fix(logs): apply filter predicates to merged in-memory call-log rows (#11082)

Validated on a combined board over tip aa128736 (incl. sibling #11081): call-logs-row-filter 4/4 green, typecheck:core clean. The merged-row predicate fix closes a real gap — in-memory (in-flight/recently-completed) rows bypassed every filter except correlationId; rowMatchesFilter() now applies search/model/provider/account/apiKey/status/combo uniformly while DB rows stay idempotent. Thank you @AndrianBalanescu!

.github/workflows/dast-smoke.yml

@@ -46,6 +46,7 @@ jobs:
         env:
           PORT: "20128"
           INJECTION_GUARD_MODE: block
+          REQUIRE_API_KEY: "false"
         run: |
           node dist/server.js > server.log 2>&1 &
           echo $! > server.pid
@@ -64,16 +65,20 @@ jobs:
           # those 302s as "the API accepted a schema-violating request" and the configured-off
           # 400 as "rejected a schema-compliant request". Documenting the flow in the spec is
           # still right (operators need it); fuzzing it is not what this smoke is for.
+          # /api/auth/login has brute-force rate limiting: repeated failed logins return 429,
+          # which Schemathesis flags as rejection of schema-compliant requests.
           schemathesis run docs/openapi.yaml --url http://localhost:20128 \
             --include-path-regex '^/v1/(chat/completions|models)$|^/api/(auth|keys)' \
-            --exclude-path-regex '^/api/auth/oidc/' \
+            --exclude-path-regex '^/api/auth/(oidc/|login)' \
             --max-examples 8 --workers 4 --checks all --max-response-time 30 \
             --request-timeout 20 --suppress-health-check all --no-color
+      - name: Install promptfoo
+        run: npm install -g promptfoo@0.122.0
       - name: promptfoo injection-guard (blocking)
         env:
           OMNIROUTE_URL: http://localhost:20128
           OMNIROUTE_API_KEY: not-needed-blocked-before-upstream
-        run: npx --yes promptfoo@latest eval -c promptfooconfig.yaml --no-cache
+        run: promptfoo eval -c promptfooconfig.yaml --no-cache
       - name: Stop server
         if: always()
         run: kill "$(cat server.pid)" || true

Dockerfile.bun

@@ -0,0 +1,89 @@
+# ── Multi-stage Dockerfile for Native Bun Runtime (web-latest-bun) ───────────
+FROM oven/bun:1.3.14-slim AS base
+WORKDIR /app
+
+RUN apt-get update \
+  && apt-get upgrade -y \
+  && apt-get install -y --no-install-recommends \
+     build-essential \
+     python3 \
+     python-is-python3 \
+     make \
+     g++ \
+     libsecret-1-0 \
+     ca-certificates \
+     curl \
+  && rm -rf /var/lib/apt/lists/*
+
+# ── Builder stage (100% Bun Native Install & Build) ─────────────────────────
+FROM base AS builder
+WORKDIR /app
+
+COPY . .
+
+# Fast Bun native package install
+RUN bun install --include=optional --quiet
+
+# Compile native better-sqlite3 Node-API addon under Bun
+RUN if [ -d "node_modules/better-sqlite3" ]; then \
+      (cd node_modules/better-sqlite3 && bunx node-gyp rebuild); \
+    fi
+
+# Fetch tls-client-node native binary if script exists
+RUN if [ -f "node_modules/tls-client-node/scripts/postinstall.js" ]; then \
+      bun node_modules/tls-client-node/scripts/postinstall.js || true; \
+    fi
+
+# Disable Turbopack for Bun builder stage (Turbopack V8 internal worker bindings require Node)
+ENV OMNIROUTE_USE_TURBOPACK=0
+
+ARG OMNIROUTE_BASE_PATH=""
+ENV OMNIROUTE_BASE_PATH=$OMNIROUTE_BASE_PATH
+
+ARG DASHBOARD_ALLOW_EMBED=""
+ENV DASHBOARD_ALLOW_EMBED=$DASHBOARD_ALLOW_EMBED
+
+ENV NEXT_TELEMETRY_DISABLED=1
+ENV NODE_ENV=production
+
+# Bun native Next.js build execution
+RUN bun run --quiet build
+
+# ── Runner stage (100% Bun Native Production Runtime) ──────────────────────
+FROM oven/bun:1.3.14-slim AS runner
+
+LABEL org.opencontainers.image.title="omniroute" \
+  org.opencontainers.image.description="Unified AI proxy — route any LLM through one endpoint (Bun Native)" \
+  org.opencontainers.image.url="https://omniroute.online" \
+  org.opencontainers.image.source="https://github.com/diegosouzapw/OmniRoute" \
+  org.opencontainers.image.licenses="MIT"
+
+WORKDIR /app
+
+RUN apt-get update \
+  && apt-get install -y --no-install-recommends \
+     libsecret-1-0 \
+     ca-certificates \
+     curl \
+  && rm -rf /var/lib/apt/lists/*
+
+ENV NODE_ENV=production
+ENV PORT=20128
+ENV HOSTNAME=0.0.0.0
+ENV OMNIROUTE_MEMORY_MB=1024
+
+ENV DATA_DIR=/app/data
+RUN mkdir -p /app/data
+
+COPY --from=builder /app/.build/next/standalone ./
+COPY --from=builder /app/node_modules/better-sqlite3 ./node_modules/better-sqlite3
+ENV OMNIROUTE_MIGRATIONS_DIR=/app/migrations
+
+COPY --from=builder /app/scripts/dev/healthcheck.mjs ./healthcheck.mjs
+
+EXPOSE 20128
+
+HEALTHCHECK --interval=30s --timeout=5s --start-period=15s --retries=3 \
+  CMD bun healthcheck.mjs || exit 1
+
+ENTRYPOINT ["bun", "bin/omniroute.mjs", "serve", "--no-open"]

README.md

@@ -1009,6 +1009,19 @@ Full table: [Docker Guide — runtime RAM](docs/guides/DOCKER_GUIDE.md#runtime-r
 > are **not supported for production**. See
 > [Docker Release Channels](docs/guides/DOCKER_GUIDE.md#release-channels).
 
+**🥟 Bun**
+
+Standard `bun install` and global installation (`bun install -g omniroute`) are supported via Bun runtime detection:
+- **Built-in `bun:sqlite`**: OmniRoute uses Bun's built-in `bun:sqlite` driver when running under Bun, falling back to `better-sqlite3` on Node.js or `sql.js`.
+- **Automatic Webpack bundler selection**: Development (`bun run dev`) and production builds (`bun run build`) automatically detect Bun and disable Turbopack in favor of Webpack to prevent native V8 binding incompatibilities.
+- **Dedicated Bun Dockerfile**: Multi-stage `Dockerfile.bun` for native Bun production deployments (`docker build -f Dockerfile.bun -t omniroute:bun .`).
+
+```bash
+# Install and run with Bun
+bun install
+bun run dev
+```
+
 **🛠️ From source**
 
 ```bash

bin/cli/commands/plugin.mjs

@@ -9,10 +9,13 @@ import { discoverPlugins } from "../plugins.mjs";
 // (instead of string-interpolating into `execSync`) prevents a malicious plugin
 // name like `foo; rm -rf ~` or `` foo`id` `` from being interpreted by the shell.
 function runNpm(args) {
-  const res = spawnSync("npm", args, { stdio: "inherit", shell: false });
+  const isBun = Boolean(process.versions.bun);
+  const pm = isBun ? "bun" : "npm";
+  const cmdArgs = isBun && args[0] === "install" ? ["add", ...args.slice(1)] : args;
+  const res = spawnSync(pm, cmdArgs, { stdio: "inherit", shell: false });
   if (res.error) throw res.error;
   if (typeof res.status === "number" && res.status !== 0) {
-    throw new Error(`npm exited with code ${res.status}`);
+    throw new Error(`${pm} exited with code ${res.status}`);
   }
 }
 

bin/cli/runtime/nativeDeps.mjs

@@ -114,30 +114,30 @@ export function isBetterSqliteBinaryValid() {
 
 export function npmInstallRuntime(pkgs, opts = {}) {
   const cwd = ensureRuntimeDir();
-  // Persist to the runtime package.json (exact version) instead of --no-save so a later
-  // install of a sibling runtime dep (e.g. systray2 from trayRuntime.ts, which writes to the
-  // same runtime dir) does not prune this package as "extraneous" — that pruning otherwise
-  // reproduces "No SQLite driver available" after a tray install removes better-sqlite3.
-  // npm 12+ defaults `allowScripts` to off, silently skipping lifecycle/install
-  // scripts (e.g. better-sqlite3's node-gyp/prebuild-install rebuild) unless the
-  // package has a matching `allowScripts` entry — and still exits 0, masking the
-  // failure (#10713). The runtime dir is a CLI-owned, non-user package.json, so
-  // explicitly allowing scripts for the packages we are installing here is safe.
-  const npmArgs = [
-    "install",
-    ...pkgs,
-    "--no-audit",
-    "--no-fund",
-    "--prefer-online",
-    "--save-exact",
-    ...pkgs.map((pkg) => `--allow-scripts=${pkg}`),
-  ];
-  // On Windows .cmd files cannot be executed without a shell; use cmd.exe /c explicitly
-  // so we never set shell:true (which would propagate env and enable injection).
   const isWin = platform() === "win32";
-  const [exe, args] = isWin ? ["cmd.exe", ["/c", "npm", ...npmArgs]] : ["npm", npmArgs];
+  const isBun = Boolean(process.versions.bun);
+
+  let exe, args, displayCmd;
+  if (isBun) {
+    const bunArgs = ["add", ...pkgs, "--trust"];
+    [exe, args] = isWin ? ["cmd.exe", ["/c", "bun", ...bunArgs]] : ["bun", bunArgs];
+    displayCmd = `bun ${bunArgs.join(" ")}`;
+  } else {
+    const npmArgs = [
+      "install",
+      ...pkgs,
+      "--no-audit",
+      "--no-fund",
+      "--prefer-online",
+      "--save-exact",
+      ...pkgs.map((pkg) => `--allow-scripts=${pkg}`),
+    ];
+    [exe, args] = isWin ? ["cmd.exe", ["/c", "npm", ...npmArgs]] : ["npm", npmArgs];
+    displayCmd = `npm ${npmArgs.join(" ")}`;
+  }
+
   if (!opts.silent) {
-    process.stdout.write(`[omniroute][runtime] npm ${npmArgs.join(" ")}\n`);
+    process.stdout.write(`[omniroute][runtime] ${displayCmd}\n`);
   }
   const res = spawnSync(exe, args, {
     cwd,

bin/cli/sqlite.mjs

@@ -5,10 +5,14 @@ import { ensureSettingsSchema, hashManagementPassword, updateSettings } from "./
 
 async function loadSqlite() {
   if (process.versions.bun) {
-    return { Database: (await import("bun:sqlite")).Database };
+    try {
+      return { Database: (await import("bun:sqlite")).Database, driver: "bun:sqlite" };
+    } catch (bunError) {
+      // fall through to better-sqlite3 if bun:sqlite fails
+    }
   }
   try {
-    return { Database: (await import("better-sqlite3")).default };
+    return { Database: (await import("better-sqlite3")).default, driver: "better-sqlite3" };
   } catch (error) {
     return { error };
   }
@@ -86,12 +90,14 @@ export function normalizeBunSqliteParams(params) {
 
 export function createSqliteNativeError(error) {
   const message = error instanceof Error ? error.message : String(error);
+  const isBun = Boolean(process.versions.bun);
+  const rebuildCmd = isBun ? "bun add better-sqlite3 --trust" : "npm rebuild better-sqlite3";
   if (message.includes("NODE_MODULE_VERSION") || message.includes("ERR_DLOPEN_FAILED")) {
     return new Error(
-      "better-sqlite3 native binding is incompatible with this Node.js runtime. " +
-        "Run `npm rebuild better-sqlite3` in the OmniRoute project and try again. " +
-        "Or run: omniroute runtime repair  " +
-        "(rebuilds into a user-writable runtime; works without a C++ toolchain)."
+      `better-sqlite3 native binding is incompatible with this runtime. ` +
+        `Run \`${rebuildCmd}\` in the OmniRoute project and try again. ` +
+        `Or run: omniroute runtime repair  ` +
+        `(rebuilds into a user-writable runtime; works without a C++ toolchain).`
     );
   }
   if (
@@ -100,10 +106,9 @@ export function createSqliteNativeError(error) {
     message.includes("Cannot find module 'better-sqlite3'")
   ) {
     return new Error(
-      "better-sqlite3 native binding could not be found (no prebuilt addon for this platform). " +
-        "This is common under `npx`, which runs a fresh, ephemeral install that never built the addon. " +
-        "Run: omniroute runtime repair  " +
-        "(rebuilds into a user-writable runtime; works without a C++ toolchain)."
+      `better-sqlite3 native binding could not be found (no prebuilt addon for this platform). ` +
+        `Run: omniroute runtime repair  ` +
+        `(rebuilds into a user-writable runtime; works without a C++ toolchain).`
     );
   }
   return error;
@@ -111,7 +116,7 @@ export function createSqliteNativeError(error) {
 
 async function openSqliteDatabase(dbPath, options = {}) {
   const loaded = await loadSqlite();
-  if (process.versions.bun) {
+  if (loaded.driver === "bun:sqlite" || (process.versions.bun && !loaded.Database)) {
     if (options.fileMustExist && !fs.existsSync(dbPath)) {
       throw new Error(`SQLite file does not exist: ${dbPath}`);
     }

bin/cli/utils/ensureAndroidCacheDir.mjs

@@ -94,10 +94,15 @@ export function ensureAndroidCacheDir(options = {}) {
  */
 export function isFatalInstrumentationHookFailure(text) {
   if (!text) return false;
-  return (
-    /Unsupported platform:\s*android/i.test(text) ||
-    /error occurred while loading instrumentation hook/i.test(text)
-  );
+  // Next.js wraps ANY throw inside instrumentation.register() with the generic
+  // "An error occurred while loading instrumentation hook:" prefix, on every
+  // platform (node_modules/next/dist/server/web/globals.js). That prefix alone
+  // therefore cannot identify the Android/Termux cache-probe failure — a bare
+  // generic instrumentation error on win32/desktop would be misreported as the
+  // Android bug and hide the real cause. Only match when the text actually
+  // carries the Android platform marker that Next's getCacheDirectory() emits.
+  // #10028
+  return /Unsupported platform:\s*android/i.test(text);
 }
 
 /**

bin/nodeRuntimeSupport.mjs

@@ -44,6 +44,18 @@ export function getSecureFloorForMajor(major) {
 }
 
 export function getNodeRuntimeSupport(version = process.versions.node) {
+  if (process.versions.bun) {
+    return {
+      nodeVersion: `bun-${process.versions.bun} (Node.js API ${version})`,
+      nodeCompatible: true,
+      reason: "supported-bun",
+      supportedRange: SUPPORTED_NODE_RANGE + " || Bun >=1.1.0",
+      supportedDisplay: SUPPORTED_NODE_DISPLAY + ", or Bun 1.1+",
+      recommendedVersion: `v${RECOMMENDED_NODE_VERSION}`,
+      minimumSecureVersion: null,
+    };
+  }
+
   const parsed = parseNodeVersion(version);
   const secureFloor = getSecureFloorForMajor(parsed.major);
   const nodeCompatible = secureFloor ? compareNodeVersions(parsed, secureFloor) >= 0 : false;

bin/omniroute.mjs

@@ -17,7 +17,12 @@
 import { existsSync, readFileSync, writeFileSync } from "node:fs";
 import { join, dirname } from "node:path";
 import { fileURLToPath, pathToFileURL } from "node:url";
-import updateNotifier from "update-notifier";
+let updateNotifier = null;
+try {
+  updateNotifier = (await import("update-notifier")).default;
+} catch {
+  // update-notifier is optional in pruned standalone environments
+}
 import { isNativeBinaryCompatible } from "../scripts/build/native-binary-compat.mjs";
 import { getNodeRuntimeSupport, getNodeRuntimeWarning } from "./nodeRuntimeSupport.mjs";
 import { getDefaultDataDir } from "./cli/data-dir.mjs";
@@ -251,8 +256,9 @@ if (shouldProvisionStorageKey(process.argv)) {
 
 // Register update notifier — checks npm once per 24h, notifies on exit via stderr.
 const _pkg = JSON.parse(readFileSync(join(ROOT, "package.json"), "utf8"));
-const _notifier = updateNotifier({ pkg: _pkg, updateCheckInterval: 1000 * 60 * 60 * 24 });
+const _notifier = updateNotifier ? updateNotifier({ pkg: _pkg, updateCheckInterval: 1000 * 60 * 60 * 24 }) : null;
 process.on("exit", () => {
+  if (!_notifier || !_notifier.update) return;
   if (process.env.OMNIROUTE_NO_UPDATE_NOTIFIER) return;
   if (process.env.CI) return;
   if (process.argv.includes("--quiet") || process.argv.includes("-q")) return;

changelog.d/features/11104-operator-error-rules.md

@@ -0,0 +1 @@
+- **feat(providers):** let operators declare per-provider error rules through `settings.providerErrorRules` instead of patching the catalog — an operator-supplied rule for a provider is consulted before the built-in `providerRuleRegistry`, receives the raw error text, and has its declared scope/cooldown/reason actually honored end to end, for any provider (declaring the rule is the opt-in — no extra allowlist entry needed). Matches are plain case-insensitive substrings (never RegExp) and bounded to 50 rules to keep the hot path safe ([#11104](https://github.com/diegosouzapw/OmniRoute/pull/11104))

changelog.d/fixes/10028-windows-instrumentation-hook.md

@@ -0,0 +1 @@
+- fix(cli): stop diagnosing every Next.js instrumentation-hook failure as the Android/Termux cache bug — only the Android "Unsupported platform: android" signal now triggers the Android hint, so a win32/desktop instrumentation error surfaces its real cause instead of a useless `mkdir -p ~/.cache` (#10028)

changelog.d/fixes/10265-command-code-provider-api.md

@@ -0,0 +1 @@
+- fix(command-code): route chat to the documented /provider/v1/chat/completions endpoint instead of the CLI-only /alpha/generate, which Command Code gates/blocks for external callers (#10265)

changelog.d/fixes/10523-servicesupervisor-port-flake.md

@@ -0,0 +1 @@
+- fix(services): isolate probeBeforeSpawn adoption tests on distinct ports to stop the order-dependent flake (#10523)

changelog.d/fixes/10986-reasoning-only-content.md

@@ -0,0 +1 @@
+- fix(command-code): surface reasoning-only output as content when a model emits no text-delta (#10986)

changelog.d/fixes/10990-v0-vercel-web-static-catalog.md

@@ -0,0 +1 @@
+- **Static model catalog for v0-vercel-web:** seed a static catalog for the v0-vercel-web web-cookie provider (v0-1.0-md, v0-1.5-lg, v0-1.5-md) so its dashboard "Available Models" / "Import from /models" UI serves a usable list instead of falling through to the route's 400 "does not support models listing" ([#10990](https://github.com/diegosouzapw/OmniRoute/issues/10990)).

changelog.d/fixes/10997-blackbox-deprecation.md

@@ -0,0 +1 @@
+- fix(providers): mark the blackbox provider deprecated — api.blackbox.ai returns HTTP 404 on every path variant (sweep 2026-08-21), so the public inference surface is dead and the catalog entry now carries a deprecation notice. ([#10997](https://github.com/diegosouzapw/OmniRoute/issues/10997))

changelog.d/fixes/11002-dify-key-validation.md

@@ -0,0 +1 @@
+- fix(providers): validate Dify keys against its native /v1/chat-messages endpoint (#11002)

changelog.d/fixes/11050-remove-ghost-webhook-events.md

@@ -0,0 +1 @@
+- **fix(webhooks):** remove 3 declared-but-never-emitted events (`provider.error`, `provider.recovered`, `combo.switched`) from `WebhookEvent` — catalog now `request.completed | request.failed | quota.exceeded | test.ping`; `POST /api/webhooks` and `PUT /api/webhooks/[id]` reject ghost values with 400; OpenAPI webhook description updated across 43 locales ([11050](https://github.com/diegosouzapw/OmniRoute/pull/11050))

changelog.d/fixes/11060-perplexity-filter.md

@@ -0,0 +1 @@
+- fix(providers): filter Perplexity model import to the Sonar family so Agent-API catalog ids stop surfacing as routable chat models (#11060)

changelog.d/fixes/11095-termux-onnx.md

@@ -0,0 +1 @@
+- fix(install): make the ONNX dependency chain optional so Termux/Android installs succeed again (#11095)

changelog.d/fixes/11101-reject-silent-validation.md

@@ -0,0 +1 @@
+- **fix(providers):** Reject silent validation degradation on provider connection patch — unknown `rateLimitOverrides` keys (e.g. a typo'd `tpm`) and empty/non-numeric values now return `400` with the rejected key list instead of being silently dropped ([#11101](https://github.com/diegosouzapw/OmniRoute/pull/11101))

changelog.d/fixes/11102-combo-suggestion-count.md

@@ -0,0 +1 @@
+- **Autopilot suggestion counter:** the combo health autopilot summary now reports `suggestionCount` (the real number of suggested actions across all issues) instead of conflating it with link counts, while keeping `actionableCount` as a deprecated alias for backward compatibility. The `run_combo_test` action now links to the dashboard with the combo id (`/dashboard/combos?test=<comboId>`) rather than the read-only API route, so operators can actually trigger a test from the UI ([#11102](https://github.com/diegosouzapw/OmniRoute/pull/11102)).

changelog.d/fixes/11103-persist-config-audit-log.md

@@ -0,0 +1 @@
+- **Config audit persistence:** persist the configuration audit trail to SQLite (`config_audit_log`) instead of an in-memory buffer capped at 1000 volatile entries, and bound its growth with `cleanupConfigAudit()` driven by the `retention.configAudit` setting (default 30 days), wired into `runAutoCleanup` ([#11103](https://github.com/diegosouzapw/OmniRoute/pull/11103)).

changelog.d/fixes/11109-stream-recovery-toolcall.md

@@ -0,0 +1 @@
+- fix(sse): resume mid-stream recovery after a _completed_ tool call — `finish_reason: "tool_calls"` is now tracked per-call instead of as a general terminal marker, so truncation of trailing prose after a fully-delivered tool call is recoverable while in-flight calls stay blocked ([#11109](https://github.com/diegosouzapw/OmniRoute/pull/11109))

changelog.d/fixes/11116-reasoning-effort-capability-discovery.md

@@ -0,0 +1 @@
+- **fix(providers):** `reasoning_effort` now learns the accepted values from a provider's own 400/422 response and clamps to the highest one instead of forwarding an unsupported `xhigh`/`max` (or a hardcoded `"high"` fallback) — fixes custom OpenAI-compatible connections and registered providers with no reasoning metadata ([#11116](https://github.com/diegosouzapw/OmniRoute/pull/11116)) — thanks @maxmad64bis

changelog.d/fixes/11144-responses-parallel-tool-calls-index.md

@@ -0,0 +1 @@
+- **fix(sse):** parallel `function_call` items in a Responses API stream (e.g. several tool calls dispatched in the same turn) now each get a stable, distinct `index`/`id` when translated to Chat Completions streaming deltas, instead of colliding on index 0 and tripping strict stream parsers with `Expected 'id' to be a string.` ([#11144](https://github.com/diegosouzapw/OmniRoute/pull/11144))

changelog.d/fixes/8864-uncloseai-noauth.md

@@ -0,0 +1 @@
+- fix(dashboard): treat UncloseAI as a no-auth provider so the connect form no longer forces a fake API key (#8864)

changelog.d/fixes/9123-search-provider-local-flag-guard-mismatch.md

@@ -0,0 +1 @@
+- fix(ssrf): make `getProviderOutboundGuard()` (used for search-provider connection validation, image generation and remote image fetch) honor the local-first default `OMNIROUTE_ALLOW_LOCAL_PROVIDER_URLS` the same way the chat validation guard already does, so a LAN-hosted SearXNG/Brave search provider works with only the LOCAL flag set instead of silently requiring `OMNIROUTE_ALLOW_PRIVATE_PROVIDER_URLS` ([#9123](https://github.com/diegosouzapw/OmniRoute/issues/9123)).

changelog.d/fixes/release-v3850-basereds-tests-i18n.md

@@ -0,0 +1 @@
+- fix(i18n): complete Vietnamese translations for recently added UI strings (#9985)

changelog.d/fixes/release-v3850-basereds.md

@@ -0,0 +1,3 @@
+- fix(api): repair broken `@/lib/db/connections` import in the usage utilization route that failed the production build (#10939 follow-up)
+- chore(docs): regenerate PROVIDER_REFERENCE and refresh README diagram SVGs to the real provider count (347)
+- chore(lint): prune ESLint suppressions orphaned on the release branch

changelog.d/maintenance/10906-critical-db-state-assertions.md

@@ -0,0 +1 @@
+- **test(db):** replace three empty `test.skip` placeholders in the critical DB-state suite with real assertions — `resetDbInstance` must swap the singleton while the on-disk row survives, the on-disk DB must open in WAL journal mode, and `db_meta` must hold the seeded `schema_version` — so a regression in any of those invariants can no longer pass as silently green ([#10906](https://github.com/diegosouzapw/OmniRoute/pull/10906))

changelog.d/maintenance/11038-filesize-baseline-fix.md

@@ -0,0 +1 @@
+- fix(quality): rebaseline file-size for modelCapabilities.ts (1016->1072) drift from merged tip fixes (#11034 et al)

changelog.d/maintenance/11053-stryker-oauth-autoimport-registration.md

@@ -0,0 +1 @@
+- fix(quality): register `tests/unit/authz/oauth-autoimport-local-only.test.ts` in stryker `tap.testFiles` (residual of #11053)

changelog.d/maintenance/vi-harimport-parity.md

@@ -0,0 +1 @@
+- fix(i18n): translate the 14 `providers.harImport*` keys into Vietnamese (parity gap left by #11069)

config/quality/dashboard-typecheck-baseline.json

@@ -1,9 +1,6 @@
 {
-  "open-sse/services/payloadRules.ts": {
-    "TS2677": 1
-  },
   "src/app/(dashboard)/dashboard/HomePageClient.tsx": {
-    "TS2339": 16
+    "TS2339": 10
   },
   "src/app/(dashboard)/dashboard/agent-skills/AgentSkillsPageClient.tsx": {
     "TS2503": 3
@@ -120,10 +117,6 @@
   "src/app/(dashboard)/dashboard/providers/[id]/components/CompatibleModelsSection.tsx": {
     "TS2741": 1
   },
-  "src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionRow.tsx": {
-    "TS2345": 3,
-    "TS2322": 1
-  },
   "src/app/(dashboard)/dashboard/providers/[id]/components/ConnectionsListPanel.tsx": {
     "TS2322": 2
   },
@@ -141,12 +134,6 @@
   "src/app/(dashboard)/dashboard/providers/[id]/components/ProviderPlaygroundPanel.tsx": {
     "TS2503": 1
   },
-  "src/app/(dashboard)/dashboard/providers/[id]/components/modals/EditConnectionModal.tsx": {
-    "TS2322": 1
-  },
-  "src/app/(dashboard)/dashboard/providers/[id]/hooks/useModelImportHandlers.ts": {
-    "TS2339": 1
-  },
   "src/app/(dashboard)/dashboard/providers/[id]/hooks/useModelVisibilityHandlers.ts": {
     "TS2339": 15
   },
@@ -190,9 +177,6 @@
   "src/lib/combos/builderDraft.ts": {
     "TS2741": 1
   },
-  "src/lib/providers/codexFastTier.ts": {
-    "TS2367": 1
-  },
   "src/lib/services/htmlRewriter.ts": {
     "TS2322": 2,
     "TS2345": 2
@@ -219,14 +203,7 @@
   "src/shared/hooks/useElectron.ts": {
     "TS2339": 19
   },
-  "src/shared/providers/webSessionCredentials.ts": {
-    "TS2353": 1,
-    "TS2322": 1
-  },
   "src/shared/schemas/cliCatalog.ts": {
     "TS2554": 2
-  },
-  "src/shared/services/opencodeConfig.ts": {
-    "TS2345": 1
   }
 }

config/quality/file-size-baseline.json

@@ -443,21 +443,24 @@
       "src/shared/components/ModelSelectModal.tsx": 1138,
       "src/shared/constants/providers/apikey/gateways.ts": 1250
     },
-    "src/app/(dashboard)/dashboard/providers/[id]/components/modals/AddApiKeyModal.tsx": 1067,
+    "src/app/(dashboard)/dashboard/providers/[id]/components/modals/AddApiKeyModal.tsx": 1080,
     "src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderConnections.ts": 1051,
     "src/shared/components/ModelSelectModal.tsx": 1138,
     "src/shared/constants/providers/apikey/gateways.ts": 1298,
     "open-sse/vendor/codex-chatgpt-web/bridge.ts": 1387,
     "_rebaseline_2026_08_11_v3850_merge_storm_provider_registry": "DRIFT do merge-storm 2026-08-11 (99 PRs mergeados no release/v3.8.50). AddApiKeyModal.tsx (PR #8949 ChatGPT Web provider) e useProviderConnections.ts/ModelSelectModal.tsx (PRs #9011 combo test-all, #9499 image combos) = UI nova legitima acima do cap; gateways.ts = god-file de catalogo de providers que cresceu com PRs #9009/#9421/#9468/#9594 (qualquer split arriscaria corromper o merge de novo — o proprio PR #9421 quebrou o arquivo); bridge.ts (PR #8949) = ponte Chromium vendored; proxyFetch.ts 1207->1220 = drift herdado de merges. Owner autorizou rebaseline com anotacao (2026-08-11).",
-    "src/lib/modelCapabilities.ts": 1016,
+    "src/lib/modelCapabilities.ts": 1072,
+    "_rebaseline_2026_08_21_11034_effort_variants": "DRIFT do tip (base-red #9985): modelCapabilities.ts 1016->1072 (+56) acumulado por PRs ja mergeadas no release/v3.8.50 — principalmente #11034 (resolve effort-variant capabilities a partir do modelo base), alem de #10963/#11040/#10987 growth dos catalogos. Tip puro ficou vermelho neste gate; rebaseline no tip por push direto (owner pre-autorizou crescimento legitimo). Nao tocou no arquivo da #11038.",
     "src/app/(dashboard)/dashboard/providers/[id]/providerPageHelpers.ts": 1014,
     "open-sse/config/imageRegistry.ts": 1034,
     "src/sse/handlers/chatHelpers.ts": 1019,
     "src/shared/middleware/chatBodyAdmission.ts": 1005,
     "_rebaseline_2026_08_20_10668_tabitoken_gateway": "#10668 (yawar-aquil) own catalog growth: src/shared/constants/providers/apikey/gateways.ts 1268->1283 (+15, entirely this PR diff -- one new tabitoken gateway entry, data lines only; base moved from 1255 to 1268 via other merges since the PR forked). Not combination drift: reproducible on the PR branch alone, so the WS5.5 release-captain rule does not apply. Extraction is not available -- the file is pure data (own header: \"Pure data; merged by apikey/index.ts via spread\") and already split into 6 family files under apikey/. Same precedent as _rebaseline_2026_08_14_imagetotext_servicekinds (#10275/#10291, gateways.ts 1250->1255, data lines only) and _rebaseline_2026_08_11_v3850_merge_storm_provider_registry (owner-authorized for this same file).",
-    "open-sse/executors/commandCode.ts": 1038,
+    "open-sse/executors/commandCode.ts": 1059,
     "_rebaseline_2026_08_21_10859_vision_bridge_catalog": "#10859 own growth (Vision Bridge fixes #10808/#10809): src/lib/modelCapabilities.ts 1006->1016 (+10, cmd/gpt-5.3-codex* text-only capability resolution) and open-sse/executors/commandCode.ts 988->1023 (+35, Command Code wire-model normalization for bare ids + reasoning field fallback for opencode-routed gateways). Cohesive bug fixes at the existing capability-resolution / executor chokepoints; not extractable mid-fix. Covered by tests/unit/model-capabilities-command-code-codex-textonly-10703.test.ts, tests/unit/command-code-vision.test.ts, tests/unit/opencode-mimo-reasoning-details-nonstream.test.ts. Pushed directly to release (own-session miss: the original rebaseline was made in a throwaway validation worktree and never landed on the PR branch or the release before merge).",
-    "_rebaseline_2026_08_21_10907_sticky_pin_clear": "#10907 own growth: open-sse/executors/commandCode.ts 1023->1038 (+15, effort-suffix sanitization threading for the sticky-pin-clear fix). Cohesive change at the existing executor chokepoint. Covered by tests/unit/command-code-executor.test.ts."
+    "_rebaseline_2026_08_21_10907_sticky_pin_clear": "#10907 own growth: open-sse/executors/commandCode.ts 1023->1038 (+15, effort-suffix sanitization threading for the sticky-pin-clear fix). Cohesive change at the existing executor chokepoint. Covered by tests/unit/command-code-executor.test.ts.",
+    "_rebaseline_2026_08_21_10986_reasoning_only_content": "#10986 own growth: open-sse/executors/commandCode.ts 1038->1059 (+21, reasoning-only content fallback — when upstream emits only reasoning-delta events and never a text-delta, surface the reasoning text as message.content in createJsonResponse and emit a synthetic content delta in createStreamResponse). Cohesive bug fix at the existing executor chokepoint (mirrors precedent style of #10907/#10859). Covered by tests/unit/command-code-executor.test.ts (2 new cases: non-stream + streaming).",
+    "_rebaseline_2026_08_21_11069_m365_har_import": "#11069 own growth: AddApiKeyModal.tsx 1073->1080 (+7 = Import .har file button for the copilot-m365-web credential modal — M365 is the only provider whose credential (access_token+chathubPath) must be extracted from a DevTools HAR WebSocket URL, added as a new modal affordance). Cohesive UI at the existing modal chokepoint; not extractable. Covered by tests/unit/m365-har-import*.test.ts."
   },
   "_rebaseline_base_2026_08_10_proxyfetch": "Base-red fix (green-prs sweep, issue #9985): open-sse/utils/proxyFetch.ts 1207 > cap 1000 — new proxied-TLS fetch helper introduced by the Fal reference-image work. Owner-authorized quick rebaseline to green; structural slim tracked for v3.9.0.",
   "_rebaseline_2026_07_27_v3849_train2": "Merge-train 2 (7 PRs) — owner-approved 2026-07-27. Single entry: chatCore.ts 4955->5006 (#8595, Responses multi-turn image compaction before the context hard-reject). Genuine irreducible growth at the existing compaction chokepoint in handleChatCore — the PR adds a last-resort retry against the concrete budget plus the estimateFinalInputTokens helper, both wired at the pre-existing call site rather than a new branch. Covered by tests/unit/8560-responses-image-compaction.test.ts (4 tests).",

config/quality/open-sse-typecheck-baseline.json

@@ -1,11 +1,4 @@
 {
-  "open-sse/handlers/chatCore/clientUsageBuffer.ts": {
-    "TS2345": 2
-  },
-  "open-sse/utils/stream.ts": {
-    "TS2345": 2,
-    "TS2322": 2
-  },
   "src/lib/guardrails/videoBridgeHelpers.ts": {
     "TS2488": 1,
     "TS2365": 2,

docs/architecture/RESILIENCE_GUIDE.md

@@ -448,14 +448,14 @@ classification rules pick the fallback `reason` and lock `scope`
 Classification rules only see full error **text** (needed to match body
 markers like `额度不足`) for providers listed in the `FULL_TEXT_RULE_PROVIDERS`
 allowlist in `providerErrorRules.ts` — currently only `"agentrouter"`. For
-every other provider, `checkFallbackError` hands `getProviderErrorRuleMatch`
-only the structured error (`{code, type}`), which is enough for
-header/status/code-based rules but blind to body-text markers. The helper
-`resolveRuleMatchBody()` performs this selection: full error text for
-allowlisted providers, the structured error otherwise. Adding a provider to
-`FULL_TEXT_RULE_PROVIDERS` is an explicit per-provider opt-in — it exists so
-that the default path for every provider not on the list stays
-byte-for-byte unchanged.
+every other **built-in catalog** provider, `checkFallbackError` hands
+`getProviderErrorRuleMatch` only the structured error (`{code, type}`), which
+is enough for header/status/code-based rules but blind to body-text markers.
+The helper `resolveRuleMatchBody()` performs this selection: full error text
+for allowlisted providers, the structured error otherwise. Adding a
+**built-in** provider to `FULL_TEXT_RULE_PROVIDERS` is an explicit per-provider
+opt-in — it exists so that the default path for every provider not on the
+list stays byte-for-byte unchanged.
 
 A rule's `scope` (`model` / `provider` / `connection`) is a separate opt-in
 from `FULL_TEXT_RULE_PROVIDERS`: `checkFallbackError` only surfaces it as
@@ -466,6 +466,31 @@ honorsRuleLockScope()` — today only `"agentrouter"`). See "Restated quota
 errors" above for what a `scope: "connection"` match actually does once a
 provider is on that allowlist.
 
+**#11104 — operator-declared rules bypass both allowlists.** An operator can
+declare a per-provider rule at runtime via `settings.providerErrorRules`
+(`open-sse/config/providerErrorRules.ts::setOperatorProviderErrorRules`)
+without editing this file. Gating an operator rule behind
+`FULL_TEXT_RULE_PROVIDERS`/`HONORS_RULE_LOCK_SCOPE_PROVIDERS` — allowlists
+meant to protect the **default** behavior of built-in catalog rules — would
+make the settings mechanism inert for every provider except the ones already
+listed there, since declaring the rule is already the operator's explicit
+opt-in. `resolveRuleMatchBody()` and `honorsRuleLockScope()` both check
+`hasOperatorRuleForProvider()` first: a provider with an operator rule gets
+the raw error text and has its declared `scope` honored, regardless of
+whether it also appears in either allowlist.
+
+**Known gap — `providerRuleRegistry` is never consulted for HTTP 400.**
+`checkFallbackError`'s `BAD_REQUEST` branch classifies status 400 entirely
+through its own pattern arrays (`MODEL_ACCESS_DENIED_PATTERNS`,
+`CONTEXT_OVERFLOW_PATTERNS`, etc. in `accountFallback.ts`) and returns before
+the `configuredRule`/`getProviderErrorRuleMatch` branch above it is reached.
+A built-in catalog rule (or an operator rule) with `status: 400` is
+syntactically valid but will never fire. No existing rule targets 400 today,
+so nothing in production is affected — but a future 400 rule needs this
+branch touched first, which is a larger change than adding a rule (it
+reclassifies 400 for every provider already relying on the pattern-array
+behavior) and is out of scope for a single-provider rule addition.
+
 ### Adding a new quota-misstating gateway
 
 1. Register one rule array in `statusRestatementRegistry`

docs/reference/FREE_TIERS.md

@@ -120,7 +120,6 @@ A 50-agent web-research pass (official docs + last-7-days news, adversarially ve
 | `firecrawl`      | caution   | Cloud API ToS has no explicit personal-proxy prohibition found, but the open-source self-hosted version is AGPL-3.0 (re… |
 | `gemini`         | caution   | ToS explicitly states the free tier is for "developers building with Google AI models for professional or business purp… |
 | `groq`           | caution   | Services Agreement §6.3 prohibits reselling, sublicensing, or distributing API access; §3.2 bars reselling/leasing acco… |
-| `hackclub`       | caution   | Service is explicitly scoped to Hack Club teen members building projects/learning; no public ToS found explicitly permi… |
 | `huggingchat`    | caution   | Hugging Face ToS does not explicitly ban personal self-hosted proxies, but supplemental terms (referenced but not fully… |
 | `huggingface`    | caution   | ToS grants a limited license to access/use the service; the document does not explicitly permit or forbid a single-user… |
 | `hyperbolic`     | caution   | ToS grants API access "solely for your own personal or internal business purposes" and explicitly prohibits licensing, … |
@@ -222,7 +221,6 @@ A 50-agent web-research pass (official docs + last-7-days news, adversarially ve
 | `duckduckgo-web` | keyless       | —                | —                  | avoid     | 6      |
 | `freemodel-dev`  | keyless       | —                | —                  | unknown   | 4      |
 | `friendliai`     | keyless       | —                | —                  | avoid     | 2      |
-| `hackclub`       | keyless       | —                | —                  | caution   | 3      |
 | `iflytek`        | keyless       | —                | —                  | avoid     | 1      |
 | `inference-net`  | keyless       | —                | —                  | caution   | 3      |
 | `liquid`         | keyless       | —                | —                  | unknown   | 1      |
@@ -280,7 +278,6 @@ A 50-agent web-research pass (official docs + last-7-days news, adversarially ve
 - **`gitlawb`** — The shipped freeNote "Free tier available" is effectively stale. The original free MiMo access was removed in May 2026; the only remaining "free" option is a temporary promotional model (Nemotron 3 U…
 - **`gitlawb-gmi`** — Partially still accurate — free tier exists but is now narrowed to a single model (Nemotron 3 Ultra) after MiMo free access was revoked in late May 2026. The shipped note "Free tier available" unders…
 - **`groq`** — The shipped freeNote "30 RPM / 14.4K RPD" is accurate only for llama-3.1-8b-instant. Most other models (including llama-3.3-70b-versatile) have a much lower 1K RPD cap. The note omits model-specific …
-- **`hackclub`** — The "30+ models" count appears accurate and still matches. The core offering remains free for Hack Club members. No evidence of tightening — still "$0 ALWAYS FREE" per the homepage. The freeNote omit…
 - **`huggingchat`** — The shipped freeNote ("Free LLM chat — no subscription required. Rate limits apply.") is partially accurate but significantly understates the restrictions. The free tier now operates on a hard $0.10/…
 - **`huggingface`** — Significantly tightened. The shipped freeNote ("Free Inference API for thousands of models") implied unlimited/generous free access, but as of mid-2025 the free tier is capped at $0.10/month in recur…
 - **`hyperbolic`** — Our shipped freeNote says "$1-5 trial credits on signup" — the $1 trial credit portion is accurate, but the "$5" figure refers to the minimum deposit required to unlock GPU rental (not free credits g…

docs/reference/PROVIDER_REFERENCE.md

@@ -213,7 +213,6 @@ Use the dashboard at `/dashboard/providers` to enable, configure, and test each
 | `glm-cn` | `glmcn` | GLM Coding (China) | API key | [link](https://open.bigmodel.cn) | — |
 | `glmt` | `glmt` | GLM Thinking | API key | [link](https://open.bigmodel.cn) | — |
 | `groq` | `groq` | Groq | API key | [link](https://groq.com) | Free tier: 30 RPM / 14.4K RPD — no credit card |
-| `hackclub` | `hc` | Hackclub AI | API key, aggregator | [link](https://ai.hackclub.com) | Sign in with your Hack Club account at ai.hackclub.com. |
 | `haiper` | `hp` | Haiper | API key, video | [link](https://haiper.ai) | Get API key at haiper.ai/haiper-api |
 | `hcnsec` | `hcnsec` | Huancheng Public API | API key | [link](https://api.hcnsec.cn) | Get API key at api.hcnsec.cn |
 | `helixmind` | `helixmind` | HelixMind | API key, aggregator | [link](https://helixmind.online) | Previously circulated 3 RPM/50 RPD and no-card claims were not confirmed during the 2026-08-02 audit; current quota and billing require account verification. |

open-sse/config/constants.ts

@@ -171,6 +171,7 @@ export const HTTP_STATUS = {
   FORBIDDEN: 403,
   NOT_FOUND: 404,
   NOT_ACCEPTABLE: 406,
+  UNPROCESSABLE_ENTITY: 422,
   REQUEST_TIMEOUT: 408,
   GONE: 410,
   RATE_LIMITED: 429,
@@ -263,11 +264,17 @@ export const PROVIDER_PROFILES = {
     circuitBreakerReset: envInt("OMNIROUTE_CIRCUIT_BREAKER_API_KEY_RESET_MS", 30000),
     // Provider-level circuit breaker (entire provider cooldown after repeated failures)
     providerFailureThreshold: envInt("OMNIROUTE_PROVIDER_BREAKER_API_KEY_FAILURE_THRESHOLD", 15), // Scaled for 500+ connections (was 5)
-    providerFailureWindowMs: envInt("OMNIROUTE_PROVIDER_BREAKER_API_KEY_FAILURE_WINDOW_MS", 1800000), // 30min window (was 20min)
+    providerFailureWindowMs: envInt(
+      "OMNIROUTE_PROVIDER_BREAKER_API_KEY_FAILURE_WINDOW_MS",
+      1800000
+    ), // 30min window (was 20min)
     providerCooldownMs: envInt("OMNIROUTE_PROVIDER_BREAKER_API_KEY_COOLDOWN_MS", 600000), // 10min cooldown when threshold reached
     degradationThreshold: envInt("OMNIROUTE_PROVIDER_BREAKER_API_KEY_DEGRADATION_THRESHOLD", 7),
     maxBackoffMultiplier: envInt("OMNIROUTE_PROVIDER_BREAKER_API_KEY_MAX_BACKOFF_MULTIPLIER", 4),
-    backoffEscalationCount: envInt("OMNIROUTE_PROVIDER_BREAKER_API_KEY_BACKOFF_ESCALATION_COUNT", 3),
+    backoffEscalationCount: envInt(
+      "OMNIROUTE_PROVIDER_BREAKER_API_KEY_BACKOFF_ESCALATION_COUNT",
+      3
+    ),
   },
   // Local providers (localhost inference backends like Ollama, LM Studio, oMLX).
   // Not yet wired into getProviderProfile() — will be used when local provider_nodes

open-sse/config/freeModelCatalog.data.ts

@@ -194,9 +194,6 @@ export const FREE_MODEL_BUDGETS: FreeModelBudget[] = [
   { provider: "groq", modelId: "openai/gpt-oss-120b", displayName: "GPT-OSS 120B", monthlyTokens: 15000000, creditTokens: 0, freeType: "recurring-daily", poolKey: "groq", tos: "caution", hardStopGuaranteed: true },
   { provider: "groq", modelId: "openai/gpt-oss-20b", displayName: "GPT-OSS 20B", monthlyTokens: 15000000, creditTokens: 0, freeType: "recurring-daily", poolKey: "groq", tos: "caution", hardStopGuaranteed: true },
   { provider: "groq", modelId: "qwen/qwen3-32b", displayName: "Qwen3 32B", monthlyTokens: 15000000, creditTokens: 0, freeType: "recurring-daily", poolKey: "groq", tos: "caution", hardStopGuaranteed: true },
-  { provider: "hackclub", modelId: "meta-llama/llama-3.3-70b-instruct", displayName: "Llama 3.3 70B", monthlyTokens: 0, creditTokens: 0, freeType: "keyless", poolKey: "hackclub", tos: "caution" },
-  { provider: "hackclub", modelId: "mistralai/mistral-7b-instruct", displayName: "Mistral 7B", monthlyTokens: 0, creditTokens: 0, freeType: "keyless", poolKey: "hackclub", tos: "caution" },
-  { provider: "hackclub", modelId: "deepseek-ai/deepseek-coder-33b", displayName: "DeepSeek Coder 33B", monthlyTokens: 0, creditTokens: 0, freeType: "keyless", poolKey: "hackclub", tos: "caution" },
   { provider: "huggingchat", modelId: "baidu/ERNIE-4.5-VL-424B-A47B-Base-PT", displayName: "ERNIE 4.5 VL 424B A47B Base PT", monthlyTokens: 500000, creditTokens: 0, freeType: "recurring-monthly", poolKey: "huggingchat", tos: "caution" },
   { provider: "huggingchat", modelId: "CohereLabs/c4ai-command-r7b-12-2024", displayName: "Command R7B 12-2024", monthlyTokens: 500000, creditTokens: 0, freeType: "recurring-monthly", poolKey: "huggingchat", tos: "caution" },
   { provider: "huggingchat", modelId: "CohereLabs/command-a-reasoning-08-2025", displayName: "Command A Reasoning 08-2025", monthlyTokens: 500000, creditTokens: 0, freeType: "recurring-monthly", poolKey: "huggingchat", tos: "caution" },

open-sse/config/opencodeZenGoSharedModels.ts

@@ -0,0 +1,16 @@
+/**
+ * Models declared identically in both the `opencode-zen` and `opencode-go` provider
+ * registries (same upstream family, opencode.ai/zen/*). Mirrors the GLM_SHARED_MODELS
+ * pattern in glmProvider.ts: one array, spread into each sibling RegistryEntry, so a
+ * metadata fix (targetFormat, supportsReasoning, ...) only has to land in one file
+ * instead of drifting out of sync across registries.
+ *
+ * Only entries that are byte-identical across both registries belong here — a model
+ * with tier-specific flags (e.g. go's effort variants, or a flag only one tier needs)
+ * stays local to that registry's own `models` array.
+ */
+export const OPENCODE_ZEN_GO_SHARED_MODELS = Object.freeze([
+  { id: "kimi-k2.7-code", name: "Kimi K2.7 Code" },
+  { id: "qwen3.5-plus", name: "Qwen3.5 Plus", targetFormat: "claude", supportsVision: false },
+  { id: "qwen3.6-plus", name: "Qwen3.6 Plus", targetFormat: "claude", supportsVision: false },
+]);

open-sse/config/providerErrorRules.ts

@@ -30,21 +30,63 @@ export type ProviderErrorRule = {
 export type ProviderErrorRuleMatch = {
   reason: ConfiguredErrorReason;
   /**
-   * Intended lock scope. #10334: this field is CONSUMED end-to-end only for
-   * providers in `HONORS_RULE_LOCK_SCOPE_PROVIDERS` (agentrouter-exclusive
-   * today, gated by `honorsRuleLockScope()`) — for those, `checkFallbackError`
-   * surfaces it as `ruleScope` on its return value for the persistence layer
-   * to honor instead of re-deriving scope from `hasPerModelQuota()`. For
-   * every other provider it remains INFORMATIONAL: `getProviderErrorRuleMatch`
-   * callers still read only `reason`/`cooldownMs`, and the actual lock scope
-   * is decided independently by each call site. Widening the allowlist is
-   * tracked as a follow-up — see `docs/architecture/RESILIENCE_GUIDE.md` §7.
+   * Intended lock scope. #10334: for a BUILT-IN catalog rule, this field is
+   * CONSUMED end-to-end only for providers in `HONORS_RULE_LOCK_SCOPE_PROVIDERS`
+   * (agentrouter-exclusive today, gated by `honorsRuleLockScope()`) — for those,
+   * `checkFallbackError` surfaces it as `ruleScope` on its return value for the
+   * persistence layer to honor instead of re-deriving scope from
+   * `hasPerModelQuota()`. For every other built-in-rule provider it remains
+   * INFORMATIONAL. #11104: an OPERATOR-declared rule (`OperatorProviderErrorRule`)
+   * is exempt from this allowlist — `honorsRuleLockScope()` always returns true
+   * when the provider has one, since the operator already opted in by declaring
+   * the rule. Widening `HONORS_RULE_LOCK_SCOPE_PROVIDERS` itself (for a new
+   * built-in catalog rule) is tracked as a follow-up — see
+   * `docs/architecture/RESILIENCE_GUIDE.md` §7.
    */
   scope: "model" | "provider" | "connection";
   /** Optional explicit cooldown; falls back to the existing per-reason defaults. */
   cooldownMs?: number;
 };
 
+/**
+ * Operator-declared per-provider error rule (settings-driven).
+ *
+ * Mirrors the catalog `ProviderErrorRule` but is data-only so an operator can
+ * add a scope/cooldown/reason override for a provider without editing this
+ * file. `match` is a plain case-insensitive SUBSTRING of the error body — never
+ * a RegExp — so an operator-supplied pattern can never introduce a ReDoS on the
+ * error-classification hot path. Bounded to <= 50 rules total by the settings
+ * schema. An operator rule is consulted BEFORE the built-in `providerRuleRegistry`
+ * and wins on the first status+substring match for a provider.
+ */
+export type OperatorProviderErrorRule = {
+  status: number;
+  match: string;
+  scope: "model" | "provider" | "connection";
+  reason?: ConfiguredErrorReason;
+  cooldownMs?: number;
+};
+
+let operatorProviderErrorRules: Record<string, OperatorProviderErrorRule[]> = {};
+
+/**
+ * Inject operator-declared rules. Called from the runtime-settings applier
+ * (`applyRuntimeSettings`) once at boot and on every settings update, with the
+ * value validated by the settings schema. Pass `undefined`/empty/null to clear.
+ * Provider keys are lowercased so lookups are case-insensitive.
+ */
+export function setOperatorProviderErrorRules(
+  rules: Record<string, OperatorProviderErrorRule[]> | undefined | null
+): void {
+  operatorProviderErrorRules = {};
+  if (!rules) return;
+  for (const [provider, list] of Object.entries(rules)) {
+    if (Array.isArray(list) && list.length > 0) {
+      operatorProviderErrorRules[provider.toLowerCase()] = list;
+    }
+  }
+}
+
 // ─── Opencode ───────────────────────────────────────────────────────────────────
 // Opencode Go uses an account-wide quota. The body usually says "rate limit
 // reached" but the presence of `x-ratelimit-remaining-requests: 0` is the
@@ -272,11 +314,21 @@ export const providerRuleRegistry = new Map<string, ProviderErrorRule[]>([
  * FULL_TEXT_RULE_PROVIDERS: that set controls what body a rule matches against
  * (input), this one controls whether the matched scope changes caller behavior
  * (output). A provider could need one without the other.
+ *
+ * Providers with an operator-declared rule (`setOperatorProviderErrorRules`)
+ * are honored too, without being added here: the allowlist exists to gate
+ * BUILT-IN catalog rules, which change default behavior for every operator
+ * running that provider — an operator rule is already an explicit, per-operator
+ * opt-in, so gating it a second time behind this list would make the settings
+ * mechanism (#11104) silently inert for every provider except the ones listed
+ * below. See `hasOperatorRuleForProvider`.
  */
 const HONORS_RULE_LOCK_SCOPE_PROVIDERS = new Set(["agentrouter"]);
 
 export function honorsRuleLockScope(provider: string | null | undefined): boolean {
-  return !!provider && HONORS_RULE_LOCK_SCOPE_PROVIDERS.has(provider.toLowerCase());
+  if (!provider) return false;
+  const key = provider.toLowerCase();
+  return HONORS_RULE_LOCK_SCOPE_PROVIDERS.has(key) || hasOperatorRuleForProvider(key);
 }
 
 /**
@@ -310,28 +362,51 @@ export function egressBucketedLockProviders(): string[] {
 }
 
 /**
- * Providers whose rules match on the FULL upstream error text.
- * checkFallbackError's rule lookup normally passes only the structured
+ * Providers whose BUILT-IN catalog rules match on the FULL upstream error
+ * text. checkFallbackError's rule lookup normally passes only the structured
  * error ({code, type} — message stripped by the combo callers), which is
  * enough for header/status/code rules but blind to body-text markers like
  * agentrouter's "额度不足". Providers in this set get the raw error text as
  * the match body instead. EXCLUSIVE allowlist by owner decision (2026-08-13):
  * adding a provider here is an explicit opt-in — the default path for every
  * other provider must remain byte-for-byte unchanged.
+ *
+ * Operator-declared rules bypass this allowlist entirely (see
+ * `hasOperatorRuleForProvider`): the operator's `match` is a literal substring
+ * of the error body by construction, so a rule that never sees body text could
+ * never match anything, defeating the point of declaring it.
  */
 const FULL_TEXT_RULE_PROVIDERS = new Set(["agentrouter"]);
 
+/**
+ * True when an operator has declared at least one rule for this provider via
+ * `settings.providerErrorRules` (injected through `setOperatorProviderErrorRules`).
+ * Presence of the rule IS the opt-in — no separate allowlist to maintain, and
+ * no widening decision needed as new operators configure new providers.
+ */
+export function hasOperatorRuleForProvider(provider: string | null | undefined): boolean {
+  if (!provider) return false;
+  const rules = operatorProviderErrorRules[provider.toLowerCase()];
+  return !!rules && rules.length > 0;
+}
+
 /**
  * Resolve the body handed to getProviderErrorRuleMatch inside
- * checkFallbackError: full error text for FULL_TEXT_RULE_PROVIDERS,
- * the structured error for everyone else.
+ * checkFallbackError: full error text for FULL_TEXT_RULE_PROVIDERS or any
+ * provider with an operator-declared rule, the structured error for everyone
+ * else.
  */
 export function resolveRuleMatchBody(
   provider: string | null | undefined,
   structuredError: unknown,
   errorText: string | null | undefined
 ): unknown {
-  if (provider && FULL_TEXT_RULE_PROVIDERS.has(provider.toLowerCase()) && errorText) {
+  if (
+    provider &&
+    (FULL_TEXT_RULE_PROVIDERS.has(provider.toLowerCase()) ||
+      hasOperatorRuleForProvider(provider)) &&
+    errorText
+  ) {
     return errorText;
   }
   return structuredError ?? null;
@@ -346,10 +421,32 @@ export function getProviderErrorRuleMatch(
   provider: string | null | undefined,
   status: number,
   headers: Headers | Record<string, string> | null | undefined,
-  body?: unknown
+  body?: unknown,
+  operatorRules?: Record<string, OperatorProviderErrorRule[]>
 ): ProviderErrorRuleMatch | null {
   if (!provider) return null;
-  const rules = providerRuleRegistry.get(provider.toLowerCase());
+  const key = provider.toLowerCase();
+
+  // Operator-declared rules win first: an operator can override any catalog
+  // rule for a provider without editing this file. `operatorRules` is the
+  // injected source (tests / direct callers); when omitted we fall back to the
+  // settings-backed cache populated by `setOperatorProviderErrorRules`.
+  const opRules = (operatorRules ?? operatorProviderErrorRules)?.[key];
+  if (opRules && opRules.length > 0) {
+    const text = typeof body === "string" ? body : JSON.stringify(body ?? "");
+    const lowered = text.toLowerCase();
+    for (const r of opRules) {
+      if (r.status === status && lowered.includes(r.match.toLowerCase())) {
+        return {
+          reason: r.reason ?? "quota_exhausted",
+          scope: r.scope,
+          cooldownMs: r.cooldownMs,
+        };
+      }
+    }
+  }
+
+  const rules = providerRuleRegistry.get(key);
   if (!rules) return null;
   // Normalize headers: accept either a `Headers` object (from `fetch()`) or
   // a plain record. Provider rules access headers via plain object indexing.

open-sse/config/providerRegistry.ts

@@ -10,6 +10,7 @@ export {
 } from "./providers/registry/alibaba/index.ts";
 export { REGISTRY } from "./providers/index.ts";
 import { REGISTRY } from "./providers/index.ts";
+import { isPrivateHost } from "@/shared/network/outboundUrlGuard";
 import {
   RegistryModel,
   REASONING_UNSUPPORTED,
@@ -132,11 +133,8 @@ export function isLocalProvider(baseUrl?: string | null): boolean {
   try {
     const url = new URL(baseUrl);
     const hostname = url.hostname;
-    // Strictly matching 172.16.0.0/12 (Docker/local) and explicitly blocking ::1 per SSRF hardening
-    return (
-      LOCAL_HOSTNAMES.has(hostname) ||
-      /^172\.(1[6-9]|2[0-9]|3[0-1])\.\d{1,3}\.\d{1,3}$/.test(hostname)
-    );
+    if (!hostname) return false;
+    return LOCAL_HOSTNAMES.has(hostname) || isPrivateHost(hostname);
   } catch {
     return false;
   }

open-sse/config/providers/index.ts

@@ -70,7 +70,6 @@ import { togetherProvider } from "./registry/together/index.ts";
 import { cohereProvider } from "./registry/cohere/index.ts";
 import { cursorProvider, cursor_apiProvider } from "./registry/cursor/index.ts";
 import { volcengineProvider } from "./registry/volcengine/index.ts";
-import { hackclubProvider } from "./registry/hackclub/index.ts";
 import { freetheaiProvider } from "./registry/freetheai/index.ts";
 import { g4f_groqProvider } from "./registry/g4f-groq/index.ts";
 import { g4f_geminiProvider } from "./registry/g4f-gemini/index.ts";
@@ -336,7 +335,6 @@ export const REGISTRY: Record<string, RegistryEntry> = {
   cursor: cursorProvider,
   "cursor-api": cursor_apiProvider,
   volcengine: volcengineProvider,
-  hackclub: hackclubProvider,
   freetheai: freetheaiProvider,
   "g4f-groq": g4f_groqProvider,
   "g4f-gemini": g4f_geminiProvider,

open-sse/config/providers/registry/blackbox/index.ts

@@ -5,6 +5,12 @@ export const blackboxProvider: RegistryEntry = {
   alias: "bb",
   format: "openai",
   executor: "default",
+  // NOTE: api.blackbox.ai returns HTTP 404 on /v1/chat/completions and /v1/models
+  // (empty body, all path variants) since sweep 2026-08-21; the public inference
+  // surface has moved to the gated enterprise.blackbox.ai/v1 endpoint. The provider
+  // is marked deprecated in src/shared/constants/providers/apikey/frontier-labs.ts —
+  // this registry entry is kept intact (registration/execution unaffected), so
+  // existing configured keys keep working if a restored/enterprise host is reachable.
   baseUrl: "https://api.blackbox.ai/v1/chat/completions",
   modelsUrl: "https://api.blackbox.ai/v1/models",
   authType: "apikey",

open-sse/config/providers/registry/cline/index.ts

@@ -27,7 +27,7 @@ export const clineProvider: RegistryEntry = {
   // the official free bucket and text-output models advertised as zero-cost.
   models: [
     {
-      id: "zai/glm-5.2",
+      id: "z-ai/glm-5.2",
       name: "GLM 5.2",
       toolCalling: true,
       supportsReasoning: true,

open-sse/config/providers/registry/command-code/index.ts

@@ -8,7 +8,11 @@ export const command_codeProvider: RegistryEntry = {
   format: "openai",
   executor: "command-code",
   baseUrl: "https://api.commandcode.ai",
-  chatPath: "/alpha/generate",
+  // Chat uses the documented /provider/v1/chat/completions (OpenAI-format)
+  // endpoint — NOT the CLI-only /alpha/generate endpoint, which Command Code
+  // version-gates and proxy-blocks for external callers (#10265). Discovery
+  // already targets the sibling /provider/v1/models endpoint.
+  chatPath: "/provider/v1/chat/completions",
   modelsUrl: "https://api.commandcode.ai/provider/v1/models",
   // The discovery response is a partial routing catalog; static registry
   // entries omitted from it can still be accepted by the gateway.

open-sse/config/providers/registry/dify/index.ts

@@ -5,7 +5,11 @@ export const difyProvider: RegistryEntry = {
   alias: "dify",
   format: "openai",
   executor: "default",
-  baseUrl: "https://api.dify.ai/v1/chat/completions",
+  // Dify does not serve /chat/completions — its native completion route is
+  // POST /v1/chat-messages (validated via the dedicated dify validator, #11002).
+  // Keep this as the bare API root so route suffixes build correctly and
+  // self-hosted instances can override the base URL per connection.
+  baseUrl: "https://api.dify.ai",
   authType: "apikey",
   authHeader: "bearer",
   models: [{ id: "auto", name: "Auto" }],

open-sse/config/providers/registry/hackclub/index.ts

@@ -1,19 +0,0 @@
-import type { RegistryEntry } from "../../shared.ts";
-
-export const hackclubProvider: RegistryEntry = {
-  id: "hackclub",
-  alias: "hc",
-  format: "openai",
-  executor: "default",
-  baseUrl: "https://ai.hackclub.com/proxy/v1/chat/completions",
-  modelsUrl: "https://ai.hackclub.com/proxy/v1/models",
-  authType: "optional",
-  authHeader: "bearer",
-  passthroughModels: true,
-  defaultContextLength: 128000,
-  models: [
-    { id: "meta-llama/llama-3.3-70b-instruct", name: "Llama 3.3 70B" },
-    { id: "mistralai/mistral-7b-instruct", name: "Mistral 7B" },
-    { id: "deepseek-ai/deepseek-coder-33b", name: "DeepSeek Coder 33B" },
-  ],
-};

open-sse/config/providers/registry/minimax/web/index.ts

@@ -16,7 +16,7 @@ export const hailuo_webProvider: RegistryEntry = {
   alias: "hailuo-web",
   format: "openai",
   executor: "hailuo-web",
-  baseUrl: "https://www.hailuo.ai",
+  baseUrl: "https://chat.minimax.io",
   authType: "apikey",
   authHeader: "bearer",
   models: HAILUO_WEB_STATIC_MODELS,

open-sse/config/providers/registry/opencode/go/index.ts

@@ -1,4 +1,5 @@
 import type { RegistryEntry } from "../../../shared.ts";
+import { OPENCODE_ZEN_GO_SHARED_MODELS } from "../../../shared.ts";
 
 export const opencode_goProvider: RegistryEntry = {
   id: "opencode-go",
@@ -23,9 +24,13 @@ export const opencode_goProvider: RegistryEntry = {
     { id: "glm-5.2", name: "GLM-5.2", supportsReasoning: true },
     { id: "glm-5.2-high", name: "GLM-5.2 (high effort)", supportsReasoning: true },
     { id: "glm-5.2-max", name: "GLM-5.2 (max effort)", supportsReasoning: true },
+
+    ...OPENCODE_ZEN_GO_SHARED_MODELS,
+    // models[0] (glm-5.2) is the dashboard default (LlmChatCard/ProviderTestSlideOver take models[0]).
+
     { id: "glm-5.1", name: "GLM-5.1" },
     { id: "glm-5", name: "GLM-5" },
-    { id: "kimi-k2.7-code", name: "Kimi K2.7 Code" },
+    // kimi-k2.7-code declared identically on opencode-zen — see OPENCODE_ZEN_GO_SHARED_MODELS.
     { id: "kimi-k2.6", name: "Kimi K2.6" },
     { id: "kimi-k2.5", name: "Kimi K2.5" },
     // #8353: Kimi K3 base + max-effort alias from the OpenCode Go registry.
@@ -89,7 +94,8 @@ export const opencode_goProvider: RegistryEntry = {
       supportsVision: false,
       supportsReasoning: true,
     },
-    { id: "qwen3.6-plus", name: "Qwen3.6 Plus", targetFormat: "claude", supportsVision: false },
+    // qwen3.6-plus / qwen3.5-plus base ids declared identically on opencode-zen — see
+    // OPENCODE_ZEN_GO_SHARED_MODELS.
     {
       id: "qwen3.6-plus-high",
       name: "Qwen3.6 Plus (high effort)",
@@ -104,7 +110,6 @@ export const opencode_goProvider: RegistryEntry = {
       supportsVision: false,
       supportsReasoning: true,
     },
-    { id: "qwen3.5-plus", name: "Qwen3.5 Plus", targetFormat: "claude", supportsVision: false },
     // #8353: hy3 is the Go-tier base id (distinct from hy3-preview / hy3-free).
     { id: "hy3", name: "Hunyuan3", contextLength: 256000, supportsReasoning: true },
     {
@@ -138,6 +143,7 @@ export const opencode_goProvider: RegistryEntry = {
       supportsVision: true,
       supportsAudio: true,
       supportsVideo: true,
+      targetFormat: "openai-responses",
     },
     {
       id: "muse-spark-1.2-contributor-minimal",
@@ -148,6 +154,7 @@ export const opencode_goProvider: RegistryEntry = {
       supportsVision: true,
       supportsAudio: true,
       supportsVideo: true,
+      targetFormat: "openai-responses",
     },
     {
       id: "muse-spark-1.2-contributor-low",
@@ -158,6 +165,7 @@ export const opencode_goProvider: RegistryEntry = {
       supportsVision: true,
       supportsAudio: true,
       supportsVideo: true,
+      targetFormat: "openai-responses",
     },
     {
       id: "muse-spark-1.2-contributor-medium",
@@ -168,6 +176,7 @@ export const opencode_goProvider: RegistryEntry = {
       supportsVision: true,
       supportsAudio: true,
       supportsVideo: true,
+      targetFormat: "openai-responses",
     },
     {
       id: "muse-spark-1.2-contributor-high",
@@ -178,6 +187,7 @@ export const opencode_goProvider: RegistryEntry = {
       supportsVision: true,
       supportsAudio: true,
       supportsVideo: true,
+      targetFormat: "openai-responses",
     },
     {
       id: "muse-spark-1.2-contributor-xhigh",
@@ -188,6 +198,7 @@ export const opencode_goProvider: RegistryEntry = {
       supportsVision: true,
       supportsAudio: true,
       supportsVideo: true,
+      targetFormat: "openai-responses",
     },
     // #8353: Grok 4.5 + effort tiers from the OpenCode Go registry.
     { id: "grok-4.5", name: "Grok 4.5", supportsReasoning: true },

open-sse/config/providers/registry/opencode/zen/index.ts

@@ -1,4 +1,5 @@
 import type { RegistryEntry } from "../../../shared.ts";
+import { OPENCODE_ZEN_GO_SHARED_MODELS } from "../../../shared.ts";
 
 export const opencode_zenProvider: RegistryEntry = {
   id: "opencode-zen",
@@ -25,6 +26,10 @@ export const opencode_zenProvider: RegistryEntry = {
       supportsReasoning: true,
       interleavedField: "reasoning_content",
     },
+
+    ...OPENCODE_ZEN_GO_SHARED_MODELS,
+    // models[0] (big-pickle) is the dashboard default; SHARED spread kept after it.
+
     { id: "gpt-5.6-sol", name: "GPT 5.6 Sol" },
     { id: "gpt-5.6-terra", name: "GPT 5.6 Terra" },
     { id: "gpt-5.6-luna", name: "GPT 5.6 Luna" },
@@ -51,7 +56,27 @@ export const opencode_zenProvider: RegistryEntry = {
     { id: "grok-4.6", name: "Grok 4.6" },
 
     // ── Muse ───────────────────────────────────────────────────
-    { id: "muse-spark-1.2", name: "Muse Spark 1.2" },
+    // Muse Spark is served by OpenCode Zen only on the OpenAI Responses API
+    // endpoint, not /chat/completions (see the opencode provider's own
+    // muse-spark entries, #10874/#10867) — this provider is a separate
+    // registry entry for the same upstream and never got the same
+    // targetFormat declaration, so requests routed here still hit
+    // /chat/completions with a mismatched or unanswerable body and the
+    // upstream returns an empty message.
+    {
+      id: "muse-spark-1.2",
+      name: "Muse Spark 1.2",
+      supportsReasoning: true,
+      targetFormat: "openai-responses",
+    },
+    // Explicit wire-format overlay of the base opencode provider's muse-spark entry
+    // (targetFormat: openai-responses). Keep in sync with base on catalog syncs.
+    {
+      id: "muse-spark-1.2-contributor-free",
+      name: "Muse Spark 1.2 Contributor Free",
+      supportsReasoning: true,
+      targetFormat: "openai-responses",
+    },
 
     // ── DeepSeek ────────────────────────────────────────────────
     { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro" },
@@ -66,7 +91,7 @@ export const opencode_zenProvider: RegistryEntry = {
 
     // ── Kimi / Moonshot ────────────────────────────────────────
     { id: "kimi-k3", name: "Kimi K3" },
-    { id: "kimi-k2.7-code", name: "Kimi K2.7 Code" },
+    // kimi-k2.7-code declared identically on opencode-go — see OPENCODE_ZEN_GO_SHARED_MODELS.
 
     // ── Qwen ───────────────────────────────────────────────────
     // Issue #2292: Qwen models return Claude-format SSE bodies even
@@ -74,8 +99,8 @@ export const opencode_zenProvider: RegistryEntry = {
     // through /messages and the Claude translator.
     // Issue #2822: These models are text-only — supportsVision: false
     // ensures combo routing skips them on image-bearing requests.
-    { id: "qwen3.5-plus", name: "Qwen3.5 Plus", targetFormat: "claude", supportsVision: false },
-    { id: "qwen3.6-plus", name: "Qwen3.6 Plus", targetFormat: "claude", supportsVision: false },
+    // qwen3.5-plus / qwen3.6-plus declared identically on opencode-go — see
+    // OPENCODE_ZEN_GO_SHARED_MODELS.
 
     // ── Free Tier ──────────────────────────────────────────────
     // #6998 (2026-07-14): upstream free tier rotated — minimax-m2.5-free,

open-sse/config/providers/shared.ts

@@ -25,6 +25,7 @@ import {
   GLMT_TIMEOUT_MS,
   GLM_SHARED_MODELS,
 } from "../glmProvider.ts";
+import { OPENCODE_ZEN_GO_SHARED_MODELS } from "../opencodeZenGoSharedModels.ts";
 import { MARITALK_DEFAULT_BASE_URL } from "../maritalk.ts";
 import {
   CURSOR_REGISTRY_VERSION,
@@ -719,6 +720,7 @@ export {
   GLM_TIMEOUT_MS,
   GLMT_TIMEOUT_MS,
   GLM_SHARED_MODELS,
+  OPENCODE_ZEN_GO_SHARED_MODELS,
   MARITALK_DEFAULT_BASE_URL,
   CURSOR_REGISTRY_VERSION,
   getAntigravityProviderHeaders,

open-sse/config/searchRegistry.ts

@@ -10,6 +10,8 @@
  * perplexity-search reuses credentials from the "perplexity" chat provider.
  */
 
+import { isProviderBlockedByIdOrAlias } from "@/shared/utils/noAuthProviders";
+
 export interface SearchProviderConfig {
   id: string;
   name: string;
@@ -394,16 +396,18 @@ export function supportsSearchType(
 /**
  * Get all search providers as a flat list
  */
-export function getAllSearchProviders(): Array<{
+export function getAllSearchProviders(blockedProviders: string[] = []): Array<{
   id: string;
   name: string;
   searchTypes: string[];
 }> {
-  return Object.values(SEARCH_PROVIDERS).map((p) => ({
-    id: p.id,
-    name: p.name,
-    searchTypes: p.searchTypes,
-  }));
+  return Object.values(SEARCH_PROVIDERS)
+    .filter((p) => !p.disabled && !isProviderBlockedByIdOrAlias(p.id, blockedProviders))
+    .map((p) => ({
+      id: p.id,
+      name: p.name,
+      searchTypes: p.searchTypes,
+    }));
 }
 
 /**

open-sse/executors/base.ts

@@ -20,6 +20,10 @@ import {
   recordLearnedThinkingCap,
   parseThinkingBudgetMax,
 } from "../services/learnedThinkingCaps.ts";
+import {
+  recordLearnedReasoningEffort,
+  parseReasoningEffortEnum,
+} from "../services/learnedReasoningEffortCaps.ts";
 import {
   getParamFilterConfig,
   addParamToBlocklist,
@@ -104,6 +108,12 @@ import {
 import { applyPeerTraceHeader } from "@/shared/resilience/peerRouting";
 import { applyClineProtocolHeaders } from "@/shared/utils/clineAuth";
 import { isProbeContext } from "@/shared/utils/probeOrigin";
+import {
+  parseAndValidatePublicUrl,
+  parseAndValidateNonMetadataUrl,
+} from "@/shared/network/outboundUrlGuard";
+import { getProviderValidationGuard } from "@/shared/network/outboundUrlGuardPolicy";
+import { isLocalProvider, isSelfHostedChatProvider } from "@/shared/constants/providers";
 // Header helpers extracted to a pure leaf; re-exported for external importers
 // (executors + tests) that import them from "./base.ts".
 export {
@@ -397,6 +407,29 @@ export class BaseExecutor {
     return fallback || this.config.baseUrl || "";
   }
 
+  /**
+   * SSRF guard for the runtime dispatch path (GHSA-4f49-hj64-448x). A persisted,
+   * caller-supplied `providerSpecificData.baseUrl` reaches the fetch() calls
+   * below, so a `manage`-scope actor (or, on a keyless install, an anonymous
+   * one) could point a provider at loopback / internal / cloud-metadata hosts
+   * and exfiltrate the stored upstream key. Mirror the provider VALIDATION
+   * guard so runtime dispatch makes the same decision the validation layer
+   * already makes: local / self-hosted providers are exempt (they legitimately
+   * use private URLs, and the OMNIROUTE_ALLOW_PRIVATE_PROVIDER_URLS opt-in still
+   * applies through the guard), and for everything else `public-only` mode
+   * blocks private + metadata while the default `block-metadata` mode blocks the
+   * cloud-metadata IMDS pivot. Throws on a blocked URL.
+   */
+  protected assertOutboundUrlAllowed(url: string): void {
+    if (!url) return;
+    if (isLocalProvider(this.provider) || isSelfHostedChatProvider(this.provider)) return;
+    if (getProviderValidationGuard() === "public-only") {
+      parseAndValidatePublicUrl(url);
+      return;
+    }
+    parseAndValidateNonMetadataUrl(url);
+  }
+
   /**
    * Alternate protocol selected on this connection, if the provider declares one
    * that matches. Centralizes the registry lookup so every call-site resolves the
@@ -615,6 +648,7 @@ export class BaseExecutor {
   async countTokens({ model, body, credentials, signal, log }: CountTokensInput) {
     const url = this.buildCountTokensUrl(model, credentials);
     if (!url) return null;
+    this.assertOutboundUrlAllowed(url); // GHSA-4f49
 
     const headers = this.buildHeaders(credentials, false);
     const requestBody =
@@ -796,6 +830,9 @@ export class BaseExecutor {
     // loop. The learned cap is also recorded process-wide via
     // recordLearnedThinkingCap so future requests skip the 400 entirely.
     let thinkingBudgetClampedMax: number | null = null;
+    // Set by the reasoning_effort 4xx clamp-and-retry below — guards the same
+    // "fires at most once per URL" invariant as thinkingBudgetClampedMax above.
+    let reasoningEffortClamped = false;
 
     for (let urlIndex = 0; urlIndex < fallbackCount; urlIndex++) {
       const requestCredentials = withForcedResponsesUpstream(
@@ -869,6 +906,9 @@ export class BaseExecutor {
         // Timeout only covers response start; stream stalls are handled downstream.
         const fetchStartTimeoutMs = this.getTimeoutMs();
         const fetchWithStartTimeout = async (requestUrl: string, requestOptions: RequestInit) => {
+          // GHSA-4f49: guard here (not only next to the first buildUrl) so retries
+          // and fallback URLs are validated too, before any bytes leave the host.
+          this.assertOutboundUrlAllowed(requestUrl);
           const timeoutController = fetchStartTimeoutMs > 0 ? new AbortController() : null;
           let timeoutId: ReturnType<typeof setTimeout> | null = null;
           if (timeoutController) {
@@ -1496,6 +1536,49 @@ export class BaseExecutor {
           }
         }
 
+        // Reasoning-effort enum 4xx clamp-and-retry (any provider/model without a
+        // declared reasoning_effort capability — custom OpenAI-compatible
+        // connections, or a registered provider the registry hasn't caught up
+        // with). Mirrors the thinking_budget clamp-and-retry above: parse the
+        // upstream-advertised accepted values, record them process-wide (so
+        // FUTURE requests clamp proactively via sanitizeReasoningEffortForProvider
+        // → getLearnedReasoningEffort), clamp the live transformedBody by
+        // re-running the sanitizer, and retry the same URL once.
+        if (
+          (response.status === HTTP_STATUS.BAD_REQUEST ||
+            response.status === HTTP_STATUS.UNPROCESSABLE_ENTITY) &&
+          !reasoningEffortClamped &&
+          transformedBody &&
+          typeof transformedBody === "object"
+        ) {
+          const errText = await response
+            .clone()
+            .text()
+            .catch(() => "");
+          const acceptedValues = parseReasoningEffortEnum(errText);
+          if (acceptedValues) {
+            reasoningEffortClamped = true;
+            const learned = recordLearnedReasoningEffort(this.provider, model, acceptedValues);
+            if (learned) {
+              transformedBody = sanitizeReasoningEffortForProvider(
+                transformedBody,
+                this.provider,
+                model,
+                log
+              );
+              let retryBody = JSON.stringify(transformedBody);
+              if (usesClaudeCodeProtocol || this.provider === "claude") {
+                retryBody = await signRequestBody(retryBody);
+              }
+              log?.info?.(
+                "REASONING_SANITIZE",
+                `Upstream ${response.status} rejected reasoning_effort on ${url} — clamped to ${learned} and retrying (learned for ${this.provider}/${model})`
+              );
+              response = await fetchWithStartTimeout(url, { ...fetchOptions, body: retryBody });
+            }
+          }
+        }
+
         // Generic reactive 400 field-downgrade; each field is stripped at most once.
         if (
           response.status === HTTP_STATUS.BAD_REQUEST &&

open-sse/executors/base/reasoningEffort.ts

@@ -8,6 +8,10 @@ import {
   getProviderModel,
   getProviderModels,
 } from "../../config/providerModels.ts";
+import {
+  getLearnedReasoningEffort,
+  REASONING_EFFORT_ORDER,
+} from "../../services/learnedReasoningEffortCaps.ts";
 
 /**
  * Sanitize reasoning_effort for providers that don't accept all values.
@@ -338,10 +342,24 @@ export function sanitizeReasoningEffortForProvider(
 
   const supportsXHigh = supportsXHighEffort(provider, modelStr);
   const supportsMax = supportsMaxEffortForProvider(provider, modelStr);
+  // Highest value we've actually seen this provider+model accept in a real
+  // upstream 4xx (learnedReasoningEffortCaps.ts) — takes priority over the
+  // static registry (which defaults to "supports everything" when there's no
+  // entry, e.g. custom OpenAI-compatible connections) and over the hardcoded
+  // "high" fallback below (which isn't always valid either).
+  const learnedCap = getLearnedReasoningEffort(provider, modelStr);
+  const learnedRank = learnedCap ? REASONING_EFFORT_ORDER.indexOf(learnedCap) : -1;
 
   // ── xhigh handling ──────────────────────────────────────────────────────
   // xhigh is OmniRoute-internal. Map it to the best effort the model accepts.
   if (effortStr === "xhigh") {
+    if (learnedCap && learnedRank < REASONING_EFFORT_ORDER.indexOf("xhigh")) {
+      log?.info?.(
+        "REASONING_SANITIZE",
+        `${provider}/${modelStr}: clamped reasoning_effort xhigh → ${learnedCap} (learned)`
+      );
+      return writeEffortValue(b, learnedCap, c);
+    }
     if (supportsXHigh) return body; // model accepts xhigh natively
     if (supportsMax) {
       log?.info?.(
@@ -366,6 +384,13 @@ export function sanitizeReasoningEffortForProvider(
   // upstream, and if it 400s the user gets a clear signal. This prevents
   // new models from being unusable for weeks until they're whitelisted (#8057).
   if (effortStr === "max") {
+    if (learnedCap && learnedRank < REASONING_EFFORT_ORDER.indexOf("max")) {
+      log?.info?.(
+        "REASONING_SANITIZE",
+        `${provider}/${modelStr}: clamped reasoning_effort max → ${learnedCap} (learned)`
+      );
+      return writeEffortValue(b, learnedCap, c);
+    }
     if (supportsMax) return body; // explicitly known to accept max
 
     // A model that explicitly advertises its accepted tiers is safe to normalize.

open-sse/executors/copilot-m365-frames.ts

@@ -30,17 +30,45 @@ export const HANDSHAKE_REQUEST = { protocol: "json", version: 1 } as const;
 export const KEEPALIVE_PING = { type: 6 } as const;
 
 /**
- * Allowed message types observed in the 2026-08 recapture of the working
- * `m365.cloud.microsoft/chat` client (#10718). The old 11-entry list is no longer
- * seen on the wire — the stale shape gets closed immediately after the type:4.
+ * Allowed message types observed in a 2026-08-21 live capture of a working
+ * `m365.cloud.microsoft/chat` session (issue: "Stream ended before producing a
+ * non-ping SSE event" on every individual/consumer M365 Copilot call). The
+ * #10718 6-entry shape above no longer produces a `type:1 target:"update"`
+ * frame at all — the socket only replies with SignalR keepalive pings and then
+ * closes, which is exactly what surfaces client-side as that generic stream
+ * error. 30 entries, up from 6.
  */
 export const ALLOWED_MESSAGE_TYPES = [
   "Chat",
   "Suggestion",
+  "InternalSearchQuery",
   "Disengaged",
-  "Progress",
-  "EndOfRequest",
   "InternalLoaderMessage",
+  "Progress",
+  "GeneratedCode",
+  "RenderCardRequest",
+  "AdsQuery",
+  "SemanticSerp",
+  "GenerateContentQuery",
+  "GenerateGraphicArt",
+  "SearchQuery",
+  "ConfirmationCard",
+  "AuthError",
+  "DeveloperLogs",
+  "TriggerPlugin",
+  "HintInvocation",
+  "MemoryUpdate",
+  "EndOfRequest",
+  "TriggerConfirmation",
+  "ResumeInvokeAction",
+  "ResumeUserInputRequest",
+  "TriggerUserInputRequest",
+  "EscapeHatch",
+  "TriggerPluginAuth",
+  "ResumePluginAuth",
+  "SideBySide",
+  "ReferencesListComplete",
+  "SwitchRespondingEndpoint",
 ] as const;
 
 /**
@@ -78,19 +106,26 @@ export const M365_ENTERPRISE_EXTRA_MESSAGE_TYPES = [
 ] as const;
 
 /**
- * Individual / EDU option sets from the 2026-08 recapture (#10718) — 14 entries.
- * The previous 25-entry consumer/MSA set (enable_msa_user, pdnascan, cwc_code_*,
- * …) is no longer observed on the wire and belongs to the shape the substrate
- * now drops silently.
+ * Individual / EDU option sets from a 2026-08-21 live capture — 34 entries, up
+ * from the #10718 14-entry shape (which itself superseded an earlier 25-entry
+ * shape). Each recapture so far has been additive/reshuffled rather than a
+ * wholesale replacement — treat this as the protocol continuing to drift, not
+ * a one-time fix; a future capture may again need to update this list.
  */
 export const M365_DEFAULT_OPTION_SETS = [
   "search_result_progress_messages_with_search_queries",
   "update_textdoc_response_after_streaming",
   "deepleo_networking_timeout_10minutes_canmore",
   "cwc_flux_image",
+  "cwc_code_interpreter",
+  "cwc_code_interpreter_amsfix",
   "cwcfluxgptv",
   "flux_v3_gptv_enable_upload_multi_image_in_turn_wo_ch",
   "gptvnorm2048",
+  "cwc_code_interpreter_citation_fix",
+  "code_interpreter_interactive_charts",
+  "cwc_code_interpreter_interactive_charts_inline_image",
+  "code_interpreter_matplotlib_patching",
   "cwc_fileupload_odb",
   "update_memory_plugin",
   "add_custom_instructions",
@@ -98,6 +133,20 @@ export const M365_DEFAULT_OPTION_SETS = [
   "flux_v3_progress_messages",
   "enable_batch_token_processing",
   "enable_gg_gpt",
+  "async_client_interaction",
+  "flux_v3_references",
+  "flux_v3_references_entities",
+  "flux_v3_references_ci",
+  "add_filestore_filetype",
+  "cwc_code_interpreter_citation_sourceannotations",
+  "cdxcwc_code_interpreter_hallucinated_url_filter",
+  "flux_v3_image_gen_enable_dimensions",
+  "flux_v3_image_gen_enable_non_watermarked_storage",
+  "flux_v3_image_gen_enable_icon_dimensions",
+  "flux_v3_image_gen_enable_system_text_with_params",
+  "flux_v3_image_gen_enable_designer_dimensions_meta_prompting_in_system_prompts",
+  "flux_v3_image_gen_enable_story",
+  "rich_responses",
 ] as const;
 
 /** Append the record separator to a JSON-serializable frame. */
@@ -433,12 +482,14 @@ export function resolveChatInvocationOverrides(tier: string | undefined): {
   }
   return {
     optionsSets: [...M365_DEFAULT_OPTION_SETS],
-    // #10718 — the 2026-08 recapture sends tone:"magic" (lowercase) on the
-    // individual/EDU surface; the old "" default is part of the dropped shape.
-    tone: "magic",
+    // 2026-08-21 capture — the individual/consumer surface now sends "Magic"
+    // (capitalized), matching the enterprise tone literal. The #10718
+    // lowercase "magic" is part of the shape that gets silently dropped.
+    tone: "Magic",
     allowedMessageTypes: ALLOWED_MESSAGE_TYPES,
-    // Omitted entirely on the individual/EDU wire (see ChatInvocationOptions).
-    disconnectBehavior: undefined,
+    // 2026-08-21 capture — disconnectBehavior:"continue" is now present on the
+    // individual/consumer wire too, not just enterprise (see ChatInvocationOptions).
+    disconnectBehavior: "continue",
   };
 }
 
@@ -467,16 +518,33 @@ export function resolveToneForModel(model: string | undefined): string | undefin
 
 /**
  * Build the `type:4` chat invocation frame body (not yet `\x1e`-terminated).
- * Mirrors the argument shape recaptured from a working `m365.cloud.microsoft/chat`
- * client in 2026-08 (#10718). Notable differences from the pre-#10718 shape: a
- * populated `clientInfo` + `productThreadType:"Office"`, a `conversationId`
- * matching the WS URL query, a rich `message` object, and no
- * `spokenTextMode` / `extraExtensionParameters` / `isSbsSupported` /
- * `renderReferencesBehindEOS` / `disconnectBehavior` — none of those are still
- * observed on the wire, and the stale shape gets closed immediately after the
- * invocation.
+ * Base shape from the #10718 recapture (populated `clientInfo` +
+ * `productThreadType:"Office"`, a `conversationId` matching the WS URL query, a
+ * rich `message` object), extended per a 2026-08-21 live capture that found the
+ * #10718 shape alone no longer produces a `type:1 target:"update"` frame — the
+ * socket only replies with keepalive pings and closes. The additions below
+ * (richer `clientInfo`, non-empty `plugins`, `extraExtensionParameters`,
+ * `isSbsSupported`, `renderReferencesBehindEOS`,
+ * `message.connectedFederatedConnections`, and `disconnectBehavior` on every
+ * tier) are exactly the fields the 2026-08-21 capture had that this shape was
+ * missing; the #10718 fields (`conversationId`, `productThreadType`,
+ * `toolChoice`, `message.attachments`) are kept as-is since removing them was
+ * not verified against a live socket.
  */
 export function buildChatInvocation(opts: ChatInvocationOptions): Record<string, unknown> {
+  const clientInfo = {
+    clientAppName: "Office",
+    clientPlatform: "mcmcopilot-web",
+    clientEntrypoint: "mcmcopilot-officeweb",
+    clientSessionId: opts.sessionId,
+    ProductCategory: "Chat",
+    clientAppType: "Web",
+    productEntryPoint: "ChatPanel",
+    deviceOS: "Windows",
+    deviceType: "Desktop",
+    clientPlatformVersion: "10",
+  };
+
   return {
     type: 4,
     target: "chat",
@@ -487,17 +555,17 @@ export function buildChatInvocation(opts: ChatInvocationOptions): Record<string,
           ? [...opts.allowedMessageTypes]
           : [...ALLOWED_MESSAGE_TYPES],
         clientCorrelationId: opts.clientCorrelationId ?? opts.traceId,
-        clientInfo: {
-          clientAppName: "Office",
-          clientPlatform: "mcmcopilot-web",
-        },
+        clientInfo,
         conversationId: opts.conversationId,
+        extraExtensionParameters: {},
         isStartOfSession: opts.isStartOfSession ?? true,
         message: {
           adaptiveCards: [],
           attachments: null,
           author: "user",
+          clientInfo,
           clientPreferences: {},
+          connectedFederatedConnections: ["dummyId"],
           entityAnnotationTypes: ["People", "File", "Event", "Email", "TeamsMessage"],
           experienceType: "Default",
           inputMethod: "Keyboard",
@@ -510,22 +578,27 @@ export function buildChatInvocation(opts: ChatInvocationOptions): Record<string,
           requestId: opts.requestId,
           text: opts.text,
         },
+        isSbsSupported: true,
         options: {},
         optionsSets: opts.optionsSets ?? [...M365_DEFAULT_OPTION_SETS],
-        plugins: opts.plugins ?? [],
+        // 2026-08-21 capture (#11069): BingWebSearch is now the universal
+        // BuiltIn plugin on individual/consumer tier; keep an opt-out override.
+        plugins: opts.plugins ?? [{ Id: "BingWebSearch", Source: "BuiltIn" }],
         ...(opts.customInstructions ? { customInstructions: opts.customInstructions } : {}),
         productThreadType: "Office",
+        renderReferencesBehindEOS: true,
         sessionId: opts.sessionId,
         sliceIds: [],
         source: "officeweb",
         streamingMode: "ConciseWithPadding",
         threadLevelGptId: {},
-        tone: opts.tone ?? "magic",
+        // 2026-08-21 capture (#11069): tone is now capitalized "Magic" on both tiers.
+        tone: opts.tone ?? "Magic",
         toolChoice: opts.toolChoice ?? null,
         traceId: opts.traceId,
-        // #8971 keeps "continue" for the enterprise tier; the individual/EDU wire
-        // omits the key, so only include it when actually set (#10718).
-        ...(opts.disconnectBehavior ? { disconnectBehavior: opts.disconnectBehavior } : {}),
+        // 2026-08-21 capture — disconnectBehavior:"continue" is sent on every
+        // tier now, not gated to enterprise as the #8971 comment described.
+        disconnectBehavior: opts.disconnectBehavior ?? "continue",
       },
     ],
   };

open-sse/executors/glm.ts

@@ -430,6 +430,7 @@ export class GlmExecutor extends DefaultExecutor {
 
     let response: Response;
     try {
+      this.assertOutboundUrlAllowed(url); // GHSA-4f49: glm has its own fetch path
       response = await fetch(url, {
         method: "POST",
         headers,

open-sse/executors/hailuo-web.ts

@@ -33,7 +33,7 @@ import { createHash } from "node:crypto";
 import { BaseExecutor, type ExecuteInput } from "./base.ts";
 import { makeExecutorErrorResult as makeErrorResult, sanitizeErrorMessage } from "../utils/error.ts";
 
-const BASE_URL = "https://www.hailuo.ai";
+const BASE_URL = "https://chat.minimax.io";
 const API_PATH = "/v4/api/chat/msg";
 const USER_AGENT =
   "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/149.0.0.0 Safari/537.36";

open-sse/executors/nlpcloud.ts

@@ -471,6 +471,7 @@ export class NlpCloudExecutor extends BaseExecutor {
     }
 
     try {
+      this.assertOutboundUrlAllowed(url); // GHSA-4f49: nlpcloud has its own fetch path
       const response = await fetch(url, {
         method: "POST",
         headers,

open-sse/executors/opencode.ts

@@ -1,6 +1,6 @@
 import { BaseExecutor, type ExecuteInput, type ProviderCredentials } from "./base.ts";
 import { PROVIDERS } from "../config/constants.ts";
-import { getModelTargetFormat } from "../config/providerModels.ts";
+import { getModelTargetFormat, PROVIDER_ID_TO_ALIAS } from "../config/providerModels.ts";
 import {
   injectReasoningContentForThinkingModel,
   isThinkingMessageModel,
@@ -125,6 +125,24 @@ export function isPremiumOpencodeModel(model: string, provider: string): boolean
   return !OPENCODE_FREE_MODELS.has(model);
 }
 
+/**
+ * Resolves the registry `targetFormat` for a model, aliasing `provider` first.
+ *
+ * `PROVIDER_MODELS` is keyed by the provider's public ALIAS (e.g. `"oc"`), not its
+ * raw registry id (e.g. `"opencode"`) — mirrors `resolveChatCoreTargetFormat()`
+ * (`handlers/chatCore/targetFormat.ts`), which already aliases before calling
+ * `getModelTargetFormat()`. Calling it with the raw id here made every entry miss
+ * silently (fell through to `"openai"`), while chatCore's own request-body
+ * translation (correctly aliased) still switched to the Responses API shape for
+ * `targetFormat:"openai-responses"` models — sending a Responses-shaped body to
+ * the `/chat/completions` URL this executor's own `buildUrl()` kept selecting.
+ * Exported for testability.
+ */
+export function resolveOpencodeTargetFormat(provider: string, model: string): string {
+  const alias = PROVIDER_ID_TO_ALIAS[provider] || provider;
+  return getModelTargetFormat(alias, model) || "openai";
+}
+
 export class OpencodeExecutor extends BaseExecutor {
   /** Delegates to `isPremiumOpencodeModel`. Exported for testability. */
   static isPremiumModel(model: string, provider: string): boolean {
@@ -193,7 +211,10 @@ export class OpencodeExecutor extends BaseExecutor {
     return pickRotatableAccount(this.accounts, this);
   }
 
-  private markCooldown(account: OpencodeAccountState, kind: "transient" | "terminal" = "transient"): void {
+  private markCooldown(
+    account: OpencodeAccountState,
+    kind: "transient" | "terminal" = "transient"
+  ): void {
     markAccountCooldown(account, kind);
   }
 
@@ -202,7 +223,7 @@ export class OpencodeExecutor extends BaseExecutor {
   }
 
   async execute(input: ExecuteInput) {
-    this._requestFormat = getModelTargetFormat(this.provider, input.model) || "openai";
+    this._requestFormat = resolveOpencodeTargetFormat(this.provider, input.model);
 
     // #8681: Gate premium opencode models behind a usable API key.
     // When the connection is keyless (no apiKey, no accessToken) and the model

open-sse/handlers/rerank.ts

@@ -199,6 +199,8 @@ export async function handleRerank({
   return_documents,
   credentials,
   connectionId = null,
+  apiKeyId = null,
+  apiKeyName = null,
 }) {
   const startTime = Date.now();
   if (!model) return errorResponse(400, "model is required");
@@ -267,10 +269,23 @@ export async function handleRerank({
 
     if (!res.ok) {
       const errData = await res.json().catch(() => ({}));
-      return errorResponse(
-        res.status,
-        errData.message || errData.error?.message || `Provider returned HTTP ${res.status}`
-      );
+      const errorMessage =
+        errData.message || errData.error?.message || `Provider returned HTTP ${res.status}`;
+      saveCallLog({
+        method: "POST",
+        path: "/v1/rerank",
+        status: res.status,
+        model: `${providerId}/${modelId}`,
+        provider: providerId,
+        connectionId: connectionId || undefined,
+        duration: Date.now() - startTime,
+        requestBody,
+        responseBody: errData,
+        error: errorMessage,
+        apiKeyId: apiKeyId || undefined,
+        apiKeyName: apiKeyName || undefined,
+      }).catch(() => {});
+      return errorResponse(res.status, errorMessage);
     }
 
     const data = await res.json();
@@ -289,10 +304,13 @@ export async function handleRerank({
       status: 200,
       model: `${providerId}/${modelId}`,
       provider: providerId,
+      connectionId: connectionId || undefined,
       duration: Date.now() - startTime,
       tokens: { prompt_tokens: 0, completion_tokens: 0 },
-      responseBody: { results_count: Array.isArray(result?.results) ? result.results.length : 0 },
-      connectionId,
+      requestBody,
+      responseBody: result,
+      apiKeyId: apiKeyId || undefined,
+      apiKeyName: apiKeyName || undefined,
     }).catch(() => {});
 
     const headers = new Headers({ ...CORS_HEADERS, "Content-Type": "application/json" });

open-sse/mcp-server/tools/memoryTools.ts

@@ -10,16 +10,24 @@ import {
 import { resolveMcpCallerApiKeyId } from "../mcpCallerIdentity.ts";
 
 /**
- * Resolve the memory owner id for an MCP tool call:
- * explicit arg wins, otherwise fall back to the authenticated caller's
- * principal id (HTTP auth headers on SSE/Streamable HTTP transports,
- * OMNIROUTE_API_KEY env var on stdio). Keeps MCP-stored memories under
- * the same owner id that chat-context memory uses, so retrieval in the
- * chat pipeline finds entries written via MCP.
+ * Resolve the memory owner id for an MCP tool call.
+ *
+ * The authenticated caller's principal ALWAYS wins over a caller-supplied
+ * `apiKeyId` — otherwise any MCP caller could read, write, or delete another
+ * principal's memories by putting a different id in the tool arguments
+ * (GHSA-cpv3-xr7r-xf8q, IDOR). The caller is resolved from the per-request HTTP
+ * auth headers on SSE / Streamable HTTP transports, or from OMNIROUTE_API_KEY on
+ * stdio. The explicit argument is only honored as a fallback when no caller can
+ * be resolved (a bare local stdio process with no configured key — already
+ * trusted), preserving the local-tooling flow. Keeps MCP-stored memories under
+ * the same owner id that chat-context memory uses, so retrieval in the chat
+ * pipeline finds entries written via MCP.
  */
 async function resolveMemoryOwnerId(explicit?: string): Promise<string> {
+  const caller = await resolveMcpCallerApiKeyId().catch(() => undefined);
+  if (caller) return caller;
   if (explicit && explicit.trim() !== "") return explicit.trim();
-  return (await resolveMcpCallerApiKeyId().catch(() => undefined)) || "mcp";
+  return "mcp";
 }
 
 export const MemorySearchSchema = z.object({

open-sse/services/accountFallback.ts

@@ -21,7 +21,7 @@ import {
   honorsRuleLockScope,
 } from "../config/providerErrorRules.ts";
 import * as rot from "./rotationConfig.ts";
-import { getPassthroughProviders, getProviderCategory } from "../config/providerRegistry.ts";
+import { getPassthroughProviders, getProviderCategory, isLocalProvider } from "../config/providerRegistry.ts";
 import {
   DEFAULT_RESILIENCE_SETTINGS,
   resolveResilienceSettings,
@@ -37,7 +37,7 @@ import {
   type FailureKind,
 } from "../../src/shared/utils/classify429";
 import { recordProviderSuccess as resetCooldownFailureCount } from "./providerCooldownTracker.ts";
-import { resolveProviderId } from "../../src/shared/constants/providers";
+import { resolveProviderId, isLocalProvider as isLocalProviderId, isSelfHostedChatProvider } from "../../src/shared/constants/providers";
 import { resolveUseUpstream429BreakerHints } from "../../src/shared/utils/providerHints";
 import { getCodexModelScope } from "../config/codexQuotaScopes.ts";
 import { getQuotaScopedModelForProvider } from "./antigravityQuotaFamily.ts";
@@ -791,12 +791,14 @@ export function hasPerModelQuota(
     return connectionPassthroughModels;
   }
   if (!provider) return false;
-  if (getCanonicalLockProvider(provider) === "antigravity") return true;
-  if (getCanonicalLockProvider(provider) === "codex") return true;
-  if (provider === "gemini" || provider === "github") return true;
-  if (provider === "antigravity" || provider === "agy") return true;
-  if (getPassthroughProviders().has(provider)) return true;
-  if (isCompatibleProvider(provider)) return true;
+  const canonicalId = resolveProviderId(provider);
+  if (getCanonicalLockProvider(canonicalId) === "antigravity") return true;
+  if (getCanonicalLockProvider(canonicalId) === "codex") return true;
+  if (canonicalId === "gemini" || canonicalId === "github") return true;
+  if (canonicalId === "antigravity" || canonicalId === "agy") return true;
+  if (getPassthroughProviders().has(canonicalId)) return true;
+  if (isCompatibleProvider(canonicalId)) return true;
+  if (isLocalProviderId(canonicalId) || isSelfHostedChatProvider(canonicalId)) return true;
   return false;
 }
 

open-sse/services/autoCombo/virtualFactory.ts

@@ -607,7 +607,7 @@ export async function prepareVirtualAutoComboInputs(
       // remaining allowance as a percentage, and a raw ">0" comparison would
       // let a reading of e.g. 0.3% (rounding noise, not real headroom) pass.
       minRemainingAllowance: 1,
-      maxStateAgeMs: (settings.autoRefreshProviderQuotaInterval ?? 180) * 1000,
+      maxStateAgeMs: (Number(settings.autoRefreshProviderQuotaInterval) || 180) * 1000,
     });
     if (strictFilteredPool !== pool) pool = strictFilteredPool;
 

open-sse/services/learnedReasoningEffortCaps.ts

@@ -0,0 +1,126 @@
+/**
+ * Learned Reasoning-Effort Caps — reactive capability memory for providers/models
+ * OmniRoute has no static registry entry for (custom OpenAI-compatible connections,
+ * or any registered provider whose registry entry carries no reasoning metadata).
+ *
+ * Same shape as `learnedThinkingCaps.ts` (thinking_budget), generalized from a
+ * numeric budget to an ordinal reasoning_effort scale: on a 4xx whose body
+ * enumerates the accepted values, `base.ts`'s executor calls
+ * `recordLearnedReasoningEffort`, which stores the highest recognized value in a
+ * module-level Map keyed "provider:model" (lowercased). Subsequent requests for
+ * the same provider+model read the cap via `getLearnedReasoningEffort` (consulted
+ * by `sanitizeReasoningEffortForProvider` in `executors/base/reasoningEffort.ts`)
+ * so the 4xx→retry round-trip is paid at most once per process per provider+model.
+ *
+ * In-memory only (same operator-accepted tradeoff as the thinking-budget cache):
+ * restart resets, the first request after a restart may re-learn at the cost of
+ * one upstream 4xx.
+ */
+
+export const REASONING_EFFORT_ORDER: readonly string[] = [
+  "none",
+  "minimal",
+  "low",
+  "medium",
+  "high",
+  "xhigh",
+  "max",
+];
+
+// key: `${provider}:${model}` lowercased → highest value known to be accepted.
+const learnedCaps = new Map<string, string>();
+
+function buildKey(provider: string | null | undefined, model: string | null | undefined): string {
+  const p = typeof provider === "string" ? provider.trim().toLowerCase() : "";
+  const m = typeof model === "string" ? model.trim().toLowerCase() : "";
+  if (!p || !m) return "";
+  return `${p}:${m}`;
+}
+
+function rankOf(value: string): number {
+  return REASONING_EFFORT_ORDER.indexOf(value);
+}
+
+/**
+ * Return the learned cap for provider+model, or null when nothing has been
+ * learned yet (no upstream 4xx recorded). Keyed case-insensitively.
+ */
+export function getLearnedReasoningEffort(
+  provider: string | null | undefined,
+  model: string | null | undefined
+): string | null {
+  const key = buildKey(provider, model);
+  if (!key) return null;
+  return learnedCaps.get(key) ?? null;
+}
+
+/**
+ * Record that `acceptedValues` is the enum the upstream advertised for
+ * provider+model, and store the highest recognized value as the learned cap.
+ * Returns the stored value, or null when `acceptedValues` contained no token
+ * from `REASONING_EFFORT_ORDER` (nothing usable to learn) or the key is unusable.
+ *
+ * Always monotonically decreases: if a cap already stored ranks lower than the
+ * newly computed highest, the stored (lower) value wins and is returned
+ * unchanged. This keeps a later, laxer-looking response (or a race between
+ * concurrent requests) from ratcheting the cap back up.
+ */
+export function recordLearnedReasoningEffort(
+  provider: string | null | undefined,
+  model: string | null | undefined,
+  acceptedValues: string[]
+): string | null {
+  const key = buildKey(provider, model);
+  if (!key) return null;
+
+  let best: string | null = null;
+  let bestRank = -1;
+  for (const raw of acceptedValues) {
+    const rank = rankOf(raw);
+    if (rank > bestRank) {
+      bestRank = rank;
+      best = raw;
+    }
+  }
+  if (best === null) return null;
+
+  const existing = learnedCaps.get(key);
+  if (existing !== undefined && rankOf(existing) <= bestRank) {
+    return existing; // already learned an equal-or-lower cap; keep it
+  }
+  learnedCaps.set(key, best);
+  return best;
+}
+
+// Matches both prose shapes observed: OVH's `@ai-sdk/openai-compatible`
+// deserializer ("expected one of `a`, `b`") and a generic vendor prose form
+// ("Supported types are a, b, and c").
+const LIST_INTRO = /(?:expected one of|supported (?:types|values) are)[:\s]*([^.]+)/i;
+
+/**
+ * Extract the upstream-advertised accepted reasoning_effort values from a 4xx
+ * error body. Returns only tokens present in REASONING_EFFORT_ORDER (unknown
+ * tokens are dropped defensively) in the order they appeared, or null when the
+ * text names no recognized enum member.
+ */
+export function parseReasoningEffortEnum(errText: unknown): string[] | null {
+  if (typeof errText !== "string" || !errText) return null;
+  const match = LIST_INTRO.exec(errText);
+  if (!match) return null;
+  const tokens = match[1]
+    .split(/,|\band\b|&/i)
+    .map((t) =>
+      t
+        .replace(/`/g, "")
+        .replace(/\([^)]*\)/g, "")
+        .trim()
+        .toLowerCase()
+    )
+    .filter((t) => t.length > 0 && REASONING_EFFORT_ORDER.includes(t));
+  return tokens.length > 0 ? tokens : null;
+}
+
+/** Test-only: clear the learned-cap Map between tests. */
+export function __test_resetLearnedReasoningEffortCaps(): void {
+  learnedCaps.clear();
+}

open-sse/services/reasoningInputPolicy.ts

@@ -1,5 +1,6 @@
 import { REGISTRY } from "../config/providerRegistry.ts";
 import type { ReasoningTransport } from "../config/providerRegistry.ts";
+import { isValidResponsesItemId } from "./responsesItemId.ts";
 
 type JsonRecord = Record<string, unknown>;
 
@@ -279,13 +280,27 @@ function sanitizeResponsesInput(
       if (!hasPlaintext && !hasOpaque && (!hasDisplaySummary(next) || stripOrphanedSummaries)) {
         continue;
       }
-      if (!hasOpaque && typeof next.id === "string") delete next.id;
+      // `id` is only worth keeping on an opaque item with a valid string value —
+      // non-opaque items don't replay their id, and a malformed value (e.g. `null`,
+      // observed on opencode/zen) must not survive either way (#11108).
+      if (!hasOpaque || !isValidResponsesItemId(next.id)) delete next.id;
+      // Some upstreams (e.g. opencode/zen) omit `summary` entirely on opaque
+      // reasoning items instead of sending an empty array. Replaying that shape
+      // verbatim trips strict Responses-API validators that require the field
+      // to be present on every `input[]` item of type `reasoning` (#11108).
+      // Plaintext-only items intentionally have no `summary` key and must stay
+      // untouched.
+      if (hasOpaque && next.summary === undefined) next.summary = [];
       filtered.push(next);
       continue;
     }
 
     const cloned = { ...record };
-    if (typeof cloned.id === "string") delete cloned.id;
+    // Strip `id` whenever present, valid or not: these items don't need a
+    // replayed server id, and a malformed one (e.g. `null`, same opencode/zen
+    // omission pattern as the reasoning branch above) must not survive either
+    // (#11108).
+    if (cloned.id !== undefined) delete cloned.id;
     filtered.push(cloned);
   }
   return filtered;

open-sse/services/responsesInputSanitizer.ts

@@ -1,3 +1,5 @@
+import { isValidResponsesItemId } from "./responsesItemId.ts";
+
 type JsonRecord = Record<string, unknown>;
 type SanitizeResponsesInputOptions = {
   dropInternalAssistantMessages?: boolean;
@@ -40,7 +42,12 @@ function sanitizeFunctionName(name: string): string {
 }
 
 function sanitizeInputItemId(record: JsonRecord): JsonRecord {
-  if (typeof record.id !== "string") return record;
+  if (record.id === undefined) return record;
+  if (!isValidResponsesItemId(record.id)) {
+    const next = { ...record };
+    delete next.id;
+    return next;
+  }
 
   const type = typeof record.type === "string" ? record.type : "";
   const expectedPrefix = SERVER_ITEM_ID_PREFIX_BY_TYPE[type];

open-sse/services/responsesItemId.ts

@@ -0,0 +1,7 @@
+// Shared by reasoningInputPolicy.ts and responsesInputSanitizer.ts: both strip a
+// Responses-API `input[]` item's `id` field when it isn't a valid string before
+// replay, so a malformed value (e.g. `null`, observed on opencode/zen) never
+// survives to trip a strict upstream with "Expected 'id' to be a string." (#11108).
+export function isValidResponsesItemId(id: unknown): id is string {
+  return typeof id === "string";
+}

open-sse/services/streamRecovery.ts

@@ -185,7 +185,21 @@ export interface OpenAiSseScan {
   text: string;
   /** True if any `choices[].delta.tool_calls` appeared — NEVER continue those. */
   sawToolCall: boolean;
-  /** True if a terminal marker (`[DONE]` or a non-null `finish_reason`) appeared. */
+  /**
+   * True only when `tool_calls` appeared in this scan AND its own
+   * `finish_reason: "tool_calls"` has NOT also appeared in the same scan — i.e. the
+   * call is still being streamed (arguments may be mid-flight). Once
+   * `finish_reason: "tool_calls"` closes it, the call is complete, not in flight: the
+   * client has the full arguments and a truncation past this point only drops
+   * trailing prose, which continuation can safely recover.
+   */
+  sawToolCallInFlight: boolean;
+  /**
+   * True if a terminal marker for the OVERALL stream appeared: `[DONE]`, or a
+   * `finish_reason` other than `"tool_calls"`. A `finish_reason: "tool_calls"` ends
+   * that one choice but is not terminal for continuation purposes — the model turn
+   * (and the client-visible SSE) is still eligible to be resumed past it.
+   */
   terminal: boolean;
   /** True if at least one OpenAI-shaped `choices[].delta` was parsed (format gate). */
   parsedOpenAi: boolean;
@@ -199,10 +213,11 @@ export interface OpenAiSseScan {
 export function scanOpenAiSseText(sse: string): OpenAiSseScan {
   let text = "";
   let sawToolCall = false;
+  let toolCallFinished = false;
   let terminal = false;
   let parsedOpenAi = false;
   if (typeof sse !== "string" || sse.length === 0) {
-    return { text, sawToolCall, terminal, parsedOpenAi };
+    return { text, sawToolCall, sawToolCallInFlight: false, terminal, parsedOpenAi };
   }
   for (const line of sse.split("\n")) {
     const trimmed = line.trimStart();
@@ -231,10 +246,17 @@ export function scanOpenAiSseText(sse: string): OpenAiSseScan {
         if (Array.isArray(toolCalls) && toolCalls.length > 0) sawToolCall = true;
       }
       const finishReason = (choice as { finish_reason?: unknown })?.finish_reason;
-      if (finishReason != null) terminal = true;
+      if (finishReason === "tool_calls") {
+        // Ends this one choice, but the overall stream/turn stays continuable —
+        // never counts as the general terminal marker (see OpenAiSseScan.terminal).
+        toolCallFinished = true;
+      } else if (finishReason != null) {
+        terminal = true;
+      }
     }
   }
-  return { text, sawToolCall, terminal, parsedOpenAi };
+  const sawToolCallInFlight = sawToolCall && !toolCallFinished;
+  return { text, sawToolCall, sawToolCallInFlight, terminal, parsedOpenAi };
 }
 
 export interface ContinuableBody {
@@ -369,7 +391,7 @@ export function createRecoverableStream(
   let emittedTail = ""; // raw SSE not yet scanned (awaiting an event boundary)
   let emittedText = ""; // assistant text already delivered to the client
   let emittedTerminal = false;
-  let emittedToolCall = false;
+  let emittedToolCallInFlight = false;
   let emittedParsedOpenAi = false;
 
   // Enqueue to the client and, when continuation is enabled, fold the chunk into the
@@ -388,7 +410,7 @@ export function createRecoverableStream(
     const scan = scanOpenAiSseText(complete);
     emittedText += scan.text;
     if (scan.terminal) emittedTerminal = true;
-    if (scan.sawToolCall) emittedToolCall = true;
+    if (scan.sawToolCallInFlight) emittedToolCallInFlight = true;
     if (scan.parsedOpenAi) emittedParsedOpenAi = true;
   };
 
@@ -402,7 +424,7 @@ export function createRecoverableStream(
     continueEnabled &&
     continuations < maxContinuations &&
     emittedParsedOpenAi &&
-    !emittedToolCall &&
+    !emittedToolCallInFlight &&
     !emittedTerminal &&
     emittedText.length > 0;
 

open-sse/translator/request/openai-responses/toResponses.ts

@@ -201,6 +201,14 @@ export function openaiToOpenAIResponsesRequest(
         input.push({
           type: "reasoning",
           content: [{ type: "reasoning_text", text: reasoning }],
+          // Strict Responses-API upstreams (e.g. opencode/zen) require `summary`
+          // on every `input[]` item of type "reasoning", plaintext or opaque —
+          // omitting it rejects the request with `input[N] missing required
+          // field summary`. This item is always freshly built from a chat
+          // client's plaintext reasoning, so there is no source summary to
+          // preserve; default to an empty array like the replay sanitizer does
+          // for opaque items in reasoningInputPolicy.ts (#11108).
+          summary: [],
         });
       }
 

open-sse/translator/response/openai-responses.ts

@@ -866,21 +866,25 @@ export function openaiResponsesToOpenAIResponse(chunk, state) {
 
 function openaiResponsesToOpenAIResponseStream(chunk, state) {
   if (!chunk) {
-    if (
-      state.currentToolCallNeedsNormalization &&
-      state.currentToolCallArgsBuffer &&
-      state.currentToolCallName
-    ) {
-      const toolSchema = state.toolSchemas?.get(state.currentToolCallName);
-      const argsToEmit = stripEmptyOptionalToolArgs(
-        state.currentToolCallArgsBuffer,
-        state.currentToolCallName,
-        toolSchema
-      );
-      const argsStr =
-        typeof argsToEmit === "string" ? argsToEmit : JSON.stringify(argsToEmit ?? {});
-      state.currentToolCallArgsBuffer = "";
-      state.currentToolCallNeedsNormalization = false;
+    // Iterate every still-open call needing schema-aware normalization, not just a
+    // single one — multiple parallel calls can each be pending here if the stream
+    // ends before their output_item.done arrives.
+    const pendingNormalized: Array<{ index: number; argsStr: string }> = [];
+    if (state.toolCallByCallId instanceof Map) {
+      for (const entry of state.toolCallByCallId.values()) {
+        if (entry.needsNormalization && entry.argsBuffer) {
+          const toolSchema = state.toolSchemas?.get(entry.name);
+          const argsToEmit = stripEmptyOptionalToolArgs(entry.argsBuffer, entry.name, toolSchema);
+          pendingNormalized.push({
+            index: entry.index,
+            argsStr: typeof argsToEmit === "string" ? argsToEmit : JSON.stringify(argsToEmit ?? {}),
+          });
+          entry.argsBuffer = "";
+          entry.needsNormalization = false;
+        }
+      }
+    }
+    if (pendingNormalized.length > 0) {
       state.finishReasonSent = true;
       state.finishReason = "tool_calls";
       const common = {
@@ -889,24 +893,21 @@ function openaiResponsesToOpenAIResponseStream(chunk, state) {
         created: state.created,
         model: state.model || "gpt-4",
       };
-      return [
-        {
-          ...common,
-          choices: [
-            {
-              index: 0,
-              delta: {
-                tool_calls: [{ index: state.toolCallIndex, function: { arguments: argsStr } }],
-              },
-              finish_reason: null,
-            },
-          ],
-        },
-        {
-          ...common,
-          choices: [{ index: 0, delta: {}, finish_reason: "tool_calls" }],
-        },
-      ];
+      const chunks: Record<string, unknown>[] = pendingNormalized.map(({ index, argsStr }) => ({
+        ...common,
+        choices: [
+          {
+            index: 0,
+            delta: { tool_calls: [{ index, function: { arguments: argsStr } }] },
+            finish_reason: null,
+          },
+        ],
+      }));
+      chunks.push({
+        ...common,
+        choices: [{ index: 0, delta: {}, finish_reason: "tool_calls" }],
+      });
+      return chunks;
     }
     // Flush: send final chunk with finish_reason
     if (!state.finishReasonSent && state.started) {
@@ -952,7 +953,23 @@ function openaiResponsesToOpenAIResponseStream(chunk, state) {
     state.chatId = `chatcmpl-${Date.now()}`;
     state.created = Math.floor(Date.now() / 1000);
     state.toolCallIndex = 0;
+    // Kept for computeFinishReason (synthesizeCompletedToolCalls.ts) compatibility —
+    // that snapshot path mutates it directly and expects it to exist. In a turn with
+    // multiple parallel calls this only ever reflects the LAST one opened/closed, so
+    // it must never be used to identify a specific call — only as the "is at least
+    // one tool call in flight this turn" signal computeFinishReason needs, which
+    // toolCallIndex > 0 already covers on its own once any call has been added.
     state.currentToolCallId = null;
+    // Per-call state keyed by call_id (replaces the old singular
+    // currentToolCallId/ArgsBuffer/Name/NeedsNormalization/Deferred fields, which
+    // assumed only one function_call could ever be in flight at a time).
+    state.toolCallByCallId = new Map();
+    // response.function_call_arguments.delta carries `item_id`/`output_index`, not
+    // `call_id` — resolve either one back to the call_id key used by
+    // toolCallByCallId (two independent reverse maps, since some upstreams omit
+    // item_id on delta events but still send output_index).
+    state.toolCallItemToCallId = new Map();
+    state.toolCallOutputIndexToCallId = new Map();
   }
 
   // Text content delta
@@ -983,22 +1000,48 @@ function openaiResponsesToOpenAIResponseStream(chunk, state) {
   // Function call started
   if (eventType === "response.output_item.added" && data.item?.type === "function_call") {
     const item = data.item;
-    state.currentToolCallId = item.call_id || fallbackToolCallId();
-    state.currentToolCallArgsBuffer = ""; // reset per-call arg buffer
-    state.currentToolCallDeferred = false;
+    const callId = item.call_id || fallbackToolCallId();
+    // Kept for computeFinishReason (synthesizeCompletedToolCalls.ts) compatibility.
+    state.currentToolCallId = callId;
+
+    const toolName = normalizeToolName(item.name);
+    // Assign this call's index NOW, at .added, not at .done — two calls opened before
+    // either closes (a genuine parallel dispatch) must never share an index. Deferred
+    // (still-nameless) calls are the one exception: they don't claim an index until
+    // .done resolves a real name, so a call that never gets one never burns a slot
+    // another call could have used.
+    let index: number | null = null;
+    if (toolName) {
+      index = state.toolCallIndex ?? 0;
+      state.toolCallIndex = index + 1;
+    }
+
+    if (!(state.toolCallByCallId instanceof Map)) state.toolCallByCallId = new Map();
+    state.toolCallByCallId.set(callId, {
+      index,
+      name: toolName,
+      argsBuffer: "",
+      deferred: !toolName,
+      needsNormalization: toolName === "Agent",
+    });
+    if (!(state.toolCallItemToCallId instanceof Map)) state.toolCallItemToCallId = new Map();
+    if (item.id) state.toolCallItemToCallId.set(item.id, callId);
+    // `output_index` is a top-level field on every Responses API streamed event
+    // (response.output_item.added/.done AND function_call_arguments.delta alike) —
+    // an identifier independent of item_id, for upstreams that omit item_id on delta
+    // events.
+    if (!(state.toolCallOutputIndexToCallId instanceof Map)) {
+      state.toolCallOutputIndexToCallId = new Map();
+    }
+    if (data.output_index != null) state.toolCallOutputIndexToCallId.set(data.output_index, callId);
 
     // Track this call_id so response.completed doesn't synthesize a duplicate
     if (!state.toolCallIdsSeen) state.toolCallIdsSeen = new Set();
-    if (state.currentToolCallId) state.toolCallIdsSeen.add(state.currentToolCallId);
+    state.toolCallIdsSeen.add(callId);
 
-    const toolName = normalizeToolName(item.name);
-    state.currentToolName = toolName; // track for schema lookup at done time
-    state.currentToolCallName = toolName;
-    state.currentToolCallNeedsNormalization = toolName === "Agent";
     if (!toolName) {
       // Some Responses providers briefly emit placeholder/empty tool names.
       // Defer emission until output_item.done in case the final name is populated there.
-      state.currentToolCallDeferred = true;
       return null;
     }
 
@@ -1013,8 +1056,8 @@ function openaiResponsesToOpenAIResponseStream(chunk, state) {
           delta: {
             tool_calls: [
               {
-                index: state.toolCallIndex,
-                id: state.currentToolCallId,
+                index,
+                id: callId,
                 type: "function",
                 function: {
                   name: toolName,
@@ -1037,11 +1080,26 @@ function openaiResponsesToOpenAIResponseStream(chunk, state) {
     const argsDelta = data.delta || "";
     if (!argsDelta) return null;
 
-    state.currentToolCallArgsBuffer = (state.currentToolCallArgsBuffer || "") + argsDelta;
-    if (state.currentToolCallDeferred || state.currentToolCallNeedsNormalization) return null;
+    // Resolve which in-flight call this delta belongs to. Try item_id first (the
+    // field the Responses API documents for this event), then output_index (also a
+    // top-level field on this event, and independent of item_id — covers upstreams
+    // that omit item_id on delta events but still send output_index). Only once both
+    // identifying fields are absent/unresolved do we fall back to guessing (the
+    // single open call, or the most recently opened one as a last resort).
+    const map = state.toolCallByCallId instanceof Map ? state.toolCallByCallId : null;
+    let callId = data.item_id ? state.toolCallItemToCallId?.get(data.item_id) : undefined;
+    if (!callId && data.output_index != null) {
+      callId = state.toolCallOutputIndexToCallId?.get(data.output_index);
+    }
+    if (!callId && map) {
+      callId = map.size === 1 ? [...map.keys()][0] : state.currentToolCallId;
+    }
+    const entry = callId ? map?.get(callId) : undefined;
+    if (!entry) return null;
 
     // #9168: buffer arguments until output_item.done for schema-aware null normalization
     // Previously emitted raw null values for optional enum fields (e.g. isolation: null).
+    entry.argsBuffer = (entry.argsBuffer || "") + argsDelta;
     return null;
   }
 
@@ -1061,13 +1119,30 @@ function openaiResponsesToOpenAIResponseStream(chunk, state) {
   // carry the complete arguments only in output_item.done (no preceding delta events).
   if (eventType === "response.output_item.done" && data.item?.type === "function_call") {
     const item = data.item;
-    const buffered = state.currentToolCallArgsBuffer || "";
-    const currentIndex = state.toolCallIndex; // capture before increment
-    const callId = item.call_id || state.currentToolCallId || fallbackToolCallId();
+    const map = state.toolCallByCallId instanceof Map ? state.toolCallByCallId : null;
+    let callId = item.call_id;
+    if (!callId && item.id) callId = state.toolCallItemToCallId?.get(item.id);
+    if (!callId) callId = state.currentToolCallId || fallbackToolCallId();
+    const trackedEntry = callId ? map?.get(callId) : undefined;
+    // Some upstreams (e.g. Codex) send the complete payload only in output_item.done,
+    // with no preceding output_item.added at all — there is no tracked entry to read an
+    // index from.
+    const entry = trackedEntry || { index: null, argsBuffer: "", deferred: false };
+
+    const buffered = entry.argsBuffer || "";
     const toolName = normalizeToolName(item.name);
+
+    // Claim (and advance) this call's index now if it wasn't assigned at .added — either
+    // a deferred call whose name has just now resolved, or a Codex-style done-only
+    // payload that never had an .added at all. A deferred call whose name is STILL empty
+    // never claims an index (nothing was ever emitted for it either way).
+    if (entry.index == null && toolName) {
+      entry.index = state.toolCallIndex ?? 0;
+      state.toolCallIndex = entry.index + 1;
+    }
+    const currentIndex = entry.index;
     const toolSchema = state.toolSchemas?.get(toolName);
     const shouldNormalizeArguments = toolName === "Agent";
-    state.currentToolCallNeedsNormalization = shouldNormalizeArguments;
 
     if (toolName && state.toolCalls instanceof Map) {
       const completedArguments =
@@ -1077,6 +1152,9 @@ function openaiResponsesToOpenAIResponseStream(chunk, state) {
         toolName,
         toolSchema
       );
+      // Keyed by index, not insertion order — readers that need call order for
+      // parallel calls closed out of order should sort by this key rather than
+      // relying on Map iteration order.
       state.toolCalls.set(currentIndex, {
         id: callId,
         index: currentIndex,
@@ -1095,17 +1173,17 @@ function openaiResponsesToOpenAIResponseStream(chunk, state) {
     if (!state.toolCallIdsSeen) state.toolCallIdsSeen = new Set();
     if (callId) state.toolCallIdsSeen.add(callId);
 
-    if (state.currentToolCallDeferred) {
-      state.currentToolCallDeferred = false;
-      state.currentToolCallArgsBuffer = "";
-      state.currentToolCallId = null;
+    // This call is fully closed — remove it from the in-flight map (bounds the map
+    // to genuinely in-flight calls, and keeps the single-open-call fallback in the
+    // function_call_arguments.delta handler correct for whichever call opens next).
+    if (map && callId) map.delete(callId);
+    if (state.currentToolCallId === callId) state.currentToolCallId = null;
 
+    if (entry.deferred) {
       if (!toolName) {
         return null;
       }
 
-      state.toolCallIndex++;
-
       const terminalArguments =
         typeof item.arguments === "string"
           ? item.arguments.length > 0
@@ -1148,12 +1226,7 @@ function openaiResponsesToOpenAIResponseStream(chunk, state) {
       };
     }
 
-    state.toolCallIndex++;
-    state.currentToolCallArgsBuffer = ""; // reset for next tool call
-    state.currentToolCallId = null;
-    const needsNormalization = state.currentToolCallNeedsNormalization === true;
-    state.currentToolCallNeedsNormalization = false;
-    state.currentToolCallName = "";
+    const needsNormalization = shouldNormalizeArguments;
 
     // Nullable omission sentinels must be normalized before any argument bytes reach the client.
     // Other tool calls retain immediate argument streaming.

open-sse/utils/streamPayloadCollector.ts

@@ -128,6 +128,75 @@ function tryParseJson(raw: string): unknown {
   }
 }
 
+/**
+ * Splits a tool_call `arguments` string that is actually multiple back-to-back JSON
+ * objects glued together with no separator, into its individual object substrings.
+ *
+ * Root cause (observed on opencode/muse-spark-1.2-contributor-free via the zen
+ * provider): some upstreams never vary `index`/`id` across a 2nd/3rd/… tool_call of
+ * the SAME name emitted in one turn, so every delta in `buildOpenAISummary` above
+ * resolves to the same accumulator key and `arguments` ends up as N JSON objects
+ * concatenated with no delimiter — invalid as a single JSON value, but each object is
+ * individually well-formed. Structural, not provider-specific: applies to whichever
+ * upstream exhibits the same index-collision streaming bug.
+ *
+ * Returns `null` when `raw` is empty, already valid single JSON, or does not scan as
+ * ≥2 back-to-back valid JSON values — callers must leave `arguments` untouched in
+ * that case (never regress a value that used to reach the client as-is).
+ */
+export function splitConcatenatedToolCallArguments(raw: string): string[] | null {
+  if (!raw) return null;
+  try {
+    JSON.parse(raw);
+    return null; // Already a single valid JSON value — nothing to split.
+  } catch {
+    // Fall through to the multi-value scan below.
+  }
+
+  const parts: string[] = [];
+  let depth = 0;
+  let inString = false;
+  let escaped = false;
+  let start = -1;
+
+  for (let i = 0; i < raw.length; i++) {
+    const ch = raw[i];
+    if (start === -1) {
+      if (ch === " " || ch === "\n" || ch === "\r" || ch === "\t") continue;
+      if (ch !== "{" && ch !== "[") return null; // Not a value boundary — bail, leave untouched.
+      start = i;
+    }
+    if (inString) {
+      if (escaped) escaped = false;
+      else if (ch === "\\") escaped = true;
+      else if (ch === '"') inString = false;
+      continue;
+    }
+    if (ch === '"') {
+      inString = true;
+      continue;
+    }
+    if (ch === "{" || ch === "[") depth++;
+    else if (ch === "}" || ch === "]") {
+      depth--;
+      if (depth === 0) {
+        parts.push(raw.slice(start, i + 1));
+        start = -1;
+      }
+    }
+  }
+  if (start !== -1 || depth !== 0 || parts.length < 2) return null;
+
+  for (const part of parts) {
+    try {
+      JSON.parse(part);
+    } catch {
+      return null; // One of the scanned segments isn't valid JSON — bail entirely.
+    }
+  }
+  return parts;
+}
+
 // ─── Per-format live reducers ────────────────────────────────────────────────
 // Each reducer mirrors the corresponding build*Summary()'s original for-loop
 // body exactly (ingest = one loop iteration, finalize = the post-loop return),
@@ -262,7 +331,28 @@ function createOpenAIReducer(fallbackModel?: string | null): SummaryReducer {
         message.reasoning_content = joinedReasoning;
       }
 
-      const finalToolCalls = [...toolCalls.values()].sort((a, b) => a.index - b.index);
+      const mergedToolCalls = [...toolCalls.values()].sort((a, b) => a.index - b.index);
+      // Expand any entry whose accumulated `arguments` turned out to be multiple
+      // concatenated JSON objects (upstream never varied index/id across repeated
+      // same-name tool_calls) into its own separate tool_calls entries.
+      const finalToolCalls: ToolCall[] = [];
+      let nextIndex = 0;
+      // Normalize tool_call indexes to contiguous 0-based (OpenAI contract).
+      for (const tc of mergedToolCalls) {
+        const splitArgs = splitConcatenatedToolCallArguments(tc.function.arguments);
+        if (!splitArgs) {
+          finalToolCalls.push({ ...tc, index: nextIndex++ });
+          continue;
+        }
+        for (const [i, args] of splitArgs.entries()) {
+          finalToolCalls.push({
+            id: tc.id ? `${tc.id}_split${i}` : null,
+            index: nextIndex++,
+            type: tc.type,
+            function: { name: tc.function.name, arguments: args },
+          });
+        }
+      }
       if (finalToolCalls.length > 0) {
         finishReason = "tool_calls";
         message.tool_calls = finalToolCalls;

package-lock.json

@@ -18,7 +18,6 @@
         "@dnd-kit/core": "^6.3.1",
         "@dnd-kit/sortable": "^10.0.0",
         "@dnd-kit/utilities": "^3.2.2",
-        "@huggingface/transformers": "^4.2.0",
         "@lobehub/icons": "^5.16.0",
         "@modelcontextprotocol/sdk": "^1.29.0",
         "@monaco-editor/react": "^4.7.0",
@@ -61,7 +60,6 @@
         "next-themes": "^0.4.6",
         "node-machine-id": "^1.1.12",
         "omniglyph": "^1.4.0",
-        "onnxruntime-node": "1.24.3",
         "open": "^11.0.1",
         "ora": "^9.4.1",
         "parse5": "^8.0.1",
@@ -156,9 +154,11 @@
       },
       "optionalDependencies": {
         "@atjsh/llmlingua-2": "3.0.0",
+        "@huggingface/transformers": "^4.2.0",
         "better-sqlite3": "^13.0.2",
         "js-tiktoken": "^1.0.20",
         "keytar": "^7.9.0",
+        "onnxruntime-node": "1.24.3",
         "sqlite-vec": "^0.1.9",
         "tls-client-node": "^0.2.0",
         "wreq-js": "^3.0.0"
@@ -4510,6 +4510,7 @@
       "resolved": "https://registry.npmjs.org/@huggingface/jinja/-/jinja-0.5.9.tgz",
       "integrity": "sha512-uWTG+l3VJRsl7EXxYizuL3P+cCPoc3cRqbWWRcQN0FhejRfbdq0RNhCmbY/YDtnTcz9icdLYuLDjsnz4d8JMuw==",
       "license": "MIT",
+      "optional": true,
       "engines": {
         "node": ">=18"
       }
@@ -4518,13 +4519,15 @@
       "version": "0.1.3",
       "resolved": "https://registry.npmjs.org/@huggingface/tokenizers/-/tokenizers-0.1.3.tgz",
       "integrity": "sha512-8rF/RRT10u+kn7YuUbUg0OF30K8rjTc78aHpxT+qJ1uWSqxT1MHi8+9ltwYfkFYJzT/oS+qw3JVfHtNMGAdqyA==",
-      "license": "Apache-2.0"
+      "license": "Apache-2.0",
+      "optional": true
     },
     "node_modules/@huggingface/transformers": {
       "version": "4.2.0",
       "resolved": "https://registry.npmjs.org/@huggingface/transformers/-/transformers-4.2.0.tgz",
       "integrity": "sha512-8BRCoBMH0XsWaEIamuR0LrJGAfftgHAfb2Vrffy0VKlSAE/MnUJ5/h/zTfEP3fDIft+nk7TqB8xXEyABGitBjQ==",
       "license": "Apache-2.0",
+      "optional": true,
       "dependencies": {
         "@huggingface/jinja": "^0.5.6",
         "@huggingface/tokenizers": "^0.1.3",
@@ -9483,30 +9486,35 @@
       "version": "1.1.2",
       "resolved": "https://registry.npmjs.org/@protobufjs/aspromise/-/aspromise-1.1.2.tgz",
       "integrity": "sha512-j+gKExEuLmKwvz3OgROXtrJ2UG2x8Ch2YZUxahh+s1F2HZ+wAceUNLkvy6zKCPVRkU++ZWQrdxsUeQXmcg4uoQ==",
+      "devOptional": true,
       "license": "BSD-3-Clause"
     },
     "node_modules/@protobufjs/base64": {
       "version": "1.1.2",
       "resolved": "https://registry.npmjs.org/@protobufjs/base64/-/base64-1.1.2.tgz",
       "integrity": "sha512-AZkcAA5vnN/v4PDqKyMR5lx7hZttPDgClv83E//FMNhR2TMcLUhfRUBHCmSl0oi9zMgDDqRUJkSxO3wm85+XLg==",
+      "devOptional": true,
       "license": "BSD-3-Clause"
     },
     "node_modules/@protobufjs/codegen": {
       "version": "2.0.5",
       "resolved": "https://registry.npmjs.org/@protobufjs/codegen/-/codegen-2.0.5.tgz",
       "integrity": "sha512-zgXFLzW3Ap33e6d0Wlj4MGIm6Ce8O89n/apUaGNB/jx+hw+ruWEp7EwGUshdLKVRCxZW12fp9r40E1mQrf/34g==",
+      "devOptional": true,
       "license": "BSD-3-Clause"
     },
     "node_modules/@protobufjs/eventemitter": {
       "version": "1.1.1",
       "resolved": "https://registry.npmjs.org/@protobufjs/eventemitter/-/eventemitter-1.1.1.tgz",
       "integrity": "sha512-vW1GmwMZNnL+gMRaovlh9yZX74kc+TTU3FObkkurpMaRtBfLP3ldjS9KQWlwZgraRE0+dheEEoAxdzcJQ8eXZg==",
+      "devOptional": true,
       "license": "BSD-3-Clause"
     },
     "node_modules/@protobufjs/fetch": {
       "version": "1.1.1",
       "resolved": "https://registry.npmjs.org/@protobufjs/fetch/-/fetch-1.1.1.tgz",
       "integrity": "sha512-GpptLrs57adMSuHi3VNj0mAF8dwh36LMaYF6XyJ6JMWlVsc+t42tm1HSEDmOs3A8fC9yyeisgLhsTVQokOZ0zw==",
+      "devOptional": true,
       "license": "BSD-3-Clause",
       "dependencies": {
         "@protobufjs/aspromise": "^1.1.1"
@@ -9516,24 +9524,28 @@
       "version": "1.0.2",
       "resolved": "https://registry.npmjs.org/@protobufjs/float/-/float-1.0.2.tgz",
       "integrity": "sha512-Ddb+kVXlXst9d+R9PfTIxh1EdNkgoRe5tOX6t01f1lYWOvJnSPDBlG241QLzcyPdoNTsblLUdujGSE4RzrTZGQ==",
+      "devOptional": true,
       "license": "BSD-3-Clause"
     },
     "node_modules/@protobufjs/path": {
       "version": "1.1.2",
       "resolved": "https://registry.npmjs.org/@protobufjs/path/-/path-1.1.2.tgz",
       "integrity": "sha512-6JOcJ5Tm08dOHAbdR3GrvP+yUUfkjG5ePsHYczMFLq3ZmMkAD98cDgcT2iA1lJ9NVwFd4tH/iSSoe44YWkltEA==",
+      "devOptional": true,
       "license": "BSD-3-Clause"
     },
     "node_modules/@protobufjs/pool": {
       "version": "1.1.0",
       "resolved": "https://registry.npmjs.org/@protobufjs/pool/-/pool-1.1.0.tgz",
       "integrity": "sha512-0kELaGSIDBKvcgS4zkjz1PeddatrjYcmMWOlAuAPwAeccUrPHdUqo/J6LiymHHEiJT5NrF1UVwxY14f+fy4WQw==",
+      "devOptional": true,
       "license": "BSD-3-Clause"
     },
     "node_modules/@protobufjs/utf8": {
       "version": "1.1.1",
       "resolved": "https://registry.npmjs.org/@protobufjs/utf8/-/utf8-1.1.1.tgz",
       "integrity": "sha512-oOAWABowe8EAbMyWKM0tYDKi8Yaox52D+HWZhAIJqQXbqe0xI/GV7FhLWqlEKreMkfDjshR5FKgi3mnle0h6Eg==",
+      "devOptional": true,
       "license": "BSD-3-Clause"
     },
     "node_modules/@radix-ui/number": {
@@ -12737,6 +12749,7 @@
       "version": "26.2.0",
       "resolved": "https://registry.npmjs.org/@types/node/-/node-26.2.0.tgz",
       "integrity": "sha512-5IviulTZeRNp2vAJ514cc/HUlY5nZ9fCbq9DMyC52BrhFZACo3nI0R7qBxhQmo/d27NFe96ur/b7Wwxklda+kg==",
+      "devOptional": true,
       "license": "MIT",
       "dependencies": {
         "undici-types": "~8.3.0"
@@ -13998,6 +14011,7 @@
       "resolved": "https://registry.npmjs.org/adm-zip/-/adm-zip-0.6.0.tgz",
       "integrity": "sha512-XleryMhbuksdKtofnWZ9Sk+4CUTbms4Mb/EU32SZwToAyZ5RgVos/ki8n+yr0LWHOGKuakbXTuuYNHLQjhddgg==",
       "license": "MIT",
+      "optional": true,
       "engines": {
         "node": ">=14.0"
       }
@@ -14971,7 +14985,8 @@
       "resolved": "https://registry.npmjs.org/boolean/-/boolean-3.2.0.tgz",
       "integrity": "sha512-d0II/GO9uf9lfUHH2BQsjxzRJZBdsjgsBiW4BvhWk/3qoKwQFjIDVN19PfX8F2D/r9PCMTtLWjYVCFrpeYUzsw==",
       "deprecated": "Package no longer supported. Contact Support at https://www.npmjs.com/support for more info.",
-      "license": "MIT"
+      "license": "MIT",
+      "optional": true
     },
     "node_modules/bottleneck": {
       "version": "2.19.5",
@@ -17935,6 +17950,7 @@
       "version": "1.1.4",
       "resolved": "https://registry.npmjs.org/define-data-property/-/define-data-property-1.1.4.tgz",
       "integrity": "sha512-rBMvIzlpA8v6E+SJZoo++HAYqsLrkg7MSfIinMPFhmkorw7X+dOXVJQs+QT69zGkzMyfDnIMN2Wid1+NbL3T+A==",
+      "devOptional": true,
       "license": "MIT",
       "dependencies": {
         "es-define-property": "^1.0.0",
@@ -17964,6 +17980,7 @@
       "version": "1.2.1",
       "resolved": "https://registry.npmjs.org/define-properties/-/define-properties-1.2.1.tgz",
       "integrity": "sha512-8QmQKqEASLd5nx0U1B1okLElbUuuttJ/AnYmRXbbbGDWh6uS208EjD4Xqq/I9wK7u0v6O08XhTWnt5XtEbR6Dg==",
+      "devOptional": true,
       "license": "MIT",
       "dependencies": {
         "define-data-property": "^1.0.1",
@@ -18064,7 +18081,8 @@
       "version": "2.1.0",
       "resolved": "https://registry.npmjs.org/detect-node/-/detect-node-2.1.0.tgz",
       "integrity": "sha512-T0NIuQpnTvFDATNuHN5roPwSBG83rFsuO+MXXH9/3N1eFbn4wcPjttvjMLEPWJ0RGUYgQE7cGgS3tNxbqCGM7g==",
-      "license": "MIT"
+      "license": "MIT",
+      "optional": true
     },
     "node_modules/detect-node-es": {
       "version": "1.1.0",
@@ -18908,7 +18926,8 @@
       "version": "4.1.1",
       "resolved": "https://registry.npmjs.org/es6-error/-/es6-error-4.1.1.tgz",
       "integrity": "sha512-Um/+FxMr9CISWh0bi5Zv0iOD+4cFh5qLeks1qhAopKVAJw3drgKbKySikp7wGhDL0HPeaja0P5ULZrxLkniUVg==",
-      "license": "MIT"
+      "license": "MIT",
+      "optional": true
     },
     "node_modules/es6-promisify": {
       "version": "7.0.0",
@@ -20400,7 +20419,8 @@
       "version": "25.9.23",
       "resolved": "https://registry.npmjs.org/flatbuffers/-/flatbuffers-25.9.23.tgz",
       "integrity": "sha512-MI1qs7Lo4Syw0EOzUl0xjs2lsoeqFku44KpngfIduHBYvzm8h2+7K8YMQh1JtVVVrUvhLpNwqVi4DERegUJhPQ==",
-      "license": "Apache-2.0"
+      "license": "Apache-2.0",
+      "optional": true
     },
     "node_modules/flatted": {
       "version": "3.4.2",
@@ -21226,6 +21246,7 @@
       "resolved": "https://registry.npmjs.org/global-agent/-/global-agent-3.0.0.tgz",
       "integrity": "sha512-PT6XReJ+D07JvGoxQMkT6qji/jVNfX/h364XHZOWeRzy64sSFr+xJ5OX7LI3b4MPQzdL4H8Y8M0xzPpsVMwA8Q==",
       "license": "BSD-3-Clause",
+      "optional": true,
       "dependencies": {
         "boolean": "^3.0.1",
         "es6-error": "^4.1.1",
@@ -21243,6 +21264,7 @@
       "resolved": "https://registry.npmjs.org/semver/-/semver-7.8.5.tgz",
       "integrity": "sha512-Y7/KDsb8LjooZpwaqGyulO6DQlksgCncchHGk+sZIY4SBvUocMBEFH5Ur1fI4dV+Jvl0w6cjvucaIi40puRioA==",
       "license": "ISC",
+      "optional": true,
       "bin": {
         "semver": "bin/semver.js"
       },
@@ -21291,6 +21313,7 @@
       "version": "1.0.4",
       "resolved": "https://registry.npmjs.org/globalthis/-/globalthis-1.0.4.tgz",
       "integrity": "sha512-DpLKbNU4WylpxJykQujfCcwYWiV/Jhm50Goo0wrVILAv5jOr9d+H+UR3PhSCD2rCCEIg0uc+G+muBTwD54JhDQ==",
+      "devOptional": true,
       "license": "MIT",
       "dependencies": {
         "define-properties": "^1.2.1",
@@ -21645,7 +21668,8 @@
       "version": "1.0.9",
       "resolved": "https://registry.npmjs.org/guid-typescript/-/guid-typescript-1.0.9.tgz",
       "integrity": "sha512-Y8T4vYhEfwJOTbouREvG+3XDsjr8E3kIr7uf+JZ0BYloFsttiHU0WfvANVsR7TxNUJa/WpCnw/Ino/p+DeBhBQ==",
-      "license": "ISC"
+      "license": "ISC",
+      "optional": true
     },
     "node_modules/hachure-fill": {
       "version": "0.5.2",
@@ -21679,6 +21703,7 @@
       "version": "1.0.2",
       "resolved": "https://registry.npmjs.org/has-property-descriptors/-/has-property-descriptors-1.0.2.tgz",
       "integrity": "sha512-55JNKuIW+vq4Ke1BjOTjM2YctQIvCT7GFzHwmfZPGo5wnrgkid0YQtnAleFSqumZm4az3n2BS+erby5ipJdgrg==",
+      "devOptional": true,
       "license": "MIT",
       "dependencies": {
         "es-define-property": "^1.0.0"
@@ -24790,7 +24815,8 @@
       "version": "5.0.1",
       "resolved": "https://registry.npmjs.org/json-stringify-safe/-/json-stringify-safe-5.0.1.tgz",
       "integrity": "sha512-ZClg6AaYvamvYEE82d3Iyd3vSSIjQ+odgjaTzRuO3s7toCdFKczob2i0zCh7JE8kWn17yvAWhUVxvqGwUalsRA==",
-      "license": "ISC"
+      "license": "ISC",
+      "optional": true
     },
     "node_modules/json5": {
       "version": "2.2.3",
@@ -26654,6 +26680,7 @@
       "resolved": "https://registry.npmjs.org/matcher/-/matcher-3.0.0.tgz",
       "integrity": "sha512-OkeDaAZ/bQCxeFAozM55PKcKU0yJMPGifLwV4Qgjitu+5MoAfSQN4lsLJeXZ1b8w0x+/Emda6MZgXS1jvsapng==",
       "license": "MIT",
+      "optional": true,
       "dependencies": {
         "escape-string-regexp": "^4.0.0"
       },
@@ -29425,6 +29452,7 @@
       "version": "1.1.1",
       "resolved": "https://registry.npmjs.org/object-keys/-/object-keys-1.1.1.tgz",
       "integrity": "sha512-NuAESUOUMrlIXOfHKzD6bpPu3tYt3xvjNdRIQ+FeT0lNb4K8WR70CaDxhuNguS2XG+GjkyMwOzsN5ZktImfhLA==",
+      "devOptional": true,
       "license": "MIT",
       "engines": {
         "node": ">= 0.4"
@@ -29632,7 +29660,8 @@
       "version": "1.24.3",
       "resolved": "https://registry.npmjs.org/onnxruntime-common/-/onnxruntime-common-1.24.3.tgz",
       "integrity": "sha512-GeuPZO6U/LBJXvwdaqHbuUmoXiEdeCjWi/EG7Y1HNnDwJYuk6WUbNXpF6luSUY8yASul3cmUlLGrCCL1ZgVXqA==",
-      "license": "MIT"
+      "license": "MIT",
+      "optional": true
     },
     "node_modules/onnxruntime-node": {
       "version": "1.24.3",
@@ -29640,6 +29669,7 @@
       "integrity": "sha512-JH7+czbc8ALA819vlTgcV+Q214/+VjGeBHDjX81+ZCD0PCVCIFGFNtT0V4sXG/1JXypKPgScQcB3ij/hk3YnTg==",
       "hasInstallScript": true,
       "license": "MIT",
+      "optional": true,
       "os": [
         "win32",
         "darwin",
@@ -29656,6 +29686,7 @@
       "resolved": "https://registry.npmjs.org/onnxruntime-web/-/onnxruntime-web-1.26.0-dev.20260416-b7804b056c.tgz",
       "integrity": "sha512-MD6Ss4GSpQBo6zqoJzyT9LRbKYs7x/JVN23FT24EcEvlqF4VuzPOeH6X38orZPKHQDbprn7K+SBpu0/mj2CQiw==",
       "license": "MIT",
+      "optional": true,
       "dependencies": {
         "flatbuffers": "^25.1.24",
         "guid-typescript": "^1.0.9",
@@ -29669,13 +29700,15 @@
       "version": "5.3.2",
       "resolved": "https://registry.npmjs.org/long/-/long-5.3.2.tgz",
       "integrity": "sha512-mNAgZ1GmyNhD7AuqnTG3/VQ26o760+ZYBPKjPvugO8+nLbYfX6TVpJPseBvopbdY+qpZ/lKUnmEc1LeZYS3QAA==",
-      "license": "Apache-2.0"
+      "license": "Apache-2.0",
+      "optional": true
     },
     "node_modules/onnxruntime-web/node_modules/onnxruntime-common": {
       "version": "1.24.0-dev.20251116-b39e144322",
       "resolved": "https://registry.npmjs.org/onnxruntime-common/-/onnxruntime-common-1.24.0-dev.20251116-b39e144322.tgz",
       "integrity": "sha512-BOoomdHYmNRL5r4iQ4bMvsl2t0/hzVQ3OM3PHD0gxeXu1PmggqBv3puZicEUVOA3AtHHYmqZtjMj9FOfGrATTw==",
-      "license": "MIT"
+      "license": "MIT",
+      "optional": true
     },
     "node_modules/open": {
       "version": "11.0.1",
@@ -30906,7 +30939,8 @@
       "version": "1.3.6",
       "resolved": "https://registry.npmjs.org/platform/-/platform-1.3.6.tgz",
       "integrity": "sha512-fnWVljUchTro6RiCFvCXBbNhJc2NijN7oIQxbwsyL0buWJPG85v81ehlHI9fXrJsMNgTofEoWIQeClKpgxFLrg==",
-      "license": "MIT"
+      "license": "MIT",
+      "optional": true
     },
     "node_modules/playwright": {
       "version": "1.62.1",
@@ -31861,6 +31895,7 @@
       "version": "7.6.5",
       "resolved": "https://registry.npmjs.org/protobufjs/-/protobufjs-7.6.5.tgz",
       "integrity": "sha512-/FPD0nUc9jH6rfFjji9IBqOz4pcSE3CsT1m7Ep6Mdb0LxSUMj8hgl6GomOvZzpNpAqqGaXA0P3VSrZLFzIhQrw==",
+      "devOptional": true,
       "hasInstallScript": true,
       "license": "BSD-3-Clause",
       "dependencies": {
@@ -31884,6 +31919,7 @@
       "version": "5.3.2",
       "resolved": "https://registry.npmjs.org/long/-/long-5.3.2.tgz",
       "integrity": "sha512-mNAgZ1GmyNhD7AuqnTG3/VQ26o760+ZYBPKjPvugO8+nLbYfX6TVpJPseBvopbdY+qpZ/lKUnmEc1LeZYS3QAA==",
+      "devOptional": true,
       "license": "Apache-2.0"
     },
     "node_modules/proxy-addr": {
@@ -33280,6 +33316,7 @@
       "resolved": "https://registry.npmjs.org/roarr/-/roarr-2.15.4.tgz",
       "integrity": "sha512-CHhPh+UNHD2GTXNYhPWLnU8ONHdI+5DI+4EYIAOaiD63rHeYlZvyh8P+in5999TTSFgUYuKUAjzRI4mdh/p+2A==",
       "license": "BSD-3-Clause",
+      "optional": true,
       "dependencies": {
         "boolean": "^3.0.1",
         "detect-node": "^2.0.4",
@@ -33655,7 +33692,8 @@
       "version": "1.0.0",
       "resolved": "https://registry.npmjs.org/semver-compare/-/semver-compare-1.0.0.tgz",
       "integrity": "sha512-YM3/ITh2MJ5MtzaM429anh+x2jiLVjqILF4m4oyQB18W7Ggea7BfqdH/wGMK7dDiMghv/6WG7znWMwUDzJiXow==",
-      "license": "MIT"
+      "license": "MIT",
+      "optional": true
     },
     "node_modules/send": {
       "version": "1.2.1",
@@ -33688,6 +33726,7 @@
       "resolved": "https://registry.npmjs.org/serialize-error/-/serialize-error-7.0.1.tgz",
       "integrity": "sha512-8I8TjW5KMOKsZQTvoxjuSIa7foAwPWGOts+6o7sgjz41/qMD9VQHEDxi6PBvK2l0MXUmqZyNpUK+T2tQaaElvw==",
       "license": "MIT",
+      "optional": true,
       "dependencies": {
         "type-fest": "^0.13.1"
       },
@@ -33703,6 +33742,7 @@
       "resolved": "https://registry.npmjs.org/type-fest/-/type-fest-0.13.1.tgz",
       "integrity": "sha512-34R7HTnG0XIJcBSn5XhDd7nNFPRcXYRZrBB2O2jdKqYODldSzBAqzsWoZYYvduky73toYS/ESqxPvkDf/F0XMg==",
       "license": "(MIT OR CC0-1.0)",
+      "optional": true,
       "engines": {
         "node": ">=10"
       },
@@ -34474,7 +34514,8 @@
       "version": "1.1.3",
       "resolved": "https://registry.npmjs.org/sprintf-js/-/sprintf-js-1.1.3.tgz",
       "integrity": "sha512-Oo+0REFV59/rz3gfJNKQiBlwfHaSESl1pcGyABQsnnIfWOFt6JNj5gCog2U6MLZ//IGYD+nA8nI+mTShREReaA==",
-      "license": "BSD-3-Clause"
+      "license": "BSD-3-Clause",
+      "optional": true
     },
     "node_modules/sql.js": {
       "version": "1.14.2",
@@ -36187,6 +36228,7 @@
       "version": "8.3.0",
       "resolved": "https://registry.npmjs.org/undici-types/-/undici-types-8.3.0.tgz",
       "integrity": "sha512-j375ScV60dom+YkPFIfTLcOiPxkN/buHz5GobjLhixFuANaNs3C9l4GmrWqejgXWJ7BbJcFYpTEUkS1Ge8bpZQ==",
+      "devOptional": true,
       "license": "MIT"
     },
     "node_modules/unicode-emoji-modifier-base": {

package.json

@@ -265,7 +265,6 @@
     "@dnd-kit/core": "^6.3.1",
     "@dnd-kit/sortable": "^10.0.0",
     "@dnd-kit/utilities": "^3.2.2",
-    "@huggingface/transformers": "^4.2.0",
     "@lobehub/icons": "^5.16.0",
     "@modelcontextprotocol/sdk": "^1.29.0",
     "@monaco-editor/react": "^4.7.0",
@@ -308,7 +307,6 @@
     "next-themes": "^0.4.6",
     "node-machine-id": "^1.1.12",
     "omniglyph": "^1.4.0",
-    "onnxruntime-node": "1.24.3",
     "open": "^11.0.1",
     "ora": "^9.4.1",
     "parse5": "^8.0.1",
@@ -343,9 +341,11 @@
   },
   "optionalDependencies": {
     "@atjsh/llmlingua-2": "3.0.0",
+    "@huggingface/transformers": "^4.2.0",
     "better-sqlite3": "^13.0.2",
     "js-tiktoken": "^1.0.20",
     "keytar": "^7.9.0",
+    "onnxruntime-node": "1.24.3",
     "sqlite-vec": "^0.1.9",
     "tls-client-node": "^0.2.0",
     "wreq-js": "^3.0.0"

scripts/build/build-next-isolated.mjs

@@ -131,12 +131,12 @@ function runNextBuild() {
 }
 
 export function resolveNextBuildBundlerFlag(baseEnv = process.env) {
-  // Turbopack is the default production bundler (Next 16 stable). Benchmarked on
-  // this codebase: 2-3x faster than the single-threaded webpack pass (17min -> 9min
-  // on a 32-core box; ~20min -> 7min on ubuntu-latest), artifact validated
-  // end-to-end (standalone smoke + e2e/package/electron CI jobs). Webpack stays as
-  // the explicit escape hatch (=0) for bundler-compat regressions.
-  return baseEnv.OMNIROUTE_USE_TURBOPACK === "0" ? "--webpack" : "--turbopack";
+  // Turbopack is the default on Node.js; on Bun or when explicitly disabled (=0),
+  // use Webpack (--webpack) to avoid Turbopack V8 internal worker API mismatches.
+  if (process.versions.bun || baseEnv.OMNIROUTE_USE_TURBOPACK === "0") {
+    return "--webpack";
+  }
+  return "--turbopack";
 }
 
 /**

scripts/check/check-supported-node-runtime.ts

@@ -15,6 +15,12 @@ if (!support.nodeCompatible) {
   process.exit(1);
 }
 
-console.log(
-  `Node.js ${support.nodeVersion} satisfies OmniRoute secure runtime policy (${support.supportedRange}).`
-);
+if (process.versions.bun) {
+  console.log(
+    `Bun ${process.versions.bun} (${support.nodeVersion}) satisfies OmniRoute secure runtime policy.`
+  );
+} else {
+  console.log(
+    `Node.js ${support.nodeVersion} satisfies OmniRoute secure runtime policy (${support.supportedRange}).`
+  );
+}

scripts/dev/run-next.mjs

@@ -83,8 +83,10 @@ const { dashboardPort } = runtimePorts;
 const hostname = process.env.HOST || "0.0.0.0";
 // Turbopack by default in dev (matches the Next 16 CLI default and the production
 // build default in build-next-isolated.mjs); OMNIROUTE_USE_TURBOPACK=0 is the
-// webpack escape hatch.
-const useTurbopack = dev && mergedEnv.OMNIROUTE_USE_TURBOPACK !== "0";
+// webpack escape hatch. Under Bun, Turbopack native V8 bindings are unavailable,
+// so Bun automatically disables Turbopack and uses Webpack.
+const isBun = Boolean(process.versions.bun);
+const useTurbopack = dev && mergedEnv.OMNIROUTE_USE_TURBOPACK !== "0" && !isBun;
 process.env.OMNIROUTE_WS_BRIDGE_SECRET ||= randomUUID();
 // Per-process secret used to prove the trusted peer-IP stamp came from this
 // server (read by the authz middleware in the same process). See peer-stamp.mjs.

skills/omni-webhooks/SKILL.md

@@ -1,12 +1,12 @@
 ---
 name: omni-webhooks
-description: Register, list, test, and remove webhook endpoints. Configure event subscriptions (request.completed, provider.error, budget.exceeded, etc.) and manage delivery retries.
+description: Register, list, test, and remove webhook endpoints. Configure event subscriptions (request.completed, request.failed, quota.exceeded, etc.) and manage delivery retries.
 ---
 <!-- generated by src/lib/agentSkills/generator.ts; manual edits will be overwritten -->
 
 ## Overview
 
-Register, list, test, and remove webhook endpoints. Configure event subscriptions (request.completed, provider.error, budget.exceeded, etc.) and manage delivery retries.
+Register, list, test, and remove webhook endpoints. Configure event subscriptions (request.completed, request.failed, quota.exceeded, etc.) and manage delivery retries.
 
 ## Authentication
 

src/app/(dashboard)/dashboard/analytics/ComboHealthTab.tsx

@@ -291,7 +291,9 @@ function ComboAutopilotPanel({ report }: { report: ComboAutopilotReport }) {
             icon="monitor_heart"
             label={t("comboHealthIssues")}
             value={report.summary.issueCount.toLocaleString()}
-            subValue={t("comboHealthActionable", { count: report.summary.actionableCount })}
+            subValue={t("comboHealthActionable", {
+              count: report.summary.suggestionCount ?? report.summary.actionableCount ?? 0,
+            })}
           />
           <MetricBlock
             icon="error"

src/app/(dashboard)/dashboard/combos/IntelligentComboPanel.tsx

@@ -12,8 +12,13 @@ import {
 } from "@/lib/combos/intelligentRouting";
 import { AI_PROVIDERS } from "@/shared/constants/providers";
 
-function getI18nOrFallback(t: any, key: string, fallback: string) {
-  if (typeof t?.has === "function" && t.has(key)) return t(key);
+function getI18nOrFallback(
+  t: any,
+  key: string,
+  fallback: string,
+  values?: Record<string, unknown>
+) {
+  if (typeof t?.has === "function" && t.has(key)) return t(key, values);
   return fallback;
 }
 
@@ -94,10 +99,9 @@ export default function IntelligentComboPanel({
       const updatedCombo = await response.json();
       onComboUpdated?.(updatedCombo);
       notify.success(
-        getI18nOrFallback(t, "modePackUpdated", "Mode pack updated to {pack}.").replace(
-          "{pack}",
-          modePackId
-        )
+        getI18nOrFallback(t, "modePackUpdated", "Mode pack updated to {pack}.", {
+          pack: modePackId,
+        }).replace("{pack}", modePackId)
       );
     } catch (error: any) {
       notify.error(error?.message || "Failed to update mode pack.");
@@ -184,10 +188,9 @@ export default function IntelligentComboPanel({
               </div>
               {savingModePack && (
                 <span className="text-[11px] text-text-muted">
-                  {getI18nOrFallback(t, "savingModePack", "Saving {pack}…").replace(
-                    "{pack}",
-                    savingModePack
-                  )}
+                  {getI18nOrFallback(t, "savingModePack", "Saving {pack}…", {
+                    pack: savingModePack,
+                  }).replace("{pack}", savingModePack)}
                 </span>
               )}
             </div>

src/app/(dashboard)/dashboard/combos/page.tsx

@@ -533,9 +533,9 @@ function getStrategyBadgeClass(strategy) {
   return "bg-blue-500/15 text-blue-600 dark:text-blue-400";
 }
 
-function getI18nOrFallback(t, key, fallback) {
+function getI18nOrFallback(t, key, fallback, values) {
   try {
-    if (typeof t.has === "function" && t.has(key)) return t(key);
+    if (typeof t.has === "function" && t.has(key)) return t(key, values);
   } catch {}
   return fallback;
 }
@@ -1565,7 +1565,8 @@ function StrategyRecommendationsPanel({ strategy, onApply, showNudge }) {
           {getI18nOrFallback(
             t,
             "recommendationsUpdated",
-            "Recommendations updated for {strategy}."
+            "Recommendations updated for {strategy}.",
+            { strategy: strategyLabel }
           ).replace("{strategy}", strategyLabel)}
         </div>
       )}

src/app/(dashboard)/dashboard/providers/[id]/components/HarImportButton.tsx

@@ -0,0 +1,165 @@
+"use client";
+
+import { useRef, useState } from "react";
+import { useTranslations } from "next-intl";
+import {
+  extractM365CredentialFromHar,
+  describeHarImportExpiry,
+  type M365HarImportResult,
+} from "@/shared/utils/m365HarImport";
+import { providerText, type ProviderMessageTranslator } from "../providerPageHelpers";
+
+type HarImporter = (text: string) => M365HarImportResult;
+
+// One entry per web-session provider that can offer HAR import. Add a new
+// key here (and its own extractor in src/shared/utils/) to support another
+// provider — the button renders nothing for any provider not listed.
+const HAR_IMPORTERS: Record<string, HarImporter> = {
+  "copilot-m365-web": extractM365CredentialFromHar,
+};
+
+const ERROR_MESSAGE_KEYS: Record<string, [string, string]> = {
+  notJson: ["harImportErrorNotJson", "That file isn't valid JSON — is it really a .har export?"],
+  noEntries: ["harImportErrorNoEntries", "This HAR has no network entries recorded."],
+  noChathubUrl: [
+    "harImportErrorNoChathubUrl",
+    "No Copilot chat connection found in this HAR. Send at least one chat message in m365.cloud.microsoft before exporting.",
+  ],
+  unparsableUrl: [
+    "harImportErrorUnparsableUrl",
+    "Found the chat connection, but couldn't read its URL.",
+  ],
+  missingFields: [
+    "harImportErrorMissingFields",
+    "Found the chat connection, but the token was missing from it.",
+  ],
+};
+
+export interface HarImportButtonProps {
+  provider: string;
+  onImport: (apiKey: string) => void;
+}
+
+export default function HarImportButton({ provider, onImport }: HarImportButtonProps) {
+  const t = useTranslations("providers") as ProviderMessageTranslator;
+  const importer = HAR_IMPORTERS[provider];
+  const fileInputRef = useRef<HTMLInputElement>(null);
+  const [state, setState] = useState<
+    | { phase: "idle" }
+    | { phase: "reading" }
+    | { phase: "error"; message: string }
+    | { phase: "success"; expiresAt: number | null }
+  >({ phase: "idle" });
+
+  if (!importer) return null;
+
+  async function handleFile(file: File | undefined) {
+    if (!file) return;
+    setState({ phase: "reading" });
+    let text: string;
+    try {
+      text = await file.text();
+    } catch {
+      setState({
+        phase: "error",
+        message: providerText(t, "harImportErrorReadFailed", "Couldn't read that file."),
+      });
+      return;
+    }
+
+    const result = importer(text);
+    if (!result.ok) {
+      const [key, fallback] = ERROR_MESSAGE_KEYS[result.error] ?? [
+        "harImportErrorUnknown",
+        "Couldn't extract a credential from that HAR file.",
+      ];
+      setState({ phase: "error", message: providerText(t, key, fallback) });
+      return;
+    }
+
+    onImport(result.apiKey);
+    setState({ phase: "success", expiresAt: result.expiresAt });
+  }
+
+  const expiry = state.phase === "success" ? describeHarImportExpiry(state.expiresAt) : null;
+  const expiryText =
+    expiry?.tone === "unknown"
+      ? providerText(t, "harImportStatusUnknownExpiry", "Imported. Couldn't read its expiry.")
+      : expiry?.tone === "bad"
+        ? providerText(
+            t,
+            "harImportStatusExpired",
+            "Imported, but this token already expired ({minutes}m ago) — export a fresh HAR.",
+            { minutes: Math.abs(expiry.minutesRemaining ?? 0) }
+          )
+        : expiry?.tone === "warn"
+          ? providerText(
+              t,
+              "harImportStatusExpiringSoon",
+              "Imported — valid for only ~{minutes}m more.",
+              { minutes: expiry.minutesRemaining ?? 0 }
+            )
+          : expiry?.tone === "ok"
+            ? providerText(t, "harImportStatusValid", "Imported — valid for ~{minutes}m.", {
+                minutes: expiry.minutesRemaining ?? 0,
+              })
+            : null;
+
+  return (
+    <div className="flex flex-col gap-1.5">
+      <div className="flex items-center gap-2">
+        <button
+          type="button"
+          onClick={() => fileInputRef.current?.click()}
+          disabled={state.phase === "reading"}
+          data-testid="har-import-button"
+          className="inline-flex items-center gap-1.5 rounded border border-border px-2.5 py-1.5 text-xs font-medium text-text-main hover:bg-surface-hover disabled:opacity-50"
+        >
+          <span className="material-symbols-outlined text-[16px]" aria-hidden="true">
+            upload_file
+          </span>
+          {state.phase === "reading"
+            ? providerText(t, "harImportButtonBusy", "Importing…")
+            : providerText(t, "harImportButtonLabel", "Import .har file")}
+        </button>
+        <span className="text-xs text-text-muted">
+          {providerText(
+            t,
+            "harImportButtonHint",
+            "Export from DevTools Network tab after sending at least one chat message."
+          )}
+        </span>
+        <input
+          ref={fileInputRef}
+          type="file"
+          accept=".har,application/json"
+          data-testid="har-import-input"
+          className="hidden"
+          onChange={(event) => {
+            void handleFile(event.target.files?.[0]);
+            event.target.value = "";
+          }}
+        />
+      </div>
+      {state.phase === "error" && (
+        <p className="text-xs text-red-600 dark:text-red-400" data-testid="har-import-error">
+          {state.message}
+        </p>
+      )}
+      {state.phase === "success" && expiryText && (
+        <p
+          className={
+            expiry?.tone === "bad"
+              ? "text-xs text-red-600 dark:text-red-400"
+              : expiry?.tone === "warn"
+                ? "text-xs text-amber-700 dark:text-amber-300"
+                : "text-xs text-emerald-700 dark:text-emerald-300"
+          }
+          data-testid="har-import-status"
+        >
+          {expiryText}
+        </p>
+      )}
+    </div>
+  );
+}

src/app/(dashboard)/dashboard/providers/[id]/components/modals/AddApiKeyModal.tsx

@@ -31,6 +31,7 @@ import {
 import { getWebSessionCredentialRequirement } from "../../webSessionCredentials";
 import { useOpenRouterPresetControl } from "../OpenRouterPresetInput";
 import WebSessionCredentialGuide from "../WebSessionCredentialGuide";
+import HarImportButton from "../HarImportButton";
 import CcCompatibleRequestDefaultsFields from "./CcCompatibleRequestDefaultsFields";
 import { buildAddProviderSpecificData } from "./connectionProviderSpecificData";
 import { getCommandCodeAuthPhaseLabel } from "./commandCodeAuthPhase";
@@ -155,7 +156,7 @@ export default function AddApiKeyModal({
     if (!isOpen || wasOpen) return;
     // On open, reset baseUrl and assign a unique default name so a second API key
     // for the same provider doesn't reuse "main" and trigger the backend
-    // name-based upsert that would silently overwrite the first connection (#6499).
+    // name-based upsert that would silently overwrite the first connection (#6499, #11033).
     setFormData((current) => ({
       ...current,
       name: computeConnectionDefaultName(existingConnectionCount),
@@ -209,13 +210,13 @@ export default function AddApiKeyModal({
         ? "Freebuff uses an authentic CLI auth token obtained via codebuff CLI login or automated harvester."
         : isWebSessionCredential
           ? getWebSessionCredentialHint(t, webSessionCredential, providerDisplayName, false)
-        : isLocalSelfHostedProvider
-          ? t("localProviderApiKeyOptionalHint", {
-              provider: localProviderMetadata?.name || providerName || provider || "",
-            })
-          : apiKeyOptional
-            ? t("apiKeyOptionalHint")
-            : undefined;
+          : isLocalSelfHostedProvider
+            ? t("localProviderApiKeyOptionalHint", {
+                provider: localProviderMetadata?.name || providerName || provider || "",
+              })
+            : apiKeyOptional
+              ? t("apiKeyOptionalHint")
+              : undefined;
   const credentialValidationFailedMessage = isWebSessionCredential
     ? providerText(
         t,
@@ -750,6 +751,12 @@ export default function AddApiKeyModal({
                 t={t}
               />
             )}
+            {provider && (
+              <HarImportButton
+                provider={provider}
+                onImport={(apiKey) => setFormData({ ...formData, apiKey })}
+              />
+            )}
             {!isNoAuthWebSessionCredential && (
               <div className="flex gap-2">
                 <Input
@@ -757,6 +764,12 @@ export default function AddApiKeyModal({
                   type="password"
                   value={formData.apiKey}
                   onChange={(e) => setFormData({ ...formData, apiKey: e.target.value })}
+                  onKeyDown={(e) => {
+                    if (e.key === "Enter" && !validating && !saving) {
+                      e.preventDefault();
+                      handleValidate();
+                    }
+                  }}
                   className="flex-1"
                   placeholder={apiCredentialPlaceholder}
                   hint={apiCredentialHint}

src/app/(dashboard)/dashboard/providers/[id]/components/modals/EditConnectionModal.tsx

@@ -49,6 +49,7 @@ import {
 import { getWebSessionCredentialRequirement } from "../../webSessionCredentials";
 import { useOpenRouterPresetControl } from "../OpenRouterPresetInput";
 import WebSessionCredentialGuide from "../WebSessionCredentialGuide";
+import HarImportButton from "../HarImportButton";
 import CcCompatibleRequestDefaultsFields from "./CcCompatibleRequestDefaultsFields";
 import { CodexConnectionFields } from "./CodexFingerprintFields";
 import { assignEditApiKeyProviderSpecificData } from "./connectionProviderSpecificData";
@@ -909,6 +910,12 @@ export default function EditConnectionModal({
                 t={t}
               />
             )}
+            {provider && (
+              <HarImportButton
+                provider={provider}
+                onImport={(apiKey) => setFormData({ ...formData, apiKey })}
+              />
+            )}
             {!isNoAuthWebSessionCredential && (
               <div className="flex gap-2">
                 <Input

src/app/(dashboard)/dashboard/providers/[id]/components/modals/computeConnectionDefaultName.ts

@@ -4,7 +4,23 @@
 // connection. Deriving a unique default from the existing connection count keeps
 // the first connection ("main") backward-compatible while giving each subsequent
 // one a distinct name ("main-2", "main-3", …).
-export function computeConnectionDefaultName(existingConnectionCount?: number): string {
-  const count = existingConnectionCount ?? 0;
+export function computeConnectionDefaultName(
+  existingConnectionCountOrConnections?: number | string[] | { name?: string }[]
+): string {
+  if (Array.isArray(existingConnectionCountOrConnections)) {
+    const names = new Set(
+      existingConnectionCountOrConnections
+        .map((item) => (typeof item === "string" ? item : item?.name ?? ""))
+        .filter(Boolean)
+    );
+    if (!names.has("main")) return "main";
+    let index = 2;
+    while (names.has(`main-${index}`)) {
+      index++;
+    }
+    return `main-${index}`;
+  }
+
+  const count = existingConnectionCountOrConnections ?? 0;
   return count <= 0 ? "main" : `main-${count + 1}`;
 }

src/app/a2a/route.ts

@@ -17,6 +17,8 @@ import { logRoutingDecision } from "@/lib/a2a/routingLogger";
 import { createA2AStream, SSE_HEADERS } from "@/lib/a2a/streaming";
 import { A2A_SKILL_HANDLERS, executeA2ATaskWithState } from "@/lib/a2a/taskExecution";
 import { getSettings } from "@/lib/db/settings";
+import { isRequireApiKeyEnabled } from "@/shared/utils/featureFlags";
+import { extractApiKey, isValidApiKey } from "@/sse/services/auth";
 
 // ============ A2A v1.0 ↔ v0.3 compatibility layer ============
 // A2A 1.0 renamed the JSON-RPC methods (message/send → SendMessage,
@@ -136,14 +138,25 @@ function tokensMatch(provided: string, expected: string): boolean {
   return timingSafeEqual(a, b);
 }
 
-function authenticate(req: NextRequest): boolean {
-  // If no API key is configured, allow all requests
-  const configuredKey = process.env.OMNIROUTE_API_KEY;
-  if (!configuredKey) return true;
+async function authenticate(req: NextRequest): Promise<boolean> {
+  // /a2a is outside the authz proxy matcher, so the REQUIRE_API_KEY posture the
+  // pipeline enforces for /v1 never ran here — the route accepted every caller
+  // whenever OMNIROUTE_API_KEY was unset, which is the shipped default
+  // (GHSA-v54m-6rm3-p565). Apply the same posture directly: when a client key is
+  // required, demand a valid OmniRoute key; otherwise honor the legacy explicit
+  // A2A key; otherwise stay keyless (the same local-first default as /v1).
+  const apiKey = extractApiKey(req);
+  if (isRequireApiKeyEnabled()) {
+    return apiKey ? await isValidApiKey(apiKey) : false;
+  }
 
-  const authHeader = req.headers.get("authorization") || "";
-  const token = authHeader.replace(/^Bearer\s+/i, "");
-  return tokensMatch(token, configuredKey);
+  const configuredKey = process.env.OMNIROUTE_API_KEY;
+  if (configuredKey) {
+    return apiKey ? tokensMatch(apiKey, configuredKey) : false;
+  }
+
+  // No API key required and none configured — allow (keyless local-first).
+  return true;
 }
 
 // ============ JSON-RPC Helpers ============
@@ -179,7 +192,7 @@ async function rejectIfA2ADisabled(id: string | number | null) {
 
 export async function POST(req: NextRequest) {
   // Auth check
-  if (!authenticate(req)) {
+  if (!(await authenticate(req))) {
     return jsonRpcError(null, -32600, "Unauthorized: missing or invalid API key");
   }
 

src/app/api/acp/agents/route.ts

@@ -1,5 +1,7 @@
 import { NextResponse } from "next/server";
 import { z } from "zod";
+
+export const dynamic = "force-dynamic";
 import {
   type CliAgentInfo,
   detectInstalledAgents,

src/app/api/admin/concurrency/route.ts

@@ -1,4 +1,5 @@
 import { NextResponse } from "next/server";
+export const dynamic = "force-dynamic";
 import { getAllRateLimitStatus } from "@omniroute/open-sse/services/rateLimitManager.ts";
 import {
   getStats as getSemaphoreStats,

src/app/api/analytics/compression/route.ts

@@ -6,6 +6,7 @@
  */
 
 import { NextResponse } from "next/server";
+export const dynamic = "force-dynamic";
 import { getCompressionAnalyticsSummary } from "@/lib/db/compressionAnalytics";
 import { requireManagementAuth } from "@/lib/api/requireManagementAuth";
 

src/app/api/auth/csrf/route.ts

@@ -1,4 +1,5 @@
 import { NextResponse } from "next/server";
+export const dynamic = "force-dynamic";
 
 import { requireManagementAuth } from "@/lib/api/requireManagementAuth";
 import { issueDashboardCsrfToken } from "@/server/authz/csrf";

src/app/api/auth/status/route.ts

@@ -1,4 +1,5 @@
 import { NextResponse } from "next/server";
+export const dynamic = "force-dynamic";
 import { cookies } from "next/headers";
 import { jwtVerify } from "jose";
 

src/app/api/batches/[id]/route.ts

@@ -1,4 +1,5 @@
 import { NextResponse } from "next/server";
+export const dynamic = "force-dynamic";
 import { getBatch } from "@/lib/localDb";
 import { requireManagementAuth } from "@/lib/api/requireManagementAuth";
 

src/app/api/batches/route.ts

@@ -1,4 +1,5 @@
 import { NextResponse } from "next/server";
+export const dynamic = "force-dynamic";
 import { listBatches } from "@/lib/localDb";
 import { requireManagementAuth } from "@/lib/api/requireManagementAuth";
 

src/app/api/cache/entries/route.ts

@@ -1,4 +1,5 @@
 import { NextRequest, NextResponse } from "next/server";
+export const dynamic = "force-dynamic";
 import { isAuthenticated } from "@/shared/utils/apiAuth";
 import {
   listSemanticCacheEntries,

src/app/api/cache/reasoning/route.ts

@@ -1,4 +1,5 @@
 import { NextRequest, NextResponse } from "next/server";
+export const dynamic = "force-dynamic";
 import { isAuthenticated } from "@/shared/utils/apiAuth";
 import {
   clearReasoningCacheAll,

src/app/api/cache/route.ts

@@ -1,4 +1,5 @@
 import { NextRequest, NextResponse } from "next/server";
+export const dynamic = "force-dynamic";
 import {
   getCacheStats,
   clearCache,