deferred = true;
-            break;
-          }
-          memberEntries.push(synthesizeNestedMember(comboName, nested));
-          continue;
-        }
-        const modelId = readStepField(step, "model");
-        if (typeof modelId !== "string" || modelId.length === 0) continue;
-        const member = rawModelById.get(modelId);
-        if (member) memberEntries.push(member);
-      }
-      if (deferred) {
-        stillPending.push(combo);
-        continue;
-      }
-      const mapped = mapComboToModelV2(combo, memberEntries, X, opts.baseURL, opts.apiFormat);
-      applyEnrichment(mapped, lookupEnrichment(combo.id, enrichment, canonicalToAlias), {
-        isCombo: true,
-      });
-      const mid = mapped.id.startsWith(X + "/") ? mapped.id.slice(X.length + 1) : mapped.id;
-      const key = X + "/" + mid;
-      if (publishedKeys.has(key)) {
-        // Intentional dedup (v1 parity): `/v1/models` pre-mirrors combos as
-        // raw entries, so the combo's friendly NAME matches the overwritten
-        // entry's model id (bare or provider-prefixed, endsWith to cover
-        // both). Only warn on a genuine accidental collision (name differs
-        // from the entry it overwrites).
-        const existingId = publishedModelIds.get(key) ?? "";
-        const friendly =
-          typeof combo.name === "string" && combo.name.trim().length > 0
-            ? combo.name.trim()
-            : combo.id;
-        const isIntentionalDedup =
-          existingId === friendly ||
-          existingId === X + "/" + friendly ||
-          existingId.endsWith("/" + friendly);
-        if (!isIntentionalDedup) {
-          const dedupeKey = `${cacheKey}::${key}`;
-          if (!warnedCombos.has(dedupeKey)) {
-            warnedCombos.add(dedupeKey);
-            log.warn(`[omniroute-v2] combo key "${key}" collides with a model id; combo wins.`);
-          }
-        }
-      }
-      draft.model.update(X, mid, (m) => {
-        assignModelFields(m, mapped, hostContract);
-      });
-      publishedKeys.add(key);
-      publishedModelIds.set(key, mapped.id);
-      comboCount += 1;
-      const lookupName =
-        typeof combo.name === "string" && combo.name.trim().length > 0
-          ? combo.name.trim()
-          : combo.id;
-      if (!resolvedByName.has(lookupName)) resolvedByName.set(lookupName, mapped);
-    }
-    if (stillPending.length === pending.length) {
-      unresolved = stillPending;
-      break;
-    }
-    unresolved = stillPending;
-    pending.length = 0;
-    pending.push(...stillPending);
-  }
-
-  if (unresolved.length > 0) {
-    log.warn(
-      `[omniroute-v2] ${unresolved.length} combo(s) could not resolve all nested combo-refs after ${MAX_COMBO_PASSES} passes; dropped to avoid over-claiming.`
-    );
-  }
-  return comboCount;
-}
-
-/**
- * Synthesize a raw-model entry from an already-resolved nested combo so a
- * parent combo's LCD folds the whole nested capability vector (context,
- * output, modalities, capabilities) instead of only direct raw members.
- * v1 parity (combo member synthesis at nested resolution time).
- */
-function synthesizeNestedMember(name: string, nested: LegacyModelV2): OmniRouteRawModelEntry {
-  const inputModalities: string[] = [];
-  if (nested.capabilities.input.text) inputModalities.push("text");
-  if (nested.capabilities.input.audio) inputModalities.push("audio");
-  if (nested.capabilities.input.image) inputModalities.push("image");
-  if (nested.capabilities.input.video) inputModalities.push("video");
-  if (nested.capabilities.input.pdf) inputModalities.push("pdf");
-  const outputModalities: string[] = [];
-  if (nested.capabilities.output.text) outputModalities.push("text");
-  if (nested.capabilities.output.audio) outputModalities.push("audio");
-  if (nested.capabilities.output.image) outputModalities.push("image");
-  if (nested.capabilities.output.video) outputModalities.push("video");
-  if (nested.capabilities.output.pdf) outputModalities.push("pdf");
-  return {
-    id: `combo-ref:${name}`,
-    context_length: nested.limit.context,
-    max_output_tokens: nested.limit.output,
-    ...(nested.limit.input !== undefined ? { max_input_tokens: nested.limit.input } : {}),
-    owned_by: "combo",
-    input_modalities: inputModalities,
-    output_modalities: outputModalities,
-    capabilities: {
-      temperature: nested.capabilities.temperature,
-      // A raw entry carries plain flags; the mapped model widens them to
-      // `boolean | { field }` (custom reasoning/thinking field). Every
-      // non-false form means the capability is present, which is all the
-      // LCD fold reads.
-      reasoning: isCapabilityEnabled(nested.capabilities.reasoning),
-      thinking: isCapabilityEnabled(nested.capabilities.interleaved),
-      attachment: nested.capabilities.attachment,
-      tool_calling: nested.capabilities.toolcall,
-    },
-  };
-}
-
-export async function publishCatalog(
-  draft: CatalogDraft,
-  opts: ResolvedOptions,
-  fetchers?: CatalogFetchers
-): Promise<PublishCounts> {
-  const X = opts.providerId;
-  const log = opts.logger ?? createLogger(opts.startupDebug ? "debug" : (opts.logLevel ?? "warn"));
-  const modelsTimeout = opts.timeouts?.models ?? opts.timeoutMs;
-  const combosTimeout = opts.timeouts?.combos ?? opts.timeoutMs;
-  // v1 parity keeps the 5s auto-combos budget when no per-endpoint value is
-  // set (P2 resolves it in index.ts; direct publishCatalog callers may only
-  // pass timeoutMs).
-  const autoCombosTimeout = opts.timeouts?.autoCombos ?? 5_000;
-  // The contract is discovered from the object the host seeds into the
-  // provider draft, which the host fills before any model is published. The
-  // verdict is then reused for every model: the model seed carries no
-  // discriminating key, and a single provider/model pair always speaks one
-  // contract.
-  let hostContract: HostContract = "unknown";
-  draft.provider.update(X, (p) => {
-    hostContract = detectHostContract(p);
-    assignProviderFields(
-      p,
-      {
-        name: opts.displayName ?? "OmniRoute",
-        api: {
-          type: "aisdk",
-          package: "@ai-sdk/openai-compatible",
-          url: ensureV1Suffix(opts.baseURL),
-        },
-        integrationID: X,
-      },
-      hostContract
-    );
-  });
-  log.debug(`[omniroute-v2] host catalog contract detected: ${hostContract}`);
-
-  const modelsFetcher = fetchers?.fetcher ?? fetchers?.models;
-  const combosFetcher = fetchers?.combosFetcher ?? fetchers?.combos;
-  const autoCombosFetcher = fetchers?.autoCombosFetcher ?? fetchers?.autoCombos;
-  const providersFetcher = fetchers?.providersFetcher ?? fetchers?.providers;
-
-  let rawModels: OmniRouteRawModelEntry[];
-  try {
-    rawModels = modelsFetcher ? await modelsFetcher(opts.baseURL, opts.apiKey, modelsTimeout) : [];
-  } catch (err) {
-    log.warn(
-      `[omniroute-v2] models fetch failed, publishing empty catalog: ${err instanceof Error ? err.message : String(err)}`
-    );
-    return { models: 0, combos: 0, autoCombos: 0 };
-  }
-
-  const visibleFilter = compileModelListFilter(opts.visibleModels);
-  const hiddenFilter = compileModelListFilter(opts.hiddenModels);
-
-  const enrichment = await resolveEnrichmentOverlay(opts, fetchers, log);
-  const canonicalToAlias = buildCanonicalToAliasMap(enrichment);
-  const canonicalDedup = canonicalDedupSet(rawModels, canonicalToAlias);
-
-  const usable = await resolveUsableAliases(
-    opts,
-    providersFetcher,
-    fetchers?.onSourceError,
-    enrichment,
-    modelsTimeout,
-    log
-  );
-
-  const rawModelById = new Map<string, OmniRouteRawModelEntry>();
-  for (const entry of rawModels) {
-    if (entry.id) rawModelById.set(entry.id, entry);
-  }
-
-  const publishedKeys = new Set<string>();
-  // Mapped model id per published key (models and combos alike). Mirrors
-  // v1's `models[comboKey]` lookup so the intentional-dedup check sees the
-  // overwritten entry's id, not just key presence.
-  const publishedModelIds = new Map<string, string>();
-  let modelCount = 0;
-  for (const entry of rawModels) {
-    if (!entry.id) continue;
-    if (canonicalDedup.has(entry.id)) continue;
-    if (usable && !isUsableRawModelId(entry.id, usable)) continue;
-    if (!passesModelAllowlist(entry.id, visibleFilter, hiddenFilter)) continue;
-    const mapped = mapRawModelToModelV2(entry, {
-      providerId: X,
-      baseURL: opts.baseURL,
-      apiFormat: opts.apiFormat,
-    });
-    applyEnrichment(mapped, lookupEnrichment(entry.id, enrichment, canonicalToAlias), {
-      providerTag: opts.providerTag !== false,
-    });
-    const mid = mapped.id.startsWith(X + "/") ? mapped.id.slice(X.length + 1) : mapped.id;
-    draft.model.update(X, mid, (m) => {
-      assignModelFields(m, mapped, hostContract);
-    });
-    publishedKeys.add(X + "/" + mid);
-    publishedModelIds.set(X + "/" + mid, mapped.id);
-    modelCount += 1;
-  }
-
-  const warnedCombos = opts.collisionWarned ?? new Set<string>();
-  const cacheKey = `${opts.baseURL}::${opts.providerId}`;
-  const comboCount = await publishCombos({
-    draft,
-    opts,
-    log,
-    providerId: X,
-    hostContract,
-    enrichment,
-    rawModelById,
-    publishedKeys,
-    publishedModelIds,
-    visibleFilter,
-    hiddenFilter,
-    usable,
-    canonicalToAlias,
-    combosFetcher,
-    combosTimeout,
-    warnedCombos,
-    cacheKey,
-  });
-  if (comboCount === undefined) return { models: modelCount, combos: 0, autoCombos: 0 };
-
-  // Migration: v1 published opencode-X; v2 publishes X bare. Sessions pinned
-  // opencode-X resolve ModelUnavailableError -- see RELEASE.md migration note.
-  // Re-publishing under "opencode-"+X here is FORBIDDEN: a double
-  // publish would double chat entries in the picker.
-
-  // Auto combos: virtual server-side entries from /api/combos/auto, keyed
-  // "auto" / "auto/<variant>" (v1 parity). Fail-open: a fetcher throw keeps
-  // models + combos and only warns - old gateways may not serve the
-  // endpoint at all (the default fetcher maps 404 to [] itself).
-  let rawAutoCombos: OmniRouteRawAutoCombo[];
-  try {
-    rawAutoCombos = autoCombosFetcher
-      ? await autoCombosFetcher(
-          opts.baseURL,
-          opts.managementReadToken ?? opts.apiKey,
-          autoCombosTimeout
-        )
-      : [];
-  } catch (err) {
-    log.warn(
-      `[omniroute-v2] auto combos fetch failed, falling back to models+combos catalog: ${err instanceof Error ? err.message : String(err)}`
-    );
-    return { models: modelCount, combos: comboCount, autoCombos: 0 };
-  }
-
-  let autoComboCount = 0;
-  for (const autoCombo of rawAutoCombos) {
-    if (!autoCombo || !autoCombo.id) continue;
-    if (autoCombo.isHidden === true) continue;
-    // Auto combos are catalog entries like any other: an id a user asked to
-    // hide must stay hidden, and an allowlist that excludes it must exclude
-    // it. They used to skip both filters entirely.
-    if (!passesModelAllowlist(autoCombo.id, visibleFilter, hiddenFilter)) continue;
-    if (usable && !isUsableRawModelId(autoCombo.id, usable)) continue;
-    const mapped = mapAutoComboToModelV2(autoCombo, X, opts.baseURL, opts.apiFormat);
-    applyEnrichment(mapped, lookupEnrichment(autoCombo.id, enrichment, canonicalToAlias), {
-      isCombo: true,
-      isAutoCombo: true,
-    });
-    const key = X + "/" + mapped.id;
-    if (publishedKeys.has(key)) {
-      const dedupeKey = `${cacheKey}::${key}`;
-      if (!warnedCombos.has(dedupeKey)) {
-        warnedCombos.add(dedupeKey);
-        log.warn(
-          `[omniroute-v2] auto combo key "${key}" collides with a model id; auto combo wins.`
-        );
-      }
-    }
-    draft.model.update(X, mapped.id, (m) => {
-      assignModelFields(m, mapped, hostContract);
-    });
-    publishedKeys.add(key);
-    publishedModelIds.set(key, mapped.id);
-    autoComboCount += 1;
-  }
-
-  return { models: modelCount, combos: comboCount, autoCombos: autoComboCount };
-}
--- a/@omniroute/opencode-plugin-v2/src/compat.ts
+++ b/@omniroute/opencode-plugin-v2/src/compat.ts
@@ -1,66 +0,0 @@
-function isObject(value: unknown): value is Record<string, unknown> {
-  return typeof value === "object" && value !== null;
-}
-
-function isTransformHolder(value: unknown): value is { transform: unknown } {
-  return isObject(value) && "transform" in value;
-}
-
-/**
- * The catalog domain is the one this plugin cannot work without. The
- * integration domain carries the credential flow and the `aisdk` domain the
- * tool-schema cleaning: a host missing either still gets its catalog, so
- * neither is asserted here — each is probed where it is used.
- */
-export function assertContext(ctx: unknown): void {
-  if (!isObject(ctx)) {
-    throw new Error("[omniroute-v2] contract breach: ctx must be an object");
-  }
-  if (!isTransformHolder(ctx.catalog) || typeof ctx.catalog.transform !== "function") {
-    throw new Error("[omniroute-v2] contract breach: ctx.catalog.transform must be a function");
-  }
-  if (!isObject(ctx.options)) {
-    throw new Error("[omniroute-v2] contract breach: ctx.options must be an object");
-  }
-}
-
-/**
- * Catalog contract spoken by the running host.
- *
- * opencode v2 is a moving target: the catalog contract changed between the
- * binary that ships today and the SDK types this package pins. Rather than
- * keying off a version list (which goes stale on the next release), the
- * contract is discovered at runtime from the object the host seeds into the
- * draft.
- *
- * - `legacy-package` — the seed carries a top-level `package` and no `api`
- *   block. Observed on `@opencode-ai/cli` 0.0.0-beta-17823, whose
- *   `Provider.Info.empty` is `{id, name, activation, package}`.
- * - `sdk-api` — the seed carries an `api` block. This is the contract of the
- *   pinned `@opencode-ai/plugin`/`@opencode-ai/sdk` types.
- * - `unknown` — neither or both. The caller publishes the superset.
- */
-export type HostContract = "legacy-package" | "sdk-api" | "unknown";
-
-export function detectHostContract(seed: unknown): HostContract {
-  if (!isObject(seed)) return "unknown";
-  const hasApi = "api" in seed;
-  const hasPackage = "package" in seed;
-  if (hasApi && !hasPackage) return "sdk-api";
-  if (hasPackage && !hasApi) return "legacy-package";
-  return "unknown";
-}
-
-/**
- * Whether to publish the legacy top-level fields (`package`, `settings`,
- * `headers`, `variants[].settings`) next to the `api`-block fields.
- *
- * A host proven to speak the legacy contract gets them because it needs them;
- * an unrecognised host gets them because the superset is the safer default
- * (both field sets have been observed to survive an unknown-key write). A host
- * that speaks the `api` contract does not, so a future strict schema cannot
- * reject the write on an excess property.
- */
-export function emitsLegacyFields(contract: HostContract): boolean {
-  return contract !== "sdk-api";
-}
--- a/@omniroute/opencode-plugin-v2/src/credentials.ts
+++ b/@omniroute/opencode-plugin-v2/src/credentials.ts
@@ -1,101 +0,0 @@
-import type { PluginContext } from "@opencode-ai/plugin/v2/promise";
-import type { Logger } from "./shared/index.js";
-
-/** Where a resolved key came from, so the failure message can name the fix. */
-export type ApiKeyOrigin = "connection" | "option" | "env" | "missing";
-
-export interface ResolvedApiKey {
-  key: string;
-  origin: ApiKeyOrigin;
-}
-
-const ENV_VAR = "OMNIROUTE_API_KEY";
-
-/**
- * `ctx.integration.connection` is newer than the `key`/`env` methods this
- * plugin registers, so a host that predates it exposes `integration` without
- * it. Probing the shape keeps the plugin loadable on both.
- */
-function connectionApi(ctx: PluginContext): PluginContext["integration"]["connection"] | undefined {
-  const connection = (ctx.integration as Partial<PluginContext["integration"]>).connection;
-  if (
-    connection === undefined ||
-    typeof connection.active !== "function" ||
-    typeof connection.resolve !== "function"
-  ) {
-    return undefined;
-  }
-  return connection;
-}
-
-/**
- * Read the credential the user stored through the host's own auth flow.
- *
- * The plugin advertises `key` and `env` methods on its integration, so a user
- * can connect it from the UI; without this lookup that connection would only
- * feed inference and the catalog fetches would still need a key pasted into
- * the config file.
- *
- * Returns `undefined` (never throws) when there is no connection, when the
- * host is too old to expose one, or when the stored credential is an OAuth
- * grant — this plugin authenticates the gateway with a bearer key, and an
- * access token from an unrelated grant is not one.
- */
-async function keyFromConnection(
-  ctx: PluginContext,
-  integrationID: string,
-  log: Logger
-): Promise<string | undefined> {
-  const connection = connectionApi(ctx);
-  if (connection === undefined) return undefined;
-  try {
-    const active = await connection.active(integrationID);
-    if (active === undefined) return undefined;
-    const credential = await connection.resolve(active);
-    if (credential === undefined) return undefined;
-    if (credential.type !== "key") {
-      log.warn(
-        `[omniroute-v2] ignoring the stored ${credential.type} credential: this plugin authenticates with an API key`
-      );
-      return undefined;
-    }
-    return credential.key.length > 0 ? credential.key : undefined;
-  } catch (err) {
-    log.warn(
-      `[omniroute-v2] could not read the stored credential: ${err instanceof Error ? err.message : String(err)}`
-    );
-    return undefined;
-  }
-}
-
-/**
- * Resolve the gateway key, preferring the credential the host holds over one
- * written in config. A key in `opencode.json` still wins over the environment
- * so an explicit per-project override keeps working.
- */
-export async function resolveApiKey(
-  ctx: PluginContext,
-  integrationID: string,
-  optionKey: string | undefined,
-  log: Logger
-): Promise<ResolvedApiKey> {
-  const stored = await keyFromConnection(ctx, integrationID, log);
-  if (stored !== undefined) return { key: stored, origin: "connection" };
-  if (optionKey !== undefined && optionKey.length > 0) return { key: optionKey, origin: "option" };
-  const fromEnv = process.env[ENV_VAR];
-  if (fromEnv !== undefined && fromEnv.length > 0) return { key: fromEnv, origin: "env" };
-  return { key: "", origin: "missing" };
-}
-
-/**
- * A missing key produces an empty catalog and no error the user can see, so
- * say it once, and name the three ways to supply one.
- */
-export function warnIfMissing(resolved: ResolvedApiKey, integrationID: string, log: Logger): void {
-  if (resolved.origin !== "missing") return;
-  log.warn(
-    `[omniroute-v2] no API key for "${integrationID}": the catalog will be empty. ` +
-      `Connect the integration from opencode, set "apiKey" in the plugin options, ` +
-      `or export ${ENV_VAR}.`
-  );
-}
--- a/@omniroute/opencode-plugin-v2/src/enrichment-report.ts
+++ b/@omniroute/opencode-plugin-v2/src/enrichment-report.ts
@@ -1,41 +0,0 @@
-import type { Logger } from "./shared/index.js";
-
-/** What the catalog loses when a given gateway source cannot be read. */
-function consequenceOf(endpoint: string): string {
-  if (endpoint.includes("/api/providers")) {
-    return "the usable-provider filter is disabled for this refresh, so unprovisioned providers stay listed";
-  }
-  return "model names, provider tags, canonical dedupe and pricing are degraded";
-}
-
-/**
- * A source the gateway refuses is not fatal — the catalog still publishes —
- * but staying quiet about it is: the picker then shows raw ids, or lists
- * providers that cannot serve, with nothing telling the user why. Say it once
- * per endpoint so a refresh loop cannot spam the log.
- *
- * `usingFallbackToken` is true when no `managementReadToken` was configured and
- * the inference key stands in for it, which is the usual reason a gateway
- * answers 401/403 on `/api/*` — the advice differs from a token that was set
- * and still got rejected.
- */
-export function createSourceErrorReporter(
-  log: Logger,
-  usingFallbackToken: boolean
-): (endpoint: string, reason: string) => void {
-  const warned = new Set<string>();
-  return (endpoint, reason) => {
-    if (warned.has(endpoint)) return;
-    warned.add(endpoint);
-    const unauthorized = reason.includes("401") || reason.includes("403");
-    const hint = !unauthorized
-      ? ""
-      : usingFallbackToken
-        ? ` These endpoints need a management token: set "managementReadToken" in the plugin options ` +
-          `(it currently falls back to "apiKey", which a gateway usually rejects here).`
-        : ` The configured "managementReadToken" was rejected — check it grants read access to /api/*.`;
-    log.warn(
-      `[omniroute-v2] gateway source ${endpoint} unavailable (${reason}): ${consequenceOf(endpoint)}.${hint}`
-    );
-  };
-}
--- a/@omniroute/opencode-plugin-v2/src/gemini-language.ts
+++ b/@omniroute/opencode-plugin-v2/src/gemini-language.ts
@@ -1,43 +0,0 @@
-import type { LanguageModelV3 } from "@ai-sdk/provider";
-import { type Logger, isGeminiModelId, sanitizeToolInputSchemas } from "./shared/index.js";
-
-type CallOptions = Parameters<LanguageModelV3["doGenerate"]>[0];
-
-/**
- * Gemini answers `400 INVALID_ARGUMENT` — for the entire request, not just the
- * offending tool — when a tool declaration carries `$schema` or
- * `additionalProperties`. Anything upstream that emits standard JSON Schema
- * therefore breaks tool calling as soon as the chain routes to Gemini. A
- * `$ref` is forwarded untouched instead: stripping it would widen the schema
- * to "accept anything", which is worse than letting the gateway answer. The
- * v1 plugin dealt with this by wrapping `fetch` and rewriting the JSON body; the
- * v2 home for it is the language model, where the tools are still structured
- * data and no re-parsing is needed.
- *
- * Returns the model untouched when it is not bound for Gemini, so the wrapper
- * costs nothing on every other chain.
- */
-export function sanitizeToolSchemasFor<T extends LanguageModelV3 | undefined>(
-  language: T,
-  modelId: string,
-  log: Logger
-): T {
-  if (language === undefined) return language;
-  if (!isGeminiModelId(modelId)) return language;
-
-  const clean = (options: CallOptions): CallOptions => {
-    const tools = sanitizeToolInputSchemas(options.tools);
-    if (tools === undefined) return options;
-    log.debug(
-      `[omniroute-v2] stripped Gemini-incompatible schema keywords from ${tools.length} tool declaration(s) for ${modelId}`
-    );
-    return { ...options, tools } as CallOptions;
-  };
-
-  // Prototype-linked so every other member of the model — including accessors
-  // and anything a future SDK version adds — keeps working untouched.
-  const wrapped: LanguageModelV3 = Object.create(language as object) as LanguageModelV3;
-  wrapped.doGenerate = (options) => language.doGenerate(clean(options));
-  wrapped.doStream = (options) => language.doStream(clean(options));
-  return wrapped as T;
-}
--- a/@omniroute/opencode-plugin-v2/src/index.ts
+++ b/@omniroute/opencode-plugin-v2/src/index.ts
@@ -1,539 +0,0 @@
-import { define, type PluginContext } from "@opencode-ai/plugin/v2/promise";
-import {
-  optionalTierFingerprint,
-  catalogContentFingerprint,
-  createLogger,
-  defaultOmniRouteAutoCombosFetcher,
-  defaultOmniRouteCombosFetcher,
-  defaultOmniRouteEnrichmentFetcher,
-  defaultOmniRouteModelsFetcher,
-  defaultOmniRouteProvidersFetcher,
-  type OmniRouteEnrichmentMap,
-  type OmniRouteProviderConnection,
-} from "./shared/index.js";
-import type {
-  OmniRouteRawAutoCombo,
-  OmniRouteRawCombo,
-  OmniRouteRawModelEntry,
-} from "./shared/index.js";
-import type { ResolvedOptions } from "./catalog.js";
-import { publishCatalog } from "./catalog.js";
-import {
-  DEFAULT_MODEL_CACHE_TTL_MS,
-  UNREACHABLE_COOLDOWN_MS,
-  memoryCacheKey,
-  readDiskSnapshot,
-  snapshotIdentityFingerprint,
-  writeDiskSnapshot,
-  type CatalogSnapshot,
-} from "./cache.js";
-import { assertContext } from "./compat.js";
-import { type ApiKeyOrigin, resolveApiKey, warnIfMissing } from "./credentials.js";
-import { createSourceErrorReporter } from "./enrichment-report.js";
-import { sanitizeToolSchemasFor } from "./gemini-language.js";
-import { PLUGIN_ID, parsePluginOptions, resolveTimeouts, type PluginOptions } from "./options.js";
-
-/**
- * A fetch result that says whether it succeeded. Returning a bare `[]` on
- * failure makes an outage indistinguishable from a gateway that legitimately
- * has no combos — and the difference decides whether the last known value
- * should be kept or dropped.
- */
-type SourceResult<T> = { ok: true; value: T } | { ok: false };
-
-interface RefreshState {
-  entries: Map<string, CatalogSnapshot>;
-  inFlight: Map<string, Promise<CatalogSnapshot>>;
-  fingerprint: string | undefined;
-  /** Digest of the optional tier, so a reload only follows a real change. */
-  optionalFingerprint: string | undefined;
-  /**
-   * When the last refresh found the gateway unreachable, skip the network
-   * until this timestamp and serve last-known-good instead. Without it every
-   * transform past TTL re-fires the full fetch suite against a gateway that
-   * just proved it cannot answer — a self-inflicted retry storm.
-   */
-  unreachableUntil: number;
-}
-
-function toResolvedOptions(parsed: PluginOptions): ResolvedOptions {
-  return {
-    providerId: parsed.providerId,
-    baseURL: parsed.baseURL,
-    apiKey: parsed.apiKey ?? process.env.OMNIROUTE_API_KEY ?? "",
-    managementReadToken: parsed.managementReadToken,
-    timeoutMs: parsed.timeoutMs,
-    timeouts: parsed.timeouts,
-    logLevel: parsed.logLevel,
-    startupDebug: parsed.startupDebug,
-    providerTag: parsed.providerTag,
-    modelCacheTtlMs:
-      typeof parsed.modelCacheTtlMs === "number" && parsed.modelCacheTtlMs > 0
-        ? parsed.modelCacheTtlMs
-        : DEFAULT_MODEL_CACHE_TTL_MS,
-    displayName: parsed.displayName,
-    apiFormat: parsed.apiFormat,
-    visibleModels: parsed.visibleModels,
-    hiddenModels: parsed.hiddenModels,
-    usableOnly: parsed.usableOnly,
-    enrichment: parsed.enrichment,
-  };
-}
-
-export default define({
-  id: PLUGIN_ID,
-  setup: async (ctx: PluginContext) => {
-    assertContext(ctx);
-    const parsed = parsePluginOptions(ctx.options);
-    const X = parsed.providerId;
-    const resolved = toResolvedOptions(parsed);
-    const timeouts = resolveTimeouts(parsed);
-    const log = createLogger(parsed.startupDebug ? "debug" : (parsed.logLevel ?? "warn"));
-    resolved.logger = log;
-    resolved.logLevel = parsed.logLevel;
-    resolved.startupDebug = parsed.startupDebug;
-    log.info(`[omniroute-v2] init providerId=${X}`);
-
-    // v1 parity port: in-memory TTL + disk snapshot. The memory key
-    // `baseURL::sha256(creds)` isolates credential tuples (prod vs
-    // staging); the TTL is checked in the transform before any fetch;
-    // concurrent calls share the refresh promise in the setup closure keyed
-    // by (providerId, baseURL); the disk snapshot feeds warm-startup and
-    // the offline fallback. The existing in-memory keep-last-good is kept.
-    const state: RefreshState = {
-      entries: new Map(),
-      inFlight: new Map(),
-      fingerprint: undefined,
-      optionalFingerprint: undefined,
-      unreachableUntil: 0,
-    };
-
-    // The credential the host holds wins over one written in config, so a
-    // user who connected the integration from the UI never has to paste a
-    // key into `opencode.json`. Reading it is async and the transforms must
-    // register synchronously, so the lookup happens on the first publish;
-    // until then the option/env key resolved above stands in.
-    const credentialsOf = (): { cacheKey: string; identityFingerprint: string } => ({
-      cacheKey: memoryCacheKey(
-        resolved.baseURL,
-        `${resolved.apiKey}\0${resolved.managementReadToken ?? resolved.apiKey}`
-      ),
-      identityFingerprint: snapshotIdentityFingerprint(
-        resolved.baseURL,
-        resolved.apiKey,
-        resolved.managementReadToken ?? resolved.apiKey
-      ),
-    });
-    let { cacheKey, identityFingerprint } = credentialsOf();
-
-    // Both keys are derived from the credential: two credentials must never
-    // share a snapshot, so they are recomputed whenever the key moves.
-    let credentialChecked = false;
-    let apiKeyOrigin: ApiKeyOrigin = resolved.apiKey.length > 0 ? "option" : "missing";
-    const ensureCredential = async (): Promise<void> => {
-      // Settled once a key is in hand: re-reading on every refresh would let
-      // a mid-session change silently repoint the snapshot keys.
-      if (credentialChecked && apiKeyOrigin !== "missing") return;
-      const next = await resolveApiKey(ctx, X, parsed.apiKey, log);
-      const moved = next.key !== resolved.apiKey;
-      resolved.apiKey = next.key;
-      apiKeyOrigin = next.origin;
-      if (moved) ({ cacheKey, identityFingerprint } = credentialsOf());
-      if (!credentialChecked) warnIfMissing(next, X, log);
-      else if (moved) log.info(`[omniroute-v2] API key picked up from the ${next.origin} source`);
-      credentialChecked = true;
-    };
-
-    const fetchModelsSafe = async (): Promise<OmniRouteRawModelEntry[]> => {
-      try {
-        return await defaultOmniRouteModelsFetcher(
-          resolved.baseURL,
-          resolved.apiKey,
-          timeouts.models
-        );
-      } catch (err) {
-        log.warn(
-          `[omniroute-v2] models fetch failed, publishing empty catalog: ${err instanceof Error ? err.message : String(err)}`
-        );
-        return [];
-      }
-    };
-    // Failures are reported once per endpoint (with the management-token hint
-    // when the inference key stands in), so a gated `/api/*` degrades loudly
-    // rather than silently. Declared before the wrappers that use it.
-    const reportSourceError = createSourceErrorReporter(
-      log,
-      resolved.managementReadToken === undefined
-    );
-    const fetchCombosSafe = async (): Promise<SourceResult<OmniRouteRawCombo[]>> => {
-      try {
-        return {
-          ok: true,
-          value: await defaultOmniRouteCombosFetcher(
-            resolved.baseURL,
-            resolved.managementReadToken ?? resolved.apiKey,
-            timeouts.combos
-          ),
-        };
-      } catch (err) {
-        const reason = err instanceof Error ? err.message : String(err);
-        reportSourceError("/api/combos", reason);
-        log.warn(`[omniroute-v2] combos fetch failed, keeping the last known combos: ${reason}`);
-        return { ok: false };
-      }
-    };
-    // Providers connections follow the same rule: gated on usableOnly (no
-    // request when false, v1 parity), soft-fail to [] so the filter degrades
-    // to keep-all instead of hiding the catalog.
-    const fetchProvidersSafe = async (): Promise<SourceResult<OmniRouteProviderConnection[]>> => {
-      if (!resolved.usableOnly) return { ok: true, value: [] };
-      try {
-        return {
-          ok: true,
-          value: await defaultOmniRouteProvidersFetcher(
-            resolved.baseURL,
-            resolved.managementReadToken ?? resolved.apiKey,
-            timeouts.models,
-            reportSourceError
-          ),
-        };
-      } catch (err) {
-        log.warn(
-          `[omniroute-v2] providers fetch failed, keeping the last known provider list: ${err instanceof Error ? err.message : String(err)}`
-        );
-        return { ok: false };
-      }
-    };
-    // Enrichment follows the same rule: gated on the option (default on,
-    // v1 parity), soft-fail to an empty map so names/pricing degrade to
-    // mapper defaults instead of hiding the catalog.
-    const fetchEnrichmentSafe = async (): Promise<SourceResult<OmniRouteEnrichmentMap>> => {
-      if (resolved.enrichment === false) return { ok: true, value: new Map() };
-      try {
-        return {
-          ok: true,
-          value: await defaultOmniRouteEnrichmentFetcher(
-            resolved.baseURL,
-            resolved.managementReadToken ?? resolved.apiKey,
-            timeouts.enrichment,
-            reportSourceError
-          ),
-        };
-      } catch (err) {
-        log.warn(
-          `[omniroute-v2] enrichment fetch failed, keeping the last known names/pricing: ${err instanceof Error ? err.message : String(err)}`
-        );
-        return { ok: false };
-      }
-    };
-    const fetchAutoCombosSafe = async (): Promise<SourceResult<OmniRouteRawAutoCombo[]>> => {
-      try {
-        return {
-          ok: true,
-          value: await defaultOmniRouteAutoCombosFetcher(
-            resolved.baseURL,
-            resolved.managementReadToken ?? resolved.apiKey,
-            timeouts.autoCombos,
-            log,
-            reportSourceError
-          ),
-        };
-      } catch (err) {
-        // The default fetcher reports the refusal itself (with the
-        // management-token hint); this warn is the fallback for injected
-        // stubs that throw without reporting.
-        const reason = err instanceof Error ? err.message : String(err);
-        log.warn(`[omniroute-v2] auto combos fetch failed, keeping the last known ones: ${reason}`);
-        return { ok: false };
-      }
-    };
-
-    /**
-     * Fetch in two tiers. Models are what a catalog *is*: without them there
-     * is nothing to publish. Everything else — combos, auto-combos, the
-     * provider list, the enrichment overlay — improves an already usable
-     * catalog, so awaiting any of them before publishing makes the catalog
-     * hostage to the slowest source: a gateway that accepts the connection
-     * and never answers one endpoint kept everything unpublished until that
-     * fetch's own timeout fired, which is longer than some hosts stay alive.
-     *
-     * The optional tier therefore keeps running after the publish and upgrades
-     * the stored snapshot when it lands, so the next transform serves the
-     * complete catalog.
-     */
-    async function refreshSnapshot(): Promise<CatalogSnapshot> {
-      // Models are what a catalog *is*; everything else improves one that
-      // already works. Combos used to sit here too, so a gateway slow to
-      // answer /api/combos held the whole picker back — the very thing the
-      // staged publish exists to prevent.
-      const essential = fetchModelsSafe();
-      const optional = Promise.all([
-        fetchCombosSafe(),
-        fetchAutoCombosSafe(),
-        fetchProvidersSafe(),
-        fetchEnrichmentSafe(),
-      ]);
-      const models = await essential;
-      const previous = state.entries.get(cacheKey);
-      // A gateway that just failed everything gets a short breather: serving
-      // last-known-good for a few seconds beats hammering it on every
-      // transform while it is down. Arms whenever the models fetch comes back
-      // empty — with or without a prior entry to serve — so a totally dead
-      // gateway stops getting hit every window. Partial degradation (models
-      // healthy, an optional tier failed) still retries normally next window.
-      if (models.length === 0) {
-        state.unreachableUntil = Date.now() + UNREACHABLE_COOLDOWN_MS;
-      }
-      // Carry every source forward until its replacement lands, and keep the
-      // old value when a fetch FAILED — but honour a gateway that legitimately
-      // returns nothing, which is a different answer from "I could not ask".
-      const snapshot: CatalogSnapshot = {
-        models,
-        combos: previous?.combos ?? [],
-        autoCombos: previous?.autoCombos ?? [],
-        providers: previous?.providers ?? [],
-        enrichment: previous?.enrichment ?? new Map(),
-        fetchedAt: Date.now(),
-      };
-      if (models.length > 0) {
-        state.entries.set(cacheKey, snapshot);
-        await writeDiskSnapshot(X, snapshot, identityFingerprint);
-      }
-      void optional.then(
-        (parts) => upgradeWithOptional(snapshot, parts),
-        (err) => {
-          // The wrappers never reject; a throw here would be a bug in them, and
-          // an unhandled rejection is a worse way to learn about it.
-          log.warn(
-            `[omniroute-v2] optional catalog sources failed unexpectedly: ${err instanceof Error ? err.message : String(err)}`
-          );
-        }
-      );
-      return snapshot;
-    }
-
-    /**
-     * Fold late optional data into the snapshot that was published without it.
-     * Skipped when a newer refresh has already replaced that snapshot, so a
-     * slow tier can never resurrect a stale catalog.
-     */
-    async function upgradeWithOptional(
-      base: CatalogSnapshot,
-      [combos, autoCombos, providers, enrichment]: [
-        SourceResult<OmniRouteRawCombo[]>,
-        SourceResult<OmniRouteRawAutoCombo[]>,
-        SourceResult<OmniRouteProviderConnection[]>,
-        SourceResult<OmniRouteEnrichmentMap>,
-      ]
-    ): Promise<void> {
-      if (state.entries.get(cacheKey) !== base) return;
-      // Per source: a success replaces (even with an empty answer — that is
-      // the gateway's answer), a failure keeps what we had.
-      const upgraded: CatalogSnapshot = {
-        ...base,
-        combos: combos.ok ? combos.value : base.combos,
-        autoCombos: autoCombos.ok ? autoCombos.value : base.autoCombos,
-        providers: providers.ok ? providers.value : base.providers,
-        enrichment: enrichment.ok ? enrichment.value : base.enrichment,
-      };
-      const unchanged =
-        upgraded.combos === base.combos &&
-        upgraded.autoCombos === base.autoCombos &&
-        upgraded.providers === base.providers &&
-        upgraded.enrichment === base.enrichment;
-      if (unchanged) return;
-      state.entries.set(cacheKey, upgraded);
-      if (upgraded.models.length > 0) {
-        await writeDiskSnapshot(X, upgraded, identityFingerprint);
-      }
-      // Reload only when the optional tier actually moved: the catalog
-      // fingerprint covers ids alone, so without this the host would rebuild
-      // its catalog once per TTL window for an identical result.
-      const optionalFingerprint = optionalTierFingerprint(
-        upgraded.autoCombos ?? [],
-        upgraded.providers ?? [],
-        upgraded.enrichment,
-        upgraded.combos
-      );
-      const optionalChanged = state.optionalFingerprint !== optionalFingerprint;
-      state.optionalFingerprint = optionalFingerprint;
-      if (optionalChanged && typeof ctx.catalog.reload === "function") {
-        try {
-          await ctx.catalog.reload();
-        } catch (err) {
-          log.warn(
-            `[omniroute-v2] catalog reload after late sources failed, keeping current catalog: ${err instanceof Error ? err.message : String(err)}`
-          );
-        }
-      }
-    }
-
-    function loadSnapshot(): Promise<CatalogSnapshot> {
-      const now = Date.now();
-      const hit = state.entries.get(cacheKey);
-      if (hit && hit.fetchedAt + resolved.modelCacheTtlMs > now) return Promise.resolve(hit);
-      // Cooldown after a total models failure: skip the network until it
-      // lapses. Serves last-known-good when one exists; otherwise the refresh
-      // below still runs (nothing to serve, no point pretending).
-      if (now < state.unreachableUntil && hit) return Promise.resolve(hit);
-      if (now >= state.unreachableUntil) state.unreachableUntil = 0;
-      const inflight = state.inFlight.get(cacheKey);
-      if (inflight) return inflight;
-      const snapshot = refreshSnapshot();
-      state.inFlight.set(cacheKey, snapshot);
-      const clear = () => {
-        if (state.inFlight.get(cacheKey) === snapshot) state.inFlight.delete(cacheKey);
-      };
-      snapshot.then(clear, clear);
-      return snapshot;
-    }
-
-    // Warm-startup: the disk snapshot is read at boot (without blocking
-    // the synchronous transform registration) to publish the last-known
-    // catalog before the first successful fetch.
-    /**
-     * Warm start: publish the last known catalog from disk before the first
-     * fetch returns. Deliberately read *after* the credential is resolved —
-     * the snapshot is keyed by the credential tuple, and resolving the host
-     * credential changes that key, so reading at setup time would look up the
-     * wrong identity and reject a perfectly good snapshot.
-     */
-    let warmLoadedFor: string | undefined;
-    const ensureWarmSnapshot = async (): Promise<void> => {
-      if (warmLoadedFor === identityFingerprint) return;
-      warmLoadedFor = identityFingerprint;
-      const warm = await readDiskSnapshot(X, identityFingerprint, log);
-      if (warm && !state.entries.has(cacheKey)) state.entries.set(cacheKey, warm);
-    };
-
-    // Fail-closed models (keep-last-good, validated): an empty models fetch
-    // (transient 500/timeout) must not wipe a known catalog. The latest
-    // non-empty entry (fresh fetch or warm disk snapshot) is replayed
-    // instead of publishing the empty set. `refreshSnapshot` never overwrites
-    // the memory entry on failure, so `entries` stays the last-known-good
-    // source — including cross-setup via the disk snapshot.
-    // Fail-open one level down, in the wrappers (never reject) and the
-    // `publishCatalog` catches — so no try/catch here.
-    const catalogRegistration = ctx.catalog.transform(async (draft) => {
-      await ensureCredential();
-      await ensureWarmSnapshot();
-      const snapshot = await loadSnapshot();
-      let effective = snapshot;
-      if (snapshot.models.length === 0) {
-        const stale = state.entries.get(cacheKey);
-        if (stale !== undefined && stale.models.length > 0) {
-          log.warn(
-            `[omniroute-v2] models fetch returned empty, keeping last-known catalog (${stale.models.length} models, ${stale.combos.length} combos)`
-          );
-          effective = stale;
-        }
-      }
-      const counts = await (async (): Promise<{
-        models: number;
-        combos: number;
-        autoCombos: number;
-      }> => {
-        // fetcher-level fail-open covers fetches; this guard covers mapper/draft throws.
-        try {
-          return await publishCatalog(draft, resolved, {
-            onSourceError: reportSourceError,
-            models: async () => effective.models,
-            combos: async () => effective.combos,
-            autoCombos: async () => effective.autoCombos,
-            providers: async () => effective.providers ?? [],
-            enrichment: async () => effective.enrichment ?? new Map(),
-          });
-        } catch (err) {
-          log.warn(
-            `[omniroute-v2] catalog publish failed, keeping current catalog: ${err instanceof Error ? err.message : String(err)}`
-          );
-          return { models: 0, combos: 0, autoCombos: 0 };
-        }
-      })();
-      void counts;
-      const fingerprint = catalogContentFingerprint(
-        effective.models,
-        effective.combos,
-        effective.autoCombos
-      );
-      const changed = state.fingerprint !== undefined && state.fingerprint !== fingerprint;
-      state.fingerprint = fingerprint;
-      if (changed && typeof ctx.catalog.reload === "function") {
-        await Promise.resolve();
-        try {
-          await ctx.catalog.reload();
-        } catch (err) {
-          log.warn(
-            `[omniroute-v2] catalog reload failed, keeping current catalog: ${err instanceof Error ? err.message : String(err)}`
-          );
-        }
-      }
-    });
-    const integrationHook = (ctx.integration as Partial<PluginContext["integration"]> | undefined)
-      ?.transform;
-    // A host that exposes the hook but throws while registering it must cost
-    // the plugin nothing but the connect action: the throw happens OUTSIDE
-    // any await, so only a call-site guard catches it (an await-guard alone
-    // would let a synchronous throw escape setup and kill the catalog).
-    let integrationRegistration: unknown;
-    if (typeof integrationHook === "function") {
-      try {
-        integrationRegistration = integrationHook((draft) => {
-          draft.update(X, (integration) => {
-            integration.name = parsed.displayName ?? "OmniRoute";
-          });
-          draft.method.update({ integrationID: X, method: { type: "key", label: "API key" } });
-          draft.method.update({
-            integrationID: X,
-            method: { type: "env", names: ["OMNIROUTE_API_KEY"] },
-          });
-        });
-      } catch (err) {
-        log.warn(
-          `[omniroute-v2] host refused the integration hook, the connect action will be missing: ${err instanceof Error ? err.message : String(err)}`
-        );
-        integrationRegistration = undefined;
-      }
-    }
-    /**
-     * `aisdk.language` is newer than the catalog domain, so a host may not
-     * expose it; the plugin must stay loadable there, minus the sanitising.
-     */
-    const languageHook = (ctx.aisdk as Partial<PluginContext["aisdk"]> | undefined)?.language;
-    // A host that rejects this registration must cost the catalog nothing: the
-    // plugin is a catalog first, and tool-schema cleaning is an extra.
-    let languageRegistration: Promise<{ dispose: () => Promise<void> }> | undefined;
-    if (parsed.geminiSanitization !== false && typeof languageHook === "function") {
-      try {
-        languageRegistration = languageHook((input) => {
-          if (input.model.providerID !== X) return;
-          input.language = sanitizeToolSchemasFor(input.language, input.model.id, log);
-        });
-      } catch (err) {
-        log.warn(
-          `[omniroute-v2] host refused the language-model hook, Gemini tool schemas will not be cleaned: ${err instanceof Error ? err.message : String(err)}`
-        );
-      }
-    }
-
-    await catalogRegistration;
-    if (integrationRegistration !== undefined) {
-      try {
-        await integrationRegistration;
-      } catch (err) {
-        log.warn(
-          `[omniroute-v2] host refused the integration hook, the connect action will be missing: ${err instanceof Error ? err.message : String(err)}`
-        );
-      }
-    }
-    if (languageRegistration !== undefined) {
-      try {
-        await languageRegistration;
-      } catch (err) {
-        log.warn(
-          `[omniroute-v2] language-model hook registration failed, Gemini tool schemas will not be cleaned: ${err instanceof Error ? err.message : String(err)}`
-        );
-      }
-    }
-  },
-});
--- a/@omniroute/opencode-plugin-v2/src/options.ts
+++ b/@omniroute/opencode-plugin-v2/src/options.ts
@@ -1,117 +0,0 @@
-import { z } from "zod";
-
-const apiFormatSchema = z
-  .object({
-    allowAnthropic: z.boolean().optional(),
-    anthropicModels: z.array(z.string()).optional(),
-    // Deprecated v1 prefix list. Accepted (warn at resolve time) so copied
-    // v1 configs keep routing; prefer anthropicModels (full IDs).
-    anthropicPrefixes: z.array(z.string()).optional(),
-  })
-  .strict();
-
-const timeoutsSchema = z
-  .object({
-    models: z.number().positive().optional(),
-    combos: z.number().positive().optional(),
-    autoCombos: z.number().positive().optional(),
-    enrichment: z.number().positive().optional(),
-  })
-  .strict();
-
-const pluginOptionsSchema = z
-  .object({
-    // Reaches a filesystem path (the on-disk catalog snapshot) and the
-    // catalog keys, so it is bounded here rather than escaped at each use.
-    providerId: z
-      .string()
-      .regex(/^[A-Za-z0-9._-]+$/, "providerId may only contain letters, digits, '.', '_' and '-'")
-      .refine((v) => v !== "." && v !== "..", "providerId cannot be a path segment")
-      .default("omniroute"),
-    baseURL: z.string().url(),
-    apiKey: z.string().optional(),
-    displayName: z.string().optional(),
-    managementReadToken: z.string().optional(),
-    timeoutMs: z.number().positive().default(10000),
-    timeouts: timeoutsSchema.optional(),
-    logLevel: z.enum(["error", "warn", "info", "debug"]).optional(),
-    startupDebug: z.boolean().optional(),
-    modelCacheTtlMs: z.number().positive().optional(),
-    visibleModels: z.array(z.string()).optional(),
-    hiddenModels: z.array(z.string()).optional(),
-    usableOnly: z.boolean().default(false),
-    // v1 parity: enrichment overlay on by default (names + pricing).
-    enrichment: z.boolean().default(true),
-    // v1 parity: strip the JSON-Schema keywords Gemini rejects from tool
-    // declarations bound for a Gemini model. On by default — leaving them in
-    // fails the whole request with 400 INVALID_ARGUMENT.
-    geminiSanitization: z.boolean().default(true),
-    // v1 parity: prefix a model's display name with the upstream provider it
-    // routes to, so the same model sold through two connections is
-    // distinguishable in the picker.
-    providerTag: z.boolean().default(true),
-    apiFormat: apiFormatSchema.optional(),
-  })
-  .strict();
-
-export type PluginOptions = z.infer<typeof pluginOptionsSchema>;
-
-/** Per-endpoint timeout defaults (v1 parity). `timeoutMs` is the global fallback. */
-export const DEFAULT_TIMEOUT_MS = 10_000 as const;
-/** Auto-combos keep the v1 5s budget; the field is resolved now for the P3 port. */
-export const DEFAULT_AUTO_COMBOS_TIMEOUT_MS = 5_000 as const;
-
-export interface EndpointTimeouts {
-  models: number;
-  combos: number;
-  autoCombos: number;
-  enrichment: number;
-}
-
-export function resolveTimeouts(
-  opts: Pick<PluginOptions, "timeoutMs" | "timeouts">
-): EndpointTimeouts {
-  const fallback =
-    typeof opts.timeoutMs === "number" && opts.timeoutMs > 0 ? opts.timeoutMs : DEFAULT_TIMEOUT_MS;
-  return {
-    models: opts.timeouts?.models ?? fallback,
-    combos: opts.timeouts?.combos ?? fallback,
-    autoCombos: opts.timeouts?.autoCombos ?? DEFAULT_AUTO_COMBOS_TIMEOUT_MS,
-    enrichment: opts.timeouts?.enrichment ?? fallback,
-  };
-}
-
-/**
- * Parse the plugin block of `opencode.json`.
- *
- * A rejected option aborts the whole plugin, and the host reports that as a
- * bare load failure with the validator's raw dump attached — which is how a
- * single mistyped key turns into a wall of JSON and an empty model picker. The
- * schema is strict on purpose (a silently ignored option is worse), so the
- * least we owe the user is a first line naming what to fix.
- */
-export function parsePluginOptions(raw: unknown): PluginOptions {
-  const result = pluginOptionsSchema.safeParse(raw);
-  if (result.success) return result.data;
-  const problems = result.error.issues.map((issue) => {
-    const at = issue.path.length > 0 ? issue.path.join(".") : "(root)";
-    const unknown = issue.code === "unrecognized_keys" ? issue.keys.join(", ") : undefined;
-    return unknown !== undefined ? `unknown option "${unknown}"` : `${at}: ${issue.message}`;
-  });
-  throw new Error(`[omniroute-v2] invalid plugin options — ${problems.join("; ")}`);
-}
-
-/**
- * The host reads the plugin id from the module, before any option is known, so
- * it cannot carry the configured provider id. Publishing two gateways from one
- * install is a `providerId` matter — that one does reach the catalog.
- */
-export const PLUGIN_ID = "omniroute-v2";
-
-export function providerIdFor(providerId: string): string {
-  return providerId;
-}
-
-export function integrationIdFor(providerId: string): string {
-  return providerId;
-}
--- a/@omniroute/opencode-plugin-v2/src/shared/auto-combos.ts
+++ b/@omniroute/opencode-plugin-v2/src/shared/auto-combos.ts
@@ -1,219 +0,0 @@
-import type { Model as ModelV2 } from "@opencode-ai/sdk/v2";
-import type { ApiFormatV2 } from "./models-map.js";
-import { resolveApiBlockV2 } from "./models-map.js";
-import { autoComboModelId, formatAutoComboName, type AutoVariant } from "./naming.js";
-
-export type { AutoVariant };
-
-/**
- * Raw shape of an auto combo entry as returned by OmniRoute's
- * `/api/combos/auto` endpoint. Auto combos are virtual -- they self-manage
- * provider selection via scoring/bandit exploration at runtime.
- *
- * Ported from the v1 plugin (`index.ts:1672-1698`); the shape is unchanged
- * so old and new gateways stay wire-compatible.
- */
-export interface OmniRouteRawAutoCombo {
-  /** Stable id (e.g. "auto", "auto/coding"). */
-  id: string;
-  /** Human-readable name (e.g. "Auto", "Auto Coding"). */
-  name?: string;
-  /** Variant key or undefined for the default auto. */
-  variant?: AutoVariant;
-  /** Provider names eligible for this auto combo. */
-  candidatePool?: string[];
-  /** Number of candidates resolved at fetch time. */
-  candidateCount?: number;
-  /** MAX of candidates' context windows, served by newer gateway builds.
-   * Absent on older servers -- the mapper falls back to a safe default. */
-  context_length?: number;
-  /** MAX of candidates' max output tokens (same provenance as context_length). */
-  max_output_tokens?: number;
-  /** Whether this auto combo should be hidden from the picker. */
-  isHidden?: boolean;
-  /** Auto-combo configuration. */
-  config?: {
-    auto?: {
-      candidatePool?: string[];
-      explorationRate?: number;
-      routerStrategy?: string;
-    };
-  };
-}
-
-/** Minimal warn sink so the fetcher never depends on the plugin logger. */
-export interface AutoCombosWarnSink {
-  warn: (message: string, ...args: unknown[]) => void;
-}
-
-/**
- * Fetcher contract for `/api/combos/auto`. Returns the list of virtual
- * auto combos the server can create. Same DI shape as the other fetchers
- * so unit tests can inject a stub instead of monkey-patching `fetch`.
- *
- * HTTP refusals (non-2xx other than 404) and network errors THROW: the caller
- * distinguishes "the gateway failed" (keep last-known) from "the gateway
- * answered empty" (publish empty). Only 404 stays soft — the endpoint does
- * not exist yet on older gateways, and that is an answer, not a failure.
- */
-export type OmniRouteAutoCombosFetcher = (
-  baseURL: string,
-  apiKey: string,
-  timeoutMs?: number,
-  logger?: AutoCombosWarnSink,
-  onSourceError?: (endpoint: string, reason: string) => void
-) => Promise<OmniRouteRawAutoCombo[]>;
-
-function trimTrailingSlashes(value: string): string {
-  let i = value.length;
-  while (i > 0 && value.charCodeAt(i - 1) === 0x2f /* "/" */) i--;
-  return i === value.length ? value : value.slice(0, i);
-}
-
-function fallbackWarn(message: string, ...args: unknown[]): void {
-  console.warn(`[omniroute-plugin] [WARN] ${message}`, ...args);
-}
-
-/**
- * Default auto combos fetcher: `GET <baseURL>/api/combos/auto`.
- *
- * 404 stays soft (endpoint not deployed yet on older gateways — an answer,
- * not a failure). Any other non-2xx or network error THROWS so the caller
- * keeps last-known instead of publishing an empty tier: a 403 behind a
- * management-token gate must not wipe the auto combos the picker had.
- * v1 parity keeps the 5s timeout budget.
- */
-export const defaultOmniRouteAutoCombosFetcher: OmniRouteAutoCombosFetcher = async (
-  baseURL,
-  apiKey,
-  timeoutMs = 5_000,
-  logger?: AutoCombosWarnSink,
-  onSourceError?: (endpoint: string, reason: string) => void
-) => {
-  if (!apiKey || !baseURL) return [];
-  const warn = logger?.warn ?? fallbackWarn;
-  const report = (reason: string): void => {
-    warn(reason);
-    onSourceError?.("/api/combos/auto", reason);
-  };
-
-  const trimmed = trimTrailingSlashes(baseURL);
-  const root = trimmed.replace(/\/v\d+$/, "");
-  const url = `${root}/api/combos/auto`;
-
-  const controller = new AbortController();
-  const timer = setTimeout(() => controller.abort(), timeoutMs);
-  try {
-    const res = await fetch(url, {
-      method: "GET",
-      headers: {
-        Authorization: `Bearer ${apiKey}`,
-        Accept: "application/json",
-      },
-      signal: controller.signal,
-    });
-    // 404 = endpoint not deployed yet -- expected during rollout
-    if (res.status === 404) {
-      warn(`/api/combos/auto not available (404) -- auto combos disabled`);
-      return [];
-    }
-    if (!res.ok) {
-      const reason = `HTTP ${res.status} ${res.statusText}`;
-      report(`/api/combos/auto refused (${reason}) -- keeping last-known auto combos`);
-      throw new Error(reason);
-    }
-    const body = (await res.json()) as unknown;
-    const rawList: unknown[] = Array.isArray(body)
-      ? body
-      : body && typeof body === "object" && Array.isArray((body as { combos?: unknown }).combos)
-        ? ((body as { combos: unknown[] }).combos as unknown[])
-        : [];
-    const out: OmniRouteRawAutoCombo[] = [];
-    for (const r of rawList) {
-      if (r && typeof r === "object" && typeof (r as { id?: unknown }).id === "string") {
-        out.push(r as OmniRouteRawAutoCombo);
-      }
-    }
-    return out;
-  } catch (err) {
-    // Network error, timeout, abort -- keep last-known, never publish empty.
-    // (The 404-soft path above returns directly and never reaches this throw.)
-    const reason = `/api/combos/auto fetch failed: ${err instanceof Error ? err.message : String(err)} -- keeping last-known auto combos`;
-    report(reason);
-    throw err instanceof Error ? err : new Error(String(err));
-  } finally {
-    clearTimeout(timer);
-  }
-};
-
-/** Fallbacks when the server does not advertise auto-combo limits (older
- * gateway builds). MUST be positive: OpenCode's overflow guard treats
- * `limit.context === 0` as "never overflow" and silently DISABLES smart
- * auto-compaction, letting the session grow until the gateway's destructive
- * history purge kicks in. */
-export const AUTO_COMBO_FALLBACK_CONTEXT = 128_000;
-export const AUTO_COMBO_FALLBACK_OUTPUT = 8_192;
-
-/**
- * Convert a raw auto combo into a `ModelV2` entry for the picker.
- * Auto combos route to capable models, so tool_call and reasoning default
- * to true. Context/output limits come from the server (MAX of the
- * candidate pool's windows); a safe positive fallback applies when the
- * server omits them. Never 0.
- */
-export function mapAutoComboToModelV2(
-  autoCombo: OmniRouteRawAutoCombo,
-  providerId: string,
-  baseURL: string,
-  apiFormat?: ApiFormatV2
-): ModelV2 {
-  const name = formatAutoComboName(autoCombo.variant, autoCombo.candidateCount);
-  const context =
-    typeof autoCombo.context_length === "number" && autoCombo.context_length > 0
-      ? autoCombo.context_length
-      : AUTO_COMBO_FALLBACK_CONTEXT;
-  const output =
-    typeof autoCombo.max_output_tokens === "number" && autoCombo.max_output_tokens > 0
-      ? autoCombo.max_output_tokens
-      : AUTO_COMBO_FALLBACK_OUTPUT;
-  return {
-    id: autoComboModelId(autoCombo.variant),
-    providerID: providerId,
-    api: resolveApiBlockV2(autoComboModelId(autoCombo.variant), baseURL, apiFormat),
-    name,
-    capabilities: {
-      temperature: true,
-      reasoning: true,
-      attachment: false,
-      toolcall: true,
-      input: {
-        text: true,
-        audio: false,
-        image: false,
-        video: false,
-        pdf: false,
-      },
-      output: {
-        text: true,
-        audio: false,
-        image: false,
-        video: false,
-        pdf: false,
-      },
-      interleaved: false,
-    },
-    cost: {
-      input: 0,
-      output: 0,
-      cache: { read: 0, write: 0 },
-    },
-    limit: {
-      context,
-      output,
-    },
-    status: "active",
-    options: {},
-    headers: {},
-    release_date: "",
-  };
-}
--- a/@omniroute/opencode-plugin-v2/src/shared/combos-map.ts
+++ b/@omniroute/opencode-plugin-v2/src/shared/combos-map.ts
@@ -1,254 +0,0 @@
-import type { Model as ModelV2 } from "@opencode-ai/sdk/v2";
-import { type ApiFormatV2, type OmniRouteRawModelEntry, resolveApiBlockV2 } from "./models-map.js";
-
-export interface OmniRouteRawComboMemberRef {
-  /** Step kind: "model" references a raw model id; "combo-ref" nests another combo. */
-  kind?: "model" | "combo-ref";
-  /** Full model id referenced by this step (when kind === "model"). */
-  model?: string;
-  /** Nested combo name (when kind === "combo-ref"). */
-  comboName?: string;
-  /** Routing weight inside the combo (0–100, advisory at LCD time). */
-  weight?: number;
-  /** Step-local label, distinct from the parent combo's display name. */
-  label?: string;
-}
-
-export interface OmniRouteRawCombo {
-  id: string;
-  name?: string;
-  /** Routing strategy. Surfaced for forward-compat but not consumed by LCD. */
-  strategy?: string;
-  /** Member step list. Only `kind: "model"` steps participate in LCD. */
-  models?: OmniRouteRawComboMemberRef[];
-  /** Hidden combos are excluded from the OC model picker. */
-  isHidden?: boolean;
-  /** When OmniRoute attaches a lifecycle hint we forward it; today it doesn't. */
-  release_date?: string;
-  /**
-   * Server-computed context window for this combo (aggregated from member
-   * models using the same logic as /v1/models). When present, the client
-   * uses this value directly instead of re-aggregating from member models.
-   *
-   * Added in 3.9.x — old servers do not send it.
-   */
-  computed_context_length?: number;
-}
-
-/**
- * Fetcher contract for `/api/combos`. Same DI shape as
- * `OmniRouteModelsFetcher` so unit tests can inject a stub instead of
- * monkey-patching global `fetch`.
- */
-export type OmniRouteCombosFetcher = (
-  baseURL: string,
-  apiKey: string,
-  timeoutMs?: number
-) => Promise<OmniRouteRawCombo[]>;
-
-function trimTrailingSlashes(value: string): string {
-  let i = value.length;
-  while (i > 0 && value.charCodeAt(i - 1) === 0x2f /* "/" */) i--;
-  return i === value.length ? value : value.slice(0, i);
-}
-
-/**
- * Default fetcher: `GET <baseURL>/api/combos` with bearer auth +
- * AbortController timeout. Accepts both the `{combos: [...]}` envelope the
- * gateway emits today and a bare-array envelope (defensive — keeps the
- * plugin working if a future OmniRoute build trims the wrapper).
- *
- * Differences from `defaultOmniRouteModelsFetcher`:
- *   - URL is `/api/combos`, NOT `/v1/combos`. The `/v1/...` namespace is the
- *     OpenAI-compatible surface (chat completions, models); combo discovery
- *     lives on the management plane under `/api/...`. We tolerate both
- *     `https://host` and `https://host/v1` baseURL forms by stripping the
- *     trailing `/v1` segment before appending `/api/combos`.
- *   - Combos endpoint requires a management-scoped API key when
- *     `REQUIRE_API_KEY` is enabled. We don't enforce that here; the
- *     gateway returns 401/403 with an actionable error which we propagate.
- *
- * Anything that isn't an object with a string `id` is filtered out silently.
- */
-export const defaultOmniRouteCombosFetcher: OmniRouteCombosFetcher = async (
-  baseURL,
-  apiKey,
-  timeoutMs = 10_000
-) => {
-  if (!apiKey) throw new Error("[omniroute-v2] apiKey required to fetch /api/combos");
-  if (!baseURL) throw new Error("[omniroute-v2] baseURL required to fetch /api/combos");
-
-  // Strip trailing slashes, then strip a trailing `/v1` so we land on the
-  // management plane. Models live under `/v1/models`; combos live under
-  // `/api/combos` from the same gateway root.
-  const trimmed = trimTrailingSlashes(baseURL);
-  const root = trimmed.replace(/\/v\d+$/, "");
-  const url = `${root}/api/combos`;
-
-  const controller = new AbortController();
-  const timer = setTimeout(() => controller.abort(), timeoutMs);
-  try {
-    const res = await fetch(url, {
-      method: "GET",
-      headers: {
-        Authorization: `Bearer ${apiKey}`,
-        Accept: "application/json",
-      },
-      signal: controller.signal,
-    });
-    if (!res.ok) {
-      throw new Error(`[omniroute-v2] GET ${url} failed: ${res.status} ${res.statusText}`);
-    }
-    const body = (await res.json()) as unknown;
-    const rawList: unknown[] = Array.isArray(body)
-      ? body
-      : body && typeof body === "object" && Array.isArray((body as { combos?: unknown }).combos)
-        ? ((body as { combos: unknown[] }).combos as unknown[])
-        : [];
-    const out: OmniRouteRawCombo[] = [];
-    for (const r of rawList) {
-      if (r && typeof r === "object" && typeof (r as { id?: unknown }).id === "string") {
-        out.push(r as OmniRouteRawCombo);
-      }
-    }
-    return out;
-  } finally {
-    clearTimeout(timer);
-  }
-};
-
-/**
- * Map a raw combo entry → `ModelV2` by computing the lowest-common-denominator
- * (LCD) of its underlying member models. The LCD policy is the only way to
- * surface a single capability vector to OpenCode without lying: if any member
- * lacks a capability, the combo as a whole cannot guarantee it.
- *
- * LCD rules:
- *   - `limit.context` = `min(...members.context_length)`.
- *   - `limit.output` = `min(...members.max_output_tokens)`.
- *   - `limit.input` = `min(...members.max_input_tokens)` ONLY when every
- *     member declares one (ModelV2.limit.input is optional — better to
- *     omit than to fabricate a min over partial data).
- *   - `capabilities.toolcall` / `reasoning` / `attachment` / `temperature`:
- *     `every(member ⇒ supports?)`. The `reasoning` axis ORs across
- *     `reasoning` and `thinking` per member before AND-ing across the
- *     combo (mirrors `mapRawModelToModelV2`). The `attachment` axis ORs
- *     across `attachment` and `vision` per member. The `temperature` axis
- *     uses default-true semantics: a member supports temperature unless
- *     it explicitly declares `temperature: false`.
- *   - `capabilities.input.*` / `output.*`: flattened AND across members'
- *     modality flags. Missing arrays default to `["text"]` (same default
- *     as `mapRawModelToModelV2`).
- *
- * Defensive: empty members array → ALL capabilities `false`, limits zero.
- * That's an intentional safety posture (you can't route through an empty
- * combo, so OC should grey it out in the picker).
- *
- * Spec mapping: `cost` zeroed; `status = "active"`;
- * `release_date = combo.release_date ?? ""`;
- * `api = LCD (all-anthropic else openai-compatible)`;
- * `name = combo.name ?? combo.id`.
- *
- * @param combo Raw `/api/combos` entry.
- * @param members Raw `/v1/models` entries for THIS combo's member ids.
- *                Caller resolves `combo.models[].model` ids; unknown ids
- *                are silently dropped before this call.
- * @param providerId OpenCode provider id (multi-instance aware).
- * @param baseURL Resolved gateway base URL for ModelV2.api.url.
- */
-export function mapComboToModelV2(
-  combo: OmniRouteRawCombo,
-  members: OmniRouteRawModelEntry[],
-  providerId: string,
-  baseURL: string,
-  apiFormat?: ApiFormatV2
-): ModelV2 {
-  // `every` over an empty array returns true (would lie about an empty
-  // combo's capabilities) — short-circuit to all-false when no members.
-  const hasMembers = members.length > 0;
-
-  const memberInMods = members.map((m) => new Set(m.input_modalities ?? ["text"]));
-  const memberOutMods = members.map((m) => new Set(m.output_modalities ?? ["text"]));
-
-  const modalityAllHave = (sets: Array<Set<string>>, key: string): boolean =>
-    hasMembers && sets.every((s) => s.has(key));
-
-  const contextValues = members
-    .map((m) => m.context_length)
-    .filter((v): v is number => typeof v === "number" && v > 0);
-  const outputValues = members
-    .map((m) => m.max_output_tokens)
-    .filter((v): v is number => typeof v === "number" && v > 0);
-  const inputValues = members
-    .map((m) => m.max_input_tokens)
-    .filter((v): v is number => typeof v === "number" && v > 0);
-
-  const everyDeclaresInput = hasMembers && inputValues.length === members.length;
-
-  const capabilities: ModelV2["capabilities"] = {
-    temperature:
-      hasMembers && members.every((m) => (m.capabilities?.temperature ?? true) !== false),
-    reasoning:
-      hasMembers &&
-      members.every((m) => Boolean(m.capabilities?.reasoning || m.capabilities?.thinking)),
-    attachment:
-      hasMembers &&
-      members.every((m) => Boolean(m.capabilities?.attachment ?? m.capabilities?.vision ?? false)),
-    toolcall: hasMembers && members.every((m) => Boolean(m.capabilities?.tool_calling ?? false)),
-    input: {
-      text: modalityAllHave(memberInMods, "text"),
-      audio: modalityAllHave(memberInMods, "audio"),
-      image: modalityAllHave(memberInMods, "image"),
-      video: modalityAllHave(memberInMods, "video"),
-      pdf: modalityAllHave(memberInMods, "pdf"),
-    },
-    output: {
-      text: modalityAllHave(memberOutMods, "text"),
-      audio: modalityAllHave(memberOutMods, "audio"),
-      image: modalityAllHave(memberOutMods, "image"),
-      video: modalityAllHave(memberOutMods, "video"),
-      pdf: modalityAllHave(memberOutMods, "pdf"),
-    },
-    interleaved: hasMembers && members.every((m) => Boolean(m.capabilities?.thinking)),
-  };
-
-  // Combos span multiple providers. Use Anthropic format only when ALL
-  // members resolve to Anthropic — otherwise fall back to OpenAI-compat
-  // (lowest common denominator that every upstream understands).
-  const comboApiBlock = (() => {
-    if (!hasMembers) return resolveApiBlockV2(combo.id, baseURL, apiFormat);
-    const allAnthropic = members.every(
-      (m) => resolveApiBlockV2(m.id, baseURL, apiFormat).id === "anthropic"
-    );
-    return allAnthropic
-      ? resolveApiBlockV2(members[0].id, baseURL, apiFormat)
-      : resolveApiBlockV2(combo.id, baseURL, apiFormat);
-  })();
-
-  return {
-    id: combo.id,
-    providerID: providerId,
-    api: comboApiBlock,
-    name: combo.name && combo.name.trim().length > 0 ? combo.name : combo.id,
-    capabilities,
-    cost: {
-      input: 0,
-      output: 0,
-      cache: { read: 0, write: 0 },
-    },
-    limit: {
-      context:
-        typeof combo.computed_context_length === "number" && combo.computed_context_length > 0
-          ? combo.computed_context_length
-          : contextValues.length > 0
-            ? Math.min(...contextValues)
-            : 0,
-      ...(everyDeclaresInput ? { input: Math.min(...inputValues) } : {}),
-      output: outputValues.length > 0 ? Math.min(...outputValues) : 0,
-    },
-    status: "active",
-    options: {},
-    headers: {},
-    release_date: combo.release_date ?? "",
-  };
-}
--- a/@omniroute/opencode-plugin-v2/src/shared/enrich.ts
+++ b/@omniroute/opencode-plugin-v2/src/shared/enrich.ts
@@ -1,606 +0,0 @@
-import type { Model as ModelV2 } from "@opencode-ai/sdk/v2";
-import { buildModelDisplayName } from "./naming.js";
-import type { FreeModelFreeType } from "./naming.js";
-
-export interface OmniRouteEnrichmentEntry {
-  /** Human-readable display name. Replaces ModelV2.name when present. */
-  name?: string;
-  /** Per-million-token cost overlay onto ModelV2.cost. */
-  pricing?: {
-    input?: number;
-    output?: number;
-    cacheRead?: number;
-    cacheWrite?: number;
-  };
-  /**
-   * Provider alias prefix seen in `/v1/models` ids (e.g. `cc`, `gemini`).
-   * Populated by `defaultOmniRouteEnrichmentFetcher` from
-   * `/api/pricing/models` keys. Drives the `usableOnly` alias↔canonical
-   * resolution.
-   */
-  providerAlias?: string;
-  /**
-   * Canonical provider id used by `/api/providers` connections (e.g.
-   * `claude`, `gemini`, `kiro`). Populated from the per-provider
-   * `entry.id` field inside `/api/pricing/models`.
-   */
-  providerCanonical?: string;
-  /**
-   * Human-readable upstream provider label (e.g. `Claude`, `Kiro`,
-   * `Windsurf`, `GitHub Models`). Populated from the per-provider
-   * `entry.name` field inside `/api/pricing/models`. Used by the
-   * `providerTag` feature to suffix `ModelV2.name` with the routing
-   * destination so the OC TUI picker can differentiate the same
-   * model id sold through different upstream connections.
-   */
-  providerDisplayName?: string;
-  /** Free-model budget type (from freeModelCatalog). */
-  freeType?: FreeModelFreeType;
-  /** Monthly token budget for recurring free models. */
-  monthlyTokens?: number;
-  /** Credit token budget for credit-based free models. */
-  creditTokens?: number;
-}
-
-/** Map keyed by full model id (possibly namespaced, e.g. `cc/claude-sonnet-4-6`). */
-export type OmniRouteEnrichmentMap = Map<string, OmniRouteEnrichmentEntry>;
-
-/**
- * Reverse-index the enrichment map from `providerCanonical → providerAlias`.
- *
- * OmniRoute's `/api/pricing/models` is keyed by short ALIAS (`cc`, `cx`,
- * `pol`). But `/v1/models` exposes some models a SECOND time under their
- * CANONICAL name (`claude/claude-opus-4-7`, `codex/gpt-5.5`,
- * `pollinations/midjourney`). Without a reverse map, those canonical
- * rows miss enrichment entirely and surface as raw ids in the picker.
- *
- * Built once per refresh from the enrichment entries themselves — no
- * hardcoded registry. Only records `canonical → alias` mappings when
- * both are present AND distinct (skips slots where alias === canonical
- * like `kiro`).
- */
-export function buildCanonicalToAliasMap(
-  enrichment: OmniRouteEnrichmentMap | undefined
-): Map<string, string> {
-  const out = new Map<string, string>();
-  if (!enrichment) return out;
-  for (const entry of enrichment.values()) {
-    const alias = typeof entry.providerAlias === "string" ? entry.providerAlias.trim() : "";
-    const canonical =
-      typeof entry.providerCanonical === "string" ? entry.providerCanonical.trim() : "";
-    if (alias.length === 0 || canonical.length === 0) continue;
-    if (alias === canonical) continue;
-    if (!out.has(canonical)) out.set(canonical, alias);
-  }
-  return out;
-}
-
-/**
- * Enrichment lookup with alias-fallback chain.
- *
- * Resolution order (first hit wins):
- *
- *   1. `enrichment.get(rawId)` — direct hit on `<prefix>/<modelId>` or
- *      bare id (the fetcher writes under both forms).
- *   2. If `rawId` is `<canonical>/<modelId>` and `canonicalToAlias` has
- *      a mapping for `canonical`, try `<alias>/<modelId>`. This rescues
- *      duplicate rows like `claude/claude-opus-4-7` (canonical) when
- *      enrichment only indexed under `cc/claude-opus-4-7` (alias).
- *   3. Bare `<modelId>` as a last resort. Already covered by step 1 in
- *      practice (fetcher writes bare keys), but kept defensive.
- *
- * Returns `undefined` when no lookup hits.
- */
-export function lookupEnrichment(
-  rawId: string,
-  enrichment: OmniRouteEnrichmentMap | undefined,
-  canonicalToAlias: Map<string, string>
-): OmniRouteEnrichmentEntry | undefined {
-  if (!enrichment) return undefined;
-  const direct = enrichment.get(rawId);
-  if (direct) return direct;
-  const slash = rawId.indexOf("/");
-  if (slash > 0) {
-    const prefix = rawId.slice(0, slash);
-    const modelId = rawId.slice(slash + 1);
-    const alias = canonicalToAlias.get(prefix);
-    if (alias && alias !== prefix) {
-      const viaAlias = enrichment.get(`${alias}/${modelId}`);
-      if (viaAlias) return viaAlias;
-    }
-    const bare = enrichment.get(modelId);
-    if (bare) return bare;
-  }
-  return undefined;
-}
-
-/**
- * Pre-pass: detect raw rows that are the CANONICAL twin of an ALIAS row
- * already in the catalog. Returns the set of canonical-keyed ids to skip
- * during the raw-model loop so each model surfaces exactly once under
- * its enriched alias key.
- *
- * Example: `/v1/models` returns BOTH `cc/claude-opus-4-7` and
- * `claude/claude-opus-4-7`. The former is enriched (alias `cc` exists
- * in `/api/pricing/models`); the latter is raw. We keep `cc/...` and
- * drop `claude/...`.
- *
- * Built once per refresh. Cheap — O(M) where M = raw model count.
- */
-export function canonicalDedupSet(
-  rawModels: ReadonlyArray<{ id: string }>,
-  canonicalToAlias: Map<string, string>
-): Set<string> {
-  const drop = new Set<string>();
-  if (canonicalToAlias.size === 0) return drop;
-  // Index every alias key present in the raw catalog.
-  const aliasKeys = new Set<string>();
-  for (const m of rawModels) {
-    if (typeof m.id === "string" && m.id.length > 0) aliasKeys.add(m.id);
-  }
-  for (const m of rawModels) {
-    if (typeof m.id !== "string" || m.id.length === 0) continue;
-    const slash = m.id.indexOf("/");
-    if (slash <= 0) continue;
-    const prefix = m.id.slice(0, slash);
-    const modelId = m.id.slice(slash + 1);
-    const alias = canonicalToAlias.get(prefix);
-    if (!alias || alias === prefix) continue;
-    // Canonical row only gets suppressed if the alias row actually
-    // exists — otherwise we'd hide the model entirely.
-    if (aliasKeys.has(`${alias}/${modelId}`)) drop.add(m.id);
-  }
-  return drop;
-}
-
-/**
- * Build a per-alias index of enrichment metadata so we can render the
- * provider prefix even for raw models that don't have their own
- * curated `/api/pricing/models` entry.
- *
- * Real example: OmniRoute's `pricing['cohere']` slot lists 10 curated
- * models but `/v1/models` also returns `cohere/rerank-multilingual-v3.0`
- * and `cohere/rerank-v4.0-fast` (not in the curated 10). Without this
- * index, those rows surface in the picker as `cohere/...` with no
- * `Cohere - ` prefix because the per-model enrichment lookup misses.
- *
- * This index records the first non-empty `providerDisplayName` seen
- * for each alias, plus the alias itself. Callers use it to synthesize
- * a minimal `OmniRouteEnrichmentEntry` whenever the direct lookup
- * misses but the raw id's prefix matches a known alias.
- *
- * Built once per refresh; first-wins on duplicate alias (matches
- * `buildCanonicalToAliasMap` semantics).
- */
-export function buildAliasIndex(
-  enrichment: OmniRouteEnrichmentMap | undefined
-): Map<string, OmniRouteEnrichmentEntry> {
-  const out = new Map<string, OmniRouteEnrichmentEntry>();
-  if (!enrichment) return out;
-  for (const entry of enrichment.values()) {
-    const alias = typeof entry.providerAlias === "string" ? entry.providerAlias.trim() : "";
-    if (alias.length === 0) continue;
-    if (out.has(alias)) {
-      // First-wins, but upgrade to the first entry that carries a
-      // non-empty providerDisplayName so the prefix renders nicely.
-      const existing = out.get(alias);
-      if (
-        existing &&
-        (!existing.providerDisplayName || existing.providerDisplayName.trim().length === 0) &&
-        typeof entry.providerDisplayName === "string" &&
-        entry.providerDisplayName.trim().length > 0
-      ) {
-        out.set(alias, entry);
-      }
-      continue;
-    }
-    out.set(alias, entry);
-  }
-  return out;
-}
-
-/**
- * Resolve a synthesised enrichment entry for `applyProviderTag` /
- * `shortProviderLabel` consumption, combining two sources:
- *
- *  1. The direct per-model enrichment match (if present).
- *  2. A per-alias fallback derived from `buildAliasIndex` — covers raw
- *     ids whose prefix matches a known alias but the specific model
- *     id wasn't curated in `/api/pricing/models`. Example:
- *     `cohere/rerank-multilingual-v3.0` falls back to the cohere slot's
- *     `providerDisplayName='Cohere'` even though that specific id
- *     isn't in the curated 10-model list.
- *
- * Returns `undefined` when neither source surfaces an alias.
- *
- * NOTE: this function is read-only over its inputs; it never mutates
- * the underlying `direct` entry. When it falls back to the alias
- * index, it constructs a fresh minimal entry exposing only the
- * provider-prefix fields (`providerAlias`, `providerCanonical`,
- * `providerDisplayName`). Other fields (name, pricing) are explicitly
- * left undefined so `applyEnrichment` won't accidentally overwrite a
- * model name with the alias-slot label.
- */
-export function resolveProviderTagEntry(
-  rawId: string,
-  direct: OmniRouteEnrichmentEntry | undefined,
-  aliasIndex: Map<string, OmniRouteEnrichmentEntry>,
-  canonicalToAlias?: Map<string, string>
-): OmniRouteEnrichmentEntry | undefined {
-  if (direct) {
-    const alias = typeof direct.providerAlias === "string" ? direct.providerAlias.trim() : "";
-    const display =
-      typeof direct.providerDisplayName === "string" ? direct.providerDisplayName.trim() : "";
-    if (alias.length > 0 || display.length > 0) return direct;
-  }
-  const slash = rawId.indexOf("/");
-  if (slash <= 0) return direct;
-  const prefix = rawId.slice(0, slash);
-  // 1. Direct alias lookup (`cohere/...` → cohere slot keyed by alias=cohere).
-  let fromAlias = aliasIndex.get(prefix);
-  // 2. Canonical fallback (`pollinations/...` → look up via alias `pol`).
-  if (!fromAlias && canonicalToAlias) {
-    const alias = canonicalToAlias.get(prefix);
-    if (alias) fromAlias = aliasIndex.get(alias);
-  }
-  if (!fromAlias) return direct;
-  // Synthesize: borrow only the provider-prefix metadata.
-  return {
-    providerAlias: fromAlias.providerAlias,
-    providerCanonical: fromAlias.providerCanonical,
-    providerDisplayName: fromAlias.providerDisplayName,
-  };
-}
-
-/**
- * Fetcher contract: resolves the enrichment overlay (display names +
- * pricing + free-tier budgets) from a running OmniRoute instance.
- */
-/**
- * Reports a source that could not be read. Enrichment stays best-effort, but
- * a caller that swallows this loses display names, provider tags, canonical
- * dedupe and pricing with no way to tell why.
- */
-export type OmniRouteEnrichmentSourceError = (endpoint: string, reason: string) => void;
-
-export type OmniRouteEnrichmentFetcher = (
-  baseURL: string,
-  apiKey: string,
-  timeoutMs?: number,
-  onSourceError?: OmniRouteEnrichmentSourceError
-) => Promise<OmniRouteEnrichmentMap>;
-
-function trimTrailingSlashes(value: string): string {
-  let i = value.length;
-  while (i > 0 && value.charCodeAt(i - 1) === 0x2f /* "/" */) i--;
-  return i === value.length ? value : value.slice(0, i);
-}
-
-/**
- * Default enrichment fetcher — pulls nice display names from
- * `GET /api/pricing/models` and merges per-million-token pricing from
- * `GET /api/pricing` (the actual pricing source — `/api/pricing/models` is
- * a catalog endpoint whose entries are `{id, name, custom}` only).
- *
- * `/api/pricing/models` shape (catalog):
- *  - `{ [providerAlias]: { id, alias, name, models: [{ id, name, custom }] } }`
- *
- * `/api/pricing` shape (pricing only):
- *  - `{ [providerAlias]: { [modelId]: { input, output, cached, reasoning, cache_creation } } }`
- *    where values are USD per million tokens.
- *
- * The two responses are joined on `(providerAlias, modelId)` and the merged
- * entries are stored under both `${providerAlias}/${modelId}` and bare
- * `${modelId}` keys so downstream lookups against either form succeed.
- *
- * Soft-fails (returns whatever was collected) on non-2xx or parse errors;
- * the two fetches are independent so one missing source still surfaces the
- * other. A third best-effort fetch attaches free-tier budgets from
- * `/api/free-tier/summary`.
- *
- * Ported from the v1 plugin (`index.ts:1906-2106`); the shared logger is
- * the only intentional difference (no plugin-contract dependency here).
- */
-export const defaultOmniRouteEnrichmentFetcher: OmniRouteEnrichmentFetcher = async (
-  baseURL,
-  apiKey,
-  timeoutMs = 10_000,
-  onSourceError
-) => {
-  const report = (endpoint: string, reason: unknown): void => {
-    onSourceError?.(endpoint, reason instanceof Error ? reason.message : String(reason));
-  };
-  const out: OmniRouteEnrichmentMap = new Map();
-  if (!baseURL || !apiKey) return out;
-  const root = trimTrailingSlashes(baseURL.replace(/\/v1\/?$/, ""));
-  const headers = {
-    Authorization: `Bearer ${apiKey}`,
-    Accept: "application/json",
-  };
-
-  // 1. Catalog with nice display names.
-  const catalogAc = new AbortController();
-  const catalogTimer = setTimeout(() => catalogAc.abort(), timeoutMs);
-  let catalogStatus = 0;
-  try {
-    const res = await fetch(`${root}/api/pricing/models`, {
-      method: "GET",
-      headers,
-      signal: catalogAc.signal,
-    });
-    catalogStatus = res.status;
-    if (res.ok) {
-      const body = (await res.json()) as unknown;
-      const providers =
-        (body as { providers?: Record<string, { models?: unknown[] }> })?.providers ??
-        (body as Record<string, { models?: unknown[] }>);
-      if (providers && typeof providers === "object") {
-        for (const [providerAlias, slot] of Object.entries(providers)) {
-          if (!slot || typeof slot !== "object") continue;
-          const models = (slot as { models?: unknown[] }).models;
-          if (!Array.isArray(models)) continue;
-          const canonicalRaw = (slot as { id?: unknown }).id;
-          const providerCanonical =
-            typeof canonicalRaw === "string" && canonicalRaw.length > 0
-              ? canonicalRaw
-              : providerAlias;
-          const slotNameRaw = (slot as { name?: unknown }).name;
-          const providerDisplayName =
-            typeof slotNameRaw === "string" && slotNameRaw.trim().length > 0
-              ? slotNameRaw.trim()
-              : undefined;
-          for (const m of models) {
-            if (!m || typeof m !== "object") continue;
-            const id = (m as { id?: unknown }).id;
-            if (typeof id !== "string" || id.length === 0) continue;
-            const name = (m as { name?: unknown }).name;
-            const entry: OmniRouteEnrichmentEntry = {
-              providerAlias,
-              providerCanonical,
-            };
-            if (providerDisplayName) entry.providerDisplayName = providerDisplayName;
-            if (typeof name === "string" && name.trim().length > 0) entry.name = name;
-            const namespaced = `${providerAlias}/${id}`;
-            if (!out.has(namespaced)) out.set(namespaced, entry);
-            // The bare id is a fallback for ids that arrive unnamespaced. It
-            // gets its OWN copy: sharing the object would let a later write
-            // for one provider — a price, typically — land on another
-            // provider's entry that happens to sell the same model id.
-            if (!out.has(id)) out.set(id, { ...entry });
-          }
-        }
-      }
-    }
-  } catch (err) {
-    // Network error, timeout, abort: nothing collected from THIS source, but
-    // the pricing fetch below may still succeed — let it try, then decide at
-    // the end whether the whole overlay failed (see the throw below).
-    report("/api/pricing/models", err);
-    catalogStatus = -1;
-  } finally {
-    clearTimeout(catalogTimer);
-  }
-  if (
-    catalogStatus !== 0 &&
-    catalogStatus !== -1 &&
-    (catalogStatus < 200 || catalogStatus >= 300)
-  ) {
-    report("/api/pricing/models", `HTTP ${catalogStatus}`);
-  }
-
-  // 2. Pricing values from /api/pricing.
-  const priceAc = new AbortController();
-  const priceTimer = setTimeout(() => priceAc.abort(), timeoutMs);
-  let priceStatus = 0;
-  try {
-    const res = await fetch(`${root}/api/pricing`, {
-      method: "GET",
-      headers,
-      signal: priceAc.signal,
-    });
-    priceStatus = res.status;
-    if (res.ok) {
-      const body = (await res.json()) as unknown;
-      if (body && typeof body === "object" && !Array.isArray(body)) {
-        for (const [providerAlias, slot] of Object.entries(body as Record<string, unknown>)) {
-          if (!slot || typeof slot !== "object" || Array.isArray(slot)) continue;
-          for (const [modelId, raw] of Object.entries(slot as Record<string, unknown>)) {
-            if (!raw || typeof raw !== "object") continue;
-            const p = raw as Record<string, unknown>;
-            const parsed: NonNullable<OmniRouteEnrichmentEntry["pricing"]> = {};
-            if (typeof p.input === "number") parsed.input = p.input;
-            if (typeof p.output === "number") parsed.output = p.output;
-            const cacheRead =
-              typeof p.cached === "number"
-                ? p.cached
-                : typeof p.cacheRead === "number"
-                  ? p.cacheRead
-                  : undefined;
-            if (typeof cacheRead === "number") parsed.cacheRead = cacheRead;
-            const cacheWrite =
-              typeof p.cache_creation === "number"
-                ? p.cache_creation
-                : typeof p.cacheWrite === "number"
-                  ? p.cacheWrite
-                  : undefined;
-            if (typeof cacheWrite === "number") parsed.cacheWrite = cacheWrite;
-            if (Object.keys(parsed).length === 0) continue;
-            const namespaced = `${providerAlias}/${modelId}`;
-            const existingNs = out.get(namespaced);
-            if (existingNs) {
-              existingNs.pricing = { ...(existingNs.pricing ?? {}), ...parsed };
-            } else {
-              out.set(namespaced, { pricing: parsed });
-            }
-            const existingBare = out.get(modelId);
-            // Only the provider that owns the bare entry may price it.
-            // Otherwise the second provider selling the same model id
-            // overwrites the first one's price, and the picker shows a cost
-            // that belongs to a different connection.
-            const bareBelongsHere =
-              existingBare === undefined || existingBare.providerAlias === undefined
-                ? true
-                : existingBare.providerAlias === providerAlias;
-            if (bareBelongsHere) {
-              if (existingBare) {
-                existingBare.pricing = { ...(existingBare.pricing ?? {}), ...parsed };
-              } else {
-                out.set(modelId, { pricing: parsed });
-              }
-            }
-          }
-        }
-      }
-    }
-  } catch (err) {
-    // Same as above: report, mark this source failed, let the remaining
-    // sources try before deciding.
-    report("/api/pricing", err);
-    priceStatus = -1;
-  } finally {
-    clearTimeout(priceTimer);
-  }
-  if (priceStatus !== 0 && priceStatus !== -1 && (priceStatus < 200 || priceStatus >= 300)) {
-    report("/api/pricing", `HTTP ${priceStatus}`);
-  }
-
-  // 3. Free model budgets from /api/free-tier/summary (best-effort).
-  const freeAc = new AbortController();
-  const freeTimer = setTimeout(() => freeAc.abort(), timeoutMs);
-  let freeStatus = 0;
-  try {
-    const res = await fetch(`${root}/api/free-tier/summary`, {
-      method: "GET",
-      headers,
-      signal: freeAc.signal,
-    });
-    freeStatus = res.status;
-    if (res.ok) {
-      const body = (await res.json()) as unknown;
-      const perModel: unknown[] =
-        body && typeof body === "object" && Array.isArray((body as { perModel?: unknown }).perModel)
-          ? ((body as { perModel: unknown[] }).perModel as unknown[])
-          : Array.isArray(body)
-            ? (body as unknown[])
-            : [];
-      for (const fm of perModel) {
-        if (!fm || typeof fm !== "object") continue;
-        const fmObj = fm as Record<string, unknown>;
-        const provider = typeof fmObj.provider === "string" ? fmObj.provider : "";
-        const modelId = typeof fmObj.modelId === "string" ? fmObj.modelId : "";
-        const freeType = typeof fmObj.freeType === "string" ? fmObj.freeType : "";
-        if (!modelId || !freeType) continue;
-        const monthlyTokens =
-          typeof fmObj.monthlyTokens === "number" ? fmObj.monthlyTokens : undefined;
-        const creditTokens =
-          typeof fmObj.creditTokens === "number" ? fmObj.creditTokens : undefined;
-        const displayName = typeof fmObj.displayName === "string" ? fmObj.displayName : "";
-        const candidates = [
-          `${provider}/${modelId}`,
-          modelId,
-          ...(displayName ? [displayName] : []),
-        ];
-        for (const key of candidates) {
-          const entry = out.get(key);
-          if (entry) {
-            entry.freeType = freeType as FreeModelFreeType;
-            if (monthlyTokens !== undefined) entry.monthlyTokens = monthlyTokens;
-            if (creditTokens !== undefined) entry.creditTokens = creditTokens;
-            break;
-          }
-        }
-      }
-    }
-  } catch (err) {
-    report("/api/free-tier/summary", err);
-    // Soft-fail; free metadata is optional.
-  } finally {
-    clearTimeout(freeTimer);
-  }
-  if (freeStatus !== 0 && (freeStatus < 200 || freeStatus >= 300)) {
-    report("/api/free-tier/summary", `HTTP ${freeStatus}`);
-  }
-
-  // A source that failed contributes nothing — but the overlay keeps its own
-  // memory per source: names collected while the catalog endpoint answered
-  // survive a later pricing outage, and prices collected while pricing
-  // answered survive a later catalog outage. Without this a single flapping
-  // source wipes the other source's good data on every refresh. So a failed
-  // catalog source throws (the caller keeps last-known) UNLESS the pricing
-  // source brought something on THIS call — then whatever was collected,
-  // names or prices, is the gateway's answer and ships as-is. (Status alone
-  // cannot decide: a 2xx pricing answer with zero priced models is still an
-  // answer, but it carries nothing to save the overlay with.)
-  const sourceFailed = (status: number): boolean =>
-    status === -1 || (status !== 0 && (status < 200 || status >= 300));
-  const catalogFailed = sourceFailed(catalogStatus);
-  const pricingBroughtSomething = !sourceFailed(priceStatus) && out.size > 0;
-  if (catalogFailed && !pricingBroughtSomething) {
-    throw new Error(
-      `enrichment catalog source failed (pricing/models: ${catalogStatus}, pricing: ${priceStatus})`
-    );
-  }
-
-  return out;
-};
-
-/**
- * Apply enrichment overlay onto a ModelV2 entry. Mutates and returns the
- * passed entry for convenience.
- */
-/** What the caller knows about the entry that the overlay itself cannot tell. */
-export interface EnrichmentDisplayContext {
-  /** Combos never carry a provider tag: they route across providers. */
-  isCombo?: boolean;
-  isAutoCombo?: boolean;
-  /** Set false to publish the bare display name, without the provider tag. */
-  providerTag?: boolean;
-}
-
-/**
- * Fold the overlay into a mapped model: display name, provider tag, free-tier
- * marker and budget, and pricing.
- *
- * The name is built rather than copied, because the gateway ships the parts
- * separately — the pricing catalog gives a display name and an upstream
- * provider label, the free-tier summary gives the budget. A picker showing
- * `Claude - [Free] Sonnet 4.6 · 1M/mo` tells the user which connection serves
- * the model and what it costs them; `claude-sonnet-4-6` tells them nothing.
- */
-export function applyEnrichment(
-  model: ModelV2,
-  enrichment: OmniRouteEnrichmentEntry | undefined,
-  context: EnrichmentDisplayContext = {}
-): ModelV2 {
-  if (!enrichment) return model;
-  const built = buildModelDisplayName({
-    rawId: model.name && model.name.length > 0 ? model.name : model.id,
-    enrichmentName: enrichment.name,
-    providerAlias: context.providerTag === false ? undefined : enrichment.providerAlias,
-    providerDisplayName: context.providerTag === false ? undefined : enrichment.providerDisplayName,
-    isFree: enrichment.freeType !== undefined,
-    freeType: enrichment.freeType,
-    monthlyTokens: enrichment.monthlyTokens,
-    creditTokens: enrichment.creditTokens,
-    isCombo: context.isCombo,
-    isAutoCombo: context.isAutoCombo,
-  });
-  if (built.trim().length > 0) {
-    model.name = built;
-  }
-  if (enrichment.pricing) {
-    if (typeof enrichment.pricing.input === "number") {
-      model.cost.input = enrichment.pricing.input;
-    }
-    if (typeof enrichment.pricing.output === "number") {
-      model.cost.output = enrichment.pricing.output;
-    }
-    if (typeof enrichment.pricing.cacheRead === "number") {
-      model.cost.cache.read = enrichment.pricing.cacheRead;
-    }
-    if (typeof enrichment.pricing.cacheWrite === "number") {
-      model.cost.cache.write = enrichment.pricing.cacheWrite;
-    }
-  }
-  return model;
-}
--- a/@omniroute/opencode-plugin-v2/src/shared/fingerprint.ts
+++ b/@omniroute/opencode-plugin-v2/src/shared/fingerprint.ts
@@ -1,127 +0,0 @@
-import { createHash } from "node:crypto";
-
-/**
- * Fingerprint the CONTENT of a catalog snapshot (not endpoint/credential
- * identity) so lazy refresh can reload-after-publish only when something
- * actually changed.
- *
- * sha256 over sorted `id + "|" + (release_date ?? "")` lines for models
- * plus sorted combo ids, joined with `\n`. Order-insensitive: two
- * snapshots with the same entries in different order hash identically.
- */
-export function catalogContentFingerprint(
-  models: { id: string; release_date?: string }[],
-  combos: { id: string }[],
-  autoCombos: { id: string }[] = []
-): string {
-  const modelLines = models
-    .map((m) => `${m.id}|${m.release_date ?? ""}`)
-    .sort()
-    .join("\n");
-  const comboLines = combos
-    .map((c) => c.id)
-    .sort()
-    .join("\n");
-  const autoLines = autoCombos
-    .map((c) => c.id)
-    .sort()
-    .join("\n");
-  return createHash("sha256").update(`${modelLines}\n${comboLines}\n${autoLines}`).digest("hex");
-}
-
-/**
- * Digest of the optional tier (auto-combos, provider connections, enrichment).
- * The catalog fingerprint covers model and combo ids only, so an overlay that
- * moves — a renamed model, a provider going unusable — leaves it unchanged.
- * Reloading on every refresh instead would ask the host to rebuild its catalog
- * once per TTL window for nothing.
- */
-export function optionalTierFingerprint(
-  autoCombos: { id: string }[],
-  providers: {
-    id?: string;
-    name?: string;
-    testStatus?: string;
-    isActive?: boolean;
-    providerDisplayName?: string;
-  }[],
-  enrichment:
-    | Map<
-        string,
-        {
-          name?: string;
-          freeType?: string;
-          providerDisplayName?: string;
-          monthlyTokens?: number;
-          creditTokens?: number;
-          pricing?: Record<string, number | undefined>;
-        }
-      >
-    | undefined,
-  combos: { id: string; name?: string; models?: unknown[] }[] = []
-): string {
-  const parts: string[] = [];
-  // Membership matters: a combo keeping its id while losing a member is a
-  // different combo to anyone picking it.
-  parts.push(
-    combos
-      .map((c) => c.id + "|" + (c.name ?? "") + "|" + String(c.models?.length ?? 0))
-      .sort()
-      .join(",")
-  );
-  parts.push(
-    autoCombos
-      .map((c) => c.id)
-      .sort()
-      .join(",")
-  );
-  // A provider going quiet or getting renamed is as visible to the user as a
-  // price move: its activity flag and display name belong in the digest.
-  parts.push(
-    providers
-      .map(
-        (p) =>
-          (p.id ?? p.name ?? "") +
-          ":" +
-          (p.testStatus ?? "") +
-          ":" +
-          String(p.isActive ?? "") +
-          ":" +
-          (p.providerDisplayName ?? "")
-      )
-      .sort()
-      .join(",")
-  );
-  if (enrichment !== undefined) {
-    const rows: string[] = [];
-    for (const [key, entry] of enrichment) {
-      // Pricing is part of what the user sees, so a price move must reach
-      // the picker without waiting for an id to change.
-      const price = entry.pricing
-        ? Object.entries(entry.pricing)
-            .map(([k, v]) => k + "=" + String(v ?? ""))
-            .sort()
-            .join(";")
-        : "";
-      rows.push(
-        key +
-          "|" +
-          (entry.name ?? "") +
-          "|" +
-          (entry.freeType ?? "") +
-          "|" +
-          (entry.providerDisplayName ?? "") +
-          "|" +
-          String(entry.monthlyTokens ?? "") +
-          ";" +
-          String(entry.creditTokens ?? "") +
-          "|" +
-          price
-      );
-    }
-    rows.sort();
-    parts.push(String(enrichment.size));
-    parts.push(rows.join("\n"));
-  }
-  return createHash("sha256").update(parts.join(" ")).digest("hex");
-}
--- a/@omniroute/opencode-plugin-v2/src/shared/gemini.ts
+++ b/@omniroute/opencode-plugin-v2/src/shared/gemini.ts
@@ -1,166 +0,0 @@
-/**
- * Gemini rejects several standard JSON-Schema keywords in tool declarations
- * and answers `400 INVALID_ARGUMENT` for the whole request when it meets one.
- * The keywords carry no meaning Gemini would honour anyway, so stripping them
- * costs nothing and is what keeps a tool-calling chain alive.
- */
-/**
- * Keywords Gemini rejects outright. `$ref` is deliberately NOT here: it
- * cannot be stripped without turning the schema into "accept anything", so
- * tools carrying one are forwarded untouched (see below). `ref` is not a
- * JSON Schema keyword at all, and stripping it by name destroys a legitimate
- * tool parameter called `ref` — a walker that cannot tell a keyword from a
- * property name mangles the schema it was meant to repair.
- */
-const REJECTED_KEYWORDS = new Set(["$schema", "additionalProperties"]);
-
-/** Keys whose value is itself a schema. */
-const SCHEMA_VALUE_KEYS = [
-  "items",
-  "additionalItems",
-  "contains",
-  "not",
-  "if",
-  "then",
-  "else",
-  "propertyNames",
-  "contentSchema",
-  "unevaluatedItems",
-  "unevaluatedProperties",
-];
-/** Keys whose value maps arbitrary NAMES to schemas — never keyword space. */
-const SCHEMA_MAP_KEYS = [
-  "properties",
-  "patternProperties",
-  "$defs",
-  "definitions",
-  "dependentSchemas",
-  "dependencies",
-];
-/** Keys whose value is a list of schemas. */
-const SCHEMA_LIST_KEYS = ["allOf", "anyOf", "oneOf", "prefixItems"];
-
-function isRecord(value: unknown): value is Record<string, unknown> {
-  return typeof value === "object" && value !== null && !Array.isArray(value);
-}
-
-/** True when any schema in the tree carries a `$ref` we cannot resolve. */
-function hasUnresolvableRef(node: unknown): boolean {
-  if (Array.isArray(node)) return node.some(hasUnresolvableRef);
-  if (!isRecord(node)) return false;
-  if ("$ref" in node) return true;
-  for (const key of SCHEMA_VALUE_KEYS) if (hasUnresolvableRef(node[key])) return true;
-  // (arrays are handled by the Array branch at the top of this function)
-  for (const key of SCHEMA_LIST_KEYS) if (hasUnresolvableRef(node[key])) return true;
-  for (const key of SCHEMA_MAP_KEYS) {
-    const map = node[key];
-    if (isRecord(map) && Object.values(map).some(hasUnresolvableRef)) return true;
-  }
-  return false;
-}
-
-/**
- * Strip the rejected keywords in place, walking only the positions where a
- * schema can appear. Property names are never treated as keywords, so a tool
- * whose parameter happens to be called `additionalProperties` keeps it.
- * Returns whether anything was removed.
- */
-function stripAtSchemaPositions(node: Record<string, unknown>): boolean {
-  let changed = false;
-  for (const keyword of REJECTED_KEYWORDS) {
-    if (keyword in node) {
-      delete node[keyword];
-      changed = true;
-    }
-  }
-  for (const key of SCHEMA_VALUE_KEYS) {
-    const child = node[key];
-    if (isRecord(child)) {
-      changed = stripAtSchemaPositions(child) || changed;
-      continue;
-    }
-    // `items` also takes the tuple form: an array of schemas, one per position.
-    if (Array.isArray(child)) {
-      for (const item of child) {
-        if (isRecord(item)) changed = stripAtSchemaPositions(item) || changed;
-      }
-    }
-  }
-  for (const key of SCHEMA_LIST_KEYS) {
-    const list = node[key];
-    if (Array.isArray(list)) {
-      for (const child of list) {
-        if (isRecord(child)) changed = stripAtSchemaPositions(child) || changed;
-      }
-    }
-  }
-  for (const key of SCHEMA_MAP_KEYS) {
-    const map = node[key];
-    if (!isRecord(map)) continue;
-    for (const child of Object.values(map)) {
-      if (isRecord(child)) changed = stripAtSchemaPositions(child) || changed;
-    }
-  }
-  return changed;
-}
-
-/**
- * Families Google actually ships, anchored on the last path segment. A plain
- * substring test also claims `gemini-compatible-proxy` and `my-gemini-wrapper`
- * — and since the sanitiser removes keywords, a false positive is not free.
- */
-const GEMINI_MODEL_ID =
-  /^gemini(?:[-_.](?:\d|pro|flash|ultra|nano|exp|thinking|embedding|live|imagen)|$)/i;
-
-/**
- * True for the routing forms a Gemini model reaches a gateway under — bare
- * (`gemini-2.5-flash`), canonical (`models/gemini-1.5-pro`) and prefixed
- * (`google-vertex/gemini-2.0`).
- */
-export function isGeminiModelId(modelId: unknown): boolean {
-  if (typeof modelId !== "string") return false;
-  const segment = modelId.split("/").pop() ?? "";
-  return GEMINI_MODEL_ID.test(segment);
-}
-
-/** The subset of an AI SDK tool declaration this module reads. */
-export interface ToolWithInputSchema {
-  readonly type?: string;
-  readonly inputSchema?: unknown;
-  readonly [key: string]: unknown;
-}
-
-/**
- * Return a copy of `tools` whose input schemas are free of the keywords Gemini
- * rejects, or `undefined` when there was nothing to strip — which lets the
- * caller forward the original array and skip the clone entirely.
- *
- * Tools this module cannot read (provider-defined tools, entries without an
- * object schema) are carried through unchanged rather than dropped: a tool the
- * sanitiser does not understand is still a tool the model needs.
- */
-export function sanitizeToolInputSchemas<T extends ToolWithInputSchema>(
-  tools: readonly T[] | undefined
-): T[] | undefined {
-  if (tools === undefined || tools.length === 0) return undefined;
-  let changed = false;
-  const out = tools.map((tool) => {
-    if (!isRecord(tool.inputSchema)) return tool;
-    // A schema carrying something uncloneable is not worth failing a request
-    // over: forward the tool untouched and let the model answer.
-    // A `$ref` cannot be stripped without turning the schema into "anything
-    // goes", and cannot be resolved here. Forward the tool untouched and let
-    // the gateway answer rather than silently widen what the model may send.
-    if (hasUnresolvableRef(tool.inputSchema)) return tool;
-    let schema: Record<string, unknown>;
-    try {
-      schema = structuredClone(tool.inputSchema) as Record<string, unknown>;
-    } catch {
-      return tool;
-    }
-    if (!stripAtSchemaPositions(schema)) return tool;
-    changed = true;
-    return { ...tool, inputSchema: schema };
-  });
-  return changed ? out : undefined;
-}
--- a/@omniroute/opencode-plugin-v2/src/shared/index.ts
+++ b/@omniroute/opencode-plugin-v2/src/shared/index.ts
@@ -1,9 +0,0 @@
-export * from "./models-map.js";
-export * from "./combos-map.js";
-export * from "./auto-combos.js";
-export * from "./naming.js";
-export * from "./enrich.js";
-export * from "./fingerprint.js";
-export * from "./logger.js";
-export * from "./usable.js";
-export * from "./gemini.js";
--- a/@omniroute/opencode-plugin-v2/src/shared/logger.ts
+++ b/@omniroute/opencode-plugin-v2/src/shared/logger.ts
@@ -1,81 +0,0 @@
-/**
- * Namespaced leveled logger shared by the OmniRoute OpenCode packages.
- *
- * Levels: error < warn < info < debug. Default: warn.
- * Ported from the v1 plugin (`logger.ts`) so both new packages share one
- * sink instead of raw `console.warn` / `console.log` calls.
- */
-
-export type LogLevel = "error" | "warn" | "info" | "debug";
-
-const LEVEL_ORDER: Record<LogLevel, number> = {
-  error: 0,
-  warn: 1,
-  info: 2,
-  debug: 3,
-};
-
-const TAG = "[omniroute-plugin]";
-
-function shouldLog(current: LogLevel, target: LogLevel): boolean {
-  return LEVEL_ORDER[current] >= LEVEL_ORDER[target];
-}
-
-let _level: LogLevel = "warn";
-
-export function setLogLevel(level: LogLevel): void {
-  _level = level;
-}
-
-export function getLogLevel(): LogLevel {
-  return _level;
-}
-
-function fmt(level: LogLevel, msg: string, tag?: string): string {
-  const prefix = tag ? `${TAG}${tag}` : TAG;
-  return `${prefix} [${level.toUpperCase()}] ${msg}`;
-}
-
-function buildLogger(getLevel: () => LogLevel) {
-  return {
-    error(msg: string, ...args: unknown[]): void {
-      if (shouldLog(getLevel(), "error")) console.error(fmt("error", msg), ...args);
-    },
-    warn(msg: string, ...args: unknown[]): void {
-      if (shouldLog(getLevel(), "warn")) console.warn(fmt("warn", msg), ...args);
-    },
-    info(msg: string, ...args: unknown[]): void {
-      if (shouldLog(getLevel(), "info")) console.warn(fmt("info", msg), ...args);
-    },
-    debug(msg: string, ...args: unknown[]): void {
-      if (shouldLog(getLevel(), "debug")) console.warn(fmt("debug", msg), ...args);
-    },
-    /** Always emit regardless of level (for critical init breadcrumbs). */
-    always(msg: string, ...args: unknown[]): void {
-      console.warn(TAG, msg, ...args);
-    },
-
-    child(tag: string) {
-      return {
-        error: (msg: string, ...args: unknown[]) =>
-          shouldLog(getLevel(), "error") && console.error(fmt("error", msg, tag), ...args),
-        warn: (msg: string, ...args: unknown[]) =>
-          shouldLog(getLevel(), "warn") && console.warn(fmt("warn", msg, tag), ...args),
-        info: (msg: string, ...args: unknown[]) =>
-          shouldLog(getLevel(), "info") && console.warn(fmt("info", msg, tag), ...args),
-        debug: (msg: string, ...args: unknown[]) =>
-          shouldLog(getLevel(), "debug") && console.warn(fmt("debug", msg, tag), ...args),
-      };
-    },
-  };
-}
-
-export type Logger = ReturnType<typeof buildLogger>;
-
-/** Create an instance-scoped logger whose level cannot be changed by other instances. */
-export function createLogger(level: LogLevel): Logger {
-  return buildLogger(() => level);
-}
-
-/** Backward-compatible module-global logger controlled by setLogLevel(). */
-export const logger: Logger = buildLogger(() => _level);
--- a/@omniroute/opencode-plugin-v2/src/shared/models-map.ts
+++ b/@omniroute/opencode-plugin-v2/src/shared/models-map.ts
@@ -1,323 +0,0 @@
-import type { Model as ModelV2 } from "@opencode-ai/sdk/v2";
-import { normaliseFreeLabel } from "./naming.js";
-
-export interface OmniRouteRawModelEntry {
-  id: string;
-  object?: string;
-  owned_by?: string;
-  root?: string | null;
-  parent?: string | null;
-  context_length?: number;
-  max_input_tokens?: number;
-  max_output_tokens?: number;
-  input_modalities?: string[];
-  output_modalities?: string[];
-  capabilities?: {
-    tool_calling?: boolean;
-    reasoning?: boolean;
-    vision?: boolean;
-    thinking?: boolean;
-    attachment?: boolean;
-    structured_output?: boolean;
-    temperature?: boolean;
-    /** Runtime-learned or synced reasoning tiers (server-gated, blind-mapped). */
-    effort_tiers?: string[];
-  };
-  release_date?: string;
-  last_updated?: string;
-  api_format?: string;
-}
-
-/**
- * Fetcher contract: returns the raw `/v1/models` entry list from a running
- * OmniRoute instance. Surfaced as a dependency so unit tests can inject a
- * stub without monkey-patching global `fetch`.
- *
- * Why we inline this instead of using `@omniroute/opencode-provider`'s
- * `fetchLiveModels`: the sibling helper returns a stripped `{id, name,
- * contextLength?}` shape that drops the `capabilities` / `*_modalities` /
- * `max_*_tokens` blocks the mapping needs for ModelV2 pass-through.
- */
-export type OmniRouteModelsFetcher = (
-  baseURL: string,
-  apiKey: string,
-  timeoutMs?: number
-) => Promise<OmniRouteRawModelEntry[]>;
-
-/**
- * Default fetcher: `GET <baseURL>/v1/models` with bearer auth + AbortController
- * timeout. Accepts both the `{object:"list", data:[…]}` envelope OmniRoute
- * emits today and a bare-array envelope (defensive — keeps the plugin
- * working if a future OmniRoute build trims the wrapper). Anything that
- * isn't an object with a string `id` is filtered out silently.
- */
-export const defaultOmniRouteModelsFetcher: OmniRouteModelsFetcher = async (
-  baseURL,
-  apiKey,
-  timeoutMs = 10_000
-) => {
-  if (!apiKey) throw new Error("[omniroute-v2] apiKey required to fetch /v1/models");
-  if (!baseURL) throw new Error("[omniroute-v2] baseURL required to fetch /v1/models");
-
-  const trimmed = trimTrailingSlashes(baseURL);
-  // Tolerate both `https://host` and `https://host/v1` forms — the gateway
-  // exposes /v1/models either way; we just don't want a double `/v1/v1`.
-  const url = /\/v\d+$/.test(trimmed) ? `${trimmed}/models` : `${trimmed}/v1/models`;
-
-  const controller = new AbortController();
-  const timer = setTimeout(() => controller.abort(), timeoutMs);
-  try {
-    const res = await fetch(url, {
-      method: "GET",
-      headers: {
-        Authorization: `Bearer ${apiKey}`,
-        Accept: "application/json",
-      },
-      signal: controller.signal,
-    });
-    if (!res.ok) {
-      throw new Error(`[omniroute-v2] GET ${url} failed: ${res.status} ${res.statusText}`);
-    }
-    const body = (await res.json()) as unknown;
-    const rawList: unknown[] = Array.isArray(body)
-      ? body
-      : body && typeof body === "object" && Array.isArray((body as { data?: unknown }).data)
-        ? ((body as { data: unknown[] }).data as unknown[])
-        : [];
-    const out: OmniRouteRawModelEntry[] = [];
-    for (const r of rawList) {
-      if (r && typeof r === "object" && typeof (r as { id?: unknown }).id === "string") {
-        out.push(r as OmniRouteRawModelEntry);
-      }
-    }
-    return out;
-  } finally {
-    clearTimeout(timer);
-  }
-};
-
-// Manual trim helpers avoid polynomial-regex CodeQL warnings on
-// user-supplied baseURL strings (string.replace(/\/+$/, "")). The same
-// behaviour, no backtracking.
-function trimTrailingSlashes(value: string): string {
-  let i = value.length;
-  while (i > 0 && value.charCodeAt(i - 1) === 0x2f /* "/" */) i--;
-  return i === value.length ? value : value.slice(0, i);
-}
-
-/**
- * Ensure a baseURL ends with `/v1` so the OpenAI-compat SDK constructs
- * `/v1/chat/completions` correctly. The Anthropic SDK does NOT want `/v1`
- * (it appends `/v1/messages` automatically), so callers should branch on
- * format first.
- */
-export function ensureV1Suffix(url: string): string {
-  const trimmed = trimTrailingSlashes(url);
-  return trimmed.endsWith("/v1") ? trimmed : `${trimmed}/v1`;
-}
-
-export interface ApiFormatV2 {
-  allowAnthropic?: boolean;
-  anthropicModels?: string[];
-  /**
-   * Deprecated v1 prefix list (default v1:
-   * `cc,claude,anthropic,kiro,kr`). Accepted for backward compatibility:
-   * prefix OR allowlist routes to anthropic, with a one-time deprecation
-   * warning pointing at `anthropicModels`. Prefer full IDs.
-   */
-  anthropicPrefixes?: string[];
-}
-
-/** Default v1 prefix list, kept so copied v1 configs keep routing. */
-export const DEFAULT_ANTHROPIC_PREFIXES_V1 = ["cc", "claude", "anthropic", "kiro", "kr"];
-
-const warnedPrefixLists = new Set<string>();
-
-function warnDeprecatedPrefixesOnce(prefixes: string[]): void {
-  const key = [...prefixes].sort().join(",");
-  if (warnedPrefixLists.has(key)) return;
-  warnedPrefixLists.add(key);
-  console.warn(
-    "[omniroute-plugin] [WARN] apiFormat.anthropicPrefixes is deprecated; convert to anthropicModels (full IDs)"
-  );
-}
-
-/**
- * The Anthropic SDK block appends `/v1/messages` itself, so it needs the
- * gateway root. A config carrying the `/v1` the OpenAI-compatible block wants
- * would otherwise produce `/v1/v1/messages`.
- */
-function stripV1Suffix(baseURL: string): string {
-  return baseURL.replace(/\/v1\/?$/, "");
-}
-
-/**
- * Resolve the API block (id + url + npm package) for a given model id.
- *
- * v2 rule: a model routes to the Anthropic SDK block when
- * `apiFormat.allowAnthropic === true` AND (its FULL id is allowlisted in
- * `apiFormat.anthropicModels` OR its prefix is listed in the deprecated
- * `apiFormat.anthropicPrefixes`, defaulting to the v1 list when prefixes
- * are absent). The deprecated path warns once per prefix list. With
- * neither allowlist nor prefix match, the model stays openai-compatible.
- */
-export function resolveApiBlockV2(
-  modelId: string,
-  baseURL: string,
-  apiFormat?: ApiFormatV2
-): { id: string; url: string; npm: string } {
-  if (apiFormat?.allowAnthropic === true) {
-    if ((apiFormat.anthropicModels ?? []).includes(modelId)) {
-      return {
-        id: "anthropic",
-        url: stripV1Suffix(trimTrailingSlashes(baseURL)),
-        npm: "@ai-sdk/anthropic",
-      };
-    }
-    const prefixes = apiFormat.anthropicPrefixes ?? DEFAULT_ANTHROPIC_PREFIXES_V1;
-    if (apiFormat.anthropicPrefixes !== undefined) warnDeprecatedPrefixesOnce(prefixes);
-    const slash = modelId.indexOf("/");
-    const prefix = slash === -1 ? modelId : modelId.slice(0, slash);
-    if (prefixes.includes(prefix)) {
-      return {
-        id: "anthropic",
-        url: stripV1Suffix(trimTrailingSlashes(baseURL)),
-        npm: "@ai-sdk/anthropic",
-      };
-    }
-  }
-  return {
-    id: "openai-compatible",
-    url: ensureV1Suffix(baseURL),
-    npm: "@ai-sdk/openai-compatible",
-  };
-}
-
-/**
- * Map a raw `/v1/models` entry → `ModelV2` (the type @opencode-ai/sdk/v2
- * exports as `Model`, re-exported by @opencode-ai/plugin as `ModelV2`).
- *
- * ModelV2 requires a much richer shape than a flat record. Concretely it
- * expects:
- *   - flat `id`, `name`, `providerID`, `api: {id,url,npm}`
- *   - nested `capabilities: { temperature, reasoning, attachment, toolcall,
- *     input:{text,audio,image,video,pdf}, output:{…}, interleaved }`
- *   - `cost: { input, output, cache:{read,write} }` (NOT optional)
- *   - `limit: { context, input?, output }`
- *   - `status: "alpha"|"beta"|"deprecated"|"active"`, `options:{}`, `headers:{}`
- *   - `release_date: string`
- *
- * Field adaptations:
- *   1. Flat `tool_call` / `reasoning` / `attachment` / `modalities`
- *      top-level fields don't exist in ModelV2 — folded into
- *      `capabilities.{toolcall, reasoning, attachment, input.*, output.*}`.
- *   2. `cost: undefined` is illegal (cost is required). OmniRoute doesn't
- *      surface pricing on /v1/models, so we emit a zeroed cost block.
- *      Downstream opencode reads this for display only — the live pricing
- *      is OmniRoute's responsibility at routing time.
- *   3. `tool_call` → `toolcall` (ModelV2 field name; one word).
- *   4. `attachment` maps from `capabilities.vision` per OmniRoute
- *      convention: vision = ability to receive image attachments. If the
- *      raw entry happens to expose an explicit `capabilities.attachment`,
- *      that wins.
- *   5. `thinking` from OmniRoute has no 1:1 ModelV2 slot. We OR it into
- *      `reasoning` so thinking-only models still surface a non-false
- *      reasoning flag.
- *   6. `last_updated` from OmniRoute has no ModelV2 slot — dropped.
- *      `release_date` lands in ModelV2.release_date with `""` fallback
- *      (the field is required as `string`).
- *   7. `temperature: true` per OmniRoute convention (OpenAI-compat mode
- *      always supports the temperature knob). If a raw entry sets
- *      `capabilities.temperature` explicitly, that wins.
- *   8. Input/output modality arrays: each known modality flips its boolean.
- *      Unknown strings (future OmniRoute additions) are ignored — when the
- *      server adds new modalities we can map them here without breaking
- *      existing entries.
- *   9. `status: "active"` — OmniRoute doesn't tier models alpha/beta on
- *      /v1/models, and opencode needs a non-deprecated status to expose
- *      the model in the picker. If a future entry surfaces an explicit
- *      lifecycle hint we can map it then.
- *  10. `options: {}` and `headers: {}` left empty — they're escape hatches
- *      for opencode users to attach per-model overrides; the provider
- *      plugin must not preempt them.
- *  11. `limit.input` is OPTIONAL on ModelV2 (the `?` modifier). We only
- *      emit it when OmniRoute supplies `max_input_tokens` — keeps the
- *      shape clean for combo entries that only carry context_length.
- */
-export function mapRawModelToModelV2(
-  raw: OmniRouteRawModelEntry,
-  ctx: { providerId: string; baseURL: string; apiFormat?: ApiFormatV2 }
-): ModelV2 {
-  const caps = raw.capabilities ?? {};
-  // effort_tiers loop: server-declared tiers become ModelV2 variants so the
-  // UI offers exactly the tiers OmniRoute vouches for (instead of opencode's
-  // invented [low, medium, high] fallback). Blind: filtering/exclusion rules
-  // live server-side. Absent/empty/malformed => key omitted ENTIRELY (an
-  // empty variants object would suppress opencode's fallback for this model).
-  const declaredTiers = Array.isArray(caps.effort_tiers)
-    ? caps.effort_tiers.filter((t): t is string => typeof t === "string" && t.length > 0)
-    : [];
-  const variants =
-    declaredTiers.length > 0
-      ? Object.fromEntries(declaredTiers.map((tier) => [tier, { reasoningEffort: tier }]))
-      : undefined;
-  const inMods = new Set(raw.input_modalities ?? ["text"]);
-  const outMods = new Set(raw.output_modalities ?? ["text"]);
-
-  return {
-    // OC's static-catalog reader parses the key on `/` to recover
-    // `(providerID, modelID)`. If the raw id is already provider-prefixed
-    // (e.g. `cc/claude-opus-4-7` from the `cc` Claude Code alias, or
-    // `nvidia/llama-3-70b` from a provider that ships prefixed ids), leave
-    // it as-is — double-prefixing breaks OC's lookup. Bare **combo** ids
-    // (`owned_by: "combo"`, e.g. `gpt-5.6-sol`) must also stay unprefixed:
-    // OpenCode looks up `-m <plugin>/<combo>` as model id `<combo>` under
-    // the plugin provider. Other bare ids still prefix with
-    // `providerId` so credentials resolve as `(omniroute, model)`.
-    id: raw.id.includes("/") || raw.owned_by === "combo" ? raw.id : `${ctx.providerId}/${raw.id}`,
-    /**
-     * Display name. Falls back to raw.id when no enrichment is available;
-     * the caller overlays `/api/pricing/models` data via enrichment when
-     * the enrichment feature is enabled.
-     */
-    name: normaliseFreeLabel(raw.id),
-    capabilities: {
-      temperature: caps.temperature ?? true,
-      reasoning: Boolean(caps.reasoning || caps.thinking),
-      attachment: Boolean(caps.attachment ?? caps.vision ?? false),
-      toolcall: Boolean(caps.tool_calling ?? false),
-      input: {
-        text: inMods.has("text"),
-        audio: inMods.has("audio"),
-        image: inMods.has("image"),
-        video: inMods.has("video"),
-        pdf: inMods.has("pdf"),
-      },
-      output: {
-        text: outMods.has("text"),
-        audio: outMods.has("audio"),
-        image: outMods.has("image"),
-        video: outMods.has("video"),
-        pdf: outMods.has("pdf"),
-      },
-      interleaved: Boolean(caps.thinking),
-    },
-    cost: {
-      input: 0,
-      output: 0,
-      cache: { read: 0, write: 0 },
-    },
-    limit: {
-      context: typeof raw.context_length === "number" ? raw.context_length : 0,
-      ...(typeof raw.max_input_tokens === "number" ? { input: raw.max_input_tokens } : {}),
-      output: typeof raw.max_output_tokens === "number" ? raw.max_output_tokens : 0,
-    },
-    ...(variants ? { variants } : {}),
-    status: "active",
-    options: {},
-    headers: {},
-    release_date: raw.release_date ?? "",
-    providerID: ctx.providerId,
-    api: resolveApiBlockV2(raw.id, ctx.baseURL, ctx.apiFormat),
-  };
-}
--- a/@omniroute/opencode-plugin-v2/src/shared/naming.ts
+++ b/@omniroute/opencode-plugin-v2/src/shared/naming.ts
@@ -1,295 +0,0 @@
-/**
- * Universal model naming template for the OmniRoute plugin.
- *
- * Naming pipeline:
- *   [tag] <provider-label><separator><display-name><suffix>
- *
- *   [Free] <provider> - <name> · <budget>     ← free model
- *   Auto: <variant> (<N>p)                     ← auto combo
- *   Combo: <name>                              ← DB combo
- *   <provider> - <name>                        ← regular model
- */
-
-// ── Constants ────────────────────────────────────────────────────────────
-
-/** Separator between provider label and model display name. */
-export const PROVIDER_TAG_SEPARATOR = " - ";
-
-/** Threshold beyond which providerDisplayName is abbreviated. */
-const PROVIDER_LABEL_MAX_CHARS = 12;
-
-/** Aliases longer than this get title-case instead of UPPER. */
-const ALIAS_UPPER_MAX_CHARS = 5;
-
-// ── Auto Combo Types ─────────────────────────────────────────────────────
-
-export type AutoVariant = "coding" | "fast" | "cheap" | "offline" | "smart" | "lkgp";
-
-export const AUTO_VARIANTS: AutoVariant[] = ["coding", "fast", "cheap", "offline", "smart", "lkgp"];
-
-export const AUTO_VARIANT_DESCRIPTIONS: Record<AutoVariant | "default", string> = {
-  default: "Best provider via scoring",
-  coding: "Quality-first for code tasks",
-  fast: "Latency-optimized routing",
-  cheap: "Cost-optimized routing",
-  offline: "Offline-friendly providers",
-  smart: "Quality-first with exploration",
-  lkgp: "Last-Known-Good-Provider routing",
-};
-
-// ── Free Model Types ─────────────────────────────────────────────────────
-
-export type FreeModelFreeType =
-  | "recurring-daily"
-  | "recurring-monthly"
-  | "recurring-credit"
-  | "one-time-initial"
-  | "keyless"
-  | "discontinued";
-
-// ── Provider Label ────────────────────────────────────────────────────────
-
-/**
- * Title-case a long, lowercase-looking alias.
- * `antigravity` → `Antigravity`
- */
-function titleCaseAlias(alias: string): string {
-  if (alias.length === 0) return alias;
-  return alias.charAt(0).toUpperCase() + alias.slice(1).toLowerCase();
-}
-
-/**
- * Pick the short label for an upstream provider.
- *
- * Rules:
- *   1. Trim `providerDisplayName`. If ≤12 chars → use verbatim.
- *   2. Alias ≤5 chars → UPPER(alias). Alias >5 → titleCase.
- *   3. Neither → undefined.
- */
-export function shortProviderLabel(
-  enrichment: { providerDisplayName?: string; providerAlias?: string } | undefined
-): string | undefined {
-  if (!enrichment) return undefined;
-  const raw =
-    typeof enrichment.providerDisplayName === "string" ? enrichment.providerDisplayName.trim() : "";
-  if (raw.length > 0 && raw.length <= PROVIDER_LABEL_MAX_CHARS) return raw;
-  const alias = typeof enrichment.providerAlias === "string" ? enrichment.providerAlias.trim() : "";
-  if (alias.length > 0) {
-    return alias.length <= ALIAS_UPPER_MAX_CHARS ? alias.toUpperCase() : titleCaseAlias(alias);
-  }
-  // Long displayName with no alias to fall back on: keep the long label
-  // rather than dropping the provider prefix entirely.
-  return raw.length > 0 ? raw : undefined;
-}
-
-// ── Free Label ────────────────────────────────────────────────────────────
-
-/**
- * Normalise display name so free-tier models get a consistent `[Free] ` prefix.
- *
- * "GPT-4.1 (Free)"          → "[Free] GPT-4.1"
- * "DeepSeek V4 Flash Free"  → "[Free] DeepSeek V4 Flash"
- * "Claude Opus 4.7"         → "Claude Opus 4.7"  (unchanged)
- */
-export function normaliseFreeLabel(name: string): string {
-  // Bounded whitespace quantifiers ({0,8}/{1,8}) avoid the polynomial-ReDoS
-  // backtracking that unbounded \s* before an anchored \s*$ would allow on
-  // attacker-influenced display names. 8 covers any realistic label spacing.
-  const cleaned = name
-    .replace(/\s{0,8}\(free\)\s{0,8}$/i, "")
-    .replace(/[\s-]{1,8}free\s{0,8}$/i, "")
-    .trim();
-  const wasFree = cleaned.length < name.trim().length;
-  if (!wasFree) return name;
-  return `[Free] ${cleaned}`;
-}
-
-// ── Free Budget Formatting ────────────────────────────────────────────────
-
-/** Scales, largest first, so the unit is chosen by descending magnitude. */
-const TOKEN_UNITS = [
-  [1e9, "B"],
-  [1e6, "M"],
-  [1e3, "K"],
-] as const;
-
-/**
- * Format a token count as a short magnitude string: `25M`, `1.5K`, `999`.
- *
- * The unit has to be picked from the value that will actually be *printed*,
- * not from the raw input. `toFixed(1)` rounds to the nearest tenth, so at the
- * K scale 999_950 and above render as `1000.0` — and by then the M branch has
- * already been skipped, producing `1000K` for a number that is `1M`. The same
- * carry turns just under a billion into `1000M`. When the rounded value reaches
- * the next scale, re-render at that scale instead.
- */
-function fmtTokens(n: number): string {
-  for (let i = 0; i < TOKEN_UNITS.length; i++) {
-    const [scale, suffix] = TOKEN_UNITS[i]!;
-    if (n < scale) continue;
-    const value = Number((n / scale).toFixed(1));
-    // `Number()` also drops a trailing `.0`, which the previous regex did.
-    if (value < 1000 || i === 0) return `${value}${suffix}`;
-    const [nextScale, nextSuffix] = TOKEN_UNITS[i - 1]!;
-    return `${Number((n / nextScale).toFixed(1))}${nextSuffix}`;
-  }
-  return String(n);
-}
-
-/**
- * Format a free model budget into a short human-readable suffix.
- *
- * recurring-daily   → "25M tokens/day"
- * recurring-monthly → "25M tokens/month"
- * recurring-credit  → "10M credits"
- * one-time-initial  → "1M credits (one-time)"
- * keyless           → "(keyless)"
- * discontinued      → "(discontinued)"
- */
-export function formatFreeBudget(params: {
-  freeType: FreeModelFreeType;
-  monthlyTokens?: number;
-  creditTokens?: number;
-}): string {
-  const { freeType, monthlyTokens = 0, creditTokens = 0 } = params;
-
-  switch (freeType) {
-    case "recurring-daily":
-      return `${fmtTokens(monthlyTokens)} tokens/day`;
-    case "recurring-monthly":
-      return `${fmtTokens(monthlyTokens)} tokens/month`;
-    case "recurring-credit":
-      return `${fmtTokens(creditTokens)} credits`;
-    case "one-time-initial":
-      return `${fmtTokens(creditTokens)} credits (one-time)`;
-    case "keyless":
-      return "(keyless)";
-    case "discontinued":
-      return "(discontinued)";
-    default:
-      return "";
-  }
-}
-
-// ── Auto Combo Naming ─────────────────────────────────────────────────────
-
-/**
- * Format auto combo display name.
- *
- * "Auto: Coding (4p)"
- * "Auto: Default (6p)"
- * "Auto" (no candidate count when unknown)
- */
-export function formatAutoComboName(
-  variant: AutoVariant | undefined,
-  candidateCount?: number
-): string {
-  const label = variant ? variant.charAt(0).toUpperCase() + variant.slice(1) : "Default";
-  const count =
-    typeof candidateCount === "number" && candidateCount > 0 ? ` (${candidateCount}p)` : "";
-  return `Auto: ${label}${count}`;
-}
-
-/**
- * Build the model ID for an auto combo entry.
- * "auto/coding", "auto/fast", "auto" (default).
- */
-export function autoComboModelId(variant: AutoVariant | undefined): string {
-  return variant ? `auto/${variant}` : "auto";
-}
-
-// ── Universal Display Name Builder ────────────────────────────────────────
-
-export interface ModelDisplayNameParams {
-  /** Raw model ID (e.g. "cc/claude-sonnet-4-6"). */
-  rawId: string;
-  /** Enrichment display name (e.g. "Claude Sonnet 4.6"). */
-  enrichmentName?: string;
-  /** Provider tag enrichment. */
-  providerAlias?: string;
-  /** Human-readable upstream provider label. */
-  providerDisplayName?: string;
-  /** Whether model is free tier. */
-  isFree?: boolean;
-  /** Free model budget info. */
-  freeType?: FreeModelFreeType;
-  /** Monthly token budget (for recurring free models). */
-  monthlyTokens?: number;
-  /** Credit token budget (for credit-based free models). */
-  creditTokens?: number;
-  /** Whether this is a combo entry (skip provider tag). */
-  isCombo?: boolean;
-  /** Whether this is an auto combo entry. */
-  isAutoCombo?: boolean;
-  /** Auto combo variant. */
-  autoVariant?: AutoVariant;
-  /** Auto combo candidate count. */
-  autoCandidateCount?: number;
-}
-
-/**
- * Build the final display name following the universal template.
- *
- * Priority:
- *   1. Auto combo → "Auto: <variant> (<N>p)"
- *   2. DB combo → "Combo: <name>"
- *   3. Free + enrichment + provider tag → "[Free] <label> - <name> · <budget>"
- *   4. Free + enrichment → "[Free] <name> · <budget>"
- *   5. Free + raw → "[Free] <rawId> · <budget>"
- *   6. Enrichment + provider tag → "<label> - <name>"
- *   7. Enrichment only → "<name>"
- *   8. Raw fallback → normaliseFreeLabel(rawId)
- */
-export function buildModelDisplayName(params: ModelDisplayNameParams): string {
-  // Auto combos
-  if (params.isAutoCombo) {
-    return formatAutoComboName(params.autoVariant, params.autoCandidateCount);
-  }
-
-  // Determine base name — strip any existing free suffix first
-  const rawBase =
-    params.enrichmentName && params.enrichmentName.trim().length > 0
-      ? params.enrichmentName
-      : params.rawId;
-  const cleanedBase = rawBase
-    .replace(/\s*\(free\)\s*$/i, "")
-    .replace(/[\s-]+free\s*$/i, "")
-    .trim();
-  const wasFree = cleanedBase.length < rawBase.trim().length;
-  const isFree = !!params.isFree || wasFree;
-
-  let baseName = cleanedBase;
-
-  // Provider tag (skip for combos)
-  if (!params.isCombo) {
-    const label = shortProviderLabel({
-      providerDisplayName: params.providerDisplayName,
-      providerAlias: params.providerAlias,
-    });
-    if (label) {
-      const prefix = `${label}${PROVIDER_TAG_SEPARATOR}`;
-      if (!baseName.startsWith(prefix)) {
-        baseName = `${prefix}${baseName}`;
-      }
-    }
-  }
-
-  // Prepend [Free] if applicable (AFTER provider tag for correct ordering)
-  if (isFree) {
-    baseName = `[Free] ${baseName}`;
-  }
-
-  // Free budget suffix
-  if (isFree && params.freeType) {
-    const budget = formatFreeBudget({
-      freeType: params.freeType,
-      monthlyTokens: params.monthlyTokens,
-      creditTokens: params.creditTokens,
-    });
-    if (budget) {
-      baseName = `${baseName} · ${budget}`;
-    }
-  }
-
-  return baseName;
-}
--- a/@omniroute/opencode-plugin-v2/src/shared/usable.ts
+++ b/@omniroute/opencode-plugin-v2/src/shared/usable.ts
@@ -1,171 +0,0 @@
-import type { OmniRouteEnrichmentMap } from "./enrich.js";
-import type { OmniRouteRawCombo } from "./combos-map.js";
-
-/** Subset of `/api/providers` connections read by the usableOnly filter. */
-export interface OmniRouteProviderConnection {
-  /** Connection UUID. */
-  id: string;
-  /** Canonical provider id, e.g. `claude`, `gemini`, `kiro`. */
-  provider: string;
-  /** Operator toggle — when false, the connection is provisioned but disabled. */
-  isActive?: boolean;
-  /** Health-check verdict — `active` means routable. */
-  testStatus?: string;
-  /** Permissive bag — additional fields pass through untouched. */
-  [k: string]: unknown;
-}
-
-export type OmniRouteProvidersFetcher = (
-  baseURL: string,
-  apiKey: string,
-  timeoutMs?: number,
-  onSourceError?: (endpoint: string, reason: string) => void
-) => Promise<OmniRouteProviderConnection[]>;
-
-export interface UsableProviderSet {
-  aliases: Set<string>;
-  canonicals: Set<string>;
-  knownAliases: Set<string>;
-}
-
-/**
- * Default providers fetcher: `GET <baseURL>/api/providers` with bearer auth
- * + AbortController timeout. Accepts the `{ connections: [...] }` envelope
- * the gateway emits today, a bare-array envelope, and a `{ data: [...] }`
- * envelope (defensive). Refusals and network errors THROW so the caller keeps
- * last-known instead of silently disabling the filter; a 2xx with an empty
- * list is the only empty answer.
- */
-export const defaultOmniRouteProvidersFetcher: OmniRouteProvidersFetcher = async (
-  baseURL,
-  apiKey,
-  timeoutMs = 10_000,
-  onSourceError
-) => {
-  const empty: OmniRouteProviderConnection[] = [];
-  if (!baseURL || !apiKey) return empty;
-  const root = baseURL.replace(/\/v1\/?$/, "").replace(/\/$/, "");
-  const url = `${root}/api/providers`;
-  const ac = new AbortController();
-  const timer = setTimeout(() => ac.abort(), timeoutMs);
-  try {
-    const res = await fetch(url, {
-      method: "GET",
-      headers: {
-        Authorization: `Bearer ${apiKey}`,
-        Accept: "application/json",
-      },
-      signal: ac.signal,
-    });
-    if (!res.ok) {
-      // A refusal (403 behind a management-token gate, 503 mid-outage) is a
-      // failure, not an empty allowlist: the caller keeps last-known instead
-      // of silently disabling the filter.
-      onSourceError?.("/api/providers", `HTTP ${res.status}`);
-      throw new Error(`HTTP ${res.status}`);
-    }
-    const body = (await res.json()) as unknown;
-    const list = Array.isArray(body)
-      ? body
-      : Array.isArray((body as { connections?: unknown[] })?.connections)
-        ? (body as { connections: unknown[] }).connections
-        : Array.isArray((body as { data?: unknown[] })?.data)
-          ? (body as { data: unknown[] }).data
-          : [];
-    const out: OmniRouteProviderConnection[] = [];
-    for (const raw of list) {
-      if (!raw || typeof raw !== "object") continue;
-      const provider = (raw as { provider?: unknown }).provider;
-      if (typeof provider !== "string" || provider.length === 0) continue;
-      const id = (raw as { id?: unknown }).id;
-      const idStr = typeof id === "string" && id.length > 0 ? id : provider;
-      out.push({ ...(raw as Record<string, unknown>), id: idStr, provider });
-    }
-    return out;
-  } catch (err) {
-    // Network error, timeout, abort: keep last-known, never silently unfilter.
-    onSourceError?.("/api/providers", err instanceof Error ? err.message : String(err));
-    throw err instanceof Error ? err : new Error(String(err));
-  } finally {
-    clearTimeout(timer);
-  }
-};
-
-/**
- * Compute the provider prefixes safe to keep. A canonical provider is usable
- * when at least one connection is not explicitly disabled (`isActive: false`)
- * and has no failing health verdict (`testStatus` present and not "active"). Aliases (e.g. `cc` -> `claude`) resolve through the enrichment
- * map, which records every alias in `knownAliases` so the downstream filter
- * decides "this prefix was in /api/pricing/models" in O(1).
- *
- * Subtract-filter semantics: callers keep prefixes unknown to BOTH tables.
- */
-export function usableProviderAliasSet(
-  connections: OmniRouteProviderConnection[],
-  enrichment: OmniRouteEnrichmentMap | undefined
-): UsableProviderSet {
-  const usableCanonicals = new Set<string>();
-  for (const c of connections) {
-    // A missing toggle means "no opinion", not "disabled": the field only
-    // exists on newer gateways, and treating its absence as a veto would hide
-    // every model behind a filter the operator never asked to tighten. Only an
-    // explicit `false` disables.
-    if (!c || c.isActive === false) continue;
-    if (typeof c.testStatus === "string" && c.testStatus !== "active") continue;
-    if (typeof c.provider === "string" && c.provider.length > 0) {
-      usableCanonicals.add(c.provider);
-    }
-  }
-  const aliases = new Set<string>();
-  const knownAliases = new Set<string>();
-  if (enrichment) {
-    for (const entry of enrichment.values()) {
-      const alias = entry.providerAlias;
-      const canonical = entry.providerCanonical;
-      if (typeof alias !== "string" || alias.length === 0) continue;
-      knownAliases.add(alias);
-      if (typeof canonical !== "string" || canonical.length === 0) continue;
-      if (usableCanonicals.has(canonical)) aliases.add(alias);
-    }
-  }
-  for (const canonical of usableCanonicals) aliases.add(canonical);
-  return { aliases, canonicals: usableCanonicals, knownAliases };
-}
-
-/**
- * Decide whether a raw `/v1/models` id passes the usableOnly filter.
- * Rules (subtract-filter, bias toward keep): no `/` -> keep; usable alias
- * or canonical -> keep; unknown to both tables -> keep; known alias but not
- * usable -> drop.
- */
-export function isUsableRawModelId(id: string, usable: UsableProviderSet): boolean {
-  const slash = id.indexOf("/");
-  if (slash <= 0) return true;
-  const prefix = id.slice(0, slash);
-  if (usable.aliases.has(prefix) || usable.canonicals.has(prefix)) return true;
-  if (usable.knownAliases.has(prefix)) return false;
-  return true;
-}
-
-/**
- * Decide whether a combo passes the usableOnly filter. A combo keeps when AT
- * LEAST ONE member maps to a usable provider; unknown prefixes keep (cannot
- * prove unroutable); combos with zero resolvable members keep.
- */
-export function isUsableCombo(combo: OmniRouteRawCombo, usable: UsableProviderSet): boolean {
-  const steps = Array.isArray(combo.models) ? combo.models : [];
-  if (steps.length === 0) return true;
-  let sawResolvableMember = false;
-  for (const step of steps) {
-    if (step?.kind === "combo-ref") continue;
-    const modelId = typeof step?.model === "string" ? step.model : "";
-    const slash = modelId.indexOf("/");
-    if (slash <= 0) continue;
-    sawResolvableMember = true;
-    const prefix = modelId.slice(0, slash);
-    if (usable.aliases.has(prefix) || usable.canonicals.has(prefix)) return true;
-    if (!usable.knownAliases.has(prefix)) return true;
-  }
-  if (!sawResolvableMember) return true;
-  return false;
-}
--- a/@omniroute/opencode-plugin-v2/tests/api-package.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/api-package.test.ts
@@ -1,93 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-import { publishCatalog } from "../src/catalog.js";
-
-const SUPPORTED_PACKAGES = new Set(["@ai-sdk/openai-compatible", "@ai-sdk/anthropic"]);
-
-function fakeDraft(): { models: Map<string, ModelV2Info>; draft: CatalogDraft } {
-  const providers = new Map<string, ProviderV2Info>();
-  const models = new Map<string, ModelV2Info>();
-  const draft = {
-    provider: {
-      list: () => [],
-      get: (id: string) => providers.get(id) as never,
-      update: (id: string, fn: (p: ProviderV2Info) => void) => {
-        const p = (providers.get(id) ?? { id }) as ProviderV2Info;
-        fn(p);
-        providers.set(id, p);
-      },
-      remove: () => {},
-    },
-    model: {
-      get: () => undefined,
-      update: (pid: string, mid: string, fn: (m: ModelV2Info) => void) => {
-        const k = pid + "/" + mid;
-        const m = (models.get(k) ?? { id: mid, providerID: pid }) as ModelV2Info;
-        fn(m);
-        models.set(k, m);
-      },
-      remove: () => {},
-      default: { get: () => undefined, set: () => {} },
-    },
-  } as CatalogDraft;
-  return { models, draft };
-}
-
-const baseOpts = {
-  providerId: "omniroute",
-  baseURL: "https://gw.example.com",
-  apiKey: "k",
-  timeoutMs: 1000,
-  modelCacheTtlMs: 300000,
-  usableOnly: false,
-};
-
-function apiPackageOf(m: ModelV2Info | undefined): string {
-  assert.ok(m, "model must be published");
-  assert.equal(m?.api.type, "aisdk");
-  if (m?.api.type !== "aisdk") throw new Error("model api must be aisdk");
-  return m.api.package;
-}
-
-describe("catalog api package (models + combos + auto-combos)", () => {
-  it("every published entry carries a non-empty supported api.package", async () => {
-    const { models, draft } = fakeDraft();
-    const res = await publishCatalog(draft, baseOpts, {
-      fetcher: async () => [{ id: "gpt-x", context_length: 128000, max_output_tokens: 4096 }],
-      combosFetcher: async () => [
-        { id: "combo-a", name: "Combo A", models: [{ kind: "model", model: "gpt-x" }] },
-      ],
-      autoCombosFetcher: async () => [{ id: "auto", candidateCount: 6 }],
-    });
-    assert.deepEqual(res, { models: 1, combos: 1, autoCombos: 1 });
-    for (const key of ["omniroute/gpt-x", "omniroute/combo-a", "omniroute/auto"]) {
-      const pkg = apiPackageOf(models.get(key));
-      assert.ok(pkg.length > 0, `${key} api.package must be non-empty`);
-      assert.ok(SUPPORTED_PACKAGES.has(pkg), `${key} api.package must be supported, got ${pkg}`);
-    }
-  });
-
-  it("auto-combos follow the same anthropic apiFormat rule as models", async () => {
-    const { models, draft } = fakeDraft();
-    await publishCatalog(
-      draft,
-      {
-        ...baseOpts,
-        apiFormat: { allowAnthropic: true, anthropicModels: ["anthropic/claude-x", "auto/coding"] },
-      },
-      {
-        fetcher: async () => [{ id: "anthropic/claude-x" }],
-        combosFetcher: async () => [],
-        autoCombosFetcher: async () => [
-          { id: "auto/coding", variant: "coding", candidateCount: 4 },
-          { id: "auto/fast", variant: "fast", candidateCount: 2 },
-        ],
-      }
-    );
-    assert.equal(apiPackageOf(models.get("omniroute/anthropic/claude-x")), "@ai-sdk/anthropic");
-    assert.equal(apiPackageOf(models.get("omniroute/auto/coding")), "@ai-sdk/anthropic");
-    assert.equal(apiPackageOf(models.get("omniroute/auto/fast")), "@ai-sdk/openai-compatible");
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/auto-combos.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/auto-combos.test.ts
@@ -1,196 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-import { publishCatalog } from "../src/catalog.js";
-
-function fakeDraft(): {
-  models: Map<string, ModelV2Info>;
-  draft: CatalogDraft;
-  warns: string[];
-  restore: () => void;
-} {
-  const providers = new Map<string, ProviderV2Info>();
-  const models = new Map<string, ModelV2Info>();
-  const warns: string[] = [];
-  const origWarn = console.warn;
-  console.warn = (...args: unknown[]) => {
-    warns.push(String(args[0]));
-  };
-  const draft = {
-    provider: {
-      list: () => [],
-      get: (id: string) => providers.get(id) as never,
-      update: (id: string, fn: (p: ProviderV2Info) => void) => {
-        const p = (providers.get(id) ?? { id }) as ProviderV2Info;
-        fn(p);
-        providers.set(id, p);
-      },
-      remove: () => {},
-    },
-    model: {
-      get: () => undefined,
-      update: (pid: string, mid: string, fn: (m: ModelV2Info) => void) => {
-        const k = pid + "/" + mid;
-        const m = (models.get(k) ?? { id: mid, providerID: pid }) as ModelV2Info;
-        fn(m);
-        models.set(k, m);
-      },
-      remove: () => {},
-      default: { get: () => undefined, set: () => {} },
-    },
-  } as CatalogDraft;
-  return {
-    models,
-    draft,
-    warns,
-    restore: () => {
-      console.warn = origWarn;
-    },
-  };
-}
-
-const baseOpts = {
-  providerId: "omniroute",
-  baseURL: "https://gw.example.com",
-  apiKey: "k",
-  timeoutMs: 1000,
-  timeouts: { autoCombos: 5000 },
-  modelCacheTtlMs: 300000,
-  usableOnly: false,
-};
-
-describe("catalog auto combos (v1 parity)", () => {
-  it("publishes auto/* ids from the auto-combos fetcher", async () => {
-    const { models, draft, warns, restore } = fakeDraft();
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: async () => [],
-        combosFetcher: async () => [],
-        autoCombosFetcher: async () => [
-          { id: "auto", variant: undefined, candidateCount: 6 },
-          { id: "auto/coding", variant: "coding", candidateCount: 4 },
-        ],
-      });
-      assert.deepEqual(res, { models: 0, combos: 0, autoCombos: 2 });
-      assert.ok(models.has("omniroute/auto"), "auto entry must be published");
-      assert.ok(models.has("omniroute/auto/coding"), "auto/coding entry must be published");
-      const coding = models.get("omniroute/auto/coding");
-      assert.equal(coding?.limit.context, 128_000);
-      assert.equal(coding?.limit.output, 8_192);
-    } finally {
-      restore();
-    }
-    assert.ok(!warns.some((w) => w.includes("auto combo") && w.includes("failed")));
-  });
-
-  it("auto-combos fetch throw stays fail-open (warn, models kept)", async () => {
-    const { models, draft, warns, restore } = fakeDraft();
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: async () => [{ id: "m1" }],
-        combosFetcher: async () => [],
-        autoCombosFetcher: async () => {
-          throw new Error("GET https://gw.example.com/api/combos/auto failed: 500 boom");
-        },
-      });
-      assert.deepEqual(res, { models: 1, combos: 0, autoCombos: 0 });
-      assert.ok(models.has("omniroute/m1"));
-    } finally {
-      restore();
-    }
-    assert.ok(warns.some((w) => w.includes("auto combos") || w.includes("auto-combos")));
-  });
-
-  it("forwards the 5s auto-combos timeout to the fetcher", async () => {
-    const seen: number[] = [];
-    const { draft, restore } = fakeDraft();
-    try {
-      await publishCatalog(
-        draft,
-        { ...baseOpts, timeoutMs: 9999, timeouts: { autoCombos: 5000 } },
-        {
-          fetcher: async () => [],
-          combosFetcher: async () => [],
-          autoCombosFetcher: async (_base, _key, timeoutMs) => {
-            seen.push(timeoutMs ?? -1);
-            return [];
-          },
-        }
-      );
-    } finally {
-      restore();
-    }
-    assert.deepEqual(seen, [5000]);
-  });
-
-  it("hidden auto combos are skipped", async () => {
-    const { models, draft, restore } = fakeDraft();
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: async () => [],
-        combosFetcher: async () => [],
-        autoCombosFetcher: async () => [
-          { id: "auto/offline", variant: "offline", isHidden: true },
-          { id: "auto/fast", variant: "fast" },
-        ],
-      });
-      assert.deepEqual(res, { models: 0, combos: 0, autoCombos: 1 });
-      assert.ok(!models.has("omniroute/auto/offline"));
-      assert.ok(models.has("omniroute/auto/fast"));
-    } finally {
-      restore();
-    }
-  });
-});
-
-describe("auto combos obey the same allowlists as everything else", () => {
-  it("hides an auto combo the user hid, and keeps the rest", async () => {
-    const { publishCatalog } = await import("../src/catalog.js");
-    const published = new Map<string, Record<string, unknown>>();
-    const draft = {
-      provider: { update: (_i: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-      model: {
-        update: (pid: string, mid: string, fn: (m: Record<string, unknown>) => void) => {
-          const e: Record<string, unknown> = { id: mid, providerID: pid };
-          fn(e);
-          published.set(`${pid}/${mid}`, e);
-        },
-      },
-    };
-    await publishCatalog(
-      draft as never,
-      {
-        providerId: "omni",
-        baseURL: "https://gw.example.com",
-        apiKey: "k",
-        timeoutMs: 1000,
-        modelCacheTtlMs: 1000,
-        usableOnly: false,
-        enrichment: new Map(),
-        hiddenModels: ["auto/best-chaos"],
-      } as never,
-      {
-        models: async () => [{ id: "m1" }],
-        combos: async () => [],
-        autoCombos: async () =>
-          [
-            { id: "auto/best-coding", variant: "best-coding", candidateCount: 3 },
-            { id: "auto/best-chaos", variant: "best-chaos", candidateCount: 2 },
-          ] as never,
-        providers: async () => [],
-        enrichment: async () => new Map(),
-      }
-    );
-    const keys = [...published.keys()];
-    assert.ok(
-      keys.some((k) => k.includes("best-coding")),
-      `the other auto combo stays, got ${JSON.stringify(keys)}`
-    );
-    assert.equal(
-      keys.some((k) => k.includes("best-chaos")),
-      false,
-      `a hidden auto combo must not reach the picker, got ${JSON.stringify(keys)}`
-    );
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/cache-ttl-snapshot.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/cache-ttl-snapshot.test.ts
@@ -1,378 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { mkdtempSync, statSync } from "node:fs";
-import { tmpdir } from "node:os";
-import { join } from "node:path";
-import plugin from "../src/index.js";
-import { DEFAULT_MODEL_CACHE_TTL_MS, diskSnapshotPath } from "../src/cache.js";
-
-interface CtxOpts {
-  providerId: string;
-  baseURL: string;
-  apiKey: string;
-  modelCacheTtlMs?: number;
-}
-
-function stubFetch(
-  counter: { models: number; combos: number; autoCombos?: number; enrichment?: number },
-  modelIds: string[]
-) {
-  return (async (url: unknown) => {
-    const href = String(url);
-    if (href.includes("/api/pricing") || href.includes("/api/free-tier")) {
-      if (counter.enrichment !== undefined) counter.enrichment += 1;
-      return { ok: true, status: 200, statusText: "OK", json: async () => ({}) };
-    }
-    if (href.includes("/api/combos/auto")) {
-      if (counter.autoCombos !== undefined) counter.autoCombos += 1;
-      return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-    }
-    if (href.includes("/api/combos")) {
-      counter.combos += 1;
-      return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-    }
-    counter.models += 1;
-    return {
-      ok: true,
-      status: 200,
-      statusText: "OK",
-      json: async () => ({ data: modelIds.map((id) => ({ id })) }),
-    };
-  }) as typeof fetch;
-}
-
-async function setupPlugin(opts: CtxOpts): Promise<{
-  callbacks: Array<(draft: unknown) => Promise<void>>;
-  reloads: { count: number };
-}> {
-  const callbacks: Array<(draft: unknown) => Promise<void>> = [];
-  const reloads = { count: 0 };
-  const ctx = {
-    options: { ...opts },
-    catalog: {
-      transform: (cb: (draft: unknown) => Promise<void>) => {
-        callbacks.push(cb);
-        return Promise.resolve({ dispose: async () => {} });
-      },
-      reload: async () => {
-        reloads.count += 1;
-      },
-    },
-    integration: { transform: () => Promise.resolve({ dispose: async () => {} }) },
-  };
-  const logs: string[] = [];
-  const warns: string[] = [];
-  const origLog = console.log;
-  const origWarn = console.warn;
-  console.log = (...args: unknown[]) => {
-    logs.push(String(args[0]));
-  };
-  console.warn = (...args: unknown[]) => {
-    warns.push(String(args[0]));
-  };
-  try {
-    await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-  } finally {
-    console.log = origLog;
-    console.warn = origWarn;
-  }
-  return { callbacks, reloads };
-}
-
-function stubDraft(): { draft: unknown; published: Map<string, Record<string, unknown>> } {
-  const published = new Map<string, Record<string, unknown>>();
-  return {
-    published,
-    draft: {
-      provider: { update: (_id: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-      model: {
-        update: (pid: string, mid: string, fn: (m: Record<string, unknown>) => void) => {
-          const entry: Record<string, unknown> = { id: mid, providerID: pid };
-          fn(entry);
-          published.set(pid + "/" + mid, entry);
-        },
-      },
-    },
-  };
-}
-
-function isolateDisk(): { dir: string; restore: () => void } {
-  const dir = mkdtempSync(join(tmpdir(), "omniroute-p1-"));
-  const prev = process.env.OPENCODE_DATA_DIR;
-  process.env.OPENCODE_DATA_DIR = dir;
-  return {
-    dir,
-    restore: () => {
-      if (prev === undefined) delete process.env.OPENCODE_DATA_DIR;
-      else process.env.OPENCODE_DATA_DIR = prev;
-    },
-  };
-}
-
-describe("plugin-v2 P1 parity: TTL 300s + disk snapshot", () => {
-  it("TTL default is 300s (v1 parity)", () => {
-    assert.equal(DEFAULT_MODEL_CACHE_TTL_MS, 300_000);
-  });
-
-  it("2nd transform within TTL -> 0 network fetches (stub counter)", async () => {
-    const disk = isolateDisk();
-    const counter = { models: 0, combos: 0, autoCombos: 0 };
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch(counter, ["m1"]);
-    try {
-      const { callbacks } = await setupPlugin({
-        providerId: "ttl-hit",
-        baseURL: "https://gw.example.com",
-        apiKey: "k-ttl",
-      });
-      const { draft, published } = stubDraft();
-      await callbacks[0](draft);
-      assert.equal(counter.models, 1);
-      assert.equal(counter.combos, 1);
-      assert.equal(counter.autoCombos, 1);
-      assert.ok(published.has("ttl-hit/m1"));
-      await callbacks[0](draft);
-      assert.equal(counter.models, 1);
-      assert.equal(counter.combos, 1);
-      assert.equal(counter.autoCombos, 1);
-    } finally {
-      globalThis.fetch = origFetch;
-      disk.restore();
-    }
-  });
-
-  it("expired TTL -> network refetch", async () => {
-    const disk = isolateDisk();
-    const counter = { models: 0, combos: 0 };
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch(counter, ["m1"]);
-    const origNow = Date.now;
-    let now = 1_000_000;
-    Date.now = () => now;
-    try {
-      const { callbacks } = await setupPlugin({
-        providerId: "ttl-expire",
-        baseURL: "https://gw.example.com",
-        apiKey: "k-expire",
-        modelCacheTtlMs: 1000,
-      });
-      const { draft } = stubDraft();
-      await callbacks[0](draft);
-      assert.equal(counter.models, 1);
-      now += 500;
-      await callbacks[0](draft);
-      assert.equal(counter.models, 1);
-      now += 1000;
-      await callbacks[0](draft);
-      assert.equal(counter.models, 2);
-    } finally {
-      Date.now = origNow;
-      globalThis.fetch = origFetch;
-      disk.restore();
-    }
-  });
-
-  it("concurrent calls share a single refresh promise", async () => {
-    const disk = isolateDisk();
-    const counter = { models: 0, combos: 0 };
-    const origFetch = globalThis.fetch;
-    let release!: () => void;
-    const gate = new Promise<void>((resolve) => {
-      release = resolve;
-    });
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      if (href.includes("/api/pricing") || href.includes("/api/free-tier")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({}) };
-      }
-      if (href.includes("/v1/models")) {
-        counter.models += 1;
-        await gate;
-        return {
-          ok: true,
-          status: 200,
-          statusText: "OK",
-          json: async () => ({ data: [{ id: "m1" }] }),
-        };
-      }
-      if (href.includes("/api/combos/auto")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      if (href.includes("/api/providers")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({}) };
-      }
-      counter.combos += 1;
-      return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-    }) as typeof fetch;
-    const origLog = console.log;
-    const origWarn = console.warn;
-    console.log = () => {};
-    console.warn = () => {};
-    try {
-      const { callbacks } = await setupPlugin({
-        providerId: "singleflight",
-        baseURL: "https://gw.example.com",
-        apiKey: "k-sf",
-      });
-      const { draft } = stubDraft();
-      const a = callbacks[0](draft);
-      const b = callbacks[0](draft);
-      release();
-      await Promise.all([a, b]);
-      assert.equal(counter.models, 1);
-    } finally {
-      console.log = origLog;
-      console.warn = origWarn;
-      globalThis.fetch = origFetch;
-      disk.restore();
-    }
-  });
-
-  it("snapshot: fetch down at boot + snapshot present -> warm catalog published", async () => {
-    const disk = isolateDisk();
-    const warmCounter = { models: 0, combos: 0 };
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch(warmCounter, ["mw"]);
-    const origLog = console.log;
-    const origWarn = console.warn;
-    console.log = () => {};
-    console.warn = () => {};
-    try {
-      const { callbacks } = await setupPlugin({
-        providerId: "warm",
-        baseURL: "https://gw.example.com",
-        apiKey: "k-warm",
-      });
-      const { draft } = stubDraft();
-      await callbacks[0](draft);
-      assert.ok(statSync(diskSnapshotPath("warm")).isFile());
-    } finally {
-      console.log = origLog;
-      console.warn = origWarn;
-      globalThis.fetch = origFetch;
-      disk.restore();
-    }
-
-    const warmBody = (await import("node:fs/promises").then((fs) =>
-      fs.readFile(join(disk.dir, "plugins", "omniroute-warm.json"), "utf8")
-    )) as string;
-    void warmBody;
-
-    const disk2prev = process.env.OPENCODE_DATA_DIR;
-    process.env.OPENCODE_DATA_DIR = disk.dir;
-    const downFetch = (async (url: unknown) => {
-      const href = String(url);
-      if (href.includes("/api/pricing") || href.includes("/api/free-tier")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({}) };
-      }
-      if (href.includes("/api/combos/auto")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      if (href.includes("/api/combos")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      return { ok: false, status: 500, statusText: "Down", json: async () => ({}) };
-    }) as typeof fetch;
-    globalThis.fetch = downFetch;
-    const warns: string[] = [];
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    console.log = () => {};
-    try {
-      const { callbacks } = await setupPlugin({
-        providerId: "warm",
-        baseURL: "https://gw.example.com",
-        apiKey: "k-warm",
-      });
-      const { draft, published } = stubDraft();
-      await callbacks[0](draft);
-      assert.ok(
-        published.has("warm/mw"),
-        `warm snapshot must publish mw, got: ${JSON.stringify([...published.keys()])}`
-      );
-    } finally {
-      console.log = origLog;
-      console.warn = origWarn;
-      globalThis.fetch = origFetch;
-      if (disk2prev === undefined) delete process.env.OPENCODE_DATA_DIR;
-      else process.env.OPENCODE_DATA_DIR = disk2prev;
-    }
-    void warns;
-  });
-
-  it("invalidation: fresh setup does not reuse the previous in-memory cache", async () => {
-    const disk = isolateDisk();
-    const counter = { models: 0, combos: 0 };
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch(counter, ["m1"]);
-    const origLog = console.log;
-    const origWarn = console.warn;
-    console.log = () => {};
-    console.warn = () => {};
-    try {
-      const first = await setupPlugin({
-        providerId: "inval",
-        baseURL: "https://gw.example.com",
-        apiKey: "k-inval",
-        modelCacheTtlMs: 1,
-      });
-      const { draft } = stubDraft();
-      await first.callbacks[0](draft);
-      assert.equal(counter.models, 1);
-      // Fresh setup = empty memory (setup closure): the stale disk warm entry
-      // expires + the refetch starts, no reuse of the previous cache.
-      const { setTimeout: sleep } = await import("node:timers/promises");
-      await sleep(5);
-      const second = await setupPlugin({
-        providerId: "inval",
-        baseURL: "https://gw.example.com",
-        apiKey: "k-inval",
-        modelCacheTtlMs: 1,
-      });
-      await second.callbacks[0](draft);
-      assert.equal(counter.models, 2);
-    } finally {
-      console.log = origLog;
-      console.warn = origWarn;
-      globalThis.fetch = origFetch;
-      disk.restore();
-    }
-  });
-});
-
-describe("the snapshot carries the display overlay across a restart", () => {
-  it("round-trips enrichment so a cold start is not a wall of raw ids", async () => {
-    const { mkdtempSync } = await import("node:fs");
-    const { tmpdir } = await import("node:os");
-    const { join } = await import("node:path");
-    const { writeDiskSnapshot, readDiskSnapshot } = await import("../src/cache.js");
-    const dir = mkdtempSync(join(tmpdir(), "omniroute-enrich-snap-"));
-    const prev = process.env.OPENCODE_DATA_DIR;
-    process.env.OPENCODE_DATA_DIR = dir;
-    try {
-      const enrichment = new Map([
-        ["cc/sonnet", { name: "Claude Sonnet 4.6", providerDisplayName: "Claude" }],
-      ]);
-      await writeDiskSnapshot(
-        "omni",
-        {
-          models: [{ id: "cc/sonnet" }],
-          combos: [],
-          autoCombos: [],
-          providers: [],
-          enrichment,
-          fetchedAt: Date.now(),
-        } as never,
-        "fp-1"
-      );
-      const back = await readDiskSnapshot("omni", "fp-1");
-      assert.ok(back, "snapshot readable");
-      assert.equal(back.enrichment?.get("cc/sonnet")?.name, "Claude Sonnet 4.6");
-      assert.equal(back.enrichment?.get("cc/sonnet")?.providerDisplayName, "Claude");
-    } finally {
-      if (prev === undefined) delete process.env.OPENCODE_DATA_DIR;
-      else process.env.OPENCODE_DATA_DIR = prev;
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/catalog.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/catalog.test.ts
@@ -1,330 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-import { publishCatalog } from "../src/catalog.js";
-
-interface FakeDraft {
-  providers: Map<string, ProviderV2Info>;
-  models: Map<string, ModelV2Info>;
-  warns: string[];
-  provider: CatalogDraft["provider"];
-  model: CatalogDraft["model"];
-}
-
-function fakeDraft(): FakeDraft {
-  const providers = new Map<string, ProviderV2Info>();
-  const models = new Map<string, ModelV2Info>();
-  return {
-    providers,
-    models,
-    warns: [],
-    provider: {
-      list: () => [],
-      get: (id: string) => providers.get(id) as never,
-      update: (id: string, fn: (p: ProviderV2Info) => void) => {
-        const p = (providers.get(id) ?? { id }) as ProviderV2Info;
-        fn(p);
-        providers.set(id, p);
-      },
-      remove: () => {},
-    },
-    model: {
-      get: () => undefined,
-      update: (pid: string, mid: string, fn: (m: ModelV2Info) => void) => {
-        const k = pid + "/" + mid;
-        const m = (models.get(k) ?? { id: mid, providerID: pid }) as ModelV2Info;
-        fn(m);
-        models.set(k, m);
-      },
-      remove: () => {},
-      default: { get: () => undefined, set: () => {} },
-    },
-  };
-}
-
-const baseOpts = {
-  providerId: "omniroute",
-  baseURL: "https://gw.example.com",
-  apiKey: "k",
-  timeoutMs: 1000,
-  modelCacheTtlMs: 300000,
-  usableOnly: false,
-};
-
-describe("catalog provider template", () => {
-  it("writes provider api package/url plus integrationID", async () => {
-    const draft = fakeDraft();
-    await publishCatalog(draft, baseOpts, {
-      fetcher: async () => [],
-      combosFetcher: async () => [],
-    });
-    const p = draft.providers.get("omniroute");
-    assert.ok(p);
-    assert.equal(p?.name, "OmniRoute");
-    assert.deepEqual(p?.api, {
-      type: "aisdk",
-      package: "@ai-sdk/openai-compatible",
-      url: "https://gw.example.com/v1",
-    });
-    assert.equal(p?.integrationID, "omniroute");
-  });
-
-  it("publishes stub models with mapped fields, counts them", async () => {
-    const draft = fakeDraft();
-    const res = await publishCatalog(draft, baseOpts, {
-      fetcher: async () => [
-        {
-          id: "gpt-x",
-          context_length: 128000,
-          max_output_tokens: 4096,
-          input_modalities: ["text", "image"],
-          capabilities: { tool_calling: true, reasoning: true },
-        },
-        { id: "plain" },
-      ],
-      combosFetcher: async () => [],
-    });
-    assert.deepEqual(res, { models: 2, combos: 0, autoCombos: 0 });
-    const m = draft.models.get("omniroute/gpt-x");
-    assert.ok(m);
-    assert.equal(m?.providerID, "omniroute");
-    assert.equal(m?.limit.context, 128000);
-    assert.equal(m?.capabilities.tools, true);
-    assert.equal(m?.status, "active");
-  });
-});
-
-describe("catalog allowlist", () => {
-  const stubModels = async () => [{ id: "cc/keep-me" }, { id: "cc/drop-me" }, { id: "bare-keep" }];
-
-  it("visible exact allowlist keeps matches, drops the rest", async () => {
-    const draft = fakeDraft();
-    const res = await publishCatalog(
-      draft,
-      { ...baseOpts, visibleModels: ["cc/keep-me"] },
-      { fetcher: stubModels, combosFetcher: async () => [] }
-    );
-    assert.equal(res.models, 1);
-    assert.ok(draft.models.has("omniroute/cc/keep-me"));
-    assert.ok(!draft.models.has("omniroute/cc/drop-me"));
-  });
-
-  it("hidden deny-wins over visible allow", async () => {
-    const draft = fakeDraft();
-    const res = await publishCatalog(
-      draft,
-      { ...baseOpts, visibleModels: ["cc/keep-me"], hiddenModels: ["cc/keep-me"] },
-      { fetcher: stubModels, combosFetcher: async () => [] }
-    );
-    assert.equal(res.models, 0);
-  });
-
-  it("bare hidden id matches any prefix via suffix rule", async () => {
-    const draft = fakeDraft();
-    const res = await publishCatalog(
-      draft,
-      { ...baseOpts, hiddenModels: ["drop-me"] },
-      { fetcher: stubModels, combosFetcher: async () => [] }
-    );
-    assert.equal(res.models, 2);
-    assert.ok(!draft.models.has("omniroute/cc/drop-me"));
-    assert.ok(draft.models.has("omniroute/bare-keep"));
-  });
-});
-
-describe("catalog fail-open", () => {
-  it("models fetch throw warns and returns zeros without throwing", async () => {
-    const draft = fakeDraft();
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: async () => {
-          throw new Error("boom 500");
-        },
-        combosFetcher: async () => [],
-      });
-      assert.deepEqual(res, { models: 0, combos: 0, autoCombos: 0 });
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.equal(warns.length, 1);
-    assert.match(warns[0], /models fetch failed/);
-    assert.ok(draft.providers.has("omniroute"));
-  });
-
-  it("combos fetch throw keeps models, warns, returns models-only counts", async () => {
-    const draft = fakeDraft();
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: async () => [{ id: "m1" }],
-        enrichmentFetcher: async () => new Map(),
-        combosFetcher: async () => {
-          throw Object.assign(new Error("Not Found"), { status: 404 });
-        },
-      });
-      assert.deepEqual(res, { models: 1, combos: 0, autoCombos: 0 });
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.equal(warns.length, 1);
-    assert.match(warns[0], /combos fetch failed/);
-    assert.ok(draft.models.has("omniroute/m1"));
-  });
-
-  it("combos fetch 403 (PROD) keeps models, warns, never rejects", async () => {
-    const draft = fakeDraft();
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: async () => [{ id: "m1" }],
-        enrichmentFetcher: async () => new Map(),
-        combosFetcher: async () => {
-          throw new Error(
-            "[omniroute-v2] GET https://gw.example.com/api/combos failed: 403 Forbidden"
-          );
-        },
-      });
-      assert.deepEqual(res, { models: 1, combos: 0, autoCombos: 0 });
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.equal(warns.length, 1);
-    assert.match(warns[0], /combos fetch failed/);
-    assert.match(warns[0], /403/);
-    assert.ok(draft.models.has("omniroute/m1"));
-  });
-
-  it("combos fetch 500 keeps models, warns, never rejects", async () => {
-    const draft = fakeDraft();
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: async () => [{ id: "m1" }],
-        enrichmentFetcher: async () => new Map(),
-        combosFetcher: async () => {
-          throw new Error(
-            "[omniroute-v2] GET https://gw.example.com/api/combos failed: 500 Internal Server Error"
-          );
-        },
-      });
-      assert.deepEqual(res, { models: 1, combos: 0, autoCombos: 0 });
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.equal(warns.length, 1);
-    assert.match(warns[0], /combos fetch failed/);
-    assert.match(warns[0], /500/);
-    assert.ok(draft.models.has("omniroute/m1"));
-  });
-
-  it("combos fetch timeout (AbortError) keeps models, warns, never rejects", async () => {
-    const draft = fakeDraft();
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: async () => [{ id: "m1" }],
-        enrichmentFetcher: async () => new Map(),
-        combosFetcher: async () => {
-          const err = new Error("This operation was aborted");
-          err.name = "AbortError";
-          throw err;
-        },
-      });
-      assert.deepEqual(res, { models: 1, combos: 0, autoCombos: 0 });
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.equal(warns.length, 1);
-    assert.match(warns[0], /combos fetch failed/);
-    assert.ok(draft.models.has("omniroute/m1"));
-  });
-});
-
-describe("catalog combo vs combo", () => {
-  it("second combo with same id warns only once", async () => {
-    const draft = fakeDraft();
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: async () => [{ id: "dupe" }],
-        enrichmentFetcher: async () => new Map(),
-        combosFetcher: async () => [
-          {
-            id: "dupe",
-            name: "Dupe Combo",
-            models: [{ kind: "model", model: "dupe" }],
-          },
-          {
-            id: "dupe",
-            name: "Dupe Combo Again",
-            models: [{ kind: "model", model: "dupe" }],
-          },
-        ],
-      });
-      assert.deepEqual(res, { models: 1, combos: 2, autoCombos: 0 });
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.equal(warns.length, 1);
-    assert.match(warns[0], /collides with a model id; combo wins/);
-    const m = draft.models.get("omniroute/dupe");
-    assert.ok(m);
-    assert.equal(m?.name, "Dupe Combo Again");
-  });
-});
-
-describe("catalog model bare vs combo", () => {
-  it("bare model id colliding with combo id warns exactly once", async () => {
-    const draft = fakeDraft();
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: async () => [{ id: "dupe" }],
-        enrichmentFetcher: async () => new Map(),
-        combosFetcher: async () => [
-          {
-            id: "dupe",
-            name: "Dupe Combo",
-            models: [{ kind: "model", model: "dupe" }],
-          },
-        ],
-      });
-      assert.deepEqual(res, { models: 1, combos: 1, autoCombos: 0 });
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.equal(warns.length, 1);
-    assert.match(warns[0], /collides with a model id; combo wins/);
-    assert.match(warns[0], /"omniroute\/dupe"/);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/compat.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/compat.test.ts
@@ -1,34 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { assertContext } from "../src/compat.js";
-
-function validContext() {
-  return {
-    options: { baseURL: "https://gw.example.com" },
-    catalog: { transform: async () => {} },
-    integration: { transform: async () => {} },
-  };
-}
-
-describe("assertContext", () => {
-  it("throws on non-object ctx", () => {
-    assert.throws(() => assertContext(null), /\[omniroute-v2\] contract breach/);
-  });
-  it("throws when catalog.transform is missing", () => {
-    const ctx = { ...validContext(), catalog: {} };
-    assert.throws(() => assertContext(ctx), /\[omniroute-v2\] contract breach/);
-  });
-  it("serves a catalog on a host that has no integration domain", () => {
-    // The integration domain carries the credential flow, not the catalog.
-    // Refusing to load without it would deny the whole plugin to a host that
-    // simply does not implement that surface yet.
-    assert.doesNotThrow(() => assertContext({ catalog: { transform: () => {} }, options: {} }));
-  });
-  it("throws when options is not an object", () => {
-    const ctx = { ...validContext(), options: undefined };
-    assert.throws(() => assertContext(ctx), /\[omniroute-v2\] contract breach/);
-  });
-  it("passes for a valid context", () => {
-    assert.doesNotThrow(() => assertContext(validContext()));
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/credentials.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/credentials.test.ts
@@ -1,128 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { PluginContext } from "@opencode-ai/plugin/v2/promise";
-import type { Logger } from "../src/shared/index.js";
-import { resolveApiKey, warnIfMissing } from "../src/credentials.js";
-
-function collectingLogger(): { log: Logger; warnings: string[] } {
-  const warnings: string[] = [];
-  const log = {
-    error: () => {},
-    warn: (m: string) => warnings.push(m),
-    info: () => {},
-    debug: () => {},
-  } as unknown as Logger;
-  return { log, warnings };
-}
-
-/** A host exposing `integration.connection`, optionally with a stored value. */
-function ctxWith(stored: unknown, opts: { withConnectionApi?: boolean } = {}): PluginContext {
-  const integration =
-    opts.withConnectionApi === false
-      ? {}
-      : {
-          connection: {
-            active: async () =>
-              stored === undefined ? undefined : { type: "credential", id: "c", label: "l" },
-            resolve: async () => stored,
-          },
-        };
-  return { integration } as unknown as PluginContext;
-}
-
-const ENV = "OMNIROUTE_API_KEY";
-
-async function withEnv<T>(value: string | undefined, fn: () => Promise<T>): Promise<T> {
-  const previous = process.env[ENV];
-  if (value === undefined) delete process.env[ENV];
-  else process.env[ENV] = value;
-  try {
-    return await fn();
-  } finally {
-    if (previous === undefined) delete process.env[ENV];
-    else process.env[ENV] = previous;
-  }
-}
-
-describe("gateway credential resolution", () => {
-  it("prefers the credential the host stored over one written in config", async () => {
-    const { log } = collectingLogger();
-    const got = await withEnv("from-env", () =>
-      resolveApiKey(ctxWith({ type: "key", key: "from-host" }), "omniroute", "from-option", log)
-    );
-    assert.deepEqual(got, { key: "from-host", origin: "connection" });
-  });
-
-  it("falls back to the configured option, then to the environment", async () => {
-    const { log } = collectingLogger();
-    const option = await withEnv("from-env", () =>
-      resolveApiKey(ctxWith(undefined), "omniroute", "from-option", log)
-    );
-    assert.deepEqual(option, { key: "from-option", origin: "option" });
-    const env = await withEnv("from-env", () =>
-      resolveApiKey(ctxWith(undefined), "omniroute", undefined, log)
-    );
-    assert.deepEqual(env, { key: "from-env", origin: "env" });
-  });
-
-  it("reports no key rather than pretending an empty one works", async () => {
-    const { log, warnings } = collectingLogger();
-    const got = await withEnv(undefined, () =>
-      resolveApiKey(ctxWith(undefined), "omniroute", undefined, log)
-    );
-    assert.deepEqual(got, { key: "", origin: "missing" });
-    warnIfMissing(got, "omniroute", log);
-    assert.equal(warnings.length, 1);
-    // The message must name every way out, or it sends the user hunting.
-    assert.match(warnings[0] ?? "", /Connect the integration/);
-    assert.match(warnings[0] ?? "", /"apiKey"/);
-    assert.match(warnings[0] ?? "", new RegExp(ENV));
-  });
-
-  it("declines an oauth credential instead of reading a token as a key", async () => {
-    const { log, warnings } = collectingLogger();
-    const got = await withEnv(undefined, () =>
-      resolveApiKey(
-        ctxWith({ type: "oauth", methodID: "m", refresh: "r", access: "a", expires: 0 }),
-        "omniroute",
-        undefined,
-        log
-      )
-    );
-    assert.equal(got.origin, "missing");
-    assert.equal(warnings.length, 1);
-    assert.match(warnings[0] ?? "", /oauth/);
-  });
-
-  it("stays loadable on a host that has no connection api", async () => {
-    const { log, warnings } = collectingLogger();
-    const got = await withEnv(undefined, () =>
-      resolveApiKey(
-        ctxWith(undefined, { withConnectionApi: false }),
-        "omniroute",
-        "from-option",
-        log
-      )
-    );
-    assert.deepEqual(got, { key: "from-option", origin: "option" });
-    assert.deepEqual(warnings, []);
-  });
-
-  it("treats a lookup that throws as no credential, not as a failure to load", async () => {
-    const { log, warnings } = collectingLogger();
-    const ctx = {
-      integration: {
-        connection: {
-          active: async () => {
-            throw new Error("store unavailable");
-          },
-          resolve: async () => undefined,
-        },
-      },
-    } as unknown as PluginContext;
-    const got = await withEnv(undefined, () => resolveApiKey(ctx, "omniroute", "from-option", log));
-    assert.deepEqual(got, { key: "from-option", origin: "option" });
-    assert.equal(warnings.length, 1);
-    assert.match(warnings[0] ?? "", /store unavailable/);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/enrichment-attribution.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/enrichment-attribution.test.ts
@@ -1,59 +0,0 @@
-import { describe, it, afterEach } from "node:test";
-import assert from "node:assert/strict";
-import { defaultOmniRouteEnrichmentFetcher } from "../src/shared/enrich.js";
-
-const realFetch = globalThis.fetch;
-afterEach(() => {
-  globalThis.fetch = realFetch;
-});
-
-/**
- * A gateway routes the same model id through several upstream providers. The
- * overlay must keep them apart: one connection's price shown on another's
- * model is worse than no price at all, because it looks authoritative.
- */
-describe("two providers selling the same model id keep their own overlay", () => {
-  it("does not let the second provider's price land on the first one's model", async () => {
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      const ok = (body: unknown) => ({
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => body,
-      });
-      if (href.includes("/api/pricing/models"))
-        return ok({
-          providers: {
-            cc: {
-              id: "claude",
-              alias: "cc",
-              name: "Claude",
-              models: [{ id: "shared", name: "From Claude" }],
-            },
-            kir: {
-              id: "kiro",
-              alias: "kir",
-              name: "Kiro",
-              models: [{ id: "shared", name: "From Kiro" }],
-            },
-          },
-        });
-      if (href.includes("/api/pricing"))
-        return ok({
-          cc: { shared: { input: 3, output: 15 } },
-          kir: { shared: { input: 99, output: 99 } },
-        });
-      return ok({});
-    }) as unknown as typeof fetch;
-
-    const map = await defaultOmniRouteEnrichmentFetcher("https://gw.example.com", "k", 1000);
-    assert.deepEqual(map.get("cc/shared")?.pricing, { input: 3, output: 15 });
-    assert.deepEqual(map.get("kir/shared")?.pricing, { input: 99, output: 99 });
-    // The bare fallback belongs to whoever claimed it first, and keeps that
-    // provider's numbers rather than the last writer's.
-    assert.equal(map.get("shared")?.providerAlias, "cc");
-    assert.deepEqual(map.get("shared")?.pricing, { input: 3, output: 15 });
-    assert.notEqual(map.get("shared"), map.get("cc/shared"), "the bare entry is its own object");
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/enrichment-render.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/enrichment-render.test.ts
@@ -1,64 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { applyEnrichment } from "../src/shared/enrich.js";
-import type { Model as ModelV2 } from "@opencode-ai/sdk/v2";
-
-function model(id: string, name = id): ModelV2 {
-  return {
-    id,
-    name,
-    cost: { input: 0, output: 0, cache: { read: 0, write: 0 } },
-  } as unknown as ModelV2;
-}
-
-describe("the enrichment overlay reaches the picker", () => {
-  it("renders the upstream provider a model routes to", () => {
-    const m = applyEnrichment(model("cc/sonnet"), {
-      name: "Claude Sonnet 4.6",
-      providerDisplayName: "Claude",
-      providerAlias: "cc",
-    });
-    assert.equal(m.name, "Claude - Claude Sonnet 4.6");
-  });
-
-  it("marks a free model and states the budget the user actually gets", () => {
-    const m = applyEnrichment(model("pol/grok"), {
-      name: "Grok 4 Fast",
-      providerDisplayName: "Pollinations",
-      freeType: "recurring-monthly" as const,
-      monthlyTokens: 1_000_000,
-    });
-    assert.match(m.name, /\[Free\]/);
-    assert.match(m.name, /Grok 4 Fast/);
-    assert.match(
-      m.name,
-      /tokens\/month/,
-      "the budget the gateway reports is stated, not just the fact it is free"
-    );
-    assert.equal(m.name, "[Free] Pollinations - Grok 4 Fast · 1M tokens/month");
-  });
-
-  it("drops the tag when the caller turns it off", () => {
-    const m = applyEnrichment(
-      model("cc/sonnet"),
-      { name: "Claude Sonnet 4.6", providerDisplayName: "Claude" },
-      { providerTag: false }
-    );
-    assert.equal(m.name, "Claude Sonnet 4.6");
-  });
-
-  it("never tags a combo: it routes across providers, not to one", () => {
-    const m = applyEnrichment(
-      model("combo-fast", "Combo Fast"),
-      { name: "Combo Fast", providerDisplayName: "Claude" },
-      { isCombo: true }
-    );
-    assert.equal(m.name, "Combo Fast");
-  });
-
-  it("still overlays pricing", () => {
-    const m = applyEnrichment(model("x"), { pricing: { input: 3, output: 15 } });
-    assert.equal(m.cost.input, 3);
-    assert.equal(m.cost.output, 15);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/enrichment-report.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/enrichment-report.test.ts
@@ -1,106 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { Logger } from "../src/shared/index.js";
-import { createSourceErrorReporter } from "../src/enrichment-report.js";
-
-function collectingLogger(): { log: Logger; warnings: string[] } {
-  const warnings: string[] = [];
-  const log = {
-    error: () => {},
-    warn: (m: string) => warnings.push(m),
-    info: () => {},
-    debug: () => {},
-  } as unknown as Logger;
-  return { log, warnings };
-}
-
-describe("degraded enrichment is reported, not swallowed", () => {
-  it("names the endpoint and what the catalog loses", () => {
-    const { log, warnings } = collectingLogger();
-    createSourceErrorReporter(log, false)("/api/pricing", "HTTP 500");
-    assert.equal(warnings.length, 1);
-    assert.match(warnings[0] ?? "", /\/api\/pricing/);
-    assert.match(warnings[0] ?? "", /HTTP 500/);
-    assert.match(warnings[0] ?? "", /pricing are degraded/);
-  });
-
-  it("points a 403 at the management token when the inference key stands in", () => {
-    const { log, warnings } = collectingLogger();
-    createSourceErrorReporter(log, true)("/api/pricing/models", "HTTP 403");
-    assert.match(warnings[0] ?? "", /managementReadToken/);
-    assert.match(warnings[0] ?? "", /falls back to "apiKey"/);
-  });
-
-  it("does not blame the fallback when a management token was configured", () => {
-    const { log, warnings } = collectingLogger();
-    createSourceErrorReporter(log, false)("/api/pricing/models", "HTTP 403");
-    assert.match(warnings[0] ?? "", /was rejected/);
-    assert.doesNotMatch(warnings[0] ?? "", /falls back/);
-  });
-
-  it("keeps a transport failure free of auth advice", () => {
-    const { log, warnings } = collectingLogger();
-    createSourceErrorReporter(log, true)("/api/pricing", "connect ECONNREFUSED");
-    assert.doesNotMatch(warnings[0] ?? "", /managementReadToken/);
-  });
-
-  it("warns once per endpoint so a refresh loop cannot spam the log", () => {
-    const { log, warnings } = collectingLogger();
-    const report = createSourceErrorReporter(log, true);
-    report("/api/pricing", "HTTP 403");
-    report("/api/pricing", "HTTP 403");
-    report("/api/free-tier/summary", "HTTP 403");
-    assert.equal(warnings.length, 2);
-  });
-});
-
-describe("the library path reports too, not only the plugin", () => {
-  it("passes a refused /api/providers up to the caller through publishCatalog", async () => {
-    const { publishCatalog } = await import("../src/catalog.js");
-    const seen: Array<[string, string]> = [];
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      const ok = (body: unknown) => ({
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => body,
-      });
-      if (href.includes("/api/providers"))
-        return { ok: false, status: 403, statusText: "Forbidden", json: async () => ({}) };
-      if (href.includes("/api/pricing") || href.includes("/api/free-tier"))
-        return { ok: false, status: 403, statusText: "Forbidden", json: async () => ({}) };
-      if (href.includes("/api/combos")) return ok({ combos: [] });
-      return ok({ data: [{ id: "m1" }] });
-    }) as unknown as typeof fetch;
-    const draft = {
-      provider: { update: (_id: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-      model: {
-        update: (_p: string, _m: string, fn: (x: Record<string, unknown>) => void) => fn({}),
-      },
-    };
-    try {
-      await publishCatalog(
-        draft as never,
-        {
-          providerId: "omni",
-          baseURL: "https://gw.example.com",
-          apiKey: "k",
-          timeoutMs: 1000,
-          modelCacheTtlMs: 1000,
-          usableOnly: true,
-        } as never,
-        { onSourceError: (endpoint, reason) => seen.push([endpoint, reason]) }
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-    const endpoints = seen.map(([e]) => e);
-    assert.ok(
-      endpoints.includes("/api/providers"),
-      `the usable filter must say it was refused, got ${JSON.stringify(endpoints)}`
-    );
-    assert.ok(endpoints.some((e) => e.startsWith("/api/pricing")));
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/enrichment.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/enrichment.test.ts
@@ -1,124 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-import type { OmniRouteEnrichmentMap } from "../src/shared/index.js";
-import { publishCatalog } from "../src/catalog.js";
-
-function fakeDraft(): { models: Map<string, ModelV2Info>; draft: CatalogDraft } {
-  const providers = new Map<string, ProviderV2Info>();
-  const models = new Map<string, ModelV2Info>();
-  const draft = {
-    provider: {
-      list: () => [],
-      get: (id: string) => providers.get(id) as never,
-      update: (id: string, fn: (p: ProviderV2Info) => void) => {
-        const p = (providers.get(id) ?? { id }) as ProviderV2Info;
-        fn(p);
-        providers.set(id, p);
-      },
-      remove: () => {},
-    },
-    model: {
-      get: () => undefined,
-      update: (pid: string, mid: string, fn: (m: ModelV2Info) => void) => {
-        const k = pid + "/" + mid;
-        const m = (models.get(k) ?? { id: mid, providerID: pid }) as ModelV2Info;
-        fn(m);
-        models.set(k, m);
-      },
-      remove: () => {},
-      default: { get: () => undefined, set: () => {} },
-    },
-  } as CatalogDraft;
-  return { models, draft };
-}
-
-const baseOpts = {
-  providerId: "omniroute",
-  baseURL: "https://gw.example.com",
-  apiKey: "k",
-  timeoutMs: 1000,
-  modelCacheTtlMs: 300000,
-  usableOnly: false,
-};
-
-const stubModels = async () => [{ id: "cc/m1", context_length: 1000 }];
-
-describe("catalog enrichment source", () => {
-  it("applies names and pricing from an injected enrichmentFetcher", async () => {
-    const { models, draft } = fakeDraft();
-    const enrichment: OmniRouteEnrichmentMap = new Map([
-      ["cc/m1", { name: "Model One", pricing: { input: 3, output: 15 } }],
-      ["m1", { name: "Model One", pricing: { input: 3, output: 15 } }],
-    ]);
-    const res = await publishCatalog(draft, baseOpts, {
-      fetcher: stubModels,
-      combosFetcher: async () => [],
-      enrichmentFetcher: async () => enrichment,
-    });
-    assert.deepEqual(res, { models: 1, combos: 0, autoCombos: 0 });
-    const m = models.get("omniroute/cc/m1");
-    assert.ok(m);
-    assert.equal(m?.name, "Model One");
-    assert.equal(m?.cost[0].input, 3);
-    assert.equal(m?.cost[0].output, 15);
-  });
-
-  it("stays fail-open when the enrichment fetch throws", async () => {
-    const { models, draft } = fakeDraft();
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const res = await publishCatalog(draft, baseOpts, {
-        fetcher: stubModels,
-        combosFetcher: async () => [],
-        enrichmentFetcher: async () => {
-          throw new Error("pricing down");
-        },
-      });
-      assert.deepEqual(res, { models: 1, combos: 0, autoCombos: 0 });
-      assert.ok(models.get("omniroute/cc/m1"));
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.ok(warns.some((w) => w.includes("enrichment fetch failed")));
-  });
-
-  it("forwards the enrichment timeout and skips the fetch when enrichment is false", async () => {
-    const { draft } = fakeDraft();
-    let seenTimeout: number | undefined;
-    await publishCatalog(
-      draft,
-      { ...baseOpts, timeouts: { enrichment: 4321 } },
-      {
-        fetcher: stubModels,
-        combosFetcher: async () => [],
-        enrichmentFetcher: async (_base, _key, timeout) => {
-          seenTimeout = timeout;
-          return new Map();
-        },
-      }
-    );
-    assert.equal(seenTimeout, 4321);
-
-    const { draft: draft2 } = fakeDraft();
-    let called = 0;
-    await publishCatalog(
-      draft2,
-      { ...baseOpts, enrichment: false },
-      {
-        fetcher: stubModels,
-        combosFetcher: async () => [],
-        enrichmentFetcher: async () => {
-          called += 1;
-          return new Map();
-        },
-      }
-    );
-    assert.equal(called, 0);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/fixtures/catalog.json
+++ b/@omniroute/opencode-plugin-v2/tests/fixtures/catalog.json
@@ -1,68 +0,0 @@
-{
-  "models": [
-    {
-      "id": "m-alpha",
-      "object": "model",
-      "owned_by": "openai",
-      "context_length": 128000,
-      "max_output_tokens": 4096,
-      "input_modalities": ["text", "image"],
-      "output_modalities": ["text"],
-      "capabilities": { "tool_calling": true, "reasoning": true }
-    },
-    {
-      "id": "m-beta",
-      "object": "model",
-      "owned_by": "openai",
-      "context_length": 64000,
-      "max_output_tokens": 2048,
-      "input_modalities": ["text"],
-      "output_modalities": ["text"],
-      "capabilities": { "tool_calling": false, "thinking": true, "temperature": false }
-    },
-    {
-      "id": "cc/m-gamma",
-      "object": "model",
-      "owned_by": "anthropic",
-      "context_length": 200000,
-      "max_output_tokens": 32000,
-      "input_modalities": ["text", "image", "pdf"],
-      "output_modalities": ["text"],
-      "capabilities": { "tool_calling": true, "reasoning": true, "vision": true }
-    },
-    {
-      "id": "local-delta",
-      "object": "model",
-      "context_length": 32768,
-      "capabilities": { "effort_tiers": ["low", "high"] }
-    },
-    {
-      "id": "good-combo",
-      "object": "model",
-      "owned_by": "omniroute",
-      "context_length": 100000,
-      "max_output_tokens": 8000,
-      "capabilities": { "tool_calling": true }
-    }
-  ],
-  "combos": [
-    {
-      "id": "combo-fast",
-      "name": "Combo Fast",
-      "strategy": "priority",
-      "models": [
-        { "kind": "model", "model": "m-alpha" },
-        { "kind": "model", "model": "cc/m-gamma" }
-      ]
-    },
-    {
-      "id": "good-combo",
-      "name": "Good Combo",
-      "strategy": "priority",
-      "models": [
-        { "kind": "model", "model": "m-beta" },
-        { "kind": "model", "model": "local-delta" }
-      ]
-    }
-  ]
-}
--- a/@omniroute/opencode-plugin-v2/tests/fixtures/v1-parity.json
+++ b/@omniroute/opencode-plugin-v2/tests/fixtures/v1-parity.json
@@ -1,301 +0,0 @@
-{
-  "_note": "Recorded output of @omniroute/opencode-plugin (v1) against tests/fixtures/catalog.json. This is the contract this package claims parity with; regenerate only when intentionally re-basing on a newer v1.",
-  "v1PluginVersion": "0.2.1",
-  "hookId": "opencode-omniroute",
-  "publishedKeys": ["combo-fast", "good-combo", "local-delta", "m-alpha", "m-beta"],
-  "comboSlugs": ["combo-fast", "good-combo"],
-  "mappedModels": {
-    "m-alpha": {
-      "id": "omniroute/m-alpha",
-      "name": "m-alpha",
-      "capabilities": {
-        "temperature": true,
-        "reasoning": true,
-        "attachment": false,
-        "toolcall": true,
-        "input": {
-          "text": true,
-          "audio": false,
-          "image": true,
-          "video": false,
-          "pdf": false
-        },
-        "output": {
-          "text": true,
-          "audio": false,
-          "image": false,
-          "video": false,
-          "pdf": false
-        },
-        "interleaved": false
-      },
-      "cost": {
-        "input": 0,
-        "output": 0,
-        "cache": {
-          "read": 0,
-          "write": 0
-        }
-      },
-      "limit": {
-        "context": 128000,
-        "output": 4096
-      },
-      "status": "active",
-      "options": {},
-      "headers": {},
-      "release_date": "",
-      "providerID": "omniroute",
-      "api": {
-        "id": "openai-compatible",
-        "url": "https://gw.example.com/v1",
-        "npm": "@ai-sdk/openai-compatible"
-      }
-    },
-    "m-beta": {
-      "id": "omniroute/m-beta",
-      "name": "m-beta",
-      "capabilities": {
-        "temperature": false,
-        "reasoning": true,
-        "attachment": false,
-        "toolcall": false,
-        "input": {
-          "text": true,
-          "audio": false,
-          "image": false,
-          "video": false,
-          "pdf": false
-        },
-        "output": {
-          "text": true,
-          "audio": false,
-          "image": false,
-          "video": false,
-          "pdf": false
-        },
-        "interleaved": true
-      },
-      "cost": {
-        "input": 0,
-        "output": 0,
-        "cache": {
-          "read": 0,
-          "write": 0
-        }
-      },
-      "limit": {
-        "context": 64000,
-        "output": 2048
-      },
-      "status": "active",
-      "options": {},
-      "headers": {},
-      "release_date": "",
-      "providerID": "omniroute",
-      "api": {
-        "id": "openai-compatible",
-        "url": "https://gw.example.com/v1",
-        "npm": "@ai-sdk/openai-compatible"
-      }
-    },
-    "local-delta": {
-      "id": "omniroute/local-delta",
-      "name": "local-delta",
-      "capabilities": {
-        "temperature": true,
-        "reasoning": false,
-        "attachment": false,
-        "toolcall": false,
-        "input": {
-          "text": true,
-          "audio": false,
-          "image": false,
-          "video": false,
-          "pdf": false
-        },
-        "output": {
-          "text": true,
-          "audio": false,
-          "image": false,
-          "video": false,
-          "pdf": false
-        },
-        "interleaved": false
-      },
-      "cost": {
-        "input": 0,
-        "output": 0,
-        "cache": {
-          "read": 0,
-          "write": 0
-        }
-      },
-      "limit": {
-        "context": 32768,
-        "output": 0
-      },
-      "variants": {
-        "low": {
-          "reasoningEffort": "low"
-        },
-        "high": {
-          "reasoningEffort": "high"
-        }
-      },
-      "status": "active",
-      "options": {},
-      "headers": {},
-      "release_date": "",
-      "providerID": "omniroute",
-      "api": {
-        "id": "openai-compatible",
-        "url": "https://gw.example.com/v1",
-        "npm": "@ai-sdk/openai-compatible"
-      }
-    },
-    "good-combo": {
-      "id": "omniroute/good-combo",
-      "name": "good-combo",
-      "capabilities": {
-        "temperature": true,
-        "reasoning": false,
-        "attachment": false,
-        "toolcall": true,
-        "input": {
-          "text": true,
-          "audio": false,
-          "image": false,
-          "video": false,
-          "pdf": false
-        },
-        "output": {
-          "text": true,
-          "audio": false,
-          "image": false,
-          "video": false,
-          "pdf": false
-        },
-        "interleaved": false
-      },
-      "cost": {
-        "input": 0,
-        "output": 0,
-        "cache": {
-          "read": 0,
-          "write": 0
-        }
-      },
-      "limit": {
-        "context": 100000,
-        "output": 8000
-      },
-      "status": "active",
-      "options": {},
-      "headers": {},
-      "release_date": "",
-      "providerID": "omniroute",
-      "api": {
-        "id": "openai-compatible",
-        "url": "https://gw.example.com/v1",
-        "npm": "@ai-sdk/openai-compatible"
-      }
-    }
-  },
-  "mappedCombos": {
-    "combo-fast": {
-      "id": "combo-fast",
-      "providerID": "omniroute",
-      "api": {
-        "id": "openai-compatible",
-        "url": "https://gw.example.com/v1",
-        "npm": "@ai-sdk/openai-compatible"
-      },
-      "name": "Combo Fast",
-      "capabilities": {
-        "temperature": true,
-        "reasoning": true,
-        "attachment": false,
-        "toolcall": true,
-        "input": {
-          "text": true,
-          "audio": false,
-          "image": true,
-          "video": false,
-          "pdf": false
-        },
-        "output": {
-          "text": true,
-          "audio": false,
-          "image": false,
-          "video": false,
-          "pdf": false
-        },
-        "interleaved": false
-      },
-      "cost": {
-        "input": 0,
-        "output": 0,
-        "cache": {
-          "read": 0,
-          "write": 0
-        }
-      },
-      "limit": {
-        "context": 128000,
-        "output": 4096
-      },
-      "status": "active",
-      "options": {},
-      "headers": {},
-      "release_date": ""
-    },
-    "good-combo": {
-      "id": "good-combo",
-      "providerID": "omniroute",
-      "api": {
-        "id": "openai-compatible",
-        "url": "https://gw.example.com/v1",
-        "npm": "@ai-sdk/openai-compatible"
-      },
-      "name": "Good Combo",
-      "capabilities": {
-        "temperature": false,
-        "reasoning": false,
-        "attachment": false,
-        "toolcall": false,
-        "input": {
-          "text": true,
-          "audio": false,
-          "image": false,
-          "video": false,
-          "pdf": false
-        },
-        "output": {
-          "text": true,
-          "audio": false,
-          "image": false,
-          "video": false,
-          "pdf": false
-        },
-        "interleaved": false
-      },
-      "cost": {
-        "input": 0,
-        "output": 0,
-        "cache": {
-          "read": 0,
-          "write": 0
-        }
-      },
-      "limit": {
-        "context": 32768,
-        "output": 2048
-      },
-      "status": "active",
-      "options": {},
-      "headers": {},
-      "release_date": ""
-    }
-  }
-}
--- a/@omniroute/opencode-plugin-v2/tests/gemini-language.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/gemini-language.test.ts
@@ -1,226 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { Logger } from "../src/shared/index.js";
-import type { LanguageModelV3 } from "@ai-sdk/provider";
-import plugin from "../src/index.js";
-import { sanitizeToolSchemasFor } from "../src/gemini-language.js";
-
-function collectingLogger(): { log: Logger; warnings: string[] } {
-  const warnings: string[] = [];
-  const log = {
-    error: () => {},
-    warn: (m: string) => warnings.push(m),
-    info: () => {},
-    debug: () => {},
-  } as unknown as Logger;
-  return { log, warnings };
-}
-
-/** Records what the underlying model was actually asked to do. */
-function recordingModel(): { model: Record<string, unknown>; seen: unknown[] } {
-  const seen: unknown[] = [];
-  const model = {
-    specificationVersion: "v3",
-    modelId: "gemini-2.5-flash",
-    provider: "omniroute",
-    doGenerate: async (options: unknown) => {
-      seen.push(options);
-      return { content: [], finishReason: "stop", usage: {}, warnings: [] };
-    },
-    doStream: async (options: unknown) => {
-      seen.push(options);
-      return { stream: new ReadableStream() };
-    },
-  };
-  return { model, seen };
-}
-
-const dirtyTools = [
-  {
-    type: "function",
-    name: "edit",
-    inputSchema: {
-      type: "object",
-      additionalProperties: false,
-      properties: { p: { type: "string" } },
-    },
-  },
-];
-
-describe("Gemini sanitising on the language model", () => {
-  it("cleans the tool schemas a Gemini model would reject, on both call paths", async () => {
-    const { log } = collectingLogger();
-    const { model, seen } = recordingModel();
-    const wrapped = sanitizeToolSchemasFor(
-      model as unknown as LanguageModelV3,
-      "gemini-2.5-flash",
-      log
-    );
-    await wrapped.doGenerate({ prompt: [], tools: structuredClone(dirtyTools) } as never);
-    await wrapped.doStream({ prompt: [], tools: structuredClone(dirtyTools) } as never);
-    assert.equal(seen.length, 2);
-    for (const options of seen) {
-      const schema = (options as { tools: Array<{ inputSchema: Record<string, unknown> }> })
-        .tools[0]!.inputSchema;
-      assert.equal("additionalProperties" in schema, false);
-      assert.equal(
-        ((schema["properties"] as Record<string, Record<string, unknown>>)["p"] ?? {})["type"],
-        "string"
-      );
-    }
-  });
-
-  it("leaves a non-Gemini model's request exactly as it was", async () => {
-    const { log } = collectingLogger();
-    const { model, seen } = recordingModel();
-    const wrapped = sanitizeToolSchemasFor(
-      model as unknown as LanguageModelV3,
-      "claude-opus-5",
-      log
-    );
-    assert.equal(wrapped, model, "a model that needs no cleaning must not even be wrapped");
-    await (wrapped as unknown as { doGenerate: (o: unknown) => Promise<unknown> }).doGenerate({
-      prompt: [],
-      tools: structuredClone(dirtyTools),
-    });
-    const schema = (seen[0] as { tools: Array<{ inputSchema: Record<string, unknown> }> }).tools[0]!
-      .inputSchema;
-    assert.equal(schema["additionalProperties"], false);
-  });
-
-  it("forwards a request with no tools untouched", async () => {
-    const { log } = collectingLogger();
-    const { model, seen } = recordingModel();
-    const wrapped = sanitizeToolSchemasFor(
-      model as unknown as LanguageModelV3,
-      "gemini-2.5-flash",
-      log
-    );
-    await wrapped.doGenerate({ prompt: [] } as never);
-    assert.deepEqual(seen[0], { prompt: [] });
-  });
-
-  it("keeps the properties the model still needs, and says once that it cleaned", async () => {
-    const { log, warnings } = collectingLogger();
-    const { model, seen } = recordingModel();
-    const wrapped = sanitizeToolSchemasFor(
-      model as unknown as LanguageModelV3,
-      "gemini-2.5-flash",
-      log
-    );
-    await wrapped.doGenerate({ prompt: [], tools: structuredClone(dirtyTools) } as never);
-    await wrapped.doGenerate({ prompt: [], tools: structuredClone(dirtyTools) } as never);
-    const schema = (seen[0] as { tools: Array<{ inputSchema: Record<string, any> }> }).tools[0]!
-      .inputSchema;
-    assert.equal(schema["properties"]["p"]["type"], "string");
-    assert.equal(warnings.length, 0, "a routine cleaning is not a warning");
-  });
-
-  it("passes the untouched model through when the host hands over nothing to wrap", () => {
-    const { log } = collectingLogger();
-    assert.equal(
-      sanitizeToolSchemasFor(undefined as LanguageModelV3 | undefined, "gemini-2.5-flash", log),
-      undefined
-    );
-  });
-});
-
-describe("Gemini sanitising is wired into the host, and only where it belongs", () => {
-  interface LanguageInput {
-    model: { id: string; providerID: string };
-    sdk: unknown;
-    options: Record<string, unknown>;
-    language?: LanguageModelV3;
-  }
-
-  function hostCtx(opts: { geminiSanitization?: boolean; withAisdk?: boolean }): {
-    ctx: Record<string, unknown>;
-    languageCallbacks: Array<(input: LanguageInput) => void | Promise<void>>;
-  } {
-    const languageCallbacks: Array<(input: LanguageInput) => void | Promise<void>> = [];
-    const registration = Promise.resolve({ dispose: async () => {} });
-    const options: Record<string, unknown> = {
-      baseURL: "https://gw.example.com",
-      providerId: "omni",
-      apiKey: "k",
-    };
-    if (opts.geminiSanitization !== undefined)
-      options["geminiSanitization"] = opts.geminiSanitization;
-    const ctx: Record<string, unknown> = {
-      options,
-      catalog: { transform: () => registration, reload: async () => {} },
-      integration: { transform: () => registration },
-    };
-    if (opts.withAisdk !== false) {
-      ctx["aisdk"] = {
-        language: (cb: (input: LanguageInput) => void | Promise<void>) => {
-          languageCallbacks.push(cb);
-          return registration;
-        },
-        sdk: () => registration,
-      };
-    }
-    return { ctx, languageCallbacks };
-  }
-
-  async function setupPlugin(ctx: Record<string, unknown>): Promise<void> {
-    const warn = console.warn;
-    const log = console.log;
-    console.warn = () => {};
-    console.log = () => {};
-    try {
-      await (plugin as unknown as { setup: (c: unknown) => Promise<void> }).setup(ctx);
-    } finally {
-      console.warn = warn;
-      console.log = log;
-    }
-  }
-
-  const bareModel = { specificationVersion: "v3", modelId: "x" } as unknown as LanguageModelV3;
-
-  it("wraps a Gemini model of this provider and leaves every other one alone", async () => {
-    const { ctx, languageCallbacks } = hostCtx({});
-    await setupPlugin(ctx);
-    assert.equal(languageCallbacks.length, 1);
-    const wrapOf = async (model: { id: string; providerID: string }) => {
-      const input: LanguageInput = { model, sdk: {}, options: {}, language: bareModel };
-      await languageCallbacks[0]!(input);
-      return input.language !== bareModel;
-    };
-    assert.equal(await wrapOf({ id: "gemini-2.5-flash", providerID: "omni" }), true);
-    assert.equal(await wrapOf({ id: "claude-opus-5", providerID: "omni" }), false);
-    assert.equal(
-      await wrapOf({ id: "gemini-2.5-flash", providerID: "some-other-provider" }),
-      false,
-      "another provider's models are none of this plugin's business"
-    );
-  });
-
-  it("registers nothing when the option is turned off", async () => {
-    const { ctx, languageCallbacks } = hostCtx({ geminiSanitization: false });
-    await setupPlugin(ctx);
-    assert.deepEqual(languageCallbacks, []);
-  });
-
-  it("still loads on a host that exposes no aisdk domain", async () => {
-    const { ctx } = hostCtx({ withAisdk: false });
-    await setupPlugin(ctx);
-  });
-
-  it("keeps the catalog when the host refuses the language hook", async () => {
-    const registration = Promise.resolve({ dispose: async () => {} });
-    const ctx: Record<string, unknown> = {
-      options: { baseURL: "https://gw.example.com", providerId: "omni", apiKey: "k" },
-      catalog: { transform: () => registration, reload: async () => {} },
-      integration: { transform: () => registration },
-      aisdk: {
-        language: () => {
-          throw new Error("host says no");
-        },
-        sdk: () => registration,
-      },
-    };
-    // Must not reject: tool-schema cleaning is an extra, the catalog is the job.
-    await setupPlugin(ctx);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/host-contract.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/host-contract.test.ts
@@ -1,164 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-import {
-  publishCatalog,
-  type BinaryCompatModel,
-  type BinaryCompatProvider,
-  type BinaryCompatVariant,
-} from "../src/catalog.js";
-import { detectHostContract, emitsLegacyFields } from "../src/compat.js";
-
-/**
- * A host seed shape. `legacy` mirrors `Provider.Info.empty` as observed on
- * `@opencode-ai/cli` 0.0.0-beta-17823; `sdk` mirrors the pinned SDK contract;
- * `bare` is a host that discloses neither.
- */
-type SeedKind = "legacy" | "sdk" | "bare";
-
-function providerSeed(id: string, kind: SeedKind): ProviderV2Info {
-  if (kind === "legacy") {
-    return { id, name: id, activation: "auto", package: "" } as unknown as ProviderV2Info;
-  }
-  if (kind === "sdk") {
-    return { id, name: id, api: { type: "aisdk", package: "", url: "" } } as ProviderV2Info;
-  }
-  return { id } as ProviderV2Info;
-}
-
-function fakeDraft(kind: SeedKind): {
-  draft: CatalogDraft;
-  providers: Map<string, ProviderV2Info>;
-  models: Map<string, ModelV2Info>;
-} {
-  const providers = new Map<string, ProviderV2Info>();
-  const models = new Map<string, ModelV2Info>();
-  const draft = {
-    provider: {
-      list: () => [],
-      get: (id: string) => providers.get(id) as never,
-      update: (id: string, fn: (p: ProviderV2Info) => void) => {
-        const p = providers.get(id) ?? providerSeed(id, kind);
-        fn(p);
-        providers.set(id, p);
-      },
-      remove: () => {},
-    },
-    model: {
-      get: () => undefined,
-      update: (pid: string, mid: string, fn: (m: ModelV2Info) => void) => {
-        const k = pid + "/" + mid;
-        const m = (models.get(k) ?? { id: mid, providerID: pid }) as ModelV2Info;
-        fn(m);
-        models.set(k, m);
-      },
-      remove: () => {},
-      default: { get: () => undefined, set: () => {} },
-    },
-  } as CatalogDraft;
-  return { draft, providers, models };
-}
-
-const baseOpts = {
-  providerId: "omniroute",
-  baseURL: "https://gw.example.com",
-  apiKey: "k",
-  timeoutMs: 1000,
-  modelCacheTtlMs: 300000,
-  usableOnly: false,
-};
-
-const rawModel = {
-  id: "af/chat-latest",
-  capabilities: { effort_tiers: ["low", "high"] },
-};
-
-async function publish(kind: SeedKind) {
-  const { draft, providers, models } = fakeDraft(kind);
-  await publishCatalog(draft, baseOpts, {
-    fetcher: async () => [rawModel],
-    combosFetcher: async () => [],
-  });
-  const provider = providers.get("omniroute");
-  const model = models.get("omniroute/af/chat-latest");
-  assert.ok(provider, "provider must be published");
-  assert.ok(model, "model must be published");
-  return { provider: provider as BinaryCompatProvider, model: model as BinaryCompatModel };
-}
-
-describe("host contract detection", () => {
-  it("reads the contract off the seeded object, not off a version", () => {
-    assert.equal(detectHostContract({ id: "x", package: "" }), "legacy-package");
-    assert.equal(detectHostContract({ id: "x", api: { type: "aisdk" } }), "sdk-api");
-    assert.equal(detectHostContract({ id: "x" }), "unknown");
-    assert.equal(detectHostContract({ id: "x", api: {}, package: "" }), "unknown");
-    assert.equal(detectHostContract(undefined), "unknown");
-    assert.equal(detectHostContract("nope"), "unknown");
-  });
-
-  it("publishes the legacy fields for every contract but the sdk one", () => {
-    assert.equal(emitsLegacyFields("legacy-package"), true);
-    assert.equal(emitsLegacyFields("unknown"), true);
-    assert.equal(emitsLegacyFields("sdk-api"), false);
-  });
-});
-
-describe("legacy-package host (cli 0.0.0-beta-17823)", () => {
-  it("publishes package and settings.baseURL on the provider", async () => {
-    const { provider } = await publish("legacy");
-    assert.equal(provider.api.type, "aisdk");
-    assert.equal(provider.package, "aisdk:@ai-sdk/openai-compatible");
-    assert.equal(provider.settings.baseURL, "https://gw.example.com/v1");
-  });
-
-  it("publishes package, settings.baseURL and headers on the model", async () => {
-    const { model } = await publish("legacy");
-    if (model.api.type !== "aisdk") throw new Error("model api must be aisdk");
-    assert.equal(model.package, `aisdk:${model.api.package}`);
-    assert.equal(model.package, "aisdk:@ai-sdk/openai-compatible");
-    assert.equal(model.settings.baseURL, model.api.url);
-    assert.deepEqual(model.headers, model.request.headers);
-  });
-
-  it("publishes each variant in both shapes", async () => {
-    const { model } = await publish("legacy");
-    const variants = model.variants as BinaryCompatVariant[];
-    assert.deepEqual(
-      variants.map((v) => v.id),
-      ["low", "high"]
-    );
-    for (const variant of variants) {
-      assert.deepEqual(variant.settings, { reasoningEffort: variant.id });
-      // The pinned-contract shape stays intact next to the legacy one.
-      assert.deepEqual(variant.body, { reasoningEffort: variant.id });
-      assert.deepEqual(variant.headers, {});
-    }
-  });
-});
-
-describe("sdk-api host", () => {
-  it("publishes the api block only, with no legacy field", async () => {
-    const { provider, model } = await publish("sdk");
-    assert.equal(provider.api.type, "aisdk");
-    assert.equal("package" in provider, false);
-    assert.equal("settings" in provider, false);
-    assert.equal("package" in model, false);
-    assert.equal("settings" in model, false);
-    assert.equal("headers" in model, false);
-    for (const variant of model.variants) {
-      assert.equal("settings" in variant, false);
-      assert.deepEqual(variant.body, { reasoningEffort: variant.id });
-    }
-  });
-});
-
-describe("undisclosed host contract", () => {
-  it("falls back to the superset so an unknown host still routes", async () => {
-    const { provider, model } = await publish("bare");
-    assert.equal(provider.package, "aisdk:@ai-sdk/openai-compatible");
-    assert.equal(model.package, "aisdk:@ai-sdk/openai-compatible");
-    assert.ok(model.settings.baseURL);
-    assert.ok(model.api);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/index.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/index.test.ts
@@ -1,244 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import plugin from "../src/index.js";
-
-interface CapturedCall {
-  kind: "catalog" | "integration";
-}
-
-interface FakeCtx {
-  options: Record<string, unknown>;
-  catalog: {
-    transform: (cb: (draft: unknown) => unknown) => Promise<{ dispose: () => Promise<void> }>;
-  };
-  integration: {
-    transform: (cb: (draft: unknown) => unknown) => Promise<{ dispose: () => Promise<void> }>;
-  };
-}
-
-function fakeCtx(options: Record<string, unknown>, seen: CapturedCall[]): FakeCtx {
-  return {
-    options,
-    catalog: {
-      transform: (cb: (draft: unknown) => unknown) => {
-        seen.push({ kind: "catalog" });
-        assert.equal(typeof cb, "function");
-        return Promise.resolve({ dispose: async () => {} });
-      },
-    },
-    integration: {
-      transform: (cb: (draft: unknown) => unknown) => {
-        seen.push({ kind: "integration" });
-        assert.equal(typeof cb, "function");
-        return Promise.resolve({ dispose: async () => {} });
-      },
-    },
-  };
-}
-
-describe("plugin-v2 entrypoint", () => {
-  it("boot line is silent by default, visible with startupDebug", async () => {
-    const seen: CapturedCall[] = [];
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(args.map(String).join(" "));
-    };
-    try {
-      await (plugin as unknown as { setup: (ctx: FakeCtx) => Promise<void> }).setup(
-        fakeCtx({ baseURL: "https://gw.example.com", providerId: "omniroute" }, seen)
-      );
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.ok(
-      !warns.some((line) => line.includes("init providerId=")),
-      `boot line must stay silent by default, got: ${JSON.stringify(warns)}`
-    );
-    assert.deepEqual(
-      seen.map((s) => s.kind),
-      ["catalog", "integration"]
-    );
-    const seen2: CapturedCall[] = [];
-    const warns2: string[] = [];
-    console.warn = (...args: unknown[]) => {
-      warns2.push(args.map(String).join(" "));
-    };
-    try {
-      await (plugin as unknown as { setup: (ctx: FakeCtx) => Promise<void> }).setup(
-        fakeCtx(
-          { baseURL: "https://gw.example.com", providerId: "omniroute", startupDebug: true },
-          seen2
-        )
-      );
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.ok(
-      warns2.some((line) => line.includes("init providerId=omniroute")),
-      `boot line must show with startupDebug, got: ${JSON.stringify(warns2)}`
-    );
-  });
-
-  it("registers transforms synchronously: captures exist without awaiting fetch", async () => {
-    const seen: CapturedCall[] = [];
-    const ctx = fakeCtx({ baseURL: "https://gw.example.com" }, seen);
-    const pending = (plugin as unknown as { setup: (ctx: FakeCtx) => Promise<void> }).setup(ctx);
-    assert.deepEqual(
-      seen.map((s) => s.kind),
-      ["catalog", "integration"]
-    );
-    await pending;
-  });
-
-  it("declares key plus env methods and no oauth in the integration transform", async () => {
-    const seen: CapturedCall[] = [];
-    const integrationCallbacks: Array<(draft: unknown) => unknown> = [];
-    const catalogCallbacks: Array<(draft: unknown) => unknown> = [];
-    const ctx: FakeCtx = {
-      options: { baseURL: "https://gw.example.com", providerId: "omniroute" },
-      catalog: {
-        transform: (cb: (draft: unknown) => unknown) => {
-          seen.push({ kind: "catalog" });
-          catalogCallbacks.push(cb);
-          return Promise.resolve({ dispose: async () => {} });
-        },
-      },
-      integration: {
-        transform: (cb: (draft: unknown) => unknown) => {
-          seen.push({ kind: "integration" });
-          integrationCallbacks.push(cb);
-          return Promise.resolve({ dispose: async () => {} });
-        },
-      },
-    };
-    const logs: string[] = [];
-    const origLog = console.log;
-    console.log = (...args: unknown[]) => {
-      logs.push(String(args[0]));
-    };
-    try {
-      await (plugin as unknown as { setup: (ctx: FakeCtx) => Promise<void> }).setup(ctx);
-    } finally {
-      console.log = origLog;
-    }
-    assert.equal(integrationCallbacks.length, 1);
-
-    interface MethodUpdate {
-      integrationID: string;
-      method: { type: string; label?: string; names?: string[] };
-    }
-    interface FakeIntegrationDraft {
-      updates: Array<{ id: string; name: string }>;
-      methods: MethodUpdate[];
-      update: (id: string, fn: (i: { name: string }) => void) => void;
-      method: { update: (input: MethodUpdate) => void };
-    }
-    const draft: FakeIntegrationDraft = {
-      updates: [],
-      methods: [],
-      update(id: string, fn: (i: { name: string }) => void) {
-        const entry = { name: "" };
-        fn(entry);
-        this.updates.push({ id, name: entry.name });
-      },
-      method: {
-        update(input: MethodUpdate) {
-          draft.methods.push(input);
-        },
-      },
-    };
-    (integrationCallbacks[0] as (draft: FakeIntegrationDraft) => unknown)(draft);
-    assert.deepEqual(draft.updates, [{ id: "omniroute", name: "OmniRoute" }]);
-    const keyMethod = draft.methods.find((m) => m.method.type === "key");
-    const envMethod = draft.methods.find((m) => m.method.type === "env");
-    assert.ok(keyMethod);
-    assert.equal(keyMethod?.integrationID, "omniroute");
-    assert.ok(envMethod);
-    assert.deepEqual(envMethod?.method.names, ["OMNIROUTE_API_KEY"]);
-    assert.ok(!draft.methods.some((m) => m.method.type === "oauth"));
-  });
-
-  it("lazy refresh: [m1] then [m1,m2] reloads once; identical runs never reload", async () => {
-    const { mkdtempSync } = await import("node:fs");
-    const { tmpdir } = await import("node:os");
-    const { join } = await import("node:path");
-    const { setTimeout: sleep } = await import("node:timers/promises");
-    const dir = mkdtempSync(join(tmpdir(), "omniroute-lazy-"));
-    const prevDataDir = process.env.OPENCODE_DATA_DIR;
-    process.env.OPENCODE_DATA_DIR = dir;
-    let modelsCall = 0;
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      if (!href.includes("/v1/models")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      modelsCall += 1;
-      const ids = modelsCall <= 1 ? [{ id: "m1" }] : [{ id: "m1" }, { id: "m2" }];
-      return { ok: true, status: 200, statusText: "OK", json: async () => ({ data: ids }) };
-    }) as typeof fetch;
-    try {
-      const catalogCallbacks: Array<(draft: unknown) => Promise<void>> = [];
-      let reloads = 0;
-      const ctx = {
-        options: {
-          baseURL: "https://gw.example.com",
-          providerId: "lazy-reload",
-          apiKey: "k-lazy",
-          modelCacheTtlMs: 1,
-        },
-        catalog: {
-          transform: (cb: (draft: unknown) => Promise<void>) => {
-            catalogCallbacks.push(cb);
-            return Promise.resolve({ dispose: async () => {} });
-          },
-          reload: async () => {
-            reloads += 1;
-          },
-        },
-        integration: {
-          transform: () => Promise.resolve({ dispose: async () => {} }),
-        },
-      };
-      const logs: string[] = [];
-      const origLog = console.log;
-      console.log = (...args: unknown[]) => {
-        logs.push(String(args[0]));
-      };
-      try {
-        await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-      } finally {
-        console.log = origLog;
-      }
-      assert.equal(catalogCallbacks.length, 1);
-      const cb = catalogCallbacks[0] as (draft: unknown) => Promise<void>;
-      const draft = {
-        provider: { update: (_id: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-        model: {
-          update: (_pid: string, _mid: string, fn: (m: Record<string, unknown>) => void) => fn({}),
-        },
-      };
-      await cb(draft);
-      assert.equal(reloads, 0, "the first publish sets the baseline, it does not reload");
-      assert.equal(modelsCall, 1);
-      await sleep(5);
-      // The optional tier lands after that first publish and brings combos and
-      // the overlay with it — one reload, so the picker shows them without
-      // waiting for the next refresh.
-      const afterFirstUpgrade = reloads;
-      assert.ok(afterFirstUpgrade <= 1, `at most one reload for the first upgrade, got ${reloads}`);
-      await cb(draft);
-      assert.equal(reloads, afterFirstUpgrade + 1, "a new model id reloads once");
-      assert.equal(modelsCall, 2);
-      await sleep(5);
-      await cb(draft);
-      assert.equal(reloads, afterFirstUpgrade + 1, "an identical run never reloads");
-      assert.equal(modelsCall, 3);
-      if (prevDataDir === undefined) delete process.env.OPENCODE_DATA_DIR;
-      else process.env.OPENCODE_DATA_DIR = prevDataDir;
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/management-token.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/management-token.test.ts
@@ -1,233 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import plugin from "../src/index.js";
-import { publishCatalog } from "../src/catalog.js";
-
-const MODELS_URL = "https://gw.example.com/v1/models";
-const COMBOS_URL = "https://gw.example.com/api/combos";
-
-function silence() {
-  const warns: string[] = [];
-  const origWarn = console.warn;
-  const origLog = console.log;
-  console.warn = (...args: unknown[]) => {
-    warns.push(String(args[0]));
-  };
-  console.log = () => {};
-  return {
-    warns,
-    restore() {
-      console.warn = origWarn;
-      console.log = origLog;
-    },
-  };
-}
-
-function setup(options: Record<string, unknown>, reload?: () => Promise<void>) {
-  const catalogCallbacks: Array<(draft: unknown) => Promise<void>> = [];
-  const ctx = {
-    options,
-    catalog: {
-      transform: (cb: (draft: unknown) => Promise<void>) => {
-        catalogCallbacks.push(cb);
-        return Promise.resolve({ dispose: async () => {} });
-      },
-      ...(reload ? { reload } : {}),
-    },
-    integration: {
-      transform: () => Promise.resolve({ dispose: async () => {} }),
-    },
-  };
-  return { catalogCallbacks, ctx };
-}
-
-function stubDraft() {
-  const published = new Map<string, Record<string, unknown>>();
-  const draft = {
-    provider: { update: (_id: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-    model: {
-      update: (pid: string, mid: string, fn: (m: Record<string, unknown>) => void) => {
-        const entry: Record<string, unknown> = { id: mid, providerID: pid };
-        fn(entry);
-        published.set(pid + "/" + mid, entry);
-      },
-    },
-  };
-  return { draft, published };
-}
-
-describe("plugin-v2 managementReadToken wiring (F1)", () => {
-  it("combos fetch uses managementReadToken while models use apiKey", async () => {
-    const seen = new Map<string, string>();
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async (url: unknown, init?: { headers?: Record<string, string> }) => {
-      const href = String(url);
-      seen.set(href, String(init?.headers?.Authorization ?? ""));
-      if (href.includes("/api/combos/auto")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      if (href.includes("/api/combos")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      return {
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => ({ data: [{ id: "m1" }] }),
-      };
-    }) as typeof fetch;
-    const guard = silence();
-    try {
-      const { catalogCallbacks, ctx } = setup({
-        baseURL: "https://gw.example.com",
-        providerId: "omniroute",
-        apiKey: "chat-key",
-        managementReadToken: "mgmt-key",
-      });
-      await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-      const { draft, published } = stubDraft();
-      await catalogCallbacks[0](draft);
-      assert.ok(published.has("omniroute/m1"));
-      assert.equal(seen.get(COMBOS_URL), "Bearer mgmt-key");
-      assert.equal(seen.get(MODELS_URL), "Bearer chat-key");
-    } finally {
-      globalThis.fetch = origFetch;
-      guard.restore();
-    }
-  });
-
-  it("combos fetch falls back to apiKey when managementReadToken is absent", async () => {
-    const seen = new Map<string, string>();
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async (url: unknown, init?: { headers?: Record<string, string> }) => {
-      const href = String(url);
-      seen.set(href, String(init?.headers?.Authorization ?? ""));
-      if (href.includes("/api/combos/auto")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      if (href.includes("/api/combos")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      return {
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => ({ data: [{ id: "m1" }] }),
-      };
-    }) as typeof fetch;
-    const guard = silence();
-    try {
-      const { catalogCallbacks, ctx } = setup({
-        baseURL: "https://gw.example.com",
-        providerId: "omniroute",
-        apiKey: "chat-key",
-      });
-      await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-      const { draft } = stubDraft();
-      await catalogCallbacks[0](draft);
-      assert.equal(seen.get(COMBOS_URL), "Bearer chat-key");
-    } finally {
-      globalThis.fetch = origFetch;
-      guard.restore();
-    }
-  });
-
-  it("publishCatalog routes combosFetcher to managementReadToken, models to apiKey", async () => {
-    const calls: Array<[string, string]> = [];
-    const draft = {
-      provider: { update: (_id: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-      model: {
-        update: (_p: string, _m: string, fn: (m: Record<string, unknown>) => void) => fn({}),
-      },
-    };
-    const res = await publishCatalog(
-      draft as never,
-      {
-        providerId: "omniroute",
-        baseURL: "https://gw.example.com",
-        apiKey: "chat-key",
-        managementReadToken: "mgmt-key",
-        timeoutMs: 1000,
-        modelCacheTtlMs: 300000,
-        usableOnly: false,
-      },
-      {
-        fetcher: async (_baseURL, token) => {
-          calls.push(["models", token]);
-          return [{ id: "m1" }];
-        },
-        combosFetcher: async (_baseURL, token) => {
-          calls.push(["combos", token]);
-          return [];
-        },
-      }
-    );
-    assert.deepEqual(res, { models: 1, combos: 0, autoCombos: 0 });
-    assert.deepEqual(calls, [
-      ["models", "chat-key"],
-      ["combos", "mgmt-key"],
-    ]);
-  });
-});
-
-describe("plugin-v2 fail-closed models (F2)", () => {
-  it("empty models fetch on 2nd refresh keeps the last-known catalog", async () => {
-    const { mkdtempSync } = await import("node:fs");
-    const { tmpdir } = await import("node:os");
-    const { join } = await import("node:path");
-    const dir = mkdtempSync(join(tmpdir(), "omniroute-f2-"));
-    const prevDataDir = process.env.OPENCODE_DATA_DIR;
-    process.env.OPENCODE_DATA_DIR = dir;
-    let modelsCall = 0;
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      if (href.includes("/api/combos/auto")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      if (href.includes("/api/combos")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      modelsCall += 1;
-      if (modelsCall === 1) {
-        return {
-          ok: true,
-          status: 200,
-          statusText: "OK",
-          json: async () => ({ data: [{ id: "m1" }] }),
-        };
-      }
-      return { ok: false, status: 500, statusText: "Error", json: async () => ({}) };
-    }) as typeof fetch;
-    const guard = silence();
-    try {
-      const { catalogCallbacks, ctx } = setup({
-        baseURL: "https://gw.example.com",
-        providerId: "f2-keep",
-        apiKey: "k-f2",
-        modelCacheTtlMs: 1,
-      });
-      await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-      const first = stubDraft();
-      await catalogCallbacks[0](first.draft);
-      assert.ok(first.published.has("f2-keep/m1"), "first refresh must publish m1");
-      const { setTimeout: sleep } = await import("node:timers/promises");
-      await sleep(5);
-      const second = stubDraft();
-      await catalogCallbacks[0](second.draft);
-      if (prevDataDir === undefined) delete process.env.OPENCODE_DATA_DIR;
-      else process.env.OPENCODE_DATA_DIR = prevDataDir;
-      assert.ok(
-        second.published.has("f2-keep/m1"),
-        "empty models fetch must reuse last-known catalog"
-      );
-      assert.ok(
-        guard.warns.some((w) => w.includes("keeping last-known catalog")),
-        `expected keep-last-known warn, got: ${JSON.stringify(guard.warns)}`
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      guard.restore();
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/nested-combos.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/nested-combos.test.ts
@@ -1,236 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-import { publishCatalog } from "../src/catalog.js";
-
-interface Captured {
-  models: Map<string, ModelV2Info>;
-  draft: CatalogDraft;
-  warns: string[];
-  restore: () => void;
-}
-
-function fakeDraft(): Captured {
-  const providers = new Map<string, ProviderV2Info>();
-  const models = new Map<string, ModelV2Info>();
-  const warns: string[] = [];
-  const origWarn = console.warn;
-  console.warn = (...args: unknown[]) => {
-    warns.push(String(args[0]));
-  };
-  const draft = {
-    provider: {
-      list: () => [],
-      get: (id: string) => providers.get(id) as never,
-      update: (id: string, fn: (p: ProviderV2Info) => void) => {
-        const p = (providers.get(id) ?? { id }) as ProviderV2Info;
-        fn(p);
-        providers.set(id, p);
-      },
-      remove: () => {},
-    },
-    model: {
-      get: () => undefined,
-      update: (pid: string, mid: string, fn: (m: ModelV2Info) => void) => {
-        const k = pid + "/" + mid;
-        const m = (models.get(k) ?? { id: mid, providerID: pid }) as ModelV2Info;
-        fn(m);
-        models.set(k, m);
-      },
-      remove: () => {},
-      default: { get: () => undefined, set: () => {} },
-    },
-  } as CatalogDraft;
-  return {
-    models,
-    draft,
-    warns,
-    restore: () => {
-      console.warn = origWarn;
-    },
-  };
-}
-
-const baseOpts = {
-  providerId: "omniroute",
-  baseURL: "https://gw.example.com",
-  apiKey: "k",
-  timeoutMs: 1000,
-  modelCacheTtlMs: 300000,
-  usableOnly: false,
-};
-
-describe("catalog nested combo refs", () => {
-  it("resolves a forward combo-ref: parent LCD follows the nested combo, not zero", async () => {
-    const c = fakeDraft();
-    try {
-      const res = await publishCatalog(c.draft, baseOpts, {
-        fetcher: async () => [
-          {
-            id: "a",
-            context_length: 100000,
-            max_output_tokens: 4000,
-            capabilities: { tool_calling: true, reasoning: true },
-          },
-          {
-            id: "b",
-            context_length: 50000,
-            max_output_tokens: 2000,
-            capabilities: { tool_calling: true, reasoning: true },
-          },
-        ],
-        combosFetcher: async () => [
-          {
-            id: "parent",
-            name: "Parent",
-            models: [
-              { kind: "combo-ref", comboName: "Child" } as never,
-              { kind: "model", model: "a" } as never,
-            ],
-          },
-          {
-            id: "child",
-            name: "Child",
-            models: [{ kind: "model", model: "b" } as never],
-          },
-        ],
-      });
-      assert.deepEqual(res, { models: 2, combos: 2, autoCombos: 0 });
-      const parent = c.models.get("omniroute/parent");
-      assert.ok(parent);
-      assert.equal(parent?.limit.context, 50000);
-      assert.equal(parent?.limit.output, 2000);
-      assert.equal(parent?.capabilities.tools, true);
-    } finally {
-      c.restore();
-    }
-  });
-
-  it("drops an unresolvable combo-ref after the fixpoint instead of claiming empty LCD", async () => {
-    const c = fakeDraft();
-    try {
-      const res = await publishCatalog(c.draft, baseOpts, {
-        fetcher: async () => [{ id: "a" }],
-        combosFetcher: async () => [
-          {
-            id: "orphan",
-            name: "Orphan",
-            models: [{ kind: "combo-ref", comboName: "Missing" } as never],
-          },
-        ],
-      });
-      assert.deepEqual(res, { models: 1, combos: 0, autoCombos: 0 });
-      assert.ok(!c.models.has("omniroute/orphan"));
-      assert.ok(c.warns.some((w) => w.includes("could not resolve")));
-    } finally {
-      c.restore();
-    }
-  });
-
-  it("resolves a two-level chain within the pass budget", async () => {
-    const c = fakeDraft();
-    try {
-      await publishCatalog(c.draft, baseOpts, {
-        fetcher: async () => [
-          {
-            id: "a",
-            context_length: 80000,
-            max_output_tokens: 3000,
-            capabilities: { tool_calling: true },
-          },
-        ],
-        combosFetcher: async () => [
-          {
-            id: "top",
-            name: "Top",
-            models: [{ kind: "combo-ref", comboName: "Mid" } as never],
-          },
-          {
-            id: "mid",
-            name: "Mid",
-            models: [{ kind: "combo-ref", comboName: "Leaf" } as never],
-          },
-          {
-            id: "leaf",
-            name: "Leaf",
-            models: [{ kind: "model", model: "a" } as never],
-          },
-        ],
-      });
-      const top = c.models.get("omniroute/top");
-      assert.ok(top);
-      assert.equal(top?.limit.context, 80000);
-    } finally {
-      c.restore();
-    }
-  });
-});
-
-describe("catalog collision dedupe", () => {
-  it("warns once per key even across repeated publishes sharing collisionWarned", async () => {
-    const c = fakeDraft();
-    try {
-      const collisionWarned = new Set<string>();
-      const args = {
-        fetcher: async () => [{ id: "dupe" }],
-        combosFetcher: async () => [
-          { id: "dupe", name: "Dupe Combo", models: [{ kind: "model", model: "dupe" } as never] },
-        ],
-      };
-      await publishCatalog(c.draft, { ...baseOpts, collisionWarned }, args);
-      await publishCatalog(c.draft, { ...baseOpts, collisionWarned }, args);
-      const hits = c.warns.filter((w) => w.includes("collides with a model id"));
-      assert.equal(hits.length, 1);
-    } finally {
-      c.restore();
-    }
-  });
-
-  it("suppresses the warning for the intentional dedup pattern", async () => {
-    const c = fakeDraft();
-    try {
-      // v1 pattern: /v1/models pre-mirrors the combo under its friendly
-      // NAME and the combo is keyed by that name, so the overwritten model
-      // id equals the combo name exactly and the draft key collides.
-      const res = await publishCatalog(c.draft, baseOpts, {
-        fetcher: async () => [{ id: "omniroute/Mirror Combo" }],
-        combosFetcher: async () => [
-          {
-            id: "Mirror Combo",
-            name: "Mirror Combo",
-            models: [{ kind: "model", model: "omniroute/Mirror Combo" } as never],
-          },
-        ],
-      });
-      assert.deepEqual(res, { models: 1, combos: 1, autoCombos: 0 });
-      assert.ok(!c.warns.some((w) => w.includes("collides")));
-      assert.equal(c.models.get("omniroute/Mirror Combo")?.name, "Mirror Combo");
-    } finally {
-      c.restore();
-    }
-  });
-
-  it("shares the dedupe set when the caller passes collisionWarned", async () => {
-    const c = fakeDraft();
-    try {
-      const collisionWarned = new Set<string>();
-      const args = {
-        fetcher: async () => [{ id: "other" }],
-        combosFetcher: async () => [
-          {
-            id: "other",
-            name: "Other Combo",
-            models: [{ kind: "model", model: "other" } as never],
-          },
-        ],
-      };
-      await publishCatalog(c.draft, { ...baseOpts, collisionWarned }, args);
-      await publishCatalog(c.draft, { ...baseOpts, collisionWarned }, args);
-      const hits = c.warns.filter((w) => w.includes("collides with a model id"));
-      assert.equal(hits.length, 1);
-    } finally {
-      c.restore();
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/options.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/options.test.ts
@@ -1,129 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import {
-  integrationIdFor,
-  parsePluginOptions,
-  PLUGIN_ID,
-  providerIdFor,
-  resolveTimeouts,
-} from "../src/options.js";
-
-describe("parsePluginOptions", () => {
-  it("applies defaults for providerId, timeoutMs, usableOnly, enrichment", () => {
-    const opts = parsePluginOptions({ baseURL: "https://gw.example.com" });
-    assert.equal(opts.providerId, "omniroute");
-    assert.equal(opts.timeoutMs, 10000);
-    assert.equal(opts.usableOnly, false);
-    assert.equal(opts.enrichment, true);
-    assert.equal(opts.modelCacheTtlMs, undefined);
-  });
-  it("accepts a positive modelCacheTtlMs (in-memory TTL cache, default 300s)", () => {
-    const opts = parsePluginOptions({ baseURL: "https://gw.example.com", modelCacheTtlMs: 60000 });
-    assert.equal(opts.modelCacheTtlMs, 60000);
-  });
-  it("rejects a non-positive modelCacheTtlMs", () => {
-    assert.throws(() =>
-      parsePluginOptions({ baseURL: "https://gw.example.com", modelCacheTtlMs: 0 })
-    );
-  });
-  it("requires baseURL", () => {
-    assert.throws(() => parsePluginOptions({}), /baseURL/);
-  });
-  it("rejects unknown top-level keys (strict)", () => {
-    assert.throws(() => parsePluginOptions({ baseURL: "https://gw.example.com", bogus: 1 }));
-  });
-  it("rejects unknown apiFormat keys (strict)", () => {
-    assert.throws(() =>
-      parsePluginOptions({
-        baseURL: "https://gw.example.com",
-        apiFormat: { bogus: ["claude"] },
-      })
-    );
-  });
-  it("accepts deprecated anthropicPrefixes (warn at resolve time, not parse time)", () => {
-    const opts = parsePluginOptions({
-      baseURL: "https://gw.example.com",
-      apiFormat: { allowAnthropic: true, anthropicPrefixes: ["cc", "claude"] },
-    });
-    assert.deepEqual(opts.apiFormat, {
-      allowAnthropic: true,
-      anthropicPrefixes: ["cc", "claude"],
-    });
-  });
-  it("passes apiFormat allowlist through (shared enforces semantics)", () => {
-    const opts = parsePluginOptions({
-      baseURL: "https://gw.example.com",
-      apiFormat: { allowAnthropic: true, anthropicModels: ["anthropic/claude-x"] },
-    });
-    assert.deepEqual(opts.apiFormat, {
-      allowAnthropic: true,
-      anthropicModels: ["anthropic/claude-x"],
-    });
-  });
-});
-
-describe("identity table", () => {
-  it("maps providerId X to provider X and integration X, under one fixed plugin id", () => {
-    assert.equal(providerIdFor("omniroute"), "omniroute");
-    assert.equal(integrationIdFor("omniroute"), "omniroute");
-    assert.equal(providerIdFor("second-gateway"), "second-gateway");
-    // The host reads the plugin id before any option exists, so it never
-    // varies with providerId.
-    assert.equal(PLUGIN_ID, "omniroute-v2");
-  });
-});
-
-describe("invalid options say what to fix", () => {
-  it("names an unknown key instead of dumping the validator output", () => {
-    assert.throws(
-      () => parsePluginOptions({ baseURL: "http://gw.example.com", modelCacheTtl: 300000 }),
-      (err: Error) => {
-        assert.match(err.message, /invalid plugin options/);
-        assert.match(err.message, /unknown option "modelCacheTtl"/);
-        return true;
-      }
-    );
-  });
-
-  it("names the offending field for a wrong type", () => {
-    assert.throws(
-      () => parsePluginOptions({ baseURL: 42 }),
-      (err: Error) => {
-        assert.match(err.message, /baseURL/);
-        return true;
-      }
-    );
-  });
-
-  it("accepts the documented option names", () => {
-    const parsed = parsePluginOptions({
-      baseURL: "http://gw.example.com",
-      modelCacheTtlMs: 300000,
-      timeouts: { models: 15000, combos: 8000 },
-      geminiSanitization: false,
-    });
-    assert.equal(parsed.modelCacheTtlMs, 300000);
-    assert.equal(resolveTimeouts(parsed).models, 15000);
-  });
-});
-
-describe("providerId is bounded because it reaches a filesystem path", () => {
-  it("rejects a traversal attempt instead of writing outside the snapshot directory", () => {
-    for (const bad of ["../../etc/cron.d/x", "a/b", "..", "."]) {
-      assert.throws(
-        () => parsePluginOptions({ baseURL: "https://gw.example.com", providerId: bad }),
-        /invalid plugin options/,
-        `providerId ${JSON.stringify(bad)} must be rejected`
-      );
-    }
-  });
-
-  it("keeps the ids a user would actually pick", () => {
-    for (const ok of ["omniroute", "omniroute-2", "gw.staging", "gw_prod"]) {
-      assert.equal(
-        parsePluginOptions({ baseURL: "https://gw.example.com", providerId: ok }).providerId,
-        ok
-      );
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/parity.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/parity.test.ts
@@ -1,224 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { createReadStream } from "node:fs";
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-import { publishCatalog } from "../src/catalog.js";
-import {
-  mapComboToModelV2 as sharedMapCombo,
-  mapRawModelToModelV2 as sharedMapModel,
-  type OmniRouteCombosFetcher,
-  type OmniRouteModelsFetcher,
-  type OmniRouteRawCombo,
-  type OmniRouteRawModelEntry,
-} from "../src/shared/index.js";
-
-interface Fixture {
-  models: OmniRouteRawModelEntry[];
-  combos: OmniRouteRawCombo[];
-}
-
-/**
- * What the v1 plugin produces for the same fixture, recorded in
- * `fixtures/v1-parity.json`. Running v1 here instead would mean importing its
- * build output from a sibling package: it only exists on a machine that has
- * built v1, so the check silently passed locally and could not run in CI at
- * all. Recording it makes the claim reviewable in the diff and reproducible
- * anywhere.
- */
-interface V1Parity {
-  v1PluginVersion: string;
-  hookId: string;
-  publishedKeys: string[];
-  comboSlugs: string[];
-  mappedModels: Record<string, unknown>;
-  mappedCombos: Record<string, unknown>;
-}
-
-async function loadFixture(): Promise<Fixture> {
-  const chunks: Buffer[] = [];
-  for await (const chunk of createReadStream(new URL("./fixtures/catalog.json", import.meta.url))) {
-    chunks.push(chunk as Buffer);
-  }
-  return JSON.parse(Buffer.concat(chunks).toString("utf8")) as Fixture;
-}
-
-async function loadV1Parity(): Promise<V1Parity> {
-  const chunks: Buffer[] = [];
-  for await (const chunk of createReadStream(
-    new URL("./fixtures/v1-parity.json", import.meta.url)
-  )) {
-    chunks.push(chunk as Buffer);
-  }
-  return JSON.parse(Buffer.concat(chunks).toString("utf8")) as V1Parity;
-}
-
-type ApiAuth = { type: "api"; key: string };
-
-function fakeDraft() {
-  const providers = new Map<string, ProviderV2Info>();
-  const models = new Map<string, ModelV2Info>();
-  return {
-    providers,
-    models,
-    provider: {
-      list: () => [],
-      get: (id: string) => providers.get(id) as never,
-      update: (id: string, fn: (p: ProviderV2Info) => void) => {
-        const p = (providers.get(id) ?? { id }) as ProviderV2Info;
-        fn(p);
-        providers.set(id, p);
-      },
-      remove: () => {},
-    },
-    model: {
-      get: () => undefined,
-      update: (pid: string, mid: string, fn: (m: ModelV2Info) => void) => {
-        const k = pid + "/" + mid;
-        const m = (models.get(k) ?? { id: mid, providerID: pid }) as ModelV2Info;
-        fn(m);
-        models.set(k, m);
-      },
-      remove: () => {},
-      default: { get: () => undefined, set: () => {} },
-    },
-  };
-}
-
-const TEST_OPTS = {
-  baseURL: "https://gw.example.com",
-  providerId: "omniroute",
-  apiKey: "parity-key",
-  timeoutMs: 1000,
-  modelCacheTtlMs: 300000,
-  usableOnly: false as const,
-};
-
-describe("v1-vs-v2 catalog parity", () => {
-  it("same fixture models publish the same key set modulo documented exclusions", async () => {
-    const fixture = await loadFixture();
-    assert.equal(fixture.models.length, 5);
-    assert.equal(fixture.combos.length, 2);
-
-    const fetcher: OmniRouteModelsFetcher = async () => fixture.models;
-    const combosFetcher: OmniRouteCombosFetcher = async () => fixture.combos;
-
-    const recorded = await loadV1Parity();
-    assert.equal(recorded.hookId, "opencode-omniroute", "v1 published under its own provider id");
-
-    const stripX = (k: string): string =>
-      k.startsWith("omniroute/") ? k.slice("omniroute/".length) : k;
-    const v1Keys = recorded.publishedKeys;
-
-    const draft = fakeDraft();
-    const counts = await publishCatalog(draft, TEST_OPTS, { fetcher, combosFetcher });
-    assert.equal(counts.models, 5);
-    assert.equal(counts.combos, 2);
-    assert.equal(counts.autoCombos, 0);
-
-    // Final converted ModelV2Info shape (legacy→info boundary in
-    // src/catalog.ts assignModelFields): api resolves to the
-    // openai-compatible AISDK block, capabilities fold tool_calling into
-    // tools, cost is zeroed (pricing lives server-side).
-    const mAlpha = draft.models.get("omniroute/m-alpha");
-    assert.ok(mAlpha, "m-alpha published in v2 draft");
-    assert.equal(mAlpha.api.type, "aisdk");
-    if (mAlpha.api.type !== "aisdk") throw new Error("m-alpha api must be aisdk");
-    assert.equal(mAlpha.api.package, "@ai-sdk/openai-compatible");
-    assert.equal(mAlpha.capabilities.tools, true);
-    assert.equal(mAlpha.cost[0].input, 0);
-
-    // Measured key shapes: v1 namespaces the friendly name
-    // (`Combo Fast` -> `omniroute/combo-fast`) and keys the colliding combo
-    // by its raw id (`good-combo` -> `omniroute/good-combo`, combo wins);
-    // v2 publishes combo ids verbatim (`omniroute/combo-fast`,
-    // `omniroute/good-combo` overwriting the raw model). v1 and v2 therefore
-    // publish the SAME model keys (modulo the slashed-id exclusion above)
-    // and the SAME non-colliding combo key; the colliding `good-combo` combo
-    // overwrites the same-named raw model on BOTH sides (warn asserted below).
-    const v1Combo = v1Keys.filter((k) => k.startsWith("combo-") || k === "good-combo").sort();
-    const v2Combo = [...draft.models.keys()]
-      .map(stripX)
-      .filter((k) => k.startsWith("combo-") || k === "good-combo")
-      .sort();
-    // slug("Combo Fast") = "combo-fast", slug("Good Combo") = "good-combo".
-    assert.deepEqual(v2Combo, ["combo-fast", "good-combo"]);
-    assert.deepEqual(v1Combo, ["combo-fast", "good-combo"]);
-    assert.deepEqual(v2Combo, v1Combo);
-    assert.deepEqual(recorded.comboSlugs, ["combo-fast", "good-combo"]);
-    const v1Models = v1Keys.filter((k) => !k.startsWith("combo-") && k !== "good-combo").sort();
-    // Anchor v1 model keys as literals (modulo the slashed-id exclusion
-    // documented above: the bare `cc/m-gamma` key fails the
-    // `startsWith("omniroute/")` filter and is dropped here).
-    assert.deepEqual(v1Models, ["local-delta", "m-alpha", "m-beta"]);
-    const v2Models = [...draft.models.keys()]
-      .map(stripX)
-      .filter((k) => !k.includes("/") && !k.startsWith("combo-") && k !== "good-combo")
-      .sort();
-    assert.deepEqual(v2Models, v1Models);
-
-    // Collision: the combo whose friendly name collides with a raw model id
-    // wins on both sides (v1 warns via injected logger, v2 via console.warn).
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const draft2 = fakeDraft();
-      await publishCatalog(draft2, TEST_OPTS, { fetcher, combosFetcher });
-      assert.ok(draft2.models.has("omniroute/good-combo"), "colliding combo key wins in v2");
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.ok(warns.some((w) => w.includes("collides with a model id; combo wins")));
-
-    // Mapper-level parity for the slashed-id exclusion: v1 and shared mappers
-    // must produce identical ModelV2 payloads for every fixture entry.
-    // (No apiFormat on either side — the parity scope is key publication,
-    // not the v1-prefix vs v2-allowlist routing rule covered by A2 tests.
-    // The `cc/m-gamma` fixture entry is therefore EXCLUDED from the mapper
-    // comparison: v1 routes it to anthropic via its default
-    // anthropicPrefixes while shared/v2 leave it openai-compatible without
-    // an explicit anthropicModels allowlist.)
-    // Mapper parity for the slashed-id exclusion: shared must still produce
-    // the payloads v1 produced. (`cc/m-gamma` is excluded: v1 routes it to
-    // anthropic through its default prefix list, while v2 leaves it
-    // openai-compatible without an explicit allowlist — covered by the
-    // apiFormat tests.)
-    for (const entry of fixture.models.filter((m) => m.id !== "cc/m-gamma")) {
-      const expected = recorded.mappedModels[entry.id];
-      assert.ok(expected, `v1 output recorded for ${entry.id}`);
-      const viaShared = sharedMapModel(entry, {
-        providerId: "omniroute",
-        baseURL: TEST_OPTS.baseURL,
-      });
-      assert.deepEqual(
-        JSON.parse(JSON.stringify({ ...viaShared, providerID: undefined })),
-        JSON.parse(JSON.stringify({ ...(expected as object), providerID: undefined })),
-        `model mapper parity for ${entry.id}`
-      );
-    }
-    const byId = new Map(fixture.models.map((m) => [m.id, m]));
-    for (const combo of fixture.combos) {
-      const members = (combo.models ?? [])
-        .filter((s) => s?.kind !== "combo-ref" && typeof s?.model === "string")
-        .map((s) => byId.get(s.model as string))
-        .filter((m): m is OmniRouteRawModelEntry => m !== undefined);
-      assert.deepEqual(
-        JSON.parse(JSON.stringify(sharedMapCombo(combo, members, "omniroute", TEST_OPTS.baseURL))),
-        recorded.mappedCombos[combo.id],
-        `combo mapper parity for ${combo.id}`
-      );
-    }
-  });
-
-  it("fixture is rejected when it drifts from the 5+2 shape (guard against silent shrink)", async () => {
-    const fixture = await loadFixture();
-    assert.ok(fixture.models.length >= 5, "fixture must keep at least 5 models");
-    assert.ok(fixture.combos.length >= 2, "fixture must keep at least 2 combos");
-  });
-});
-
-void (0 as unknown as ApiAuth);
-void (0 as unknown as CatalogDraft);
--- a/@omniroute/opencode-plugin-v2/tests/publish-guard.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/publish-guard.test.ts
@@ -1,137 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { mkdtempSync } from "node:fs";
-import { tmpdir } from "node:os";
-import { join } from "node:path";
-import plugin from "../src/index.js";
-
-// Guard around `publishCatalog` in the catalog transform: fetcher-level
-// fail-open covers fetch rejections, but a mapper throw or a host throw in
-// `draft.update` would reject the transform callback (unhandled rejection).
-// The guard must warn + resolve instead.
-describe("plugin-v2 publish guard (mapper/draft throws)", () => {
-  function isolateDisk(): () => void {
-    const dir = mkdtempSync(join(tmpdir(), "omniroute-guard-"));
-    const prev = process.env.OPENCODE_DATA_DIR;
-    process.env.OPENCODE_DATA_DIR = dir;
-    return () => {
-      if (prev === undefined) delete process.env.OPENCODE_DATA_DIR;
-      else process.env.OPENCODE_DATA_DIR = prev;
-    };
-  }
-  function setupCtx(): {
-    catalogCallbacks: Array<(draft: unknown) => Promise<void>>;
-    ctx: Record<string, unknown>;
-  } {
-    const catalogCallbacks: Array<(draft: unknown) => Promise<void>> = [];
-    const ctx = {
-      options: { baseURL: "https://gw.example.com", providerId: "omniroute", apiKey: "k" },
-      catalog: {
-        transform: (cb: (draft: unknown) => Promise<void>) => {
-          catalogCallbacks.push(cb);
-          return Promise.resolve({ dispose: async () => {} });
-        },
-      },
-      integration: {
-        transform: () => Promise.resolve({ dispose: async () => {} }),
-      },
-    };
-    return { catalogCallbacks, ctx };
-  }
-
-  function stubFetch(): typeof fetch {
-    return (async (url: unknown) => {
-      const href = String(url);
-      if (href.includes("/api/combos/auto")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      if (href.includes("/api/combos")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      return {
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => ({ data: [{ id: "m1" }] }),
-      };
-    }) as typeof fetch;
-  }
-
-  async function silenceConsole<T>(fn: () => Promise<T>): Promise<{ result: T; warns: string[] }> {
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    const origLog = console.log;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    console.log = () => {};
-    try {
-      const result = await fn();
-      return { result, warns };
-    } finally {
-      console.warn = origWarn;
-      console.log = origLog;
-    }
-  }
-
-  it("host throw in draft.model.update: callback resolves + warn, never rejects", async () => {
-    const restoreDisk = isolateDisk();
-    const { catalogCallbacks, ctx } = setupCtx();
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch();
-    try {
-      const { warns } = await silenceConsole(async () => {
-        await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-        assert.equal(catalogCallbacks.length, 1);
-        const draft = {
-          provider: { update: (_id: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-          model: {
-            update: () => {
-              throw new Error("host boom");
-            },
-          },
-        };
-        // MUST resolve — without the guard this rejects with "host boom".
-        await catalogCallbacks[0](draft);
-      });
-      assert.ok(
-        warns.some((w) => w.includes("catalog publish failed") && w.includes("host boom")),
-        `expected a publish-guard warn, got: ${JSON.stringify(warns)}`
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-
-  it("host throw in draft.provider.update: callback resolves + warn, never rejects", async () => {
-    const restoreDisk = isolateDisk();
-    const { catalogCallbacks, ctx } = setupCtx();
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch();
-    try {
-      const { warns } = await silenceConsole(async () => {
-        await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-        const draft = {
-          provider: {
-            update: () => {
-              throw new Error("provider host boom");
-            },
-          },
-          model: {
-            update: (_pid: string, _mid: string, fn: (m: Record<string, unknown>) => void) =>
-              fn({}),
-          },
-        };
-        await catalogCallbacks[0](draft);
-      });
-      assert.ok(
-        warns.some((w) => w.includes("catalog publish failed")),
-        `expected a publish-guard warn, got: ${JSON.stringify(warns)}`
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/refresh-failopen.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/refresh-failopen.test.ts
@@ -1,214 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import plugin from "../src/index.js";
-
-// RED: reproduces the PROD unhandled rejection — combos 403 must not escape
-// the catalog transform. Today `loadSnapshot()` awaits
-// `Promise.all([models, combos])` with no catch, so a 403 combos fetch
-// rejects the snapshot promise and the rejection propagates out of the
-// `ctx.catalog.transform` callback (fail-open in `publishCatalog` is
-// bypassed because injected fetchers return the already-rejected data).
-describe("plugin-v2 fail-open refresh (PROD 403 combos)", () => {
-  let diskSeq = 0;
-  async function isolateDisk(): Promise<() => void> {
-    const { mkdtempSync } = await import("node:fs");
-    const { tmpdir } = await import("node:os");
-    const { join } = await import("node:path");
-    diskSeq += 1;
-    const dir = mkdtempSync(join(tmpdir(), `omniroute-fo-${diskSeq}-`));
-    const prev = process.env.OPENCODE_DATA_DIR;
-    process.env.OPENCODE_DATA_DIR = dir;
-    return () => {
-      if (prev === undefined) delete process.env.OPENCODE_DATA_DIR;
-      else process.env.OPENCODE_DATA_DIR = prev;
-    };
-  }
-  function setupCtx(opts: {
-    combosStatus: number;
-    modelsStatus?: number;
-    reloads: { count: number };
-  }): {
-    catalogCallbacks: Array<(draft: unknown) => Promise<void>>;
-    ctx: Record<string, unknown>;
-  } {
-    const catalogCallbacks: Array<(draft: unknown) => Promise<void>> = [];
-    const ctx = {
-      options: {
-        baseURL: "https://gw.example.com",
-        providerId: "fo-" + String(opts.combosStatus) + "-" + String(opts.modelsStatus ?? 200),
-        apiKey: "k-fo-" + String(opts.combosStatus),
-      },
-      catalog: {
-        transform: (cb: (draft: unknown) => Promise<void>) => {
-          catalogCallbacks.push(cb);
-          return Promise.resolve({ dispose: async () => {} });
-        },
-        reload: async () => {
-          opts.reloads.count += 1;
-        },
-      },
-      integration: {
-        transform: () => Promise.resolve({ dispose: async () => {} }),
-      },
-    };
-    return { catalogCallbacks, ctx };
-  }
-
-  function stubFetch(opts: { combosStatus: number; modelsStatus?: number }): typeof fetch {
-    const modelsStatus = opts.modelsStatus ?? 200;
-    return (async (url: unknown) => {
-      const href = String(url);
-      if (href.includes("/api/combos/auto")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      if (href.includes("/api/combos")) {
-        return {
-          ok: opts.combosStatus === 200,
-          status: opts.combosStatus,
-          statusText: opts.combosStatus === 403 ? "Forbidden" : "Error",
-          json: async () => ({ combos: [] }),
-        };
-      }
-      return {
-        ok: modelsStatus === 200,
-        status: modelsStatus,
-        statusText: "OK",
-        json: async () => ({ data: [{ id: "m1" }] }),
-      };
-    }) as typeof fetch;
-  }
-
-  function stubDraft(): {
-    draft: unknown;
-    published: Map<string, Record<string, unknown>>;
-  } {
-    const published = new Map<string, Record<string, unknown>>();
-    const draft = {
-      provider: { update: (_id: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-      model: {
-        update: (pid: string, mid: string, fn: (m: Record<string, unknown>) => void) => {
-          const entry: Record<string, unknown> = { id: mid, providerID: pid };
-          fn(entry);
-          published.set(pid + "/" + mid, entry);
-        },
-      },
-    };
-    return { draft, published };
-  }
-
-  async function silenceConsole<T>(fn: () => Promise<T>): Promise<{ result: T; warns: string[] }> {
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    const origLog = console.log;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    console.log = () => {};
-    try {
-      const result = await fn();
-      return { result, warns };
-    } finally {
-      console.warn = origWarn;
-      console.log = origLog;
-    }
-  }
-
-  it("combos 403: catalog callback resolves (models-only + warn), never rejects", async () => {
-    const restoreDisk = await isolateDisk();
-    const reloads = { count: 0 };
-    const { catalogCallbacks, ctx } = setupCtx({ combosStatus: 403, reloads });
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch({ combosStatus: 403 });
-    try {
-      const { warns } = await silenceConsole(async () => {
-        await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-        assert.equal(catalogCallbacks.length, 1);
-        const { draft, published } = stubDraft();
-        // MUST resolve — today it rejects with the 403 error.
-        await catalogCallbacks[0](draft);
-        const key = [...published.keys()].find((k) => k.endsWith("/m1"));
-        assert.ok(
-          key,
-          `models-only fallback must publish m1, got: ${JSON.stringify([...published.keys()])}`
-        );
-      });
-      assert.ok(
-        warns.some((w) => w.includes("combos") && w.includes("403")),
-        `expected a combos 403 warn, got: ${JSON.stringify(warns)}`
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-
-  it("combos 500: catalog callback resolves (models-only + warn), never rejects", async () => {
-    const restoreDisk = await isolateDisk();
-    const reloads = { count: 0 };
-    const { catalogCallbacks, ctx } = setupCtx({ combosStatus: 500, reloads });
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch({ combosStatus: 500 });
-    try {
-      const { warns } = await silenceConsole(async () => {
-        await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-        const { draft, published } = stubDraft();
-        await catalogCallbacks[0](draft);
-        const key = [...published.keys()].find((k) => k.endsWith("/m1"));
-        assert.ok(
-          key,
-          `models-only fallback must publish m1, got: ${JSON.stringify([...published.keys()])}`
-        );
-      });
-      assert.ok(
-        warns.some((w) => w.includes("combos")),
-        `expected a combos warn, got: ${JSON.stringify(warns)}`
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-
-  it("combos timeout (abort): catalog callback resolves, never rejects", async () => {
-    const restoreDisk = await isolateDisk();
-    const reloads = { count: 0 };
-    const { catalogCallbacks, ctx } = setupCtx({ combosStatus: 200, reloads });
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      if (href.includes("/api/combos/auto")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      if (href.includes("/api/combos")) {
-        const err = new Error("This operation was aborted");
-        err.name = "AbortError";
-        throw err;
-      }
-      return {
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => ({ data: [{ id: "m1" }] }),
-      };
-    }) as typeof fetch;
-    try {
-      const { warns } = await silenceConsole(async () => {
-        await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-        const { draft, published } = stubDraft();
-        await catalogCallbacks[0](draft);
-        const key = [...published.keys()].find((k) => k.endsWith("/m1"));
-        assert.ok(
-          key,
-          `models-only fallback must publish m1, got: ${JSON.stringify([...published.keys()])}`
-        );
-      });
-      assert.ok(
-        warns.some((w) => w.includes("combos")),
-        `expected a combos warn, got: ${JSON.stringify(warns)}`
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-anthropic-prefixes.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-anthropic-prefixes.test.ts
@@ -1,165 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { mapRawModelToModelV2, resolveApiBlockV2 } from "../src/shared/models-map.js";
-import { parsePluginOptions } from "../src/options.js";
-import { publishCatalog } from "../src/catalog.js";
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-
-const GW = "https://gw.example.com";
-const PREFIXES = ["cc", "claude", "anthropic", "kiro", "kr"];
-
-describe("deprecated anthropicPrefixes", () => {
-  it("routes a default prefix id to anthropic with a deprecation warning", async () => {
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const out = resolveApiBlockV2("cc/claude-x", GW, {
-        allowAnthropic: true,
-        anthropicPrefixes: [...PREFIXES],
-      });
-      assert.equal(out.id, "anthropic");
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.ok(warns.some((w) => w.includes("deprecated") && w.includes("anthropicModels")));
-  });
-
-  it("warns only once per process for the same prefix list", async () => {
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      resolveApiBlockV2("cc/a", GW, { allowAnthropic: true, anthropicPrefixes: ["cc"] });
-      resolveApiBlockV2("cc/b", GW, { allowAnthropic: true, anthropicPrefixes: ["cc"] });
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.equal(warns.filter((w) => w.includes("deprecated")).length, 1);
-  });
-
-  it("documents the known claude/openai-compatible-model edge: prefix wins", async () => {
-    // v1 parity keeps prefix matching, so a model literally named
-    // `claude/openai-compatible-model` routes to anthropic when its prefix
-    // is listed. Documented (not silently fixed): operators hitting this
-    // rare name should move the id to the allowlist instead.
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const out = resolveApiBlockV2("claude/openai-compatible-model", GW, {
-        allowAnthropic: true,
-        anthropicPrefixes: [...PREFIXES],
-        anthropicModels: ["anthropic/claude-x"],
-      });
-      assert.equal(out.id, "anthropic");
-      const mapped = mapRawModelToModelV2(
-        { id: "claude/openai-compatible-model" },
-        {
-          providerId: "omniroute",
-          baseURL: GW,
-          apiFormat: {
-            allowAnthropic: true,
-            anthropicPrefixes: [...PREFIXES],
-            anthropicModels: ["anthropic/claude-x"],
-          },
-        }
-      );
-      assert.equal(mapped.api.id, "anthropic");
-    } finally {
-      console.warn = origWarn;
-    }
-  });
-
-  it("prefix OR allowlist routes to anthropic; neither means openai-compatible", async () => {
-    const origWarn = console.warn;
-    console.warn = () => {};
-    try {
-      assert.equal(
-        resolveApiBlockV2("kiro/m", GW, { allowAnthropic: true, anthropicPrefixes: ["kiro"] }).id,
-        "anthropic"
-      );
-      assert.equal(
-        resolveApiBlockV2("anthropic/claude-x", GW, {
-          allowAnthropic: true,
-          anthropicModels: ["anthropic/claude-x"],
-        }).id,
-        "anthropic"
-      );
-      assert.equal(resolveApiBlockV2("gpt-x", GW).id, "openai-compatible");
-    } finally {
-      console.warn = origWarn;
-    }
-  });
-
-  it("copied v1 config routes anthropic and warns deprecation through publishCatalog", async () => {
-    const providers = new Map<string, ProviderV2Info>();
-    const models = new Map<string, ModelV2Info>();
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const parsed = parsePluginOptions({
-        baseURL: GW,
-        apiFormat: { allowAnthropic: true, anthropicPrefixes: [...PREFIXES, "v1copy"] },
-      });
-      const draft = {
-        provider: {
-          list: () => [],
-          get: (id: string) => providers.get(id) as never,
-          update: (id: string, fn: (p: ProviderV2Info) => void) => {
-            const p = (providers.get(id) ?? { id }) as ProviderV2Info;
-            fn(p);
-            providers.set(id, p);
-          },
-          remove: () => {},
-        },
-        model: {
-          get: () => undefined,
-          update: (pid: string, mid: string, fn: (m: ModelV2Info) => void) => {
-            const k = pid + "/" + mid;
-            const m = (models.get(k) ?? { id: mid, providerID: pid }) as ModelV2Info;
-            fn(m);
-            models.set(k, m);
-          },
-          remove: () => {},
-          default: { get: () => undefined, set: () => {} },
-        },
-      } as CatalogDraft;
-      const res = await publishCatalog(
-        draft,
-        {
-          providerId: "omniroute",
-          baseURL: GW,
-          apiKey: "k",
-          timeoutMs: 1000,
-          modelCacheTtlMs: 300000,
-          usableOnly: false,
-          apiFormat: parsed.apiFormat,
-        },
-        {
-          fetcher: async () => [{ id: "cc/claude-x" }],
-          combosFetcher: async () => [],
-          enrichmentFetcher: async () => new Map(),
-        }
-      );
-      assert.deepEqual(res, { models: 1, combos: 0, autoCombos: 0 });
-      const m = models.get("omniroute/cc/claude-x");
-      assert.ok(m);
-      if (m?.api.type !== "aisdk") throw new Error("model api must be aisdk");
-      assert.equal(m?.api.id, "anthropic");
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.ok(warns.some((w) => w.includes("deprecated") && w.includes("anthropicModels")));
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-auto-combos.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-auto-combos.test.ts
@@ -1,196 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import {
-  defaultOmniRouteAutoCombosFetcher,
-  mapAutoComboToModelV2,
-} from "../src/shared/auto-combos.js";
-
-// A fetch stub that hangs until the caller aborts: proves the AbortController
-// wiring fires. Rejects with an AbortError like undici does on abort.
-function hangingFetch(): typeof fetch {
-  return ((url: unknown, init?: { signal?: AbortSignal }) => {
-    return new Promise((_resolve, reject) => {
-      const signal = init?.signal;
-      if (signal?.aborted) {
-        reject(Object.assign(new Error("This operation was aborted"), { name: "AbortError" }));
-        return;
-      }
-      signal?.addEventListener("abort", () => {
-        reject(Object.assign(new Error("This operation was aborted"), { name: "AbortError" }));
-      });
-    }) as Promise<Response>;
-  }) as unknown as typeof fetch;
-}
-
-function stubFetch(res: { status: number; statusText?: string; body?: unknown }): typeof fetch {
-  return (async () => ({
-    ok: res.status >= 200 && res.status < 300,
-    status: res.status,
-    statusText: res.statusText ?? (res.status === 404 ? "Not Found" : "Error"),
-    json: async () => res.body,
-  })) as unknown as typeof fetch;
-}
-
-function silenceWarn(): { warns: string[]; restore: () => void } {
-  const warns: string[] = [];
-  const origWarn = console.warn;
-  console.warn = (...args: unknown[]) => {
-    warns.push(String(args[0]));
-  };
-  return {
-    warns,
-    restore() {
-      console.warn = origWarn;
-    },
-  };
-}
-
-describe("defaultOmniRouteAutoCombosFetcher", () => {
-  it("404 returns [] with a warn (old gateway without the endpoint stays fail-open)", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch({ status: 404 });
-    const guard = silenceWarn();
-    try {
-      const res = await defaultOmniRouteAutoCombosFetcher("https://gw.example.com", "k", 5000);
-      assert.deepEqual(res, []);
-    } finally {
-      globalThis.fetch = origFetch;
-      guard.restore();
-    }
-    assert.ok(
-      guard.warns.some((w) => w.includes("/api/combos/auto") && w.includes("404")),
-      `expected a 404 warn, got: ${JSON.stringify(guard.warns)}`
-    );
-  });
-
-  it("500 throws (with warn + token-hint report) so the caller keeps last-known", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch({ status: 500, statusText: "Internal Server Error" });
-    const guard = silenceWarn();
-    const reported: string[] = [];
-    try {
-      await assert.rejects(
-        defaultOmniRouteAutoCombosFetcher("https://gw.example.com", "k", 5000, undefined, (e, r) =>
-          reported.push(`${e} ${r}`)
-        ),
-        /HTTP 500/
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      guard.restore();
-    }
-    assert.ok(
-      guard.warns.some((w) => w.includes("/api/combos/auto") && w.includes("500")),
-      `expected a 500 warn, got: ${JSON.stringify(guard.warns)}`
-    );
-    assert.ok(
-      reported.some((r) => r.includes("/api/combos/auto")),
-      `expected a per-endpoint report, got: ${JSON.stringify(reported)}`
-    );
-  });
-
-  it("a network throw rejects instead of resolving empty", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = hangingFetch();
-    const guard = silenceWarn();
-    try {
-      await assert.rejects(
-        defaultOmniRouteAutoCombosFetcher("https://gw.example.com", "k", 20),
-        /abort|aborted/i
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      guard.restore();
-    }
-    assert.ok(
-      guard.warns.some((w) => w.includes("/api/combos/auto")),
-      `expected an abort warn, got: ${JSON.stringify(guard.warns)}`
-    );
-  });
-
-  it("accepts the {combos:[...]} envelope and filters entries without a string id", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch({
-      status: 200,
-      body: { combos: [{ id: "auto/coding", name: "Auto Coding" }, { name: "no-id" }] },
-    });
-    const guard = silenceWarn();
-    try {
-      const res = await defaultOmniRouteAutoCombosFetcher("https://gw.example.com", "k", 5000);
-      assert.deepEqual(res, [{ id: "auto/coding", name: "Auto Coding" }]);
-    } finally {
-      globalThis.fetch = origFetch;
-      guard.restore();
-    }
-  });
-});
-
-describe("mapAutoComboToModelV2", () => {
-  it("falls back to 128k context / 8k output when the server omits limits (never 0)", () => {
-    const m = mapAutoComboToModelV2(
-      { id: "auto/coding", name: "Auto Coding", variant: "coding" },
-      "omniroute",
-      "https://gw.example.com"
-    );
-    assert.equal(m.id, "auto/coding");
-    assert.equal(m.limit.context, 128_000);
-    assert.equal(m.limit.output, 8_192);
-    assert.equal(m.capabilities.toolcall, true);
-    assert.equal(m.capabilities.reasoning, true);
-  });
-
-  it("uses server limits when positive, falls back when zero or negative", () => {
-    const served = mapAutoComboToModelV2(
-      { id: "auto", name: "Auto", context_length: 200000, max_output_tokens: 16000 },
-      "omniroute",
-      "https://gw.example.com"
-    );
-    assert.equal(served.limit.context, 200000);
-    assert.equal(served.limit.output, 16000);
-    const zeroed = mapAutoComboToModelV2(
-      { id: "auto", name: "Auto", context_length: 0, max_output_tokens: -1 },
-      "omniroute",
-      "https://gw.example.com"
-    );
-    assert.equal(zeroed.limit.context, 128_000);
-    assert.equal(zeroed.limit.output, 8_192);
-  });
-
-  it("defaults the model id to the auto variant key (auto, auto/coding)", () => {
-    const def = mapAutoComboToModelV2({ id: "whatever" }, "omniroute", "https://gw.example.com");
-    assert.equal(def.id, "auto");
-    const coding = mapAutoComboToModelV2(
-      { id: "auto/coding", variant: "coding" },
-      "omniroute",
-      "https://gw.example.com"
-    );
-    assert.equal(coding.id, "auto/coding");
-  });
-
-  it("stamps the openai-compatible api block by default", () => {
-    const m = mapAutoComboToModelV2(
-      { id: "auto", candidateCount: 3 },
-      "omniroute",
-      "https://gw.example.com"
-    );
-    assert.deepEqual(m.api, {
-      id: "openai-compatible",
-      url: "https://gw.example.com/v1",
-      npm: "@ai-sdk/openai-compatible",
-    });
-  });
-
-  it("routes an allowlisted auto id to anthropic (same rule as models)", () => {
-    const m = mapAutoComboToModelV2(
-      { id: "auto/coding", variant: "coding" },
-      "omniroute",
-      "https://gw.example.com",
-      { allowAnthropic: true, anthropicModels: ["auto/coding"] }
-    );
-    assert.deepEqual(m.api, {
-      id: "anthropic",
-      url: "https://gw.example.com",
-      npm: "@ai-sdk/anthropic",
-    });
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-combos-map.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-combos-map.test.ts
@@ -1,77 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { mapComboToModelV2 } from "../src/shared/combos-map.js";
-
-const member = (id: string, caps = {}) => ({
-  id,
-  context_length: 100000,
-  max_output_tokens: 2000,
-  capabilities: caps,
-});
-
-describe("mapComboToModelV2", () => {
-  it("rolls up LCD: min context/output, AND of toolcall", () => {
-    const m = mapComboToModelV2(
-      {
-        id: "combo-a",
-        name: "A",
-        models: [
-          { kind: "model", model: "a" },
-          { kind: "model", model: "b" },
-        ],
-      },
-      [
-        {
-          ...member("a"),
-          context_length: 100000,
-          max_output_tokens: 2000,
-          capabilities: { tool_calling: true },
-        },
-        {
-          ...member("b"),
-          context_length: 50000,
-          max_output_tokens: 1000,
-          capabilities: { tool_calling: false },
-        },
-      ],
-      "omniroute",
-      "https://gw.example.com"
-    );
-    assert.equal(m.limit.context, 50000);
-    assert.equal(m.limit.output, 1000);
-    assert.equal(m.capabilities.toolcall, false);
-    assert.equal(m.api.id, "openai-compatible");
-  });
-  it("empty members short-circuit to all-false capabilities", () => {
-    const m = mapComboToModelV2({ id: "combo-empty" }, [], "omniroute", "https://gw.example.com");
-    assert.equal(m.capabilities.toolcall, false);
-    assert.equal(m.capabilities.reasoning, false);
-  });
-  it("stamps the api block via resolveApiBlockV2 (openai-compatible by default)", () => {
-    const m = mapComboToModelV2(
-      { id: "combo-a", models: [{ kind: "model", model: "a" }] },
-      [{ ...member("a") }],
-      "omniroute",
-      "https://gw.example.com"
-    );
-    assert.deepEqual(m.api, {
-      id: "openai-compatible",
-      url: "https://gw.example.com/v1",
-      npm: "@ai-sdk/openai-compatible",
-    });
-  });
-  it("routes an allowlisted combo id to anthropic even with mixed members", () => {
-    const m = mapComboToModelV2(
-      { id: "combo-a", models: [{ kind: "model", model: "gpt-x" }] },
-      [{ ...member("gpt-x") }],
-      "omniroute",
-      "https://gw.example.com",
-      { allowAnthropic: true, anthropicModels: ["combo-a"] }
-    );
-    assert.deepEqual(m.api, {
-      id: "anthropic",
-      url: "https://gw.example.com",
-      npm: "@ai-sdk/anthropic",
-    });
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-enrich.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-enrich.test.ts
@@ -1,47 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import {
-  applyEnrichment,
-  buildCanonicalToAliasMap,
-  canonicalDedupSet,
-  lookupEnrichment,
-  type OmniRouteEnrichmentMap,
-} from "../src/shared/enrich.js";
-import { mapRawModelToModelV2 } from "../src/shared/models-map.js";
-
-const enrichment: OmniRouteEnrichmentMap = new Map([
-  ["cc/model-x", { name: "Model X", providerAlias: "cc", providerCanonical: "claude" }],
-  ["model-x", { name: "Model X", providerAlias: "cc", providerCanonical: "claude" }],
-]);
-
-describe("lookupEnrichment", () => {
-  it("resolves a canonical id via the alias fallback chain", () => {
-    const canonicalToAlias = buildCanonicalToAliasMap(enrichment);
-    assert.equal(canonicalToAlias.get("claude"), "cc");
-    const found = lookupEnrichment("claude/model-x", enrichment, canonicalToAlias);
-    assert.equal(found?.name, "Model X");
-  });
-});
-
-describe("applyEnrichment", () => {
-  it("overlays the enrichment display name onto the model", () => {
-    const model = mapRawModelToModelV2(
-      { id: "cc/model-x" },
-      { providerId: "omniroute", baseURL: "https://gw.example.com" }
-    );
-    applyEnrichment(model, { name: "Model X" });
-    assert.equal(model.name, "Model X");
-  });
-});
-
-describe("canonicalDedupSet", () => {
-  it("drops the canonical twin when the alias row exists", () => {
-    const canonicalToAlias = buildCanonicalToAliasMap(enrichment);
-    const drop = canonicalDedupSet(
-      [{ id: "cc/model-x" }, { id: "claude/model-x" }],
-      canonicalToAlias
-    );
-    assert.ok(drop.has("claude/model-x"));
-    assert.ok(!drop.has("cc/model-x"));
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-enrichment-fetcher.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-enrichment-fetcher.test.ts
@@ -1,125 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { defaultOmniRouteEnrichmentFetcher } from "../src/shared/enrich.js";
-
-function stubFetch(handler: (url: string) => unknown): typeof fetch {
-  return (async (url: unknown) => {
-    const body = handler(String(url));
-    return {
-      ok: true,
-      status: 200,
-      statusText: "OK",
-      json: async () => body,
-    };
-  }) as unknown as typeof fetch;
-}
-
-describe("defaultOmniRouteEnrichmentFetcher", () => {
-  it("merges display names and pricing from the two catalog sources", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch((url) =>
-      url.endsWith("/api/pricing/models")
-        ? {
-            cc: {
-              id: "claude",
-              alias: "cc",
-              name: "Claude",
-              models: [{ id: "m1", name: "Model One" }],
-            },
-          }
-        : url.endsWith("/api/pricing")
-          ? { cc: { m1: { input: 3, output: 15, cached: 0.3, cache_creation: 3.75 } } }
-          : { perModel: [] }
-    );
-    try {
-      const map = await defaultOmniRouteEnrichmentFetcher("https://gw.example.com", "k", 1000);
-      const namespaced = map.get("cc/m1");
-      assert.ok(namespaced);
-      assert.equal(namespaced?.name, "Model One");
-      assert.equal(namespaced?.providerAlias, "cc");
-      assert.equal(namespaced?.providerCanonical, "claude");
-      assert.deepEqual(namespaced?.pricing, {
-        input: 3,
-        output: 15,
-        cacheRead: 0.3,
-        cacheWrite: 3.75,
-      });
-      assert.deepEqual(map.get("m1")?.pricing, namespaced?.pricing);
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-
-  it("soft-fails per source: a dead pricing endpoint still returns catalog names", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      if (href.endsWith("/api/pricing")) throw new Error("boom");
-      return {
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => ({
-          cc: { id: "claude", models: [{ id: "m1", name: "Model One" }] },
-        }),
-      };
-    }) as unknown as typeof fetch;
-    try {
-      const map = await defaultOmniRouteEnrichmentFetcher("https://gw.example.com", "k", 1000);
-      assert.equal(map.get("cc/m1")?.name, "Model One");
-      assert.equal(map.get("cc/m1")?.pricing, undefined);
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-
-  it("soft-fails per source: a dead catalog endpoint still returns pricing", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      if (href.endsWith("/api/pricing/models")) {
-        return { ok: false, status: 500, statusText: "err", json: async () => ({}) };
-      }
-      return {
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => ({ cc: { m1: { input: 1, output: 2 } } }),
-      };
-    }) as unknown as typeof fetch;
-    try {
-      const map = await defaultOmniRouteEnrichmentFetcher("https://gw.example.com", "k", 1000);
-      assert.deepEqual(map.get("cc/m1")?.pricing, { input: 1, output: 2 });
-      assert.equal(map.get("cc/m1")?.name, undefined);
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-
-  it("attaches free-tier budgets from the third source", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch((url) =>
-      url.endsWith("/api/pricing/models")
-        ? { cc: { id: "claude", models: [{ id: "m1", name: "Model One" }] } }
-        : url.endsWith("/api/pricing")
-          ? {}
-          : {
-              perModel: [
-                { provider: "cc", modelId: "m1", freeType: "monthly", monthlyTokens: 100 },
-              ],
-            }
-    );
-    try {
-      const map = await defaultOmniRouteEnrichmentFetcher("https://gw.example.com", "k", 1000);
-      assert.equal(map.get("cc/m1")?.freeType, "monthly");
-      assert.equal(map.get("cc/m1")?.monthlyTokens, 100);
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-
-  it("returns an empty map without credentials instead of throwing", async () => {
-    const map = await defaultOmniRouteEnrichmentFetcher("https://gw.example.com", "", 1000);
-    assert.equal(map.size, 0);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-enrichment-source-errors.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-enrichment-source-errors.test.ts
@@ -1,56 +0,0 @@
-import { describe, it, afterEach } from "node:test";
-import assert from "node:assert/strict";
-import { defaultOmniRouteEnrichmentFetcher } from "../src/shared/enrich.js";
-
-const realFetch = globalThis.fetch;
-afterEach(() => {
-  globalThis.fetch = realFetch;
-});
-
-describe("enrichment source failures are reported", () => {
-  it("names each endpoint a gateway refuses, then throws so the caller keeps last-known", async () => {
-    globalThis.fetch = (async () =>
-      new Response("forbidden", { status: 403 })) as unknown as typeof globalThis.fetch;
-    const seen: Array<[string, string]> = [];
-    await assert.rejects(
-      defaultOmniRouteEnrichmentFetcher("https://gw.example.com", "k", 1000, (endpoint, reason) =>
-        seen.push([endpoint, reason])
-      ),
-      /enrichment (catalog source|sources) failed/
-    );
-    assert.deepEqual(seen.map(([endpoint]) => endpoint).sort(), [
-      "/api/free-tier/summary",
-      "/api/pricing",
-      "/api/pricing/models",
-    ]);
-    for (const [, reason] of seen) assert.match(reason, /403/);
-  });
-
-  it("reports a transport failure with its cause, then throws", async () => {
-    globalThis.fetch = (async () => {
-      throw new Error("connect ECONNREFUSED");
-    }) as unknown as typeof globalThis.fetch;
-    const seen: string[] = [];
-    await assert.rejects(
-      defaultOmniRouteEnrichmentFetcher("https://gw.example.com", "k", 1000, (_e, reason) =>
-        seen.push(reason)
-      ),
-      /enrichment (catalog source|sources) failed/
-    );
-    assert.equal(seen.length, 3);
-    for (const reason of seen) assert.match(reason, /ECONNREFUSED/);
-  });
-
-  it("stays silent when every source answers", async () => {
-    globalThis.fetch = (async () =>
-      new Response("{}", {
-        status: 200,
-        headers: { "content-type": "application/json" },
-      })) as unknown as typeof globalThis.fetch;
-    const seen: string[] = [];
-    await defaultOmniRouteEnrichmentFetcher("https://gw.example.com", "k", 1000, (e) =>
-      seen.push(e)
-    );
-    assert.deepEqual(seen, []);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-fetch-timeout.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-fetch-timeout.test.ts
@@ -1,49 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { defaultOmniRouteModelsFetcher } from "../src/shared/models-map.js";
-import { defaultOmniRouteCombosFetcher } from "../src/shared/combos-map.js";
-
-// A fetch stub that hangs until the caller aborts: proves the AbortController
-// wiring fires. Rejects with an AbortError like undici does on abort.
-function hangingFetch(): typeof fetch {
-  return ((url: unknown, init?: { signal?: AbortSignal }) => {
-    return new Promise((_resolve, reject) => {
-      const signal = init?.signal;
-      if (signal?.aborted) {
-        reject(Object.assign(new Error("This operation was aborted"), { name: "AbortError" }));
-        return;
-      }
-      signal?.addEventListener("abort", () => {
-        reject(Object.assign(new Error("This operation was aborted"), { name: "AbortError" }));
-      });
-    }) as Promise<Response>;
-  }) as unknown as typeof fetch;
-}
-
-describe("shared default fetchers abort on timeout", () => {
-  it("models fetcher aborts a hanging request after timeoutMs", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = hangingFetch();
-    try {
-      await assert.rejects(
-        defaultOmniRouteModelsFetcher("https://gw.example.com", "k", 20),
-        /aborted/i
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-
-  it("combos fetcher aborts a hanging request after timeoutMs", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = hangingFetch();
-    try {
-      await assert.rejects(
-        defaultOmniRouteCombosFetcher("https://gw.example.com", "k", 20),
-        /aborted/i
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-fingerprint.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-fingerprint.test.ts
@@ -1,62 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { catalogContentFingerprint, optionalTierFingerprint } from "../src/shared/fingerprint.js";
-
-describe("catalogContentFingerprint", () => {
-  it("returns the same hash for two identical catalogs", () => {
-    const models = [
-      { id: "b", release_date: "2026-01-02" },
-      { id: "a", release_date: "2026-01-01" },
-    ];
-    const combos = [{ id: "combo-b" }, { id: "combo-a" }];
-    assert.equal(
-      catalogContentFingerprint(models, combos),
-      catalogContentFingerprint([...models].reverse(), [...combos].reverse())
-    );
-  });
-  it("returns a different hash when one model id changes", () => {
-    const combos = [{ id: "combo-a" }];
-    const before = catalogContentFingerprint([{ id: "a" }], combos);
-    const after = catalogContentFingerprint([{ id: "a2" }], combos);
-    assert.notEqual(before, after);
-  });
-  it("returns a different hash when the auto-combos set changes", () => {
-    const models = [{ id: "a" }];
-    const combos = [{ id: "combo-a" }];
-    const before = catalogContentFingerprint(models, combos, [{ id: "auto" }]);
-    const after = catalogContentFingerprint(models, combos, [
-      { id: "auto" },
-      { id: "auto/coding" },
-    ]);
-    assert.notEqual(before, after);
-  });
-});
-
-describe("optionalTierFingerprint", () => {
-  it("moves on a pricing-only change, so stale prices reach the picker", () => {
-    const priced = (input: number) =>
-      new Map([["cc/m1", { name: "M1", pricing: { input, output: 1 } }]]);
-    assert.notEqual(
-      optionalTierFingerprint([], [], priced(3)),
-      optionalTierFingerprint([], [], priced(99))
-    );
-  });
-
-  it("moves when a combo loses a member without changing id", () => {
-    const one = [{ id: "combo-a", name: "A", models: [{ model: "m1" }] }];
-    const two = [{ id: "combo-a", name: "A", models: [{ model: "m1" }, { model: "m2" }] }];
-    assert.notEqual(
-      optionalTierFingerprint([], [], undefined, one),
-      optionalTierFingerprint([], [], undefined, two)
-    );
-  });
-
-  it("moves when a provider goes quiet or gets renamed", () => {
-    const active = [{ id: "c1", testStatus: "active", isActive: true }];
-    const quiet = [{ id: "c1", testStatus: "active", isActive: false }];
-    assert.notEqual(
-      optionalTierFingerprint([], active, undefined),
-      optionalTierFingerprint([], quiet, undefined)
-    );
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-gemini.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-gemini.test.ts
@@ -1,144 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { isGeminiModelId, sanitizeToolInputSchemas } from "../src/shared/gemini.js";
-
-describe("Gemini tool-schema sanitising", () => {
-  it("recognises the routing forms a Gemini model arrives under", () => {
-    for (const id of [
-      "gemini-2.5-flash",
-      "models/gemini-1.5-pro",
-      "google-vertex/gemini-2.0",
-      "GEMINI-PRO",
-    ]) {
-      assert.equal(isGeminiModelId(id), true, id);
-    }
-    for (const id of ["gpt-5", "claude-opus-5", "", "gemma-2-9b"]) {
-      assert.equal(isGeminiModelId(id), false, id);
-    }
-  });
-
-  it("strips the keywords Gemini rejects, however deep they sit", () => {
-    const tools = [
-      {
-        type: "function" as const,
-        name: "edit",
-        inputSchema: {
-          $schema: "https://json-schema.org/draft-07/schema",
-          type: "object",
-          additionalProperties: false,
-          properties: {
-            path: { type: "string" },
-            nested: {
-              type: "object",
-              additionalProperties: true,
-              properties: { deep: { type: "string" } },
-            },
-            list: { type: "array", items: [{ type: "object", additionalProperties: false }] },
-          },
-        },
-      },
-    ];
-    const out = sanitizeToolInputSchemas(tools);
-    assert.notEqual(out, undefined, "a payload that needed cleaning must come back changed");
-    const schema = out![0]!.inputSchema as Record<string, any>;
-    assert.equal("$schema" in schema, false);
-    assert.equal("additionalProperties" in schema, false);
-    assert.equal("additionalProperties" in schema["properties"]["nested"], false);
-    assert.equal("additionalProperties" in schema["properties"]["list"]["items"][0], false);
-    // What the schema means must survive the cleaning.
-    assert.equal(schema["properties"]["path"]["type"], "string");
-    assert.equal(schema["properties"]["nested"]["properties"]["deep"]["type"], "string");
-  });
-
-  it("leaves the caller's tools untouched", () => {
-    const tools = [
-      {
-        type: "function" as const,
-        name: "t",
-        inputSchema: { type: "object", additionalProperties: false },
-      },
-    ];
-    sanitizeToolInputSchemas(tools);
-    assert.equal((tools[0]!.inputSchema as Record<string, unknown>)["additionalProperties"], false);
-  });
-
-  it("reports nothing to do rather than cloning a clean payload", () => {
-    assert.equal(
-      sanitizeToolInputSchemas([
-        { type: "function" as const, name: "t", inputSchema: { type: "object", properties: {} } },
-      ]),
-      undefined
-    );
-    assert.equal(sanitizeToolInputSchemas(undefined), undefined);
-    assert.equal(sanitizeToolInputSchemas([]), undefined);
-  });
-
-  it("walks past a provider tool it cannot read without dropping it", () => {
-    const tools = [
-      { type: "provider" as const, id: "p.search", name: "search", args: {} },
-      {
-        type: "function" as const,
-        name: "t",
-        inputSchema: { type: "object", additionalProperties: false },
-      },
-    ];
-    const out = sanitizeToolInputSchemas(tools as never);
-    assert.equal(out?.length, 2);
-    assert.deepEqual(out![0], tools[0]);
-  });
-});
-
-describe("the sanitiser repairs schemas without mangling them", () => {
-  it("keeps a tool parameter that happens to be named like a keyword", () => {
-    const out = sanitizeToolInputSchemas([
-      {
-        type: "function" as const,
-        name: "checkout",
-        inputSchema: {
-          type: "object",
-          additionalProperties: false,
-          required: ["ref"],
-          properties: {
-            ref: { type: "string", description: "a git ref" },
-            additionalProperties: { type: "boolean" },
-          },
-        },
-      },
-    ]);
-    const schema = out![0]!.inputSchema as Record<string, any>;
-    // The keyword goes; the parameters keep their names, or `required` would
-    // point at a property the model can no longer send.
-    assert.equal("additionalProperties" in schema, false);
-    assert.deepEqual(Object.keys(schema["properties"]).sort(), ["additionalProperties", "ref"]);
-    assert.deepEqual(schema["required"], ["ref"]);
-  });
-
-  it("forwards a $ref schema untouched rather than widening it to anything", () => {
-    assert.equal(
-      sanitizeToolInputSchemas([
-        { type: "function" as const, name: "t", inputSchema: { $ref: "#/$defs/x" } },
-      ]),
-      undefined
-    );
-  });
-
-  it("recognises the Gemini families, and only those", () => {
-    for (const id of [
-      "gemini",
-      "gemini-2.5-flash",
-      "models/gemini-1.5-pro",
-      "google-vertex/gemini-2.0",
-      "GEMINI-3-PRO",
-    ]) {
-      assert.equal(isGeminiModelId(id), true, id);
-    }
-    for (const id of [
-      "gemini-compatible-proxy",
-      "my-gemini-wrapper",
-      "openai/gpt-5-gemini-eval",
-      "gemma-2-9b",
-    ]) {
-      assert.equal(isGeminiModelId(id), false, id);
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-logger.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-logger.test.ts
@@ -1,130 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { createLogger, getLogLevel } from "../src/shared/logger.js";
-
-function capture(): {
-  warns: string[];
-  errors: string[];
-  restore: () => void;
-} {
-  const warns: string[] = [];
-  const errors: string[] = [];
-  const origWarn = console.warn;
-  const origError = console.error;
-  console.warn = (...args: unknown[]) => {
-    warns.push(args.map(String).join(" "));
-  };
-  console.error = (...args: unknown[]) => {
-    errors.push(args.map(String).join(" "));
-  };
-  return {
-    warns,
-    errors,
-    restore() {
-      console.warn = origWarn;
-      console.error = origError;
-    },
-  };
-}
-
-describe("shared leveled logger (v1 parity)", () => {
-  it("default level is warn", () => {
-    assert.equal(getLogLevel(), "warn");
-  });
-
-  it("warn logger emits error+warn, suppresses info+debug", () => {
-    const log = createLogger("warn");
-    const cap = capture();
-    try {
-      log.error("e");
-      log.warn("w");
-      log.info("i");
-      log.debug("d");
-    } finally {
-      cap.restore();
-    }
-    assert.equal(cap.errors.length, 1);
-    assert.equal(cap.warns.length, 1);
-  });
-
-  it("error logger emits only error", () => {
-    const log = createLogger("error");
-    const cap = capture();
-    try {
-      log.error("e");
-      log.warn("w");
-      log.info("i");
-      log.debug("d");
-    } finally {
-      cap.restore();
-    }
-    assert.equal(cap.errors.length, 1);
-    assert.equal(cap.warns.length, 0);
-  });
-
-  it("info logger emits error+warn+info, suppresses debug", () => {
-    const log = createLogger("info");
-    const cap = capture();
-    try {
-      log.error("e");
-      log.warn("w");
-      log.info("i");
-      log.debug("d");
-    } finally {
-      cap.restore();
-    }
-    assert.equal(cap.errors.length, 1);
-    assert.equal(cap.warns.length, 2);
-  });
-
-  it("debug logger emits everything", () => {
-    const log = createLogger("debug");
-    const cap = capture();
-    try {
-      log.error("e");
-      log.warn("w");
-      log.info("i");
-      log.debug("d");
-    } finally {
-      cap.restore();
-    }
-    assert.equal(cap.errors.length, 1);
-    assert.equal(cap.warns.length, 3);
-  });
-
-  it("lines carry the namespace tag and the uppercased level", () => {
-    const log = createLogger("debug");
-    const cap = capture();
-    try {
-      log.warn("hello");
-      log.error("boom");
-    } finally {
-      cap.restore();
-    }
-    assert.match(cap.warns[0], /\[omniroute-plugin\] \[WARN\] hello/);
-    assert.match(cap.errors[0], /\[omniroute-plugin\] \[ERROR\] boom/);
-  });
-
-  it("child loggers append their tag after the namespace", () => {
-    const log = createLogger("debug").child("[v2]");
-    const cap = capture();
-    try {
-      log.info("hello");
-    } finally {
-      cap.restore();
-    }
-    assert.match(cap.warns[0], /\[omniroute-plugin\]\[v2\] \[INFO\] hello/);
-  });
-
-  it("always() emits regardless of level", () => {
-    const log = createLogger("error");
-    const cap = capture();
-    try {
-      log.always("breadcrumb");
-    } finally {
-      cap.restore();
-    }
-    assert.equal(cap.warns.length, 1);
-    assert.match(cap.warns[0], /breadcrumb/);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-models-map.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-models-map.test.ts
@@ -1,94 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { mapRawModelToModelV2, resolveApiBlockV2 } from "../src/shared/models-map.js";
-
-describe("resolveApiBlockV2", () => {
-  it("routes unknown ids to openai-compatible with /v1 suffix", () => {
-    assert.deepEqual(resolveApiBlockV2("gpt-x", "https://gw.example.com"), {
-      id: "openai-compatible",
-      url: "https://gw.example.com/v1",
-      npm: "@ai-sdk/openai-compatible",
-    });
-  });
-  it("does not double the /v1 suffix", () => {
-    assert.equal(
-      resolveApiBlockV2("gpt-x", "https://gw.example.com/v1").url,
-      "https://gw.example.com/v1"
-    );
-  });
-  it("routes allowlisted full id to anthropic even with claude-like prefix elsewhere", () => {
-    const out = resolveApiBlockV2("anthropic/claude-x", "https://gw.example.com/", {
-      allowAnthropic: true,
-      anthropicModels: ["anthropic/claude-x"],
-    });
-    assert.deepEqual(out, {
-      id: "anthropic",
-      url: "https://gw.example.com",
-      npm: "@ai-sdk/anthropic",
-    });
-  });
-  it("keeps a non-allowlisted id on openai-compatible when prefixes are emptied", () => {
-    const out = resolveApiBlockV2("claude/openai-compatible-model", "https://gw.example.com", {
-      allowAnthropic: true,
-      anthropicModels: ["anthropic/claude-x"],
-      anthropicPrefixes: [],
-    });
-    assert.equal(out.id, "openai-compatible");
-  });
-  it("routes a deprecated-prefix id to anthropic with a deprecation warning", () => {
-    const warns: string[] = [];
-    const origWarn = console.warn;
-    console.warn = (...args: unknown[]) => {
-      warns.push(String(args[0]));
-    };
-    try {
-      const out = resolveApiBlockV2("cc/claude-x", "https://gw.example.com", {
-        allowAnthropic: true,
-        anthropicPrefixes: ["cc"],
-      });
-      assert.equal(out.id, "anthropic");
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.ok(warns.some((w) => w.includes("deprecated")));
-  });
-});
-
-describe("mapRawModelToModelV2", () => {
-  it("prefixes bare ids, keeps slashed ids, maps capabilities", () => {
-    const m = mapRawModelToModelV2(
-      {
-        id: "gpt-x",
-        context_length: 128000,
-        max_output_tokens: 4096,
-        input_modalities: ["text", "image"],
-        capabilities: { tool_calling: true, reasoning: true },
-      },
-      { providerId: "omniroute", baseURL: "https://gw.example.com" }
-    );
-    assert.equal(m.id, "omniroute/gpt-x");
-    assert.equal(m.providerID, "omniroute");
-    assert.equal(m.capabilities.toolcall, true);
-    assert.equal(m.capabilities.reasoning, true);
-    assert.equal(m.capabilities.input.image, true);
-    assert.equal(m.limit.context, 128000);
-    assert.equal(m.limit.output, 4096);
-    assert.equal(m.api.id, "openai-compatible");
-    assert.equal(m.status, "active");
-  });
-  it("maps effort_tiers to variants, omits key when absent", () => {
-    const withTiers = mapRawModelToModelV2(
-      { id: "r1", capabilities: { effort_tiers: ["low", "high"] } },
-      { providerId: "omniroute", baseURL: "https://gw.example.com" }
-    );
-    assert.deepEqual(withTiers.variants, {
-      low: { reasoningEffort: "low" },
-      high: { reasoningEffort: "high" },
-    });
-    const without = mapRawModelToModelV2(
-      { id: "r2" },
-      { providerId: "omniroute", baseURL: "https://gw.example.com" }
-    );
-    assert.ok(!("variants" in without));
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-naming.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-naming.test.ts
@@ -1,75 +0,0 @@
-/**
- * Tests for `formatFreeBudget` (@omniroute/opencode-plugin/src/naming.ts):
- * formats a free-tier model's budget info into a short human-readable
- * suffix, branching on `freeType`.
- */
-
-import { test } from "node:test";
-import assert from "node:assert/strict";
-
-import { formatFreeBudget, type FreeModelFreeType } from "../src/shared/naming.js";
-
-test("formatFreeBudget: recurring-daily formats tokens/day", () => {
-  assert.equal(
-    formatFreeBudget({ freeType: "recurring-daily", monthlyTokens: 25_000_000 }),
-    "25M tokens/day"
-  );
-});
-
-test("formatFreeBudget: recurring-monthly formats tokens/month", () => {
-  assert.equal(
-    formatFreeBudget({ freeType: "recurring-monthly", monthlyTokens: 1_000_000 }),
-    "1M tokens/month"
-  );
-});
-
-test("formatFreeBudget: recurring-credit formats credits", () => {
-  assert.equal(
-    formatFreeBudget({ freeType: "recurring-credit", creditTokens: 10_000_000 }),
-    "10M credits"
-  );
-});
-
-test("formatFreeBudget: one-time-initial formats credits with (one-time) suffix", () => {
-  assert.equal(
-    formatFreeBudget({ freeType: "one-time-initial", creditTokens: 1_000_000 }),
-    "1M credits (one-time)"
-  );
-});
-
-test("formatFreeBudget: keyless has no token/credit args", () => {
-  assert.equal(formatFreeBudget({ freeType: "keyless" }), "(keyless)");
-});
-
-test("formatFreeBudget: discontinued has no token/credit args", () => {
-  assert.equal(formatFreeBudget({ freeType: "discontinued" }), "(discontinued)");
-});
-
-test("formatFreeBudget: missing token/credit counts default to 0", () => {
-  assert.equal(formatFreeBudget({ freeType: "recurring-daily" }), "0 tokens/day");
-});
-
-test("formatFreeBudget: unrecognised freeType falls through to the default branch", () => {
-  // `freeType` is populated from catalog data at runtime, so a value the
-  // build doesn't know about is reachable even though TypeScript treats the
-  // `default:` arm as dead code for a well-typed caller.
-  assert.equal(formatFreeBudget({ freeType: "some-future-type" as FreeModelFreeType }), "");
-});
-
-test("formatFreeBudget: sub-1K token count is not abbreviated", () => {
-  assert.equal(
-    formatFreeBudget({ freeType: "recurring-daily", monthlyTokens: 500 }),
-    "500 tokens/day"
-  );
-});
-
-test("formatFreeBudget: the 999_999 rounding wart is fixed — promotes to 1M", () => {
-  // `toFixed(1)` rounds 999999/1e3 up to "1000.0" before the `>= 1e6` threshold
-  // check has a chance to apply. fmtTokens now promotes a rounded-up "1000" in
-  // any unit to the next unit up, so this correctly reads "1M" instead of the
-  // old "1000K" wart.
-  assert.equal(
-    formatFreeBudget({ freeType: "recurring-daily", monthlyTokens: 999_999 }),
-    "1M tokens/day"
-  );
-});
--- a/@omniroute/opencode-plugin-v2/tests/shared-usable.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/shared-usable.test.ts
@@ -1,248 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import {
-  defaultOmniRouteProvidersFetcher,
-  isUsableCombo,
-  isUsableRawModelId,
-  usableProviderAliasSet,
-} from "../src/shared/usable.js";
-import type { OmniRouteEnrichmentMap } from "../src/shared/enrich.js";
-
-function entry(
-  alias: string,
-  canonical: string
-): [string, { providerAlias: string; providerCanonical: string }] {
-  return [`${alias}/model-x`, { providerAlias: alias, providerCanonical: canonical }];
-}
-
-function enrichmentOf(...pairs: Array<[string, string]>): OmniRouteEnrichmentMap {
-  return new Map(pairs.map(([alias, canonical]) => entry(alias, canonical)));
-}
-
-function stubFetch(handler: (url: unknown) => unknown): typeof fetch {
-  return (async (url: unknown) => handler(url)) as unknown as typeof fetch;
-}
-
-describe("defaultOmniRouteProvidersFetcher envelopes", () => {
-  it("reads the { connections: [...] } envelope", async () => {
-    const origFetch = globalThis.fetch;
-    const seen: string[] = [];
-    globalThis.fetch = stubFetch((url) => {
-      seen.push(String(url));
-      return {
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => ({
-          connections: [{ id: "c1", provider: "claude", isActive: true, testStatus: "active" }],
-        }),
-      };
-    });
-    try {
-      const out = await defaultOmniRouteProvidersFetcher("https://gw.example.com/v1/", "k");
-      assert.equal(seen[0], "https://gw.example.com/api/providers");
-      assert.deepEqual(
-        out.map((c) => c.provider),
-        ["claude"]
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-
-  it("reads a bare array envelope", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch(() => ({
-      ok: true,
-      status: 200,
-      statusText: "OK",
-      json: async () => [{ id: "c1", provider: "gemini", isActive: true, testStatus: "active" }],
-    }));
-    try {
-      const out = await defaultOmniRouteProvidersFetcher("https://gw.example.com", "k");
-      assert.deepEqual(
-        out.map((c) => c.provider),
-        ["gemini"]
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-
-  it("reads the { data: [...] } envelope", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch(() => ({
-      ok: true,
-      status: 200,
-      statusText: "OK",
-      json: async () => ({
-        data: [{ id: "c1", provider: "kiro", isActive: true, testStatus: "active" }],
-      }),
-    }));
-    try {
-      const out = await defaultOmniRouteProvidersFetcher("https://gw.example.com", "k");
-      assert.deepEqual(
-        out.map((c) => c.provider),
-        ["kiro"]
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-
-  it("throws on non-2xx so the caller keeps last-known instead of unfiltering", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch(() => ({
-      ok: false,
-      status: 500,
-      statusText: "Error",
-      json: async () => ({}),
-    }));
-    try {
-      await assert.rejects(
-        defaultOmniRouteProvidersFetcher("https://gw.example.com", "k"),
-        /HTTP 500/
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-
-  it("throws on network failure so the caller keeps last-known", async () => {
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch(() => {
-      throw new Error("down");
-    });
-    try {
-      await assert.rejects(defaultOmniRouteProvidersFetcher("https://gw.example.com", "k"), /down/);
-    } finally {
-      globalThis.fetch = origFetch;
-    }
-  });
-
-  it("soft-fails to [] without baseURL or apiKey", async () => {
-    assert.deepEqual(await defaultOmniRouteProvidersFetcher("", "k"), []);
-    assert.deepEqual(await defaultOmniRouteProvidersFetcher("https://gw.example.com", ""), []);
-  });
-});
-
-describe("usableProviderAliasSet", () => {
-  it("maps usable canonicals to aliases via enrichment", () => {
-    const usable = usableProviderAliasSet(
-      [{ id: "c1", provider: "claude", isActive: true, testStatus: "active" }],
-      enrichmentOf(["cc", "claude"])
-    );
-    assert.ok(usable.aliases.has("cc"));
-    assert.ok(usable.aliases.has("claude"));
-    assert.ok(usable.canonicals.has("claude"));
-    assert.ok(usable.knownAliases.has("cc"));
-  });
-
-  it("ignores inactive or unhealthy connections", () => {
-    const usable = usableProviderAliasSet(
-      [
-        { id: "c1", provider: "claude", isActive: false, testStatus: "active" },
-        { id: "c2", provider: "gemini", isActive: true, testStatus: "error" },
-      ],
-      enrichmentOf(["cc", "claude"], ["gm", "gemini"])
-    );
-    assert.equal(usable.aliases.size, 0);
-    assert.equal(usable.canonicals.size, 0);
-    assert.ok(usable.knownAliases.has("cc"));
-    assert.ok(usable.knownAliases.has("gm"));
-  });
-
-  it("treats a missing toggle as no opinion, not disabled", () => {
-    // Older gateways predate the `isActive` field: reading its absence as a
-    // veto would hide every model behind a filter the operator never asked
-    // to tighten. Only an explicit `false` disables.
-    const usable = usableProviderAliasSet(
-      [{ id: "c1", provider: "claude", testStatus: "active" }],
-      enrichmentOf(["cc", "claude"])
-    );
-    assert.ok(usable.aliases.has("cc"));
-    assert.ok(usable.canonicals.has("claude"));
-  });
-});
-
-describe("isUsableRawModelId subtract-filter", () => {
-  const usable = usableProviderAliasSet(
-    [{ id: "c1", provider: "claude", isActive: true, testStatus: "active" }],
-    enrichmentOf(["cc", "claude"], ["dead", "legacy"])
-  );
-
-  it("keeps ids with a usable alias prefix", () => {
-    assert.equal(isUsableRawModelId("cc/claude-opus-4-7", usable), true);
-  });
-
-  it("keeps ids with a usable canonical prefix", () => {
-    assert.equal(isUsableRawModelId("claude/sonnet-4", usable), true);
-  });
-
-  it("keeps ids with an unknown prefix", () => {
-    assert.equal(isUsableRawModelId("agentrouter/mystery", usable), true);
-  });
-
-  it("drops ids with a known-but-not-provisioned prefix", () => {
-    assert.equal(isUsableRawModelId("dead/legacy-model", usable), false);
-  });
-
-  it("keeps ids without a prefix", () => {
-    assert.equal(isUsableRawModelId("bare-model", usable), true);
-  });
-});
-
-describe("isUsableCombo", () => {
-  const usable = usableProviderAliasSet(
-    [{ id: "c1", provider: "claude", isActive: true, testStatus: "active" }],
-    enrichmentOf(["cc", "claude"], ["dead", "legacy"])
-  );
-
-  it("keeps a combo with one usable member", () => {
-    assert.equal(
-      isUsableCombo(
-        {
-          id: "c",
-          models: [
-            { kind: "model", model: "dead/legacy" },
-            { kind: "model", model: "cc/x" },
-          ],
-        },
-        usable
-      ),
-      true
-    );
-  });
-
-  it("drops a combo whose members are all known-but-not-provisioned", () => {
-    assert.equal(
-      isUsableCombo(
-        {
-          id: "c",
-          models: [
-            { kind: "model", model: "dead/a" },
-            { kind: "model", model: "dead/b" },
-          ],
-        },
-        usable
-      ),
-      false
-    );
-  });
-
-  it("keeps a combo with an unknown-prefix member", () => {
-    assert.equal(
-      isUsableCombo({ id: "c", models: [{ kind: "model", model: "agentrouter/mystery" }] }, usable),
-      true
-    );
-  });
-
-  it("keeps combos with no resolvable members", () => {
-    assert.equal(isUsableCombo({ id: "c", models: [] }, usable), true);
-    assert.equal(isUsableCombo({ id: "c" }, usable), true);
-    assert.equal(
-      isUsableCombo({ id: "c", models: [{ kind: "combo-ref", comboName: "nested" }] }, usable),
-      true
-    );
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/smoke-types.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/smoke-types.test.ts
@@ -1,93 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { CatalogDraft, PluginContext } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-import plugin from "../src/index.js";
-
-describe("v2 contract smoke", () => {
-  it("default export has string id and function setup", () => {
-    assert.equal(typeof (plugin as { id: unknown }).id, "string");
-    assert.equal(typeof (plugin as { setup: unknown }).setup, "function");
-  });
-  it("setup registers transforms against a structurally-real ctx", async () => {
-    const seen: string[] = [];
-    const ctx = {
-      options: { baseURL: "https://gw.example.com", providerId: "omniroute" },
-      catalog: {
-        transform: async () => {
-          seen.push("catalog.transform");
-          return { dispose: async () => {} };
-        },
-        reload: async () => {},
-      },
-      integration: {
-        transform: async () => {
-          seen.push("integration.transform");
-          return { dispose: async () => {} };
-        },
-        reload: async () => {},
-        connection: { active: async () => undefined, resolve: async () => undefined },
-      },
-      agent: { transform: async () => ({ dispose: async () => {} }), reload: async () => {} },
-      command: { transform: async () => ({ dispose: async () => {} }), reload: async () => {} },
-      reference: { transform: async () => ({ dispose: async () => {} }), reload: async () => {} },
-      skill: { transform: async () => ({ dispose: async () => {} }), reload: async () => {} },
-      aisdk: {
-        sdk: async () => ({ dispose: async () => {} }),
-        language: async () => ({ dispose: async () => {} }),
-      },
-      plugin: { add: async () => {}, remove: async () => {} },
-    } satisfies PluginContext;
-    await (plugin as { setup: (c: PluginContext) => Promise<void> }).setup(ctx);
-    assert.deepEqual(seen, ["catalog.transform", "integration.transform"]);
-  });
-  it("publishCatalog writes into a real CatalogDraft without proxy breakage", async () => {
-    const { publishCatalog } = await import("../src/catalog.js");
-    const written: { provider?: string; models: string[] } = { models: [] };
-    const draft: CatalogDraft = {
-      provider: {
-        list: () => [],
-        get: () => undefined,
-        update: (id: string, fn: (p: ProviderV2Info) => void) => {
-          written.provider = id;
-          const p = {
-            id,
-            name: "",
-            api: { type: "aisdk", package: "" },
-            request: { headers: {}, body: {} },
-          } as ProviderV2Info;
-          fn(p);
-        },
-        remove: () => {},
-      },
-      model: {
-        get: () => undefined,
-        update: (providerID: string, modelID: string, fn: (d: ModelV2Info) => void) => {
-          written.models.push(providerID + "/" + modelID);
-          const d = { id: modelID, providerID } as ModelV2Info;
-          fn(d);
-        },
-        remove: () => {},
-        default: { get: () => undefined, set: () => {} },
-      },
-    };
-    const res = await publishCatalog(
-      draft,
-      {
-        providerId: "omniroute",
-        baseURL: "https://gw.example.com",
-        apiKey: "k",
-        timeoutMs: 1000,
-        modelCacheTtlMs: 300000,
-        usableOnly: false,
-      },
-      {
-        models: async () => [{ id: "m1", context_length: 1000 }],
-        combos: async () => [],
-      }
-    );
-    assert.equal(res.models, 1);
-    assert.equal(written.provider, "omniroute");
-    assert.deepEqual(written.models, ["omniroute/m1"]);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/snapshot-stale-entries.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/snapshot-stale-entries.test.ts
@@ -1,219 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { mkdtempSync, writeFileSync, mkdirSync } from "node:fs";
-import { tmpdir } from "node:os";
-import { join } from "node:path";
-import { createHash } from "node:crypto";
-import plugin from "../src/index.js";
-import { diskSnapshotPath, snapshotIdentityFingerprint } from "../src/cache.js";
-import { legacyApiToInfoApi } from "../src/catalog.js";
-
-function isolateDisk(): { dir: string; restore: () => void } {
-  const dir = mkdtempSync(join(tmpdir(), "omniroute-snapfix-"));
-  const prev = process.env.OPENCODE_DATA_DIR;
-  process.env.OPENCODE_DATA_DIR = dir;
-  return {
-    dir,
-    restore: () => {
-      if (prev === undefined) delete process.env.OPENCODE_DATA_DIR;
-      else process.env.OPENCODE_DATA_DIR = prev;
-    },
-  };
-}
-
-function setupCtx(providerId: string): {
-  callbacks: Array<(draft: unknown) => Promise<void>>;
-  ctx: Record<string, unknown>;
-} {
-  const callbacks: Array<(draft: unknown) => Promise<void>> = [];
-  const ctx = {
-    options: {
-      providerId,
-      baseURL: "https://gw.example.com",
-      apiKey: "k-snapfix",
-    },
-    catalog: {
-      transform: (cb: (draft: unknown) => Promise<void>) => {
-        callbacks.push(cb);
-        return Promise.resolve({ dispose: async () => {} });
-      },
-      reload: async () => {},
-    },
-    integration: { transform: () => Promise.resolve({ dispose: async () => {} }) },
-  };
-  return { callbacks, ctx };
-}
-
-function stubDraft(): { draft: unknown; published: Map<string, Record<string, unknown>> } {
-  const published = new Map<string, Record<string, unknown>>();
-  return {
-    published,
-    draft: {
-      provider: { update: (_id: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-      model: {
-        update: (pid: string, mid: string, fn: (m: Record<string, unknown>) => void) => {
-          const entry: Record<string, unknown> = { id: mid, providerID: pid };
-          fn(entry);
-          published.set(pid + "/" + mid, entry);
-        },
-      },
-    },
-  };
-}
-
-async function silenceConsole<T>(fn: () => Promise<T>): Promise<{ result: T; warns: string[] }> {
-  const warns: string[] = [];
-  const origWarn = console.warn;
-  const origLog = console.log;
-  console.warn = (...args: unknown[]) => {
-    warns.push(String(args[0]));
-  };
-  console.log = () => {};
-  try {
-    const result = await fn();
-    return { result, warns };
-  } finally {
-    console.warn = origWarn;
-    console.log = origLog;
-  }
-}
-
-function downFetch(): typeof fetch {
-  return (async (url: unknown) => {
-    const href = String(url);
-    if (href.includes("/api/pricing") || href.includes("/api/free-tier")) {
-      return { ok: true, status: 200, statusText: "OK", json: async () => ({}) };
-    }
-    if (href.includes("/api/combos/auto")) {
-      return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-    }
-    if (href.includes("/api/combos")) {
-      return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-    }
-    return { ok: false, status: 500, statusText: "Down", json: async () => ({}) };
-  }) as typeof fetch;
-}
-
-const fingerprint = snapshotIdentityFingerprint("https://gw.example.com", "k-snapfix", "k-snapfix");
-
-describe("plugin-v2 snapshot stale-entry filter", () => {
-  it("snapshot with 2 entries without api block + 1 valid: only the valid one is published + warn emitted", async () => {
-    const disk = isolateDisk();
-    const providerId = "snapfix-mixed";
-    mkdirSync(join(disk.dir, "plugins"), { recursive: true });
-    writeFileSync(
-      diskSnapshotPath(providerId),
-      JSON.stringify({
-        v: 2,
-        identityFingerprint: fingerprint,
-        // Two pre-mapped entries with a broken api block (missing npm) plus
-        // one plain raw entry (no api block: synthesized at publish time).
-        models: [
-          { id: "stale-a", api: {} },
-          { id: "stale-b", api: { npm: "" } },
-          { id: "good-1", context_length: 128000 },
-        ],
-        combos: [],
-        autoCombos: [],
-        providers: [],
-        writtenAt: Date.now(),
-      })
-    );
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = downFetch();
-    try {
-      const { callbacks, ctx } = setupCtx(providerId);
-      const { warns } = await silenceConsole(async () => {
-        await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-        const { draft, published } = stubDraft();
-        await callbacks[0](draft);
-        assert.ok(
-          published.has(`${providerId}/good-1`),
-          `valid entry must be published, got: ${JSON.stringify([...published.keys()])}`
-        );
-        assert.ok(
-          ![...published.keys()].some((k) => k.includes("stale")),
-          `stale entries must be dropped, got: ${JSON.stringify([...published.keys()])}`
-        );
-      });
-      assert.ok(
-        warns.some((w) => w.includes("dropping 2 stale snapshot entries without api block")),
-        `expected stale-drop warn, got: ${JSON.stringify(warns)}`
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      disk.restore();
-    }
-  });
-
-  it("snapshot without version (v1 format): ignored entirely, fail-open to fresh fetch", async () => {
-    const disk = isolateDisk();
-    const providerId = "snapfix-unversioned";
-    mkdirSync(join(disk.dir, "plugins"), { recursive: true });
-    writeFileSync(
-      diskSnapshotPath(providerId),
-      JSON.stringify({
-        identityFingerprint: "whatever",
-        rawModels: [{ id: "ancient" }],
-        rawCombos: [],
-        writtenAt: Date.now(),
-      })
-    );
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      if (href.includes("/api/pricing") || href.includes("/api/free-tier")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({}) };
-      }
-      if (href.includes("/api/combos/auto")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      if (href.includes("/api/combos")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      return {
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => ({ data: [{ id: "fresh-1" }] }),
-      };
-    }) as typeof fetch;
-    try {
-      const { callbacks, ctx } = setupCtx(providerId);
-      await silenceConsole(async () => {
-        await (plugin as unknown as { setup: (ctx: unknown) => Promise<void> }).setup(ctx);
-        const { draft, published } = stubDraft();
-        await callbacks[0](draft);
-        assert.ok(
-          published.has(`${providerId}/fresh-1`),
-          `fresh fetch must win over unversioned snapshot, got: ${JSON.stringify([...published.keys()])}`
-        );
-        assert.ok(
-          ![...published.keys()].some((k) => k.includes("ancient")),
-          `unversioned snapshot must be ignored, got: ${JSON.stringify([...published.keys()])}`
-        );
-      });
-    } finally {
-      globalThis.fetch = origFetch;
-      disk.restore();
-    }
-  });
-
-  it("legacyApiToInfoApi throws on missing api.npm (fail-fast, publish guard turns it into a warn)", () => {
-    assert.throws(
-      () => legacyApiToInfoApi(undefined as unknown as { id: string; npm: string; url: string }),
-      /without an api block/
-    );
-    assert.throws(
-      () =>
-        legacyApiToInfoApi({ id: "openai-compatible", url: "https://x/v1" } as unknown as {
-          id: string;
-          npm: string;
-          url: string;
-        }),
-      /without an api block/
-    );
-    // Sanity: sha256 helper used above matches the plugin identity scheme.
-    assert.equal(createHash("sha256").update("x").digest("hex").length, 64);
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/staged-refresh.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/staged-refresh.test.ts
@@ -1,452 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import plugin from "../src/index.js";
-
-// A catalog that only appears once the slowest optional source has answered is
-// a catalog that never appears at all on a host that exits first: an
-// unreachable `/api/combos/auto` kept models, combos and everything else
-// unpublished until its own timeout fired. Models and combos must reach the
-// draft as soon as they are known; the optional sources upgrade the snapshot
-// when they land.
-describe("plugin-v2 staged refresh: optional sources never gate the publish", () => {
-  let seq = 0;
-
-  async function isolateDisk(): Promise<() => void> {
-    const { mkdtempSync } = await import("node:fs");
-    const { tmpdir } = await import("node:os");
-    const { join } = await import("node:path");
-    seq += 1;
-    const dir = mkdtempSync(join(tmpdir(), `omniroute-staged-${seq}-`));
-    const prev = process.env.OPENCODE_DATA_DIR;
-    process.env.OPENCODE_DATA_DIR = dir;
-    return () => {
-      if (prev === undefined) delete process.env.OPENCODE_DATA_DIR;
-      else process.env.OPENCODE_DATA_DIR = prev;
-    };
-  }
-
-  function setupCtx(
-    providerId: string,
-    reloads: { count: number }
-  ): {
-    catalogCallbacks: Array<(draft: unknown) => Promise<void>>;
-    ctx: Record<string, unknown>;
-  } {
-    const catalogCallbacks: Array<(draft: unknown) => Promise<void>> = [];
-    const ctx = {
-      options: { baseURL: "https://gw.example.com", providerId, apiKey: "k-" + providerId },
-      catalog: {
-        transform: (cb: (draft: unknown) => Promise<void>) => {
-          catalogCallbacks.push(cb);
-          return Promise.resolve({ dispose: async () => {} });
-        },
-        reload: async () => {
-          reloads.count += 1;
-        },
-      },
-      integration: { transform: () => Promise.resolve({ dispose: async () => {} }) },
-    };
-    return { catalogCallbacks, ctx };
-  }
-
-  function stubDraft(): { draft: unknown; published: Map<string, Record<string, unknown>> } {
-    const published = new Map<string, Record<string, unknown>>();
-    const draft = {
-      provider: { update: (_id: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-      model: {
-        update: (pid: string, mid: string, fn: (m: Record<string, unknown>) => void) => {
-          const entry: Record<string, unknown> = { id: mid, providerID: pid };
-          fn(entry);
-          published.set(pid + "/" + mid, entry);
-        },
-      },
-    };
-    return { draft, published };
-  }
-
-  /**
-   * `/api/combos/auto` never answers and never honours the abort signal — the
-   * shape of a gateway that accepts the connection and then goes quiet.
-   */
-  function stubFetch(opts: {
-    autoCombosHangs: boolean;
-    combosHangs?: boolean;
-    enrichmentDelayMs?: number;
-  }): typeof fetch {
-    return (async (url: unknown) => {
-      const href = String(url);
-      const ok = (body: unknown) => ({
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => body,
-      });
-      if (href.includes("/api/combos/auto")) {
-        if (opts.autoCombosHangs) return await new Promise(() => {});
-        return ok({ combos: [] });
-      }
-      if (href.includes("/api/combos")) {
-        // A gateway that accepts the connection and then goes quiet on the
-        // combos endpoint: models must still publish without waiting for it.
-        if (opts.combosHangs) return await new Promise(() => {});
-        return ok({ combos: [] });
-      }
-      if (href.includes("/api/pricing/models")) {
-        if (opts.enrichmentDelayMs !== undefined) {
-          await new Promise((r) => setTimeout(r, opts.enrichmentDelayMs));
-        }
-        return ok({
-          omni: {
-            id: "omni",
-            alias: "omni",
-            name: "Omni",
-            models: [{ id: "m1", name: "Model One" }],
-          },
-        });
-      }
-      if (href.includes("/api/pricing")) return ok({});
-      if (href.includes("/api/free-tier/summary")) return ok({});
-      return ok({ data: [{ id: "m1" }] });
-    }) as typeof fetch;
-  }
-
-  async function withSilentConsole<T>(fn: () => Promise<T>): Promise<T> {
-    const warn = console.warn;
-    const log = console.log;
-    console.warn = () => {};
-    console.log = () => {};
-    try {
-      return await fn();
-    } finally {
-      console.warn = warn;
-      console.log = log;
-    }
-  }
-
-  it("publishes models while an optional source is still hanging", async () => {
-    const restoreDisk = await isolateDisk();
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch({ autoCombosHangs: true });
-    const reloads = { count: 0 };
-    const { catalogCallbacks, ctx } = setupCtx("staged-hang", reloads);
-    try {
-      await withSilentConsole(async () => {
-        await (plugin as unknown as { setup: (c: unknown) => Promise<void> }).setup(ctx);
-        const { draft, published } = stubDraft();
-        const done = catalogCallbacks[0]!(draft);
-        const raced = await Promise.race([
-          done.then(() => "published" as const),
-          new Promise<"timeout">((r) => setTimeout(() => r("timeout"), 1500)),
-        ]);
-        assert.equal(
-          raced,
-          "published",
-          "the publish must not wait on a source that never answers"
-        );
-        assert.ok([...published.keys()].some((k) => k.endsWith("/m1")));
-      });
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-
-  it("applies an optional source that lands after the publish, on the next transform", async () => {
-    const restoreDisk = await isolateDisk();
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch({ autoCombosHangs: false, enrichmentDelayMs: 120 });
-    const reloads = { count: 0 };
-    const { catalogCallbacks, ctx } = setupCtx("staged-late", reloads);
-    try {
-      await withSilentConsole(async () => {
-        await (plugin as unknown as { setup: (c: unknown) => Promise<void> }).setup(ctx);
-        const first = stubDraft();
-        await catalogCallbacks[0]!(first.draft);
-        const early = [...first.published.values()].find((m) => m["id"] === "m1");
-        assert.ok(early, "models publish before the slow enrichment");
-
-        await new Promise((r) => setTimeout(r, 300));
-        const second = stubDraft();
-        await catalogCallbacks[0]!(second.draft);
-        const late = [...second.published.values()].find((m) => m["id"] === "m1");
-        // The overlay is rendered, not just stored: the provider label the
-        // gateway ships alongside the display name reaches the picker.
-        assert.equal(
-          late?.["name"],
-          "Omni - Model One",
-          "the late enrichment must reach the catalog, provider tag included"
-        );
-      });
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-
-  it("does not ask the host to reload when the overlay came back identical", async () => {
-    const restoreDisk = await isolateDisk();
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch({ autoCombosHangs: false });
-    const reloads = { count: 0 };
-    const { catalogCallbacks, ctx } = setupCtx("staged-stable", reloads);
-    try {
-      await withSilentConsole(async () => {
-        await (plugin as unknown as { setup: (c: unknown) => Promise<void> }).setup(ctx);
-        for (let i = 0; i < 3; i++) {
-          const d = stubDraft();
-          await catalogCallbacks[0]!(d.draft);
-          await new Promise((r) => setTimeout(r, 60));
-        }
-      });
-      assert.ok(
-        reloads.count <= 1,
-        `an unchanged overlay must not trigger a reload per refresh, got ${reloads.count}`
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-
-  it("a refresh after the TTL keeps the overlay instead of downgrading the picker", async () => {
-    const restoreDisk = await isolateDisk();
-    const origFetch = globalThis.fetch;
-    // Enrichment answers once, then goes away: the second refresh must not
-    // strip the names the first one obtained.
-    let enrichCalls = 0;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      const ok = (body: unknown) => ({
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => body,
-      });
-      if (href.includes("/api/combos/auto")) return ok({ combos: [] });
-      if (href.includes("/api/combos")) return ok({ combos: [] });
-      if (href.includes("/api/pricing/models")) {
-        enrichCalls += 1;
-        if (enrichCalls > 1) {
-          return { ok: false, status: 503, statusText: "Unavailable", json: async () => ({}) };
-        }
-        return ok({
-          omni: {
-            id: "omni",
-            alias: "omni",
-            name: "Omni",
-            models: [{ id: "m1", name: "Model One" }],
-          },
-        });
-      }
-      if (href.includes("/api/pricing")) return ok({});
-      if (href.includes("/api/free-tier/summary")) return ok({});
-      return ok({ data: [{ id: "m1" }] });
-    }) as unknown as typeof fetch;
-    const reloads = { count: 0 };
-    const { catalogCallbacks, ctx } = setupCtx("staged-ttl", reloads);
-    (ctx["options"] as Record<string, unknown>)["modelCacheTtlMs"] = 1;
-    try {
-      await withSilentConsole(async () => {
-        await (plugin as unknown as { setup: (c: unknown) => Promise<void> }).setup(ctx);
-        await catalogCallbacks[0]!(stubDraft().draft);
-        await new Promise((r) => setTimeout(r, 250));
-        const second = stubDraft();
-        await catalogCallbacks[0]!(second.draft);
-        const m1 = [...second.published.values()].find((m) => m["id"] === "m1");
-        assert.equal(
-          m1?.["name"],
-          "Omni - Model One",
-          "the second refresh must keep the name the first one resolved"
-        );
-      });
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-
-  it("publishes models while a hanging /api/combos holds nothing back", async () => {
-    // Combos used to sit on the critical path (Promise.all with models), so a
-    // gateway slow on /api/combos held the whole picker back. Regression pin:
-    // models publish even when combos never answers.
-    const restoreDisk = await isolateDisk();
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch({ autoCombosHangs: false, combosHangs: true });
-    const reloads = { count: 0 };
-    const { catalogCallbacks, ctx } = setupCtx("staged-combos-hang", reloads);
-    try {
-      await withSilentConsole(async () => {
-        await (plugin as unknown as { setup: (c: unknown) => Promise<void> }).setup(ctx);
-        const { draft, published } = stubDraft();
-        const done = catalogCallbacks[0]!(draft);
-        const raced = await Promise.race([
-          done.then(() => "published" as const),
-          new Promise<"timeout">((r) => setTimeout(() => r("timeout"), 1500)),
-        ]);
-        assert.equal(
-          raced,
-          "published",
-          "models must publish without waiting for a hanging /api/combos"
-        );
-        assert.ok([...published.keys()].some((k) => k.endsWith("/m1")));
-        // "staged-combos-hang" contains "combo" as a substring — filter on the
-        // model id suffix instead: no published model id may start with a
-        // combo prefix.
-        assert.equal(
-          [...published.keys()].filter((k) => /\/combo/i.test(k)).length,
-          0,
-          `no combos known yet — models-only on the first publish is correct, got ${JSON.stringify([...published.keys()])}`
-        );
-      });
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-
-  it("keeps names/pricing on the third transform when enrichment stays down", async () => {
-    // The old all-empty guard is gone by design (it also blocked genuine
-    // removals); the per-source failure signal replaces it. A persistent 503
-    // on the overlay must keep last-known names on EVERY later transform,
-    // not just the second one — this is the regression pin for the throw
-    // instead of soft-fail change.
-    const restoreDisk = await isolateDisk();
-    const origFetch = globalThis.fetch;
-    let enrichCalls = 0;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      const ok = (body: unknown) => ({
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => body,
-      });
-      if (href.includes("/api/combos/auto")) return ok({ combos: [] });
-      if (href.includes("/api/combos")) return ok({ combos: [] });
-      if (href.includes("/api/pricing/models")) {
-        enrichCalls += 1;
-        if (enrichCalls > 1) {
-          return { ok: false, status: 503, statusText: "Unavailable", json: async () => ({}) };
-        }
-        return ok({
-          omni: {
-            id: "omni",
-            alias: "omni",
-            name: "Omni",
-            models: [{ id: "m1", name: "Model One" }],
-          },
-        });
-      }
-      if (href.includes("/api/pricing")) return ok({});
-      if (href.includes("/api/free-tier/summary")) return ok({});
-      return ok({ data: [{ id: "m1" }] });
-    }) as unknown as typeof fetch;
-    const reloads = { count: 0 };
-    const { catalogCallbacks, ctx } = setupCtx("staged-enrich-down", reloads);
-    (ctx["options"] as Record<string, unknown>)["modelCacheTtlMs"] = 1;
-    try {
-      await withSilentConsole(async () => {
-        await (plugin as unknown as { setup: (c: unknown) => Promise<void> }).setup(ctx);
-        await catalogCallbacks[0]!(stubDraft().draft);
-        await new Promise((r) => setTimeout(r, 250));
-        await catalogCallbacks[0]!(stubDraft().draft);
-        await new Promise((r) => setTimeout(r, 250));
-        const third = stubDraft();
-        await catalogCallbacks[0]!(third.draft);
-        const m1 = [...third.published.values()].find((m) => m["id"] === "m1");
-        assert.equal(
-          m1?.["name"],
-          "Omni - Model One",
-          `a persistently failing overlay must not wipe names, got ${JSON.stringify(m1?.["name"])}`
-        );
-      });
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-
-  it("serves last-known without refetching inside the unreachable window", async () => {
-    // A gateway that answers nothing at all gets a short breather instead of
-    // a full fetch suite on every transform: the cooldown arms only once a
-    // total failure is confirmed (no models AND a prior entry exists to
-    // serve), and transforms inside the window must not issue new requests.
-    const restoreDisk = await isolateDisk();
-    const origFetch = globalThis.fetch;
-    let modelCalls = 0;
-    let down = false;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      const ok = (body: unknown) => ({
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => body,
-      });
-      if (href.includes("/api/combos/auto")) return ok({ combos: [] });
-      if (href.includes("/api/combos")) return ok({ combos: [] });
-      if (href.includes("/api/pricing") || href.includes("/api/free-tier")) return ok({});
-      modelCalls += 1;
-      if (down) {
-        return { ok: false, status: 500, statusText: "Down", json: async () => ({}) };
-      }
-      return ok({ data: [{ id: "m1" }] });
-    }) as unknown as typeof fetch;
-    const reloads = { count: 0 };
-    const { catalogCallbacks, ctx } = setupCtx("staged-unreachable", reloads);
-    (ctx["options"] as Record<string, unknown>)["modelCacheTtlMs"] = 1;
-    try {
-      await withSilentConsole(async () => {
-        await (plugin as unknown as { setup: (c: unknown) => Promise<void> }).setup(ctx);
-        // First transform: gateway healthy, entry stored.
-        await catalogCallbacks[0]!(stubDraft().draft);
-        await new Promise((r) => setTimeout(r, 250));
-        // Gateway goes down only now: the next transform fails totally while
-        // a prior entry exists, arming the cooldown.
-        down = true;
-        await new Promise((r) => setTimeout(r, 10));
-        await catalogCallbacks[0]!(stubDraft().draft);
-        const afterArming = modelCalls;
-        assert.ok(afterArming >= 2, "the failing transform tries the network once");
-        await catalogCallbacks[0]!(stubDraft().draft);
-        assert.equal(modelCalls, afterArming, "a transform inside the cooldown must not refetch");
-      });
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-
-  it("throws on a synchronously refusing integration hook, catalog intact", async () => {
-    // A host whose integration.transform throws while registering must cost
-    // the plugin the connect action only — setup resolves, the catalog
-    // callback is registered, and the throw is warned, not propagated.
-    const restoreDisk = await isolateDisk();
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = stubFetch({ autoCombosHangs: false });
-    const catalogCallbacks: Array<(draft: unknown) => Promise<void>> = [];
-    const ctx = {
-      options: { baseURL: "https://gw.example.com", providerId: "staged-integ", apiKey: "k" },
-      catalog: {
-        transform: (cb: (draft: unknown) => Promise<void>) => {
-          catalogCallbacks.push(cb);
-          return Promise.resolve({ dispose: async () => {} });
-        },
-      },
-      integration: {
-        transform: () => {
-          throw new Error("host says no");
-        },
-      },
-    };
-    try {
-      await withSilentConsole(async () => {
-        await (plugin as unknown as { setup: (c: unknown) => Promise<void> }).setup(ctx);
-        assert.equal(catalogCallbacks.length, 1, "the catalog still registers");
-      });
-    } finally {
-      globalThis.fetch = origFetch;
-      restoreDisk();
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/timeouts-logger.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/timeouts-logger.test.ts
@@ -1,161 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import { parsePluginOptions, resolveTimeouts } from "../src/options.js";
-import { publishCatalog } from "../src/catalog.js";
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-
-function fakeDraft(): CatalogDraft {
-  const providers = new Map<string, ProviderV2Info>();
-  const models = new Map<string, ModelV2Info>();
-  return {
-    provider: {
-      list: () => [],
-      get: (id: string) => providers.get(id) as never,
-      update: (id: string, fn: (p: ProviderV2Info) => void) => {
-        const p = (providers.get(id) ?? { id }) as ProviderV2Info;
-        fn(p);
-        providers.set(id, p);
-      },
-      remove: () => {},
-    },
-    model: {
-      get: () => undefined,
-      update: (pid: string, mid: string, fn: (m: ModelV2Info) => void) => {
-        const k = pid + "/" + mid;
-        const m = (models.get(k) ?? { id: mid, providerID: pid }) as ModelV2Info;
-        fn(m);
-        models.set(k, m);
-      },
-      remove: () => {},
-      default: { get: () => undefined, set: () => {} },
-    },
-  } as CatalogDraft;
-}
-
-const BER = "https://gw.example.com";
-
-describe("plugin-v2 P2 parity: per-endpoint timeouts", () => {
-  it("timeoutMs default is 10s (v1 parity over the old 30s)", () => {
-    const opts = parsePluginOptions({ baseURL: BER });
-    assert.equal(opts.timeoutMs, 10000);
-  });
-
-  it("endpoint timeouts default to the global timeout", () => {
-    const opts = parsePluginOptions({ baseURL: BER });
-    assert.deepEqual(resolveTimeouts(opts), {
-      models: 10000,
-      combos: 10000,
-      autoCombos: 5000,
-      enrichment: 10000,
-    });
-  });
-
-  it("endpoint overrides win over the global timeout", () => {
-    const opts = parsePluginOptions({
-      baseURL: BER,
-      timeoutMs: 3000,
-      timeouts: { models: 1111, combos: 2222, autoCombos: 3333, enrichment: 4444 },
-    });
-    assert.deepEqual(resolveTimeouts(opts), {
-      models: 1111,
-      combos: 2222,
-      autoCombos: 3333,
-      enrichment: 4444,
-    });
-  });
-
-  it("publishCatalog forwards per-endpoint timeouts (slow models stub aborts at 20ms)", async () => {
-    const seen: number[] = [];
-    const t0 = Date.now();
-    const slowModels = async (_base: string, _key: string, timeoutMs?: number) => {
-      seen.push(timeoutMs ?? -1);
-      await new Promise<void>((_resolve, reject) => {
-        setTimeout(() => {
-          reject(Object.assign(new Error("This operation was aborted"), { name: "AbortError" }));
-        }, timeoutMs ?? 10000);
-      });
-      return [{ id: "m1" }];
-    };
-    const draft = fakeDraft();
-    const origWarn = console.warn;
-    console.warn = () => {};
-    try {
-      const res = await publishCatalog(
-        draft,
-        {
-          providerId: "omniroute",
-          baseURL: BER,
-          apiKey: "k",
-          // A per-endpoint value must win over the global one: the 20ms
-          // models budget is what this test asserts, not the 10s fallback.
-          timeoutMs: 10000,
-          timeouts: { models: 20, combos: 10000 },
-          modelCacheTtlMs: 300000,
-          usableOnly: false,
-        },
-        { fetcher: slowModels, combosFetcher: async () => [] }
-      );
-      assert.deepEqual(res, { models: 0, combos: 0, autoCombos: 0 });
-    } finally {
-      console.warn = origWarn;
-    }
-    assert.deepEqual(seen, [20]);
-    assert.ok(Date.now() - t0 < 2000, "slow fetcher must abort near 20ms, not hang");
-  });
-
-  it("logger option accepts logLevel plus startupDebug (default warn)", () => {
-    assert.equal(parsePluginOptions({ baseURL: BER }).logLevel, undefined);
-    assert.equal(parsePluginOptions({ baseURL: BER }).startupDebug, undefined);
-    assert.equal(parsePluginOptions({ baseURL: BER, logLevel: "debug" }).logLevel, "debug");
-    assert.equal(parsePluginOptions({ baseURL: BER, startupDebug: true }).startupDebug, true);
-    assert.throws(() => parsePluginOptions({ baseURL: BER, logLevel: "verbose" }));
-  });
-
-  it("startupDebug silences the boot line at default, shows it with startupDebug", async () => {
-    const plugin = (await import("../src/index.js")).default as unknown as {
-      setup: (ctx: unknown) => Promise<void>;
-    };
-    async function bootLogLines(options: Record<string, unknown>): Promise<string[]> {
-      const warns: string[] = [];
-      const origWarn = console.warn;
-      console.warn = (...args: unknown[]) => {
-        warns.push(String(args[0]));
-      };
-      const ctx = {
-        options,
-        catalog: {
-          transform: () => Promise.resolve({ dispose: async () => {} }),
-        },
-        integration: {
-          transform: () => Promise.resolve({ dispose: async () => {} }),
-        },
-      };
-      try {
-        await plugin.setup(ctx);
-      } finally {
-        console.warn = origWarn;
-      }
-      return warns;
-    }
-    const quiet = await bootLogLines({
-      baseURL: BER,
-      providerId: "p2-quiet",
-      apiKey: "k",
-    });
-    assert.ok(
-      !quiet.some((line) => line.includes("init providerId=")),
-      `boot line must stay silent by default, got: ${JSON.stringify(quiet)}`
-    );
-    const loud = await bootLogLines({
-      baseURL: BER,
-      providerId: "p2-loud",
-      apiKey: "k",
-      startupDebug: true,
-    });
-    assert.ok(
-      loud.some((line) => line.includes("init providerId=p2-loud")),
-      `boot line must show with startupDebug, got: ${JSON.stringify(loud)}`
-    );
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/usable-only.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/usable-only.test.ts
@@ -1,222 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import type { CatalogDraft } from "@opencode-ai/plugin/v2/promise";
-import type { ModelV2Info, ProviderV2Info } from "@opencode-ai/sdk/v2/types";
-import { publishCatalog } from "../src/catalog.js";
-import { parsePluginOptions } from "../src/options.js";
-
-function fakeDraft(): { models: Map<string, ModelV2Info>; draft: CatalogDraft } {
-  const providers = new Map<string, ProviderV2Info>();
-  const models = new Map<string, ModelV2Info>();
-  const draft = {
-    provider: {
-      list: () => [],
-      get: (id: string) => providers.get(id) as never,
-      update: (id: string, fn: (p: ProviderV2Info) => void) => {
-        const p = (providers.get(id) ?? { id }) as ProviderV2Info;
-        fn(p);
-        providers.set(id, p);
-      },
-      remove: () => {},
-    },
-    model: {
-      get: () => undefined,
-      update: (pid: string, mid: string, fn: (m: ModelV2Info) => void) => {
-        const k = pid + "/" + mid;
-        const m = (models.get(k) ?? { id: mid, providerID: pid }) as ModelV2Info;
-        fn(m);
-        models.set(k, m);
-      },
-      remove: () => {},
-      default: { get: () => undefined, set: () => {} },
-    },
-  } as CatalogDraft;
-  return { models, draft };
-}
-
-function silence() {
-  const origWarn = console.warn;
-  const origLog = console.log;
-  console.warn = () => {};
-  console.log = () => {};
-  return {
-    restore() {
-      console.warn = origWarn;
-      console.log = origLog;
-    },
-  };
-}
-function enrichmentOf(...pairs: Array<[string, string]>) {
-  return new Map(
-    pairs.map(([alias, canonical]) => [
-      `${alias}/model-x`,
-      { providerAlias: alias, providerCanonical: canonical },
-    ])
-  );
-}
-
-const baseOpts = {
-  providerId: "omniroute",
-  baseURL: "https://gw.example.com",
-  apiKey: "k",
-  timeoutMs: 1000,
-  modelCacheTtlMs: 300000,
-};
-
-describe("usableOnly option default", () => {
-  it("defaults to false", () => {
-    const opts = parsePluginOptions({ baseURL: "https://gw.example.com" });
-    assert.equal(opts.usableOnly, false);
-  });
-});
-
-describe("catalog usableOnly gating", () => {
-  const stubModels = async () => [
-    { id: "cc/keep-me" },
-    { id: "dead/drop-me" },
-    { id: "agentrouter/mystery" },
-  ];
-
-  it("usableOnly=false performs zero providers fetches", async () => {
-    const { draft, models } = fakeDraft();
-    let providersCalls = 0;
-    const res = await publishCatalog(
-      draft,
-      { ...baseOpts, usableOnly: false },
-      {
-        fetcher: stubModels,
-        combosFetcher: async () => [],
-        providersFetcher: async () => {
-          providersCalls += 1;
-          return [];
-        },
-      }
-    );
-    assert.equal(providersCalls, 0);
-    assert.equal(res.models, 3);
-    assert.ok(models.has("omniroute/cc/keep-me"));
-    assert.ok(models.has("omniroute/dead/drop-me"));
-  });
-
-  it("usableOnly=true drops known-but-not-provisioned models, keeps unknown prefixes", async () => {
-    const { draft, models } = fakeDraft();
-    const res = await publishCatalog(
-      draft,
-      {
-        ...baseOpts,
-        usableOnly: true,
-        enrichment: enrichmentOf(["cc", "claude"], ["dead", "legacy"]),
-      },
-      {
-        fetcher: stubModels,
-        combosFetcher: async () => [],
-        providersFetcher: async () => [
-          { id: "c1", provider: "claude", isActive: true, testStatus: "active" },
-        ],
-      }
-    );
-    assert.equal(res.models, 2);
-    assert.ok(models.has("omniroute/cc/keep-me"));
-    assert.ok(!models.has("omniroute/dead/drop-me"));
-    assert.ok(models.has("omniroute/agentrouter/mystery"));
-  });
-
-  it("usableOnly=true with providers soft-fail keeps the full catalog", async () => {
-    const { draft, models } = fakeDraft();
-    const res = await publishCatalog(
-      draft,
-      { ...baseOpts, usableOnly: true },
-      {
-        fetcher: stubModels,
-        combosFetcher: async () => [],
-        providersFetcher: async () => {
-          throw new Error("down");
-        },
-      }
-    );
-    assert.equal(res.models, 3);
-    assert.ok(models.has("omniroute/dead/drop-me"));
-  });
-
-  it("usableOnly=true filters combos by member usability", async () => {
-    const { draft, models } = fakeDraft();
-    const res = await publishCatalog(
-      draft,
-      {
-        ...baseOpts,
-        usableOnly: true,
-        enrichment: enrichmentOf(["cc", "claude"], ["dead", "legacy"]),
-      },
-      {
-        fetcher: async () => [{ id: "cc/x" }, { id: "dead/y" }],
-        combosFetcher: async () => [
-          { id: "good", name: "Good", models: [{ kind: "model", model: "cc/x" }] },
-          { id: "bad", name: "Bad", models: [{ kind: "model", model: "dead/y" }] },
-        ],
-        providersFetcher: async () => [
-          { id: "c1", provider: "claude", isActive: true, testStatus: "active" },
-        ],
-      }
-    );
-    assert.equal(res.combos, 1);
-    assert.ok(models.has("omniroute/good"));
-    assert.ok(!models.has("omniroute/bad"));
-  });
-
-  it("usableOnly=false issues no providers request through setup (gating)", async () => {
-    const seen: string[] = [];
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async (url: unknown) => {
-      const href = String(url);
-      seen.push(href);
-      if (href.includes("/api/combos/auto")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      if (href.includes("/api/combos")) {
-        return { ok: true, status: 200, statusText: "OK", json: async () => ({ combos: [] }) };
-      }
-      return {
-        ok: true,
-        status: 200,
-        statusText: "OK",
-        json: async () => ({ data: [{ id: "m1" }] }),
-      };
-    }) as typeof fetch;
-    const guard = silence();
-    try {
-      const plugin = (await import("../src/index.js")).default as unknown as {
-        setup: (ctx: unknown) => Promise<void>;
-      };
-      const catalogCallbacks: Array<(draft: unknown) => Promise<void>> = [];
-      await plugin.setup({
-        options: { baseURL: "https://gw.example.com", providerId: "usable-gate" },
-        catalog: {
-          transform: (cb: (draft: unknown) => Promise<void>) => {
-            catalogCallbacks.push(cb);
-            return Promise.resolve({ dispose: async () => {} });
-          },
-        },
-        integration: { transform: () => Promise.resolve({ dispose: async () => {} }) },
-      });
-      const { mkdtempSync } = await import("node:fs");
-      const { tmpdir } = await import("node:os");
-      const { join } = await import("node:path");
-      const prevDataDir = process.env.OPENCODE_DATA_DIR;
-      process.env.OPENCODE_DATA_DIR = mkdtempSync(join(tmpdir(), "omniroute-usable-"));
-      try {
-        const { draft } = fakeDraft();
-        await catalogCallbacks[0](draft as never);
-      } finally {
-        if (prevDataDir === undefined) delete process.env.OPENCODE_DATA_DIR;
-        else process.env.OPENCODE_DATA_DIR = prevDataDir;
-      }
-      assert.ok(
-        !seen.some((href) => href.includes("/api/providers")),
-        `no providers fetch expected, got: ${JSON.stringify(seen)}`
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      guard.restore();
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tests/warm-snapshot-identity.test.ts
+++ b/@omniroute/opencode-plugin-v2/tests/warm-snapshot-identity.test.ts
@@ -1,88 +0,0 @@
-import { describe, it } from "node:test";
-import assert from "node:assert/strict";
-import plugin from "../src/index.js";
-import { snapshotIdentityFingerprint, writeDiskSnapshot } from "../src/cache.js";
-
-/**
- * The recommended setup stores the gateway key in the host's credential store,
- * so the key the plugin ends up using is not the one its options carry. The
- * disk snapshot is keyed by that credential: reading it before the credential
- * is resolved looks up the wrong identity and throws away a usable catalog —
- * exactly when it is needed, on a cold start against an unreachable gateway.
- */
-describe("warm snapshot is read under the credential actually in use", () => {
-  it("serves the snapshot written for the host credential, gateway down", async () => {
-    const { mkdtempSync } = await import("node:fs");
-    const { tmpdir } = await import("node:os");
-    const { join } = await import("node:path");
-    const dir = mkdtempSync(join(tmpdir(), "omniroute-warm-id-"));
-    const prevDir = process.env.OPENCODE_DATA_DIR;
-    process.env.OPENCODE_DATA_DIR = dir;
-    const origFetch = globalThis.fetch;
-    globalThis.fetch = (async () => {
-      throw new Error("gateway unreachable");
-    }) as unknown as typeof fetch;
-    const warn = console.warn;
-    const log = console.log;
-    console.warn = () => {};
-    console.log = () => {};
-    try {
-      const baseURL = "https://gw.example.com";
-      const hostKey = "key-from-the-host-store";
-      await writeDiskSnapshot(
-        "warmid",
-        {
-          models: [{ id: "m-snap" }],
-          combos: [],
-          autoCombos: [],
-          providers: [],
-          fetchedAt: Date.now(),
-        } as never,
-        snapshotIdentityFingerprint(baseURL, hostKey, hostKey)
-      );
-
-      const published = new Map<string, Record<string, unknown>>();
-      const callbacks: Array<(draft: unknown) => Promise<void>> = [];
-      const registration = Promise.resolve({ dispose: async () => {} });
-      const ctx = {
-        options: { baseURL, providerId: "warmid", apiKey: "key-written-in-the-config" },
-        catalog: {
-          transform: (cb: (d: unknown) => Promise<void>) => {
-            callbacks.push(cb);
-            return registration;
-          },
-          reload: async () => {},
-        },
-        integration: {
-          transform: () => registration,
-          connection: {
-            active: async () => ({ type: "credential", id: "c", label: "l" }),
-            resolve: async () => ({ type: "key", key: hostKey }),
-          },
-        },
-      };
-      await (plugin as unknown as { setup: (c: unknown) => Promise<void> }).setup(ctx);
-      const draft = {
-        provider: { update: (_i: string, fn: (p: Record<string, unknown>) => void) => fn({}) },
-        model: {
-          update: (pid: string, mid: string, fn: (m: Record<string, unknown>) => void) => {
-            const e: Record<string, unknown> = { id: mid, providerID: pid };
-            fn(e);
-            published.set(`${pid}/${mid}`, e);
-          },
-        },
-      };
-      await callbacks[0]!(draft);
-      assert.ok(
-        [...published.keys()].some((k) => k.endsWith("/m-snap")),
-        `the snapshot must survive the credential switch, published: ${JSON.stringify([...published.keys()])}`
-      );
-    } finally {
-      globalThis.fetch = origFetch;
-      console.warn = warn;
-      console.log = log;
-      if (prevDir === undefined) delete process.env.OPENCODE_DATA_DIR;
-      else process.env.OPENCODE_DATA_DIR = prevDir;
-    }
-  });
-});
--- a/@omniroute/opencode-plugin-v2/tsconfig.json
+++ b/@omniroute/opencode-plugin-v2/tsconfig.json
@@ -1,24 +0,0 @@
-{
-  "compilerOptions": {
-    "target": "ES2022",
-    "module": "NodeNext",
-    "moduleResolution": "NodeNext",
-    "strict": true,
-    "declaration": true,
-    "outDir": "dist",
-    "skipLibCheck": true,
-    "isolatedModules": true,
-    "forceConsistentCasingInFileNames": true,
-    "types": [
-      "node"
-    ],
-    "lib": [
-      "ES2022"
-    ],
-    "esModuleInterop": true
-  },
-  "include": [
-    "src/**/*.ts",
-    "tests/**/*.ts"
-  ]
-}
--- a/@omniroute/opencode-plugin-v2/tsup.config.ts
+++ b/@omniroute/opencode-plugin-v2/tsup.config.ts
@@ -1,16 +0,0 @@
-import { defineConfig } from "tsup";
-
-export default defineConfig({
-  entry: ["src/index.ts"],
-  format: ["esm"],
-  dts: true,
-  clean: true,
-  sourcemap: false,
-  splitting: false,
-  treeshake: false,
-  target: "node22",
-  outDir: "dist",
-  minify: false,
-  cjsInterop: false,
-  noExternal: ["zod"],
-});
--- a/@omniroute/opencode-plugin/package-lock.json
+++ b/@omniroute/opencode-plugin/package-lock.json
@@ -1745,9 +1745,9 @@
      }
    },
    "node_modules/toml": {
-      "version": "4.3.0",
-      "resolved": "https://registry.npmjs.org/toml/-/toml-4.3.0.tgz",
-      "integrity": "sha512-lVb8X9BsPVuH0M4BKeS91tXAmJvCjQ5UIyAbQFaxkKGyUFK2RPkhwaFSQH8vbpl1d23eu/IBH+dwVMHWaq9A5A==",
+      "version": "4.1.1",
+      "resolved": "https://registry.npmjs.org/toml/-/toml-4.1.1.tgz",
+      "integrity": "sha512-EBJnVBr3dTXdA89WVFoAIPUqkBjxPMwRqsfuo1r240tKFHXv3zgca4+NJib/h6TyvGF7vOawz0jGuryJCdNHrw==",
      "dev": true,
      "license": "MIT",
      "engines": {
--- a/@omniroute/opencode-plugin/package.json
+++ b/@omniroute/opencode-plugin/package.json
@@ -23,7 +23,7 @@
  "scripts": {
    "build": "tsup",
    "clean": "rm -rf dist",
-    "test": "node --import tsx/esm --test tests/scaffold.test.ts tests/auth.test.ts tests/options-schema.test.ts tests/multi-instance.test.ts tests/fetch-interceptor.test.ts tests/telemetry.test.ts tests/provider.test.ts tests/gemini-sanitize.test.ts tests/combos.test.ts tests/config-shim.test.ts tests/features.test.ts tests/feature-defaults.test.ts tests/usable-combo.test.ts tests/disk-snapshot-perms.test.ts tests/fork-features.test.ts tests/auto-combo-context.test.ts tests/provider-id-routing.test.ts tests/management-read-token.test.ts tests/auto-sync.test.ts tests/model-allowlist.test.ts tests/log-level.test.ts tests/effort-tier-variants.test.ts tests/naming.test.ts tests/free-budget-magnitude.test.ts tests/models-fetcher.test.ts",
+    "test": "node --import tsx/esm --test tests/scaffold.test.ts tests/auth.test.ts tests/options-schema.test.ts tests/multi-instance.test.ts tests/fetch-interceptor.test.ts tests/provider.test.ts tests/gemini-sanitize.test.ts tests/combos.test.ts tests/config-shim.test.ts tests/features.test.ts tests/feature-defaults.test.ts tests/usable-combo.test.ts tests/disk-snapshot-perms.test.ts tests/fork-features.test.ts tests/auto-combo-context.test.ts tests/provider-id-routing.test.ts tests/management-read-token.test.ts tests/auto-sync.test.ts tests/model-allowlist.test.ts tests/log-level.test.ts tests/effort-tier-variants.test.ts tests/naming.test.ts tests/free-budget-magnitude.test.ts",
    "prepublishOnly": "npm run clean && npm run build && npm test"
  },
  "keywords": [
@@ -68,7 +68,6 @@
    "typescript": "^5.9.3"
  },
  "overrides": {
-    "esbuild": "^0.28.1",
-    "toml": "^4.1.2"
+    "esbuild": "^0.28.1"
  }
 }
--- a/@omniroute/opencode-plugin/src/index.ts
+++ b/@omniroute/opencode-plugin/src/index.ts
@@ -75,7 +75,6 @@ import {
  AUTO_VARIANT_DESCRIPTIONS,
  type FreeModelFreeType,
 } from "./naming.js";
-import { applyOmniRouteInferenceTelemetry } from "./telemetry.js";

 /**
 * Minimal leveled logger sink accepted by the default fetchers and the static
@@ -1200,7 +1199,7 @@ export type OmniRouteModelsFetcher = (
 export const defaultOmniRouteModelsFetcher: OmniRouteModelsFetcher = async (
  baseURL,
  apiKey,
-  timeoutMs = 30_000
+  timeoutMs = 10_000
 ) => {
  if (!apiKey) throw new Error("@omniroute/opencode-plugin: apiKey required to fetch /v1/models");
  if (!baseURL) throw new Error("@omniroute/opencode-plugin: baseURL required to fetch /v1/models");
@@ -1222,12 +1221,9 @@ export const defaultOmniRouteModelsFetcher: OmniRouteModelsFetcher = async (
      signal: controller.signal,
    });
    if (!res.ok) {
-      const err = new Error(
+      throw new Error(
        `@omniroute/opencode-plugin: GET ${url} failed: ${res.status} ${res.statusText}`
-      ) as Error & { statusCode: number; status: number };
-      err.statusCode = res.status;
-      err.status = res.status;
-      throw err;
+      );
    }
    const body = (await res.json()) as unknown;
    const rawList: unknown[] = Array.isArray(body)
@@ -3481,7 +3477,7 @@ export function createOmniRouteProviderHook(

      // ── Combo LCD across nested combo-refs (T-NN) ───────────────────────
      // Combos can nest other combos via `kind: "combo-ref"` members
-      // (e.g. MASTER-LIGHT contains LEGACY, KIRO, Opecode Zen FREE). The
+      // (e.g. MASTER-LIGHT contains OldLLM, KIRO, Opecode Zen FREE). The
      // nested combo's own `limit.context` is computed below in this same
      // loop, so we need a fixpoint iteration: if a combo-ref points at a
      // combo not yet processed, defer this combo and try again after the
@@ -3770,8 +3766,6 @@ export function createOmniRouteFetchInterceptor(config: {
    baseOrigin = baseUrl.origin;
    const basePath = ensureV1Suffix(baseUrl.pathname);
    inferencePaths.add(`${basePath}/chat/completions`);
-    inferencePaths.add(`${basePath}/responses`);
-    inferencePaths.add(`${basePath}/messages`);
    inferencePaths.add(`${basePath}/models`);
  } catch {
    // Credential-attached base URLs are not schema-validated. A malformed
@@ -3815,7 +3809,7 @@ export function createOmniRouteFetchInterceptor(config: {
      headers.set("Content-Type", "application/json");
    }

-    return applyOmniRouteInferenceTelemetry(await fetch(input, { ...init, headers }));
+    return fetch(input, { ...init, headers });
  };
 }

@@ -4501,7 +4495,7 @@ export function buildStaticProviderEntry(
  // ── Combo LCD across nested combo-refs (T-NN mirror) ─────────────────
  // Mirror of the dynamic-catalog fixpoint iteration: combos can nest
  // other combos via `kind: "combo-ref"` members (e.g. MASTER-LIGHT
-  // contains LEGACY, KIRO, Opecode Zen FREE). The nested combo's own
+  // contains OldLLM, KIRO, Opecode Zen FREE). The nested combo's own
  // capabilities and limits are computed in this same loop, so we need
  // a fixpoint pass: if a combo-ref points at a combo not yet processed,
  // defer this combo and try again after the sibling combos catch up.
@@ -5404,7 +5398,7 @@ export function createOmniRouteConfigHook(
        // exact warn message so per-endpoint fallbacks are preserved.
        const doModels = async (): Promise<void> => {
          try {
-            localRawModels = await fetcher(baseURL, apiKey, 30_000);
+            localRawModels = await fetcher(baseURL, apiKey, 10_000);
          } catch (err) {
            logAt(
              "error",
--- a/@omniroute/opencode-plugin/src/telemetry.ts
+++ b/@omniroute/opencode-plugin/src/telemetry.ts
@@ -1,249 +0,0 @@
-/**
- * Map gateway-reported OmniRoute inference telemetry onto the JSON/SSE
- * payload OpenCode already consumes. Prefer headers / usage fields from the
- * gateway. Never invent tok/s from tokens / latency (that includes TTFT).
- */
-export type OmniRouteInferenceTelemetry = {
-  costUsd?: number;
-  tokensIn?: number;
-  tokensOut?: number;
-  tokensPerSecond?: number;
-  ttftMs?: number;
-  latencyMs?: number;
-  model?: string;
-  provider?: string;
-};
-
-const HEADER = {
-  cost: "x-omniroute-response-cost",
-  tokensIn: "x-omniroute-tokens-in",
-  tokensOut: "x-omniroute-tokens-out",
-  tokensPerSecond: "x-omniroute-tokens-per-second",
-  ttftMs: "x-omniroute-ttft-ms",
-  latencyMs: "x-omniroute-latency-ms",
-  model: "x-omniroute-model",
-  provider: "x-omniroute-provider",
-} as const;
-
-function isRecord(value: unknown): value is Record<string, unknown> {
-  return typeof value === "object" && value !== null && !Array.isArray(value);
-}
-
-function readFiniteNumber(raw: string | null): number | undefined {
-  if (raw == null) return undefined;
-  const trimmed = raw.trim();
-  if (trimmed === "") return undefined;
-  const parsed = Number(trimmed);
-  return Number.isFinite(parsed) ? parsed : undefined;
-}
-
-function readPositiveNumber(raw: string | null): number | undefined {
-  const parsed = readFiniteNumber(raw);
-  if (parsed === undefined || parsed <= 0) return undefined;
-  return parsed;
-}
-
-function readNonNegativeInt(raw: string | null): number | undefined {
-  const parsed = readFiniteNumber(raw);
-  if (parsed === undefined || parsed < 0) return undefined;
-  return Math.round(parsed);
-}
-
-function readToken(raw: string | null): string | undefined {
-  if (raw == null) return undefined;
-  const trimmed = raw.trim();
-  return trimmed === "" ? undefined : trimmed;
-}
-
-export function parseOmniRouteInferenceTelemetry(headers: Headers): OmniRouteInferenceTelemetry {
-  const out: OmniRouteInferenceTelemetry = {};
-  const cost = readFiniteNumber(headers.get(HEADER.cost));
-  if (cost !== undefined && cost >= 0) out.costUsd = cost;
-  const tokensIn = readNonNegativeInt(headers.get(HEADER.tokensIn));
-  if (tokensIn !== undefined) out.tokensIn = tokensIn;
-  const tokensOut = readNonNegativeInt(headers.get(HEADER.tokensOut));
-  if (tokensOut !== undefined) out.tokensOut = tokensOut;
-  const tps = readPositiveNumber(headers.get(HEADER.tokensPerSecond));
-  if (tps !== undefined) out.tokensPerSecond = tps;
-  const ttft = readPositiveNumber(headers.get(HEADER.ttftMs));
-  if (ttft !== undefined) out.ttftMs = ttft;
-  const latency = readPositiveNumber(headers.get(HEADER.latencyMs));
-  if (latency !== undefined) out.latencyMs = latency;
-  const model = readToken(headers.get(HEADER.model));
-  if (model) out.model = model;
-  const provider = readToken(headers.get(HEADER.provider));
-  if (provider) out.provider = provider;
-  return out;
-}
-
-function telemetryFromUsage(usage: Record<string, unknown>): OmniRouteInferenceTelemetry {
-  const out: OmniRouteInferenceTelemetry = {};
-  const tps = usage.tokens_per_second;
-  if (typeof tps === "number" && Number.isFinite(tps) && tps > 0) {
-    out.tokensPerSecond = tps;
-  }
-  const ttft = usage.ttft_ms;
-  if (typeof ttft === "number" && Number.isFinite(ttft) && ttft > 0) {
-    out.ttftMs = ttft;
-  }
-  return out;
-}
-
-function mergeTelemetry(
-  base: OmniRouteInferenceTelemetry,
-  extra: OmniRouteInferenceTelemetry,
-): OmniRouteInferenceTelemetry {
-  return {
-    ...base,
-    ...Object.fromEntries(Object.entries(extra).filter(([, value]) => value !== undefined)),
-  };
-}
-
-function isInferencePayload(payload: Record<string, unknown>): boolean {
-  return (
-    isRecord(payload.usage) ||
-    Array.isArray(payload.choices) ||
-    payload.object === "chat.completion" ||
-    payload.object === "response" ||
-    payload.type === "message" ||
-    Array.isArray(payload.output)
-  );
-}
-
-function attachToUsage(
-  usage: Record<string, unknown>,
-  telemetry: OmniRouteInferenceTelemetry,
-): Record<string, unknown> {
-  const next = { ...usage };
-  if (
-    telemetry.tokensPerSecond !== undefined &&
-    (typeof next.tokens_per_second !== "number" || next.tokens_per_second <= 0)
-  ) {
-    next.tokens_per_second = telemetry.tokensPerSecond;
-  }
-  if (telemetry.ttftMs !== undefined && (typeof next.ttft_ms !== "number" || next.ttft_ms <= 0)) {
-    next.ttft_ms = telemetry.ttftMs;
-  }
-  if (telemetry.costUsd !== undefined && typeof next.cost !== "number") {
-    next.cost = telemetry.costUsd;
-  }
-  return next;
-}
-
-export function attachOmniRouteTelemetryToPayload(
-  payload: unknown,
-  telemetry: OmniRouteInferenceTelemetry,
-): unknown {
-  if (!isRecord(payload) || !isInferencePayload(payload)) {
-    return payload;
-  }
-  const next: Record<string, unknown> = { ...payload };
-  if (telemetry.model) {
-    next.model = telemetry.model;
-  }
-  if (isRecord(next.usage)) {
-    next.usage = attachToUsage(next.usage, mergeTelemetry(telemetry, telemetryFromUsage(next.usage)));
-  }
-  if (isRecord(next.response) && isRecord(next.response.usage)) {
-    next.response = {
-      ...next.response,
-      usage: attachToUsage(
-        next.response.usage,
-        mergeTelemetry(telemetry, telemetryFromUsage(next.response.usage)),
-      ),
-    };
-  }
-  return next;
-}
-
-export function attachOmniRouteTelemetryToSseLine(
-  line: string,
-  telemetry: OmniRouteInferenceTelemetry,
-): string {
-  const trimmed = line.trim();
-  if (!trimmed.startsWith("data:")) {
-    return line;
-  }
-  const jsonText = trimmed.slice("data:".length).trim();
-  if (!jsonText.startsWith("{")) {
-    return line;
-  }
-  try {
-    const parsed = JSON.parse(jsonText) as unknown;
-    const updated = attachOmniRouteTelemetryToPayload(parsed, telemetry);
-    if (updated === parsed) {
-      return line;
-    }
-    const prefix = line.slice(0, line.indexOf(jsonText));
-    const suffix = line.endsWith("\r") ? "\r" : "";
-    return `${prefix}${JSON.stringify(updated)}${suffix}`;
-  } catch {
-    return line;
-  }
-}
-
-export async function applyOmniRouteInferenceTelemetry(response: Response): Promise<Response> {
-  const telemetry = parseOmniRouteInferenceTelemetry(response.headers);
-  const contentType = response.headers.get("content-type") ?? "";
-  if (contentType.includes("text/event-stream") && response.body) {
-    return new Response(mapSseBody(response.body, telemetry), {
-      status: response.status,
-      statusText: response.statusText,
-      headers: response.headers,
-    });
-  }
-  if (!contentType.includes("json")) {
-    return response;
-  }
-  const text = await response.text();
-  let parsed: unknown;
-  try {
-    parsed = JSON.parse(text);
-  } catch {
-    return new Response(text, {
-      status: response.status,
-      statusText: response.statusText,
-      headers: response.headers,
-    });
-  }
-  const next = attachOmniRouteTelemetryToPayload(parsed, telemetry);
-  if (next === parsed) {
-    return new Response(text, {
-      status: response.status,
-      statusText: response.statusText,
-      headers: response.headers,
-    });
-  }
-  return new Response(JSON.stringify(next), {
-    status: response.status,
-    statusText: response.statusText,
-    headers: response.headers,
-  });
-}
-
-function mapSseBody(
-  body: ReadableStream<Uint8Array>,
-  telemetry: OmniRouteInferenceTelemetry,
-): ReadableStream<Uint8Array> {
-  const decoder = new TextDecoder();
-  const encoder = new TextEncoder();
-  let pending = "";
-  let live = { ...telemetry };
-  return body.pipeThrough(
-    new TransformStream<Uint8Array, Uint8Array>({
-      transform(chunk, controller) {
-        pending += decoder.decode(chunk, { stream: true });
-        const lines = pending.split("\n");
-        pending = lines.pop() ?? "";
-        for (const line of lines) {
-          controller.enqueue(encoder.encode(`${attachOmniRouteTelemetryToSseLine(line, live)}\n`));
-        }
-      },
-      flush(controller) {
-        if (pending.length > 0) {
-          controller.enqueue(encoder.encode(attachOmniRouteTelemetryToSseLine(pending, live)));
-        }
-      },
-    }),
-  );
-}
--- a/@omniroute/opencode-plugin/tests/combos.test.ts
+++ b/@omniroute/opencode-plugin/tests/combos.test.ts
@@ -641,13 +641,13 @@ test("models(): combos fetcher receives the resolved baseURL + apiKey", async ()

 test("models(): nested combo-ref context is the min of nested + raw members", async () => {
  // Top-level combo MASTER-LIGHT has 1 raw model (claude-primary, 200k)
-  // and 2 combo-refs: LEGACY (8k member) and KIRO (32k member). The OLD
+  // and 2 combo-refs: OldLLM (8k member) and KIRO (32k member). The OLD
  // plugin would advertise 200k (only the raw model); the fix should
  // make it advertise 8k (the bottleneck across the member graph).
  const modelsFetcher = stubModelsFetcher([
    MODEL_PRIMARY,
    {
-      id: "legacy-member-1",
+      id: "oldllm-member-1",
      context_length: 8_000,
      max_output_tokens: 4_000,
      capabilities: {
@@ -677,9 +677,9 @@ test("models(): nested combo-ref context is the min of nested + raw members", as
  ]);
  const combosFetcher = stubCombosFetcher([
    {
-      id: "legacy",
-      name: "LEGACY",
-      models: [{ id: "s1", kind: "model", model: "legacy-member-1", weight: 100 }],
+      id: "oldllm",
+      name: "OldLLM",
+      models: [{ id: "s1", kind: "model", model: "oldllm-member-1", weight: 100 }],
    },
    {
      id: "kiro",
@@ -691,7 +691,7 @@ test("models(): nested combo-ref context is the min of nested + raw members", as
      name: "MASTER-LIGHT",
      models: [
        { id: "r1", kind: "model", model: "claude-primary", weight: 50 },
-        { id: "r2", kind: "combo-ref", comboName: "LEGACY", weight: 25 },
+        { id: "r2", kind: "combo-ref", comboName: "OldLLM", weight: 25 },
        { id: "r3", kind: "combo-ref", comboName: "KIRO", weight: 25 },
      ],
    },
@@ -706,6 +706,6 @@ test("models(): nested combo-ref context is the min of nested + raw members", as
  assert.equal(
    masterLight.limit.context,
    8_000,
-    `expected 8_000 (LEGACY bottleneck), got ${masterLight.limit.context}`
+    `expected 8_000 (OldLLM bottleneck), got ${masterLight.limit.context}`
  );
 });
--- a/@omniroute/opencode-plugin/tests/models-fetcher.test.ts
+++ b/@omniroute/opencode-plugin/tests/models-fetcher.test.ts
@@ -1,45 +0,0 @@
-import test from "node:test";
-import assert from "node:assert/strict";
-import { defaultOmniRouteModelsFetcher } from "../src/index.js";
-
-test("defaultOmniRouteModelsFetcher attaches statusCode on HTTP 401", async () => {
-  const original = globalThis.fetch;
-  globalThis.fetch = (async () =>
-    new Response(JSON.stringify({ error: "authentication expired" }), {
-      status: 401,
-      statusText: "Unauthorized",
-    })) as typeof fetch;
-  try {
-    await assert.rejects(
-      () => defaultOmniRouteModelsFetcher("https://gateway.example/v1", "test-key"),
-      (err: unknown) => {
-        assert.ok(err instanceof Error);
-        const rec = err as Error & { statusCode?: number; status?: number };
-        assert.equal(rec.statusCode, 401);
-        assert.equal(rec.status, 401);
-        assert.match(rec.message, /401/);
-        return true;
-      },
-    );
-  } finally {
-    globalThis.fetch = original;
-  }
-});
-
-test("defaultOmniRouteModelsFetcher default timeout is 30s", async () => {
-  const original = globalThis.fetch;
-  let signal: AbortSignal | undefined;
-  globalThis.fetch = (async (_input, init) => {
-    signal = init?.signal ?? undefined;
-    return new Response(JSON.stringify({ object: "list", data: [] }), {
-      status: 200,
-      headers: { "Content-Type": "application/json" },
-    });
-  }) as typeof fetch;
-  try {
-    await defaultOmniRouteModelsFetcher("https://gateway.example/v1", "test-key");
-    assert.equal(signal instanceof AbortSignal, true);
-  } finally {
-    globalThis.fetch = original;
-  }
-});
--- a/@omniroute/opencode-plugin/tests/telemetry.test.ts
+++ b/@omniroute/opencode-plugin/tests/telemetry.test.ts
@@ -1,103 +0,0 @@
-import test from "node:test";
-import assert from "node:assert/strict";
-import {
-  applyOmniRouteInferenceTelemetry,
-  attachOmniRouteTelemetryToPayload,
-  attachOmniRouteTelemetryToSseLine,
-  parseOmniRouteInferenceTelemetry,
-} from "../src/telemetry.js";
-
-test("parseOmniRouteInferenceTelemetry: copies cost, tokens, tok/s, winning model", () => {
-  const headers = new Headers({
-    "X-OmniRoute-Response-Cost": "0.0123",
-    "X-OmniRoute-Tokens-In": "10",
-    "X-OmniRoute-Tokens-Out": "200",
-    "X-OmniRoute-Tokens-Per-Second": "100.5",
-    "X-OmniRoute-Ttft-Ms": "300",
-    "X-OmniRoute-Latency-Ms": "2300",
-    "X-OmniRoute-Model": "winner-model",
-    "X-OmniRoute-Provider": "openai",
-  });
-  const got = parseOmniRouteInferenceTelemetry(headers);
-  assert.equal(got.costUsd, 0.0123);
-  assert.equal(got.tokensIn, 10);
-  assert.equal(got.tokensOut, 200);
-  assert.equal(got.tokensPerSecond, 100.5);
-  assert.equal(got.ttftMs, 300);
-  assert.equal(got.model, "winner-model");
-  assert.equal(got.provider, "openai");
-});
-
-test("parseOmniRouteInferenceTelemetry: omits tok/s when header missing (do not invent from latency)", () => {
-  const headers = new Headers({
-    "X-OmniRoute-Tokens-Out": "200",
-    "X-OmniRoute-Latency-Ms": "2000",
-  });
-  const got = parseOmniRouteInferenceTelemetry(headers);
-  assert.equal(got.tokensPerSecond, undefined);
-  assert.equal(got.tokensOut, 200);
-  const payload = attachOmniRouteTelemetryToPayload(
-    { object: "chat.completion", usage: { prompt_tokens: 10, completion_tokens: 200 } },
-    got,
-  ) as { usage: { tokens_per_second?: number } };
-  assert.equal(payload.usage.tokens_per_second, undefined);
-});
-
-test("attachOmniRouteTelemetryToPayload: writes usage.tokens_per_second and winning model", () => {
-  const got = attachOmniRouteTelemetryToPayload(
-    {
-      object: "chat.completion",
-      model: "combo/auto",
-      usage: { prompt_tokens: 10, completion_tokens: 200 },
-    },
-    { tokensPerSecond: 80, ttftMs: 250, costUsd: 0, model: "gpt-winner" },
-  ) as {
-    model: string;
-    usage: { tokens_per_second: number; ttft_ms: number; cost: number };
-  };
-  assert.equal(got.model, "gpt-winner");
-  assert.equal(got.usage.tokens_per_second, 80);
-  assert.equal(got.usage.ttft_ms, 250);
-  assert.equal(got.usage.cost, 0);
-});
-
-test("attachOmniRouteTelemetryToPayload: does not mutate /v1/models catalog JSON", () => {
-  const catalog = { object: "list", data: [{ id: "m1" }] };
-  const got = attachOmniRouteTelemetryToPayload(catalog, {
-    tokensPerSecond: 99,
-    model: "should-not-apply",
-  });
-  assert.deepEqual(got, catalog);
-});
-
-test("attachOmniRouteTelemetryToSseLine: patches terminal usage data line", () => {
-  const line =
-    'data: {"object":"chat.completion.chunk","usage":{"completion_tokens":200}}';
-  const got = attachOmniRouteTelemetryToSseLine(line, { tokensPerSecond: 50 });
-  assert.match(got, /"tokens_per_second":50/);
-  assert.match(got, /^data: /);
-});
-
-test("applyOmniRouteInferenceTelemetry: JSON response gets header tok/s", async () => {
-  const response = new Response(
-    JSON.stringify({
-      object: "chat.completion",
-      model: "combo/auto",
-      usage: { prompt_tokens: 1, completion_tokens: 20 },
-    }),
-    {
-      headers: {
-        "Content-Type": "application/json",
-        "X-OmniRoute-Tokens-Per-Second": "40",
-        "X-OmniRoute-Model": "winner",
-      },
-    },
-  );
-  const next = await applyOmniRouteInferenceTelemetry(response);
-  const body = JSON.parse(await next.text()) as {
-    model: string;
-    usage: { tokens_per_second: number };
-  };
-  assert.equal(body.model, "winner");
-  assert.equal(body.usage.tokens_per_second, 40);
-});
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -56,9 +56,9 @@ Repository map and Reference Documentation sections below.
 | Translators   | `open-sse/translator/`  | Format conversion (OpenAI↔Claude↔Gemini)                                                                                                                                  |
 | Transformer   | `open-sse/transformer/` | Responses API ↔ Chat Completions                                                                                                                                          |
 | Services      | `open-sse/services/`    | Combo routing, rate limits, caching, etc                                                                                                                                  |
-| Database      | `src/lib/db/`           | SQLite domain modules (169 migrations)                                                                                                                                    |
+| Database      | `src/lib/db/`           | SQLite domain modules (160 migrations)                                                                                                                                    |
 | Domain/Policy | `src/domain/`           | Policy engine, cost rules, fallback logic                                                                                                                                 |
-| MCP Server    | `open-sse/mcp-server/`  | 110 tools (45 canonical + memory/skill/GitHub/pool/gamification/plugin/Notion/Obsidian/local-corpus/RTK modules), 3 transports (stdio / SSE / Streamable HTTP), 33 scopes |
+| MCP Server    | `open-sse/mcp-server/`  | 110 tools (44 canonical + memory/skill/GitHub/pool/gamification/plugin/Notion/Obsidian/local-corpus/RTK modules), 3 transports (stdio / SSE / Streamable HTTP), 33 scopes |
 | A2A Server    | `src/lib/a2a/`          | JSON-RPC 2.0 agent protocol                                                                                                                                               |
 | Skills        | `src/lib/skills/`       | Extensible skill framework                                                                                                                                                |
 | Memory        | `src/lib/memory/`       | Persistent conversational memory                                                                                                                                          |
@@ -83,7 +83,7 @@ Client → /v1/chat/completions (Next.js route)

 API routes follow a consistent pattern: `Route → CORS preflight → Zod body validation → Optional auth (extractApiKey/isValidApiKey) → API key policy enforcement → Handler delegation (open-sse)`. No global Next.js middleware — interception is route-specific.

-**Combo routing** (`open-sse/services/combo.ts`): 19 public strategies (priority, weighted, fill-first, round-robin, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline). Each target calls `handleSingleModel()` which wraps `handleChatCore()` with per-target error handling and circuit breaker checks. The `fusion` strategy is the exception: it fans out to a panel of models in parallel, then a judge model synthesizes one final answer (`open-sse/services/fusion.ts`). See `docs/routing/AUTO-COMBO.md` for the 16-factor Auto-Combo scoring + the full strategy table and `docs/architecture/RESILIENCE_GUIDE.md` for the 3 resilience layers.
+**Combo routing** (`open-sse/services/combo.ts`): 19 public strategies (priority, weighted, fill-first, round-robin, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline). Each target calls `handleSingleModel()` which wraps `handleChatCore()` with per-target error handling and circuit breaker checks. The `fusion` strategy is the exception: it fans out to a panel of models in parallel, then a judge model synthesizes one final answer (`open-sse/services/fusion.ts`). See `docs/routing/AUTO-COMBO.md` for the 15-factor Auto-Combo scoring + the full strategy table and `docs/architecture/RESILIENCE_GUIDE.md` for the 3 resilience layers.

 ---

@@ -110,36 +110,26 @@ upstream/service level, so one unhealthy provider does not slow down every reque
 - Shared wrappers: `open-sse/services/accountFallback.ts`
 - Persisted state table: `domain_circuit_breakers`

-**States** (4 — `src/shared/utils/circuitBreaker.ts`):
+**States**:

 - `CLOSED`: normal traffic is allowed.
- `DEGRADED`: early-warning band — failures crossed the degradation threshold but not the
-  breaker threshold yet; traffic still flows, dashboards show the warning.
 - `OPEN`: provider is temporarily blocked; callers get a provider-circuit-open response
  or combo routing skips to another target.
 - `HALF_OPEN`: reset timeout has elapsed; allow a probe request. Success closes the
  breaker, failure opens it again.

-**Defaults** (`open-sse/config/constants.ts` → `PROVIDER_PROFILES`, consumed via
-`DEFAULT_RESILIENCE_SETTINGS.providerBreaker` in `src/lib/resilience/settings.ts` →
-`getCircuitBreaker(provider, …)` in `src/sse/handlers/chatHelpers.ts`). The whole-provider
-breaker runs on `circuitBreakerThreshold` / `circuitBreakerReset`:
+**Defaults** (`open-sse/config/constants.ts` → `PROVIDER_PROFILES`). Two thresholds live side by
+side — do not confuse them:

-| Profile | degrades at | opens at (`circuitBreakerThreshold`) | reset (`circuitBreakerReset`) |
-| ------- | ----------: | -----------------------------------: | ----------------------------: |
-| OAuth   |         `5` |                                  `8` |                         `60s` |
-| API key |         `7` |                                 `12` |                         `30s` |
-| Local   |   (derived) |                                  `2` |                         `15s` |
+| Profile | `providerFailureThreshold` (whole provider) | `providerCooldownMs` | `circuitBreakerThreshold` (one connection) | `circuitBreakerReset` |
+| ------- | ------------------------------------------: | -------------------: | -----------------------------------------: | --------------------: |
+| OAuth   |                                        `10` |               `5min` |                                        `8` |                 `60s` |
+| API key |                                        `15` |              `10min` |                                       `12` |                 `30s` |
+| Local   |                                         `2` |               `1min` |                                        `2` |                 `15s` |

-`PROVIDER_PROFILES` also defines `providerFailureThreshold` (10/15/2),
-`providerFailureWindowMs` (15/30/5 min) and `providerCooldownMs` (5/10/1 min): these power the
-**window gate of the opt-in global Provider Cooldown** (`PROVIDER_COOLDOWN_ENABLED`, default
-off) — a provider-level entry in `open-sse/services/providerCooldownTracker.ts` only counts as
-cooling after `providerFailureThreshold` failures inside `providerFailureWindowMs`, and then
-cools for `providerCooldownMs`. They are NOT the live breaker's thresholds — do not tune them
-expecting breaker behavior. Every default is overridable through the
-`OMNIROUTE_PROVIDER_BREAKER_*` and `OMNIROUTE_CIRCUIT_BREAKER_*` env vars; the
-runtime-accurate reference table lives in `docs/architecture/RESILIENCE_GUIDE.md`.
+The provider-level thresholds were scaled up for deployments with 500+ connections (OAuth was
+`3`, API key was `5`); every default is overridable through the `OMNIROUTE_PROVIDER_BREAKER_*`
+and `OMNIROUTE_CIRCUIT_BREAKER_*` env vars.

 Only provider-level failure statuses should trip the provider breaker:

@@ -252,7 +242,7 @@ Read the nearest `AGENTS.md` and the linked deep-dive before making a non-trivia
 | Streaming request handling         | `open-sse/handlers/`                                    | [`docs/architecture/ARCHITECTURE.md`](docs/architecture/ARCHITECTURE.md)                                                                         |
 | Provider execution and translation | `open-sse/executors/`, `open-sse/translator/`           | [`docs/architecture/CODEBASE_DOCUMENTATION.md`](docs/architecture/CODEBASE_DOCUMENTATION.md)                                                     |
 | Routing and resilience             | `open-sse/services/`                                    | [`open-sse/services/AGENTS.md`](open-sse/services/AGENTS.md), [`docs/routing/AUTO-COMBO.md`](docs/routing/AUTO-COMBO.md)                         |
-| Database and migrations            | `src/lib/db/`, `src/lib/db/migrations/`                 | [`src/lib/db/AGENTS.md`](src/lib/db/AGENTS.md)                                                                                                   |
+| Database and migrations            | `src/lib/db/`, `db/migrations/`                         | [`src/lib/db/AGENTS.md`](src/lib/db/AGENTS.md)                                                                                                   |
 | Domain policy                      | `src/domain/`                                           | [`docs/architecture/ARCHITECTURE.md`](docs/architecture/ARCHITECTURE.md)                                                                         |
 | MCP and A2A                        | `open-sse/mcp-server/`, `src/lib/a2a/`                  | [`docs/frameworks/MCP-SERVER.md`](docs/frameworks/MCP-SERVER.md), [`docs/frameworks/A2A-SERVER.md`](docs/frameworks/A2A-SERVER.md)               |
 | Agent features                     | `src/lib/{acp,memory,skills,cloudAgent}/`               | [`docs/frameworks/AGENT_PROTOCOLS_GUIDE.md`](docs/frameworks/AGENT_PROTOCOLS_GUIDE.md), [`docs/frameworks/SKILLS.md`](docs/frameworks/SKILLS.md) |
@@ -264,13 +254,13 @@ Read the nearest `AGENTS.md` and the linked deep-dive before making a non-trivia
 ## File placement & repo-root hygiene

 - **Test files**: ALL unit tests, integration tests, ecosystem tests, or Vitest files MUST strictly be placed within the `tests/` directory (e.g., `tests/unit/`, `tests/integration/`). NEVER create test files in the project root (`/`).
- **Scripts and utilities**: ALL maintenance, debugging, generation, or experimental scripts (`.cjs`, `.mjs`, `.js`, `.ts`) MUST be placed strictly inside one of the `scripts/` subfolders (`build/`, `dev/`, `check/`, `docs/`, `i18n/`, `ad-hoc/`, `quality/`, `release/`, `ci/`, `ops/`, `perf/`, `research/`, `sre/`, `vps/`, `homolog/`, `packs/`, `skills/`, `test/`, `cli/`, `compression/`, `compression-eval/`, `devin-bridge/`, `docker/`, `features/`, `router-eval/`). One-shot or experimental code goes under `scripts/ad-hoc/`. NEVER dump loose scripts in the project root (`/`) or the top-level `scripts/` folder.
+- **Scripts and utilities**: ALL maintenance, debugging, generation, or experimental scripts (`.cjs`, `.mjs`, `.js`, `.ts`) MUST be placed strictly inside one of the `scripts/` subfolders (`build/`, `dev/`, `check/`, `docs/`, `i18n/`, `ad-hoc/`, `quality/`, `release/`, `ci/`, `ops/`, `perf/`, `research/`, `sre/`, `vps/`, `homolog/`, `raycast/`, `skills/`, `test/`, `cli/`, `compression/`, `compression-eval/`, `devin-bridge/`, `docker/`, `features/`, `router-eval/`). One-shot or experimental code goes under `scripts/ad-hoc/`. NEVER dump loose scripts in the project root (`/`) or the top-level `scripts/` folder.

 **The project root MUST ONLY contain:**

 - Configuration files (`vitest.config.ts`, `next.config.mjs`, `eslint.config.mjs`, `tsconfig*.json`, `playwright.config.ts`, `prettier.config.mjs`, `postcss.config.mjs`, `sonar-project.properties`, `fly.toml`, `docker-compose*.yml`, `Dockerfile`)
 - Dependency files (`package.json`, `package-lock.json`)
- Documentation files (`README.md`, `CHANGELOG.md`, `ROADMAP.md`, `LICENSE`, `AGENTS.md`, `CLAUDE.md`, `GEMINI.md`, `CONTRIBUTING.md`, `SECURITY.md`, `CODE_OF_CONDUCT.md`, `llm.txt`)
+- Documentation files (`README.md`, `CHANGELOG.md`, `ROADMAP.md`, `LICENSE`, `AGENTS.md`, `CLAUDE.md`, `GEMINI.md`, `CONTRIBUTING.md`, `SECURITY.md`, `CODE_OF_CONDUCT.md`, `llm.txt`, `Tuto_Qdrant.md`)
 - CI/CD files and ignore definitions (`.gitignore`, `.dockerignore`, `.npmignore`, `.npmrc`, `.node-version`, `.nvmrc`, `.env.example`)

 When creating _any_ validation tests or one-off logic scripts, default to `scripts/ad-hoc/` or `tests/unit/` according to your goals. Do not pollute the `/` root context.
@@ -299,7 +289,8 @@ When creating _any_ validation tests or one-off logic scripts, default to `scrip
 ### Database

 - **Always** go through `src/lib/db/` domain modules — **never** write raw SQL in routes or handlers
- **Never** barrel-import from `localDb.ts` — import specific `src/lib/db/*` modules
+- **Never** add logic to `src/lib/localDb.ts` (re-export layer only)
+- **Never** barrel-import from `localDb.ts` — import specific `db/` modules instead
 - DB singleton: `getDbInstance()` from `src/lib/db/core.ts` (WAL journaling)
 - Migrations: `src/lib/db/migrations/` — versioned SQL files, idempotent, run in transactions

@@ -343,7 +334,6 @@ Documentation must describe verified behavior, not plausible behavior.

 ### Adding a New Provider

-0. Check `docs/reference/REMOVED_PROVIDERS.md` first — providers removed at their operator's request must never be reintroduced (guarded by `tests/unit/removed-providers-blocklist.test.ts`)
 1. Register in `src/shared/constants/providers.ts` (Zod-validated at load)
 2. Add executor in `open-sse/executors/` if custom logic needed (extend `BaseExecutor`)
 3. Add translator in `open-sse/translator/` if non-OpenAI format
@@ -365,18 +355,19 @@ Documentation must describe verified behavior, not plausible behavior.
 1. Create `src/lib/db/yourModule.ts` — import `getDbInstance` from `./core.ts`
 2. Export CRUD functions for your domain table(s)
 3. Add migration in `src/lib/db/migrations/` if new tables needed
-4. Write tests
+4. Re-export from `src/lib/localDb.ts` (add to the re-export list only)
+5. Write tests

 ### Adding a New MCP Tool

 1. Add tool definition in `open-sse/mcp-server/tools/` with Zod input schema + async handler
 2. Register in tool set (wired by `createMcpServer()`)
 3. Assign to appropriate scope(s)
-4. Write tests (tool invocation logged to the `mcp_tool_audit` table)
+4. Write tests (tool invocation logged to `mcp_audit` table)

 ### Adding a New A2A Skill

-1. Create skill in `src/lib/a2a/skills/` (6 already exist: smart-routing, quota-management, provider-discovery, cost-analysis, health-report, list-capabilities)
+1. Create skill in `src/lib/a2a/skills/` (5 already exist: smart-routing, quota-management, provider-discovery, cost-analysis, health-report)
 2. Skill receives task context (messages, metadata) → returns structured result
 3. Register in `A2A_SKILL_HANDLERS` in `src/lib/a2a/taskExecution.ts`
 4. Expose in `src/app/.well-known/agent.json/route.ts` (Agent Card)
@@ -385,7 +376,7 @@ Documentation must describe verified behavior, not plausible behavior.

 ### Adding a New Cloud Agent

-1. Create agent class in `src/lib/cloudAgent/agents/` extending `CloudAgentBase` (4 already exist: codex-cloud, devin, jules, cursor-cloud)
+1. Create agent class in `src/lib/cloudAgent/agents/` extending `CloudAgentBase` (3 already exist: codex-cloud, devin, jules)
 2. Implement `createTask`, `getStatus`, `approvePlan`, `sendMessage`, `listSources`
 3. Register in `src/lib/cloudAgent/registry.ts`
 4. Add OAuth/credentials handling if needed (`src/lib/oauth/providers/`)
@@ -396,7 +387,7 @@ Documentation must describe verified behavior, not plausible behavior.
 1. Create installer in `src/lib/services/installers/{name}.ts` modeled on `ninerouter.ts` (use `runNpm` from `installers/utils.ts` — no shell interpolation, hard rule #13).
 2. Register the service in `src/lib/services/bootstrap.ts` (add to `SERVICES[]` array and extend `buildSpawnArgsFactory()`).
 3. Add a DB seed row for the new service in `src/lib/db/migrations/` (`version_manager` table, `status='not_installed'`, `auto_start=0`).
-4. Create 8 API endpoints under `src/app/api/services/{name}/` (`_lib.ts`, `install`, `start`, `stop`, `restart`, `update`, `status`, `auto-start`, `auto-restart-adopted`). All delegate errors through `createErrorResponse()`. The shared `logs` endpoint is already wired via `[name]/logs/route.ts`.
+4. Create 7 API endpoints under `src/app/api/services/{name}/` (`_lib.ts`, `install`, `start`, `stop`, `restart`, `update`, `status`, `auto-start`). All delegate errors through `createErrorResponse()`. The shared `logs` endpoint is already wired via `[name]/logs/route.ts`.
 5. Verify `/api/services/` is in `LOCAL_ONLY_API_PREFIXES` in `src/server/authz/routeGuard.ts`; add a test asserting `isLocalOnlyPath()` returns `true` for the new prefix if you add one (hard rule #17).
 6. Add a UI tab in `src/app/(dashboard)/dashboard/providers/services/tabs/` reusing `ServiceStatusCard`, `ServiceLifecycleButtons`, `ServiceLogsPanel`.
 7. Document in `docs/frameworks/EMBEDDED-SERVICES.md` (update §1 service table + §4 API reference) and `docs/openapi.yaml`.
@@ -408,9 +399,6 @@ Documentation must describe verified behavior, not plausible behavior.
 - Eval suite: `src/lib/evals/` → docs: `docs/frameworks/EVALS.md`
 - Skill (sandbox): `src/lib/skills/` → docs: `docs/frameworks/SKILLS.md`
 - Webhook event: `src/lib/webhookDispatcher.ts` → docs: `docs/frameworks/WEBHOOKS.md`
- Log-export destination: add `src/lib/logExport/destinations/<name>.ts` + one line in
-  `src/lib/logExport/registry.ts` → docs: `docs/frameworks/LOG-EXPORT.md`. The runner, REST layer
-  and dashboard form all read the registry, so nothing else changes.

 ---

@@ -423,7 +411,7 @@ For any non-trivial change, read the matching deep-dive first:
 | Repo navigation                               | `docs/architecture/REPOSITORY_MAP.md`                   |
 | Architecture                                  | `docs/architecture/ARCHITECTURE.md`                     |
 | Engineering reference                         | `docs/architecture/CODEBASE_DOCUMENTATION.md`           |
-| Auto-Combo (16-factor scoring, 19 strategies) | `docs/routing/AUTO-COMBO.md`                            |
+| Auto-Combo (15-factor scoring, 19 strategies) | `docs/routing/AUTO-COMBO.md`                            |
 | Resilience (3 mechanisms)                     | `docs/architecture/RESILIENCE_GUIDE.md`                 |
 | Reasoning replay                              | `docs/routing/REASONING_REPLAY.md`                      |
 | Skills framework                              | `docs/frameworks/SKILLS.md`                             |
@@ -436,7 +424,6 @@ For any non-trivial change, read the matching deep-dive first:
 | Evals                                         | `docs/frameworks/EVALS.md`                              |
 | Compliance / audit                            | `docs/security/COMPLIANCE.md`                           |
 | Webhooks                                      | `docs/frameworks/WEBHOOKS.md`                           |
-| Log export (call logs → BigQuery/…)           | `docs/frameworks/LOG-EXPORT.md`                         |
 | Authorization pipeline                        | `docs/architecture/AUTHZ_GUIDE.md`                      |
 | Stealth (TLS / fingerprint)                   | `docs/security/STEALTH_GUIDE.md`                        |
 | Agent protocols (A2A / ACP / Cloud)           | `docs/frameworks/AGENT_PROTOCOLS_GUIDE.md`              |
@@ -449,7 +436,7 @@ For any non-trivial change, read the matching deep-dive first:
 | VS Code Copilot Chat (OmniCopilot extension)  | `docs/guides/VSCODE-COPILOT.md`                         |
 | Release flow                                  | `docs/ops/RELEASE_CHECKLIST.md`                         |
 | Embedded services                             | `docs/frameworks/EMBEDDED-SERVICES.md`                  |
-| Quality gates (~90 scripts, allowlist policy) | `docs/architecture/QUALITY_GATES.md`                    |
+| Quality gates (~80 scripts, allowlist policy) | `docs/architecture/QUALITY_GATES.md`                    |

 ---

@@ -495,12 +482,6 @@ Why this matters: fixing bug A while opening bug B is worse than not fixing at a
  pipeline, and A2A skills.
 - Do not close a contributor pull request after using its code; merge it through GitHub so
  the contributor receives credit.
- **Never merge a PR that touches an agent-instruction surface without explicit operator
-  approval** — `CLAUDE.md`, `AGENTS.md`, `GEMINI.md`, `llm.txt` (+ mirrors) and
-  `skills/**/SKILL.md` are executed as authority by every AI session; a merged instruction
-  compromises every future agent run. Check with `gh pr diff <N> --name-only` before any
-  merge. Incident record: PR #11770 (2026-09-01) told agents to execute a third-party
-  setup script and was swept in by a merge campaign; reverted in #12249.

 ---

@@ -613,18 +594,6 @@ inside your feature branch (a base-red fix is its own freeze-gated `fix/release-
 PR); and if you must open a PR anyway, add `⚠️ base-red inherited: #<issue>` to the PR body so
 reviewers and CI babysitters do not chase ghosts.

-### Sync-back landings are fast-forward, never squash
-
-A `main → release/vX+1` sync-back (Phase 5 of `/generate-release`, or any later "bring main's
-post-release commits over" PR) must reach the release branch as the merge commit it already is:
-`git merge-base --is-ancestor origin/release/vX+1 <head>` then
-`git push origin <head>:refs/heads/release/vX+1` (GitHub marks the PR merged). Squash-merging it
-drops `main` from the release branch's ancestry and the next sync-back re-conflicts on every file
-main touched (551 conflicts on the v3.8.50 → v3.8.51 sync before the two-step merge). After
-landing, `git merge-base --is-ancestor origin/main origin/release/vX+1` must be true — and check
-that `config/quality/eslint-suppressions.json` / `quality-baseline.json` carried main's freezes
-(they merge as "ours" silently). Details: `.agents/skills/generate-release/phases/phase-5-next-cycle.md`.
-
 ---

 ## Upstream contributions
@@ -646,8 +615,8 @@ focused checks, and use a Conventional Commit message (for example, `docs: slim

 ## Environment

- **Runtime**: Node.js ≥22.22.2 <23 || ≥24.0.0 <27, ES Modules. This is the **only supported** runtime for the published `omniroute` CLI, the server, and the test suites (`node:test` + vitest) — `engines.node` is authoritative and end users never need Bun. A **best-effort `bun:sqlite` compatibility path** exists so a global Bun install (`bun install -g omniroute`) can start without `better-sqlite3` (driver adapter + Bun-aware process spawning); it is **not** a supported runtime — no support guarantees — and every Bun-specific runtime change MUST preserve the Node driver/fallback chain and ship a Bun test (`test:bun:db`) or an explicit reason why the path is Node-only.
- **Bun (build/dev script runner + compatibility smoke only)**: Bun `1.4.0` is pinned as an **exact devDependency** (provisioned through the existing `npm ci` via the lockfile's `@oven/bun-*` platform binaries — no `setup-bun`/ad-hoc install). It is used **only** to execute a small, allow-listed set of TypeScript **gate/generator scripts** (replacing `node --import tsx` for startup speed): the CI checks `check:provider-consistency`, `check:compression-budget`, `check:known-symbols`, and the non-CI `gen:provider-reference`, `bench:compression` — plus the focused `test:bun:db` compatibility smoke suite for the best-effort `bun:sqlite` path. **Do NOT** widen Bun to `npm install`, the build (`build:cli*`), `check:pack-artifact`, the supported published runtime, or the main test runners — those stay on Node. Any new Bun-invoking gate/generator script must be validated byte-identical against its `node --import tsx` output first. After pulling the lockfile change, run `npm install` so `bun` resolves locally (a stale `node_modules` will fail those scripts with `bun: not found`).
+- **Runtime**: Node.js ≥22.0.0 <23 || ≥24.0.0 <27, ES Modules. This is the **only supported** runtime for the published `omniroute` CLI, the server, and the test suites (`node:test` + vitest) — `engines.node` is authoritative and end users never need Bun. A **best-effort `bun:sqlite` compatibility path** exists so a global Bun install (`bun install -g omniroute`) can start without `better-sqlite3` (driver adapter + Bun-aware process spawning); it is **not** a supported runtime — no support guarantees — and every Bun-specific runtime change MUST preserve the Node driver/fallback chain and ship a Bun test (`test:bun:db`) or an explicit reason why the path is Node-only.
+- **Bun (build/dev script runner + compatibility smoke only)**: Bun `1.3.14` is pinned as an **exact devDependency** (provisioned through the existing `npm ci` via the lockfile's `@oven/bun-*` platform binaries — no `setup-bun`/ad-hoc install). It is used **only** to execute a small, allow-listed set of TypeScript **gate/generator scripts** (replacing `node --import tsx` for startup speed): the CI checks `check:provider-consistency`, `check:compression-budget`, `check:known-symbols`, and the non-CI `gen:provider-reference`, `bench:compression` — plus the focused `test:bun:db` compatibility smoke suite for the best-effort `bun:sqlite` path. **Do NOT** widen Bun to `npm install`, the build (`build:cli*`), `check:pack-artifact`, the supported published runtime, or the main test runners — those stay on Node. Any new Bun-invoking gate/generator script must be validated byte-identical against its `node --import tsx` output first. After pulling the lockfile change, run `npm install` so `bun` resolves locally (a stale `node_modules` will fail those scripts with `bun: not found`).
 - **TypeScript**: 6.0+, target ES2022, module esnext, resolution bundler
 - **Path aliases**: `@/*` → `src/`, `@omniroute/open-sse` → `open-sse/`, `@omniroute/open-sse/*` → `open-sse/*`
 - **Default port**: 20128 (API + dashboard on same port)
@@ -659,12 +628,12 @@ focused checks, and use a Conventional Commit message (for example, `docs: slim

 ## Quality Gates & Ratchets

-OmniRoute has **~90 quality-gate scripts** (`scripts/check/` + `scripts/quality/`) wired
+OmniRoute has **~80 quality-gate scripts** (`scripts/check/` + `scripts/quality/`) wired
 across **9 gate-running jobs** in `.github/workflows/ci.yml` (`lint`, `quality-gate`,
 `quality-extended`, `docs-sync-strict`, `i18n-ui-coverage`, `i18n`, `pr-test-policy`,
 `test-vitest`, `sonarqube`), plus the `quality.yml` fast-gates job (PR→`release/**`) and
-5 quality nightly workflows (`nightly-property`, `nightly-resilience`,
-`nightly-llm-security`, `nightly-mutation`, `nightly-schemathesis`). Full inventory, per-job breakdown, and operational
+3 nightly workflows (`nightly-property`, `nightly-resilience`, `nightly-llm-security`;
+`nightly-mutation` once merged). Full inventory, per-job breakdown, and operational
 procedures are in [`docs/architecture/QUALITY_GATES.md`](docs/architecture/QUALITY_GATES.md).

 **Quick reference:**
@@ -676,10 +645,6 @@ procedures are in [`docs/architecture/QUALITY_GATES.md`](docs/architecture/QUALI
  `npm run quality:ratchet -- --update` when a metric genuinely improves.
 - Job `test-vitest` runs `npm run test:vitest` (MCP tools, autoCombo, cache) — blocking.
  `test:vitest:ui` has been blocking since PR #7127.
- **Velocity phase (2026-08-30 → v4.0)**: every numeric baseline is loosened by 20% and
-  `--require-tighten` is advisory (`quality-baseline.json` → `_policy`); the nightly
-  `baseline-headroom` job tracks how much of the budget is left in the issue
-  "📈 Baseline headroom". See `docs/architecture/QUALITY_GATES.md` → "Velocity phase".

 **Allowlist policy (short form):** Fix the cause; use the allowlist only for pre-existing
 violations you cannot fix in the same PR. Add a comment with justification + issue number.
@@ -691,7 +656,7 @@ the stale-enforcement added in Fase 6A.3.
 ## Hard Rules

 1. Never commit secrets or credentials
-2. Never barrel-import from `localDb.ts` — import specific `src/lib/db/*` modules
+2. Never add logic to `localDb.ts`
 3. Never use `eval()` / `new Function()` / implied eval
 4. Never commit directly to `main`
 5. Never write raw SQL in routes — use `src/lib/db/` modules
@@ -753,13 +718,3 @@ The dashboard is reachable at the operator's chosen URL/port (default `http://lo
 - **Local VPS / shared dev environments**: ask the operator for the URL and current credentials — they live in their personal vault, NOT in this repo.

 > Any credential observed in a previous version of this file was a non-production demo value; treat it as compromised and do not reuse it.
-
-<!-- BEGIN:nextjs-agent-rules -->
-
-# This is NOT the Next.js you know
-
-This version has breaking changes — APIs, conventions, and file structure may all differ from your training data. Read the relevant guide in `node_modules/next/dist/docs/` (resolved from this file's directory; in monorepos the `next` package may not be visible from the repo root) before writing any code. Heed deprecation notices.
-
-This block is written and re-added by `next dev` — verify at `node_modules/next/dist/server/lib/generate-agent-files.js`. Removing it from a diff only re-creates the uncommitted change; committing it with your work keeps the tree clean.
-
-<!-- END:nextjs-agent-rules -->
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -73,9 +73,6 @@ npm run dev
 npm run build    # next build → .build/next/ then assembleStandalone → dist/
 npm run start

-# Fast backend/API-only compile for contributor changes
-npm run build:contributor
-
 # Release build (clean rebuild + HEAD sentinel — required for deploy)
 npm run build:release   # rm -rf .build dist && build + writes dist/BUILD_SHA

@@ -83,10 +80,6 @@ npm run build:release   # rm -rf .build dist && build + writes dist/BUILD_SHA
 PORT=20128 NEXT_PUBLIC_BASE_URL=http://localhost:20128 npm run dev
 ```

-The contributor build performs compile-only validation: it does not assemble the standalone
-distribution or build optional native packaging assets. Use the regular production build when
-you need to validate the shippable bundle.
-
 ### Build Output Layout

 | Directory | Contents                                                                  | Tracked |
@@ -107,11 +100,6 @@ npm run build
 `npm run build:release` additionally cleans both directories first and writes
 `dist/BUILD_SHA` (= `git rev-parse --short HEAD`) as a deploy integrity sentinel.

-`npm run build:contributor` uses the backend-only build profile. It temporarily stubs
-dashboard UI files while building, keeps API route handlers, and restores the original files
-after the build. Use `npm run build` for changes that affect the dashboard UI or for full
-release validation; the contributor profile is not a replacement for the release build.
-
 > **VPS deploy note:** the remote image directory `/usr/lib/node_modules/omniroute/app/`
 > is unchanged. The deploy skills rsync the contents of `dist/` into it.
 > Only the in-repo build output path moved (`app/` → `dist/`).
@@ -177,13 +165,6 @@ npm run test:all
 # Single test file (Node.js native test runner — most tests use this)
 node --import tsx/esm --test tests/unit/your-file.test.ts

-# Only the unit tests impacted by your change (same TIA selector as the CI gate, #8084)
-npm run test:scoped            # changes in the last commit (or the working tree)
-npm run test:scoped:staged     # staged changes only — pairs well with a pre-commit run
-npm run test:scoped:full       # rebuild the import-graph map first (after adding/moving files)
-# Exit 1 + "run the full suite" means a hub file (tsconfig, package.json, …) or an
-# unmapped source changed — the selector fails safe, it never silently skips.
-
 # Vitest (MCP server, autoCombo, cache)
 npm run test:vitest

@@ -320,7 +301,7 @@ src/                        # TypeScript (.ts / .tsx)
 open-sse/                   # @omniroute/open-sse workspace
 ├── executors/              # 89 executor implementation modules
 ├── handlers/               # 11 request handlers (chat, responses, embeddings, images, etc.)
-├── mcp-server/             # MCP server (110 unique tools, 3 transports, 33 scopes)
+├── mcp-server/             # MCP server (107 unique tools, 3 transports, 32 scopes)
 ├── services/               # 178 top-level services (combo, autoCombo, rateLimitManager, etc.)
 ├── translator/             # Format translators (OpenAI ↔ Claude ↔ Gemini ↔ Responses ↔ Ollama)
 ├── transformer/            # Responses API transformer
--- a/64
+++ b/64
@@ -103,12 +103,25 @@ RUN test -f package-lock.json \
 # node-gyp comes from npm's own bundled copy (deterministic, already in the image)
 # instead of `npx --yes`, which would install an arbitrary registry version
 # on-demand and run its lifecycle scripts (Sonar docker:S6505).
+#
+# tls-client-node (chatgpt-web/claude-web/grok-web/lmarena/perplexity-web TLS
+# impersonation) hits the same --ignore-scripts wall: its own postinstall.js
+# fetches a platform .so/.dylib/.dll from the bogdanfinn/tls-client GitHub
+# Releases API and is never invoked when npm ci skips lifecycle scripts. Unlike
+# better-sqlite3 above, that script never throws on failure — it only
+# `console.warn`s and exits 0 — so a rate-limited or offline build would
+# otherwise succeed silently with an empty bin/ and only fail at first request
+# in production (TlsClientUnavailableError, #7802). Run it explicitly here so
+# a broken/rate-limited fetch fails the BUILD loudly instead of shipping a
+# broken image.
 RUN --mount=type=cache,id=s/92ca8a61-c1ba-421f-a389-d48ac7258c2d-npm-cache,target=/root/.npm \
  npm ci --include=optional --no-audit --no-fund --legacy-peer-deps --ignore-scripts \
  && (cd node_modules/better-sqlite3 \
      && node /usr/local/lib/node_modules/npm/node_modules/node-gyp/bin/node-gyp.js rebuild) \
  && node -e "require('better-sqlite3')(':memory:').close()" \
-  && node -e "const wreq=require('wreq-js'); if(typeof wreq.createTransport!=='function') process.exit(1)"
+  && node node_modules/tls-client-node/scripts/postinstall.js \
+  && (test -n "$(find node_modules/tls-client-node/bin -mindepth 1 -print -quit 2>/dev/null)" \
+      || (echo "tls-client-node native binary missing after postinstall — GitHub API fetch likely rate-limited or failed (#7802)" >&2 && exit 1))

 # Build with Turbopack (stable in Next 16, the repo default). The v3.8.27-era
 # TurbopackInternalError panic ("entered unreachable code: there must be a path to a
@@ -171,29 +184,19 @@ ENV NODE_OPTIONS="--max-old-space-size=${OMNIROUTE_BUILD_MEMORY_MB}"
 # silently leaving no standalone bundle. Next derives the worker count from
 # CIRCLE_NODE_TOTAL (workers = N-1). (#10060)
 #
-# Lowered 8 → 3 (7 workers → 2) in #11419, then 3 → 2 (2 workers → 1) in #7518.
-# Every page-data worker inherits NODE_OPTIONS above, so the ceiling is per
-# PROCESS, not per build: 7 workers on a 16 GB GitHub runner (ubuntu-24.04 /
-# ubuntu-24.04-arm, 4 vCPU) exhausted the host and buildkit failed the whole
-# step with `ResourceExhausted: ... cannot allocate memory`. The compile phase
-# always finished ("✓ Compiled successfully in 4.2min"); the kernel killed the
-# build right after "Collecting page data using N workers".
-#
-# #11419's first fix (8 → 3) modeled the per-worker peak as an INFERENCE
-# (2560 MB, guessed from "7 workers didn't fit") and assumed the parent
-# process's RSS tracked the V8 heap ceiling. Both assumptions were wrong: a
-# live VPS reproduction (issue #7518, dmesg OOM-killer report) measured the
-# real per-process RSS directly at ~4.5 GB, independent of the NODE_OPTIONS
-# heap flag (Turbopack itself is native/Rust, outside the V8 heap) — and it
-# applies to the parent process too, not just workers. 2 workers (3 processes
-# × 4.5 GB = 13.5 GB) still didn't fit the 12.288 GB (75%) budget on a 16 GB
-# runner, matching the still-live publish failures after #11419 merged. 1
-# worker (2 processes × 4.5 GB = 9 GB) fits with headroom to spare.
-# tests/unit/docker-build-memory-budget.test.ts does the arithmetic against
-# the measured figure and fails if either knob is raised past what a 16 GB
-# runner holds. Override for a big builder: `--build-arg
-# OMNIROUTE_BUILD_WORKERS=8`.
-ARG OMNIROUTE_BUILD_WORKERS=2
+# Lowered 8 → 3 (7 workers → 2). Every page-data worker inherits NODE_OPTIONS
+# above, so the ceiling is per PROCESS, not per build: 7 workers on a 16 GB
+# GitHub runner (ubuntu-24.04 / ubuntu-24.04-arm, 4 vCPU) exhausted the host and
+# buildkit failed the whole step with `ResourceExhausted: ... cannot allocate
+# memory`. The compile phase always finished ("✓ Compiled successfully in
+# 4.2min"); the kernel killed the build right after "Collecting page data using
+# 7 workers". It was intermittent for a while and went 100% on 2026-08-22, which
+# is what a threshold being crossed by ordinary codebase growth looks like.
+# tests/unit/docker-build-memory-budget.test.ts does the arithmetic and fails if
+# either knob is raised past what a 16 GB runner holds. 2 workers also stops
+# oversubscribing the runner's 4 vCPU, which 7 did. Override for a big builder:
+# `--build-arg OMNIROUTE_BUILD_WORKERS=8`.
+ARG OMNIROUTE_BUILD_WORKERS=3
 ENV CIRCLE_NODE_TOTAL=${OMNIROUTE_BUILD_WORKERS}

 COPY . ./
@@ -331,18 +334,7 @@ RUN --mount=type=cache,id=s/92ca8a61-c1ba-421f-a389-d48ac7258c2d-apt-cache,targe
  && git config --system url."https://github.com/".insteadOf "ssh://git@github.com/"

 # Install CLI tools globally. Separate layer from apt for better cache reuse.
-# Pinned to exact versions per Diego's diagnosis in #12576 — floating
-# `@latest` causes two CI failures:
-#   1. `openclaw` ships a breaking major ~weekly; overnight builds silently
-#      advance to a version that no longer matches the tested combo stack.
-#   2. `codex` / `claude-code` dev pre-releases (`@next`, dist-tags) mutate
-#      API surface without notice; reproducible builds need a SHA-pinned dev
-#      build, not the floating `@latest`.
 RUN --mount=type=cache,id=s/92ca8a61-c1ba-421f-a389-d48ac7258c2d-npm-cache,target=/root/.npm \
-  npm install -g --no-audit --no-fund \
-    @openai/codex@0.153.2 \
-    @anthropic-ai/claude-code@2.1.260 \
-    droid@0.212.0 \
-    openclaw@2026.9.1
+  npm install -g --no-audit --no-fund @openai/codex @anthropic-ai/claude-code droid openclaw@latest

 USER node
--- a/Show More
+++ b/Show More