build(deps): bump github/codeql-action from 4.37.8 to 4.37.9

Bumps [github/codeql-action](https://github.com/github/codeql-action) from 4.37.8 to 4.37.9.
- [Release notes](https://github.com/github/codeql-action/releases)
- [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md)
- [Commits](https://github.com/github/codeql-action/compare/v4.37.8...v4.37.9)

---
updated-dependencies:
- dependency-name: github/codeql-action
  dependency-version: 4.37.9
  dependency-type: direct:production
  update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>

.github/dependabot.yml

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

.github/workflows/docker-publish.yml

@@ -535,7 +535,7 @@ jobs:
       - name: Upload Trivy SARIF to Security tab
         if: needs.prepare.outputs.version != 'main'
         continue-on-error: true
-        uses: github/codeql-action/upload-sarif@v4.37.8
+        uses: github/codeql-action/upload-sarif@v4.37.9
         with:
           sarif_file: trivy-results.sarif
           category: trivy-image

AGENTS.md

@@ -83,7 +83,7 @@ Client → /v1/chat/completions (Next.js route)
 
 API routes follow a consistent pattern: `Route → CORS preflight → Zod body validation → Optional auth (extractApiKey/isValidApiKey) → API key policy enforcement → Handler delegation (open-sse)`. No global Next.js middleware — interception is route-specific.
 
-**Combo routing** (`open-sse/services/combo.ts`): 19 public strategies (priority, weighted, fill-first, round-robin, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline). Each target calls `handleSingleModel()` which wraps `handleChatCore()` with per-target error handling and circuit breaker checks. The `fusion` strategy is the exception: it fans out to a panel of models in parallel, then a judge model synthesizes one final answer (`open-sse/services/fusion.ts`). See `docs/routing/AUTO-COMBO.md` for the 15-factor Auto-Combo scoring + the full strategy table and `docs/architecture/RESILIENCE_GUIDE.md` for the 3 resilience layers.
+**Combo routing** (`open-sse/services/combo.ts`): 19 public strategies (priority, weighted, fill-first, round-robin, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline). Each target calls `handleSingleModel()` which wraps `handleChatCore()` with per-target error handling and circuit breaker checks. The `fusion` strategy is the exception: it fans out to a panel of models in parallel, then a judge model synthesizes one final answer (`open-sse/services/fusion.ts`). See `docs/routing/AUTO-COMBO.md` for the 16-factor Auto-Combo scoring + the full strategy table and `docs/architecture/RESILIENCE_GUIDE.md` for the 3 resilience layers.
 
 ---
 
@@ -422,7 +422,7 @@ For any non-trivial change, read the matching deep-dive first:
 | Repo navigation                               | `docs/architecture/REPOSITORY_MAP.md`                   |
 | Architecture                                  | `docs/architecture/ARCHITECTURE.md`                     |
 | Engineering reference                         | `docs/architecture/CODEBASE_DOCUMENTATION.md`           |
-| Auto-Combo (15-factor scoring, 19 strategies) | `docs/routing/AUTO-COMBO.md`                            |
+| Auto-Combo (16-factor scoring, 19 strategies) | `docs/routing/AUTO-COMBO.md`                            |
 | Resilience (3 mechanisms)                     | `docs/architecture/RESILIENCE_GUIDE.md`                 |
 | Reasoning replay                              | `docs/routing/REASONING_REPLAY.md`                      |
 | Skills framework                              | `docs/frameworks/SKILLS.md`                             |

README.md

@@ -49,11 +49,11 @@
 
 [![Discord](https://img.shields.io/badge/Discord-5865F2?style=for-the-badge&logo=discord&logoColor=white)](https://discord.gg/U47eFqAXCn)
 [![Telegram](https://img.shields.io/badge/Telegram-26A5E4?style=for-the-badge&logo=telegram&logoColor=white)](https://t.me/omnirouteOficial)
-[![WhatsApp Global](https://img.shields.io/badge/WhatsApp_Global-25D366?style=for-the-badge&logo=whatsapp&logoColor=white)](https://chat.whatsapp.com/JI7cDQ1GyaiDHhVBpLxf8b?mode=gi_t)
+[![WhatsApp Global](https://img.shields.io/badge/WhatsApp_Global-25D366?style=for-the-badge&logo=whatsapp&logoColor=white)](https://chat.whatsapp.com/FvuCbrpZmQ6I85n2vW5QIC?s=cl&p=a&mlu=4)
 [![WhatsApp Brasil](https://img.shields.io/badge/WhatsApp_Brasil-25D366?style=for-the-badge&logo=whatsapp&logoColor=white)](https://chat.whatsapp.com/KWgatljAjmbELQory59Oti?s=cl&p=a&mlu=4)
 [![Website](https://img.shields.io/badge/Website-omniroute.online-blue?logo=google-chrome&logoColor=white)](https://omniroute.online)
 
-**Questions, provider tips, roadmap & support → [Discord](https://discord.gg/U47eFqAXCn) · [Telegram](https://t.me/omnirouteOficial) · WhatsApp [🌍 Global](https://chat.whatsapp.com/JI7cDQ1GyaiDHhVBpLxf8b?mode=gi_t) / [🇧🇷 Brasil](https://chat.whatsapp.com/LTSpdFhXTxjH4R6CCNiKWz)**
+**Questions, provider tips, roadmap & support → [Discord](https://discord.gg/U47eFqAXCn) · [Telegram](https://t.me/omnirouteOficial) · WhatsApp [🌍 Global](https://chat.whatsapp.com/FvuCbrpZmQ6I85n2vW5QIC?s=cl&p=a&mlu=4) / [🇧🇷 Brasil](https://chat.whatsapp.com/KWgatljAjmbELQory59Oti?s=cl&p=a&mlu=4) / [Portal](https://portal.sthub.com.br/communities/groups/st-hub/channels/Omniroute-World-8kRjmK)**
 
 <br/>
 
@@ -431,7 +431,7 @@ All **19** strategies — mix & match per combo step:
   <tr>
     <td align="center">17</td>
     <td nowrap><code>auto</code></td>
-    <td>15-factor live scoring across every connection 🤖</td>
+    <td>16-factor live scoring across every connection 🤖</td>
   </tr>
   <tr>
     <td align="center">18</td>
@@ -445,7 +445,7 @@ All **19** strategies — mix & match per combo step:
   </tr>
 </table>
 
-<sub>The Auto-Combo engine scores every candidate on **15 factors** (health, quota, cost, latency, task fit, quality, session availability…) — see [`docs/routing/AUTO-COMBO.md`](docs/routing/AUTO-COMBO.md).</sub>
+<sub>The Auto-Combo engine scores every candidate on **16 factors** (health, quota, cost, latency, task fit, quality, session availability…) — see [`docs/routing/AUTO-COMBO.md`](docs/routing/AUTO-COMBO.md).</sub>
 
 ##
 
@@ -1183,10 +1183,10 @@ Métricas canônicas em 2026-08-24: **1.029 vídeos únicos** · **11.132.922 vi
 | 🐙 **GitHub** — follow for releases & tips | [@diegosouzapw](https://github.com/diegosouzapw)                                                                          |
 | 💬 **Discord**                             | [discord.gg/U47eFqAXCn](https://discord.gg/U47eFqAXCn)                                                                    |
 | ✈️ **Telegram**                            | [t.me/omnirouteOficial](https://t.me/omnirouteOficial)                                                                    |
-| 🟢 **WhatsApp — 🌍 Global**                | [join the group](https://chat.whatsapp.com/FvuCbrpZmQ6I85n2vW5QIC?s=cl&p=a&mlu=4)                                        |
+| 🟢 **WhatsApp — 🌍 Global**                | [join the group](https://chat.whatsapp.com/FvuCbrpZmQ6I85n2vW5QIC?s=cl&p=a&mlu=4)                                         |
 | 🟢 **WhatsApp — 🇧🇷 Brasil**                | [entrar no grupo](https://chat.whatsapp.com/KWgatljAjmbELQory59Oti?s=cl&p=a&mlu=4)                                        |
 | 🌍 **Website**                             | [omniroute.online](https://omniroute.online)                                                                              |
-| 🌍 **🌍StHub OmniRoute Community (free)** | [portal sthub](https://portal.sthub.com.br/communities/groups/st-hub/channels/Omniroute-World-8kRjmK)                     |
+| 🌍 **🌍StHub OmniRoute Community (free)**  | [portal sthub](https://portal.sthub.com.br/communities/groups/st-hub/channels/Omniroute-World-8kRjmK)                     |
 | 📦 **Source code**                         | [github.com/diegosouzapw/OmniRoute](https://github.com/diegosouzapw/OmniRoute)                                            |
 | 🐛 **Report a bug**                        | [open an issue](https://github.com/diegosouzapw/OmniRoute/issues) — attach `npm run system-info` output                   |
 | 🤝 **Contribute**                          | [CONTRIBUTING.md](CONTRIBUTING.md) · [Branching & Release Model](docs/ops/BRANCHING_MODEL.md) · pick a `good first issue` |
@@ -1269,7 +1269,7 @@ Métricas canônicas em 2026-08-24: **1.029 vídeos únicos** · **11.132.922 vi
   <tr><td nowrap><b><a href="docs/compression/COMPRESSION_RULES_FORMAT.md">Compression Rules Format</a></b></td><td>JSON rule-pack schemas for Caveman and RTK filters</td></tr>
   <tr><td nowrap><b><a href="docs/compression/COMPRESSION_LANGUAGE_PACKS.md">Compression Language Packs</a></b></td><td>Language detection and Caveman rule-pack authoring</td></tr>
   <tr><td nowrap><b><a href="docs/architecture/RESILIENCE_GUIDE.md">Resilience Guide</a></b></td><td>Circuit breakers, cooldowns, queue, anti-thundering herd, TLS spoofing</td></tr>
-  <tr><td nowrap><b><a href="docs/routing/AUTO-COMBO.md">Auto-Combo Engine</a></b></td><td>15-factor scoring, mode packs, self-healing</td></tr>
+  <tr><td nowrap><b><a href="docs/routing/AUTO-COMBO.md">Auto-Combo Engine</a></b></td><td>16-factor scoring, mode packs, self-healing</td></tr>
   <tr><td nowrap><b><a href="docs/ops/PROXY_GUIDE.md">Proxy Guide</a></b></td><td>3-level proxy system, 1proxy marketplace, registry CRUD</td></tr>
   <tr><td nowrap><b><a href="docs/reference/FREE_TIERS.md">Free Tiers</a></b></td><td>Consolidated directory: 38 documented recurring pools / 446 cataloged free-tier entries</td></tr>
   <tr><td nowrap><b><a href="docs/guides/FEATURES.md">Features Gallery</a></b></td><td>Visual dashboard tour with screenshots</td></tr>

changelog.d/features/12218-rankings-reliability-sort.md

@@ -0,0 +1 @@
+- **feat(rankings):** order Free Provider Rankings by what each provider actually served — `GET /api/free-provider-rankings?sortBy=reliability` and a "Most reliable first" toggle on the page. Providers with too few calls to state a success rate keep their score order below the measured ones; the default order is unchanged ([#12218](https://github.com/diegosouzapw/OmniRoute/pull/12218)).

changelog.d/features/12317-reliability-scoring-factor.md

@@ -0,0 +1 @@
+- **feat(auto-combo):** Auto-Combo scoring can now weigh how often a provider/model has actually succeeded. The engine already carried that number on every candidate — 24 hours of usage history behind a ten-sample floor, real-time metrics otherwise — and the scoring function never read it, while the weight table described `stability` as if it did. `reliability` (`1 - failureRate`, with the same field precedence and the same rate-bounding the speed ranking already uses, so a corrupt reading means "nothing observed" rather than "fails every call") is now a declared factor shipping at weight `0`, so routing is unchanged until an operator gives it one, and the `stability` description now matches what that factor computes ([#12317](https://github.com/diegosouzapw/OmniRoute/pull/12317))

changelog.d/features/12319-strict-zero-cost-exclusion-reason.md

@@ -0,0 +1 @@
+- **feat(routing):** With `freeAccessPolicy: "strict"`, the read-only candidate listing (`GET /v1/auto-combo/{channel}/candidates`) no longer hides the candidates the zero-cost guard excludes — the same read-only transparency the resilience filter already honours (#9133). Each candidate now carries `freeAccessExclusion` saying why it would be kept out, and it tells an exhausted allowance apart from a quota reading that never arrived or went stale, which used to look identical from the outside. Routing is unchanged: the listing reports, it never enforces. The separate `excludeTosAvoid` guard still drops its candidates without a reason; that gap is now documented rather than closed ([#12319](https://github.com/diegosouzapw/OmniRoute/pull/12319))

changelog.d/features/12320-radar-catalog-limits-training.md

@@ -0,0 +1 @@
+- **feat(radar):** The Radar catalog table now shows two facts it was already receiving from the feed and dropping on the floor: the per-model rate limits (requests and tokens, per minute and per day) in a new column, and a badge when a provider's terms state it may train on the prompts you send. A limit of zero renders as zero rather than "rate-only" — for a ceiling those are opposite facts — and a model with no training statement gets no badge, because an absent statement is not a guarantee ([#12320](https://github.com/diegosouzapw/OmniRoute/pull/12320))

changelog.d/fixes/12151-passthrough-usage-estimation.md

@@ -0,0 +1 @@
+- **fix(sse):** passthrough streams now estimate usage on finish when upstream closes without usage even with `stream_options.include_usage` — avoids `0 tokens / 0%` for providers that stay silent (and correctly handles trailing empty-choices usage) ([#12151](https://github.com/diegosouzapw/OmniRoute/pull/12151))

changelog.d/fixes/12177-combo-clear-agent-features.md

@@ -0,0 +1,4 @@
+- **fix(combos):** clearing an agent feature in the combos editor now persists — unchecking
+  context cache protection, or emptying the system message or tool filter, sends an explicit
+  `null` instead of dropping the field from the `PUT` body, which the update merge read as
+  "leave unchanged" ([#12177](https://github.com/diegosouzapw/OmniRoute/pull/12177)) — thanks @foreveryh

changelog.d/fixes/12215-free-tier-overlay-freshness.md

@@ -0,0 +1 @@
+- **fix(free-tier):** `/api/free-tier/summary` no longer computes its totals from a Radar feed built before the catalog the running release ships. When the cached feed is older — or carries no build date at all — the route answers from the shipped catalog, resolved through the operator's local model state so disabled and tombstoned models stay out of the numbers ([#12215](https://github.com/diegosouzapw/OmniRoute/pull/12215)).

changelog.d/fixes/12242-passthrough-402-per-model-lockout.md

@@ -0,0 +1 @@
+- **fix(resilience):** a 402 on a single paid model of a passthrough/gateway provider (e.g. `kilo-gateway`, `ollama-cloud`) no longer terminalizes the whole connection with a never-auto-recovered `credits_exhausted` status — only the paid model is locked out, so free models on the same key keep serving. 402 variant of [#3027](https://github.com/diegosouzapw/OmniRoute/issues/3027). Single-credential providers are unaffected — a 402 there is still treated as the key being genuinely out of credit ([#5239](https://github.com/diegosouzapw/OmniRoute/issues/5239) / [#10616](https://github.com/diegosouzapw/OmniRoute/issues/10616)) ([#12242](https://github.com/diegosouzapw/OmniRoute/issues/12242)) — thanks @brick30llc-ctrl

changelog.d/fixes/12314-scoring-weights-coverage.md

@@ -0,0 +1 @@
+- **fix(combo):** Expose the two Auto-Combo scoring factors nobody could set — the combo validation schema and the dashboard weight sliders both declared 13 of the scorer's 15 factors, so `connectionDensity` (spreads load across a provider's connections) and `quality` were dropped on save and offered nowhere. The sliders also shipped their own default table that differed from the engine's on every non-zero factor and summed to 1.05, so the percentages shown next to them added up to 105%. Both lists now match `DEFAULT_WEIGHTS`, and a test keeps them there. Note that a combo whose stored `weights` omitted the two keys was effectively running with them at zero and the other thirteen renormalized upward; it now runs with the engine's intended distribution, so its routing does shift ([#12314](https://github.com/diegosouzapw/OmniRoute/pull/12314))

changelog.d/fixes/12318-free-tiers-confidence-claim.md

@@ -0,0 +1 @@
+- **fix(docs):** The free-tier reference no longer says its numbers come "confidence tagged per row" — no catalog entry carries a confidence tag and the API serves none, so every figure on that page is an estimate of the same, unstated quality. The page now states what an entry does vouch for: an independently documented hard stop (set by hand with the source in a comment, never defaulted to `true`) and a prompt-training disclosure, both with live counts the `check:docs-counts` gate keeps honest ([#12318](https://github.com/diegosouzapw/OmniRoute/pull/12318))

changelog.d/fixes/12321-usage-fetcher-registration.md

@@ -0,0 +1 @@
+- **fix(usage):** `adobe-firefly` and `firefly` have had a working usage fetcher since Adobe Firefly landed, but neither was ever added to the registration list, so the provider-plugin manifest, `genericQuotaFetcher` and the free-access quota cache all reported them as having no usage support — while `USAGE_SUPPORTED_PROVIDERS` said the opposite. Both are now declared, which also means their credit balance is fetched like any other declared provider's: `registerGenericQuotaFetchers` now registers a generic quota fetcher for them, and `resolveFreeAccessState` no longer returns early. A test holds the registration list to the dispatcher's switch in both directions, which is what the module's own docstring already asked for in prose ([#12321](https://github.com/diegosouzapw/OmniRoute/pull/12321))

changelog.d/maintenance/12316-mode-packs-doc-gate.md

@@ -0,0 +1 @@
+- **docs(auto-combo):** The mode pack table in `docs/routing/AUTO-COMBO.md` now lists all six shipped packs with every weight each one sets, replacing a four-pack table whose numbers had also drifted from the source. It states plainly that no pack sets `quality`, so selecting any pack silences the observed-quality signal. Six more documents that quote the scoring factor count joined the `check:docs-counts` gate, which caught five stale claims — including one naming nine factors that do not exist — and two stale mode pack counts. The dashboard routing panel, which offered four of the six packs and labelled the default strategy "6-Factor Scoring", is now covered by a test; the two packs it was missing are `reliability-first` and `chaos-mode`, the latter labelled as the fault-injection profile it is rather than as one more routing preference ([#12316](https://github.com/diegosouzapw/OmniRoute/pull/12316))

docs/architecture/ARCHITECTURE.md

@@ -370,14 +370,18 @@ Key capabilities:
   **auto**, lkgp, context-optimized, context-relay, **fusion**, plus a fallback path) —
   auto is the headline addition in v3.8.0; `fusion` (panel fan-out + judge synthesis,
   `open-sse/services/fusion.ts`) is new in v3.8.36.
-- **9-factor scoring**: cost, latency p95, success rate, quota headroom, lockout
-  proximity, breaker state, recent failures, model availability, and tag affinity.
+- **16-factor scoring**: quota, health, inverse cost, inverse latency, task fit and
+  ten more. The canonical table of factors and their default weights lives in
+  [`docs/routing/AUTO-COMBO.md`](../routing/AUTO-COMBO.md) — restating it here would
+  give it a second place to go stale.
 - **Virtual factory** materializes ephemeral combos when no matching named combo
   exists, sourcing candidates from healthy active provider connections.
 - **Auto prefixes**: `auto/coding`, `auto/cheap`, `auto/fast`, `auto/offline`,
   `auto/smart`, `auto/lkgp` — each backed by a tuned weight profile.
-- **4 mode packs**: coding, fast, cheap, smart — shipped as preset weight
-  configurations callable from the dashboard.
+- **6 mode packs**: `ship-fast`, `cost-saver`, `quality-first`, `offline-friendly`,
+  `reliability-first` and `chaos-mode` — preset weight configurations callable from
+  the dashboard. (Not to be confused with the `auto/*` prefixes above, which are
+  request-time variants.)
 
 For full algorithmic detail (factor formulas, weight tuning), see
 [`docs/routing/AUTO-COMBO.md`](../routing/AUTO-COMBO.md).

docs/architecture/REPOSITORY_MAP.md

@@ -402,24 +402,24 @@ open-sse/
 
 ### Subsystem deep-dives
 
-| Doc                        | Purpose                                                             |
-| -------------------------- | ------------------------------------------------------------------- |
-| `MCP-SERVER.md`            | MCP server: 110 tools, 3 transports, 33 scopes, REST endpoints      |
-| `A2A-SERVER.md`            | A2A v0.3: JSON-RPC, 6 skills, REST helpers, agent card              |
-| `AGENT_PROTOCOLS_GUIDE.md` | Unified guide: A2A vs ACP vs Cloud Agents                           |
-| `CLOUD_AGENT.md`           | Codex Cloud / Devin / Jules orchestration                           |
-| `SKILLS.md`                | Skills framework (built-in + marketplace + SkillsSH + sandbox)      |
-| `RADAR.md`                 | Radar free-model catalog overlay (`RADAR_ENABLED`, off by default)  |
-| `MEMORY.md`                | Memory system (SQLite FTS5 + Qdrant)                                |
-| `EVALS.md`                 | Eval framework (suites, runs, rubrics)                              |
-| `GUARDRAILS.md`            | PII masker, prompt injection, vision bridge                         |
-| `COMPLIANCE.md`            | Audit log, retention, noLog opt-out                                 |
-| `WEBHOOKS.md`              | HMAC-signed webhook delivery                                        |
-| `REASONING_REPLAY.md`      | Hybrid memory/SQLite cache for `reasoning_content`                  |
-| `AUTHZ_GUIDE.md`           | Authorization pipeline (`classify` → `policies` → `enforce`)        |
-| `RESILIENCE_GUIDE.md`      | Circuit breaker + cooldown + model lockout                          |
-| `STEALTH_GUIDE.md`         | TLS fingerprinting (JA3/JA4), Claude Code CCH, MITM cert            |
-| `AUTO-COMBO.md`            | Auto Combo engine (9-factor scoring, 4 mode packs, virtual factory) |
+| Doc                        | Purpose                                                              |
+| -------------------------- | -------------------------------------------------------------------- |
+| `MCP-SERVER.md`            | MCP server: 110 tools, 3 transports, 33 scopes, REST endpoints       |
+| `A2A-SERVER.md`            | A2A v0.3: JSON-RPC, 6 skills, REST helpers, agent card               |
+| `AGENT_PROTOCOLS_GUIDE.md` | Unified guide: A2A vs ACP vs Cloud Agents                            |
+| `CLOUD_AGENT.md`           | Codex Cloud / Devin / Jules orchestration                            |
+| `SKILLS.md`                | Skills framework (built-in + marketplace + SkillsSH + sandbox)       |
+| `RADAR.md`                 | Radar free-model catalog overlay (`RADAR_ENABLED`, off by default)   |
+| `MEMORY.md`                | Memory system (SQLite FTS5 + Qdrant)                                 |
+| `EVALS.md`                 | Eval framework (suites, runs, rubrics)                               |
+| `GUARDRAILS.md`            | PII masker, prompt injection, vision bridge                          |
+| `COMPLIANCE.md`            | Audit log, retention, noLog opt-out                                  |
+| `WEBHOOKS.md`              | HMAC-signed webhook delivery                                         |
+| `REASONING_REPLAY.md`      | Hybrid memory/SQLite cache for `reasoning_content`                   |
+| `AUTHZ_GUIDE.md`           | Authorization pipeline (`classify` → `policies` → `enforce`)         |
+| `RESILIENCE_GUIDE.md`      | Circuit breaker + cooldown + model lockout                           |
+| `STEALTH_GUIDE.md`         | TLS fingerprinting (JA3/JA4), Claude Code CCH, MITM cert             |
+| `AUTO-COMBO.md`            | Auto Combo engine (16-factor scoring, 6 mode packs, virtual factory) |
 
 ### Compression
 

docs/architecture/RESILIENCE_GUIDE.md

@@ -652,4 +652,4 @@ default `test:integration`, chaos and heap self-skip (without `RUN_CHAOS_INT`/`-
 
 - [Architecture Guide](./ARCHITECTURE.md) — System architecture and internals
 - [User Guide](../guides/USER_GUIDE.md) — Providers, combos, CLI integration
-- [Auto-Combo Engine](../routing/AUTO-COMBO.md) — 13-factor scoring, mode packs
+- [Auto-Combo Engine](../routing/AUTO-COMBO.md) — 16-factor scoring, mode packs

docs/diagrams/auto-combo-scoring.mmd

@@ -1,13 +1,13 @@
-%% Auto-Combo 15-factor scoring
+%% Auto-Combo 16-factor scoring
 %% Reflects: open-sse/services/autoCombo/scoring.ts (DEFAULT_WEIGHTS, sum = 1.0)
 %% v3.8.50
-%% svg-title: OmniRoute Auto-Combo 15-factor scoring
+%% svg-title: OmniRoute Auto-Combo 16-factor scoring
 %% svg-description: Flow from an incoming request through eligible candidates, the 15 weighted scoring factors, descending score sort, top-N selection, and sequential dispatch.
 flowchart TB
   Request["Incoming request"] --> Candidates["Eligible candidates<br/>(provider × model × account)"]
   Candidates --> Score["Compute composite score<br/>per candidate"]
 
-  subgraph Factors["15-factor scoring weights (sum = 1.0)"]
+  subgraph Factors["16-factor scoring weights (sum = 1.0)"]
     f1["quota (0.1429)"]
     f2["health (0.1605)"]
     f3["costInv (0.1429)"]

docs/diagrams/strategies-grid.svg

@@ -95,7 +95,7 @@
 <rect width="173" height="158" rx="10" fill="#161b22" stroke="#ffffff" stroke-opacity="0.07" stroke-width="1"/>
 <text x="12" y="21" font-family="Consolas, 'Courier New', monospace" font-size="11" fill="#a78bfa">auto</text>
 <circle cx="20" cy="79" r="4" fill="none" stroke="#c9d1d9" stroke-width="1.6"/><circle cx="20" cy="79" r="1.6" fill="#c9d1d9"/><path d="M 26,79 C 62,79 84,67.5 112,67.5" fill="none" stroke="#8b5cf6" stroke-opacity="0.55" stroke-width="1.6"/><rect x="116" y="38.0" width="26" height="11" rx="2.5" fill="#1c2330" stroke="#ffffff" stroke-opacity="0.10" stroke-width="1"/><text x="147" y="46.5" font-family="Consolas, 'Courier New', monospace" font-size="8.5" fill="#71717a">72</text><rect x="116" y="62.0" width="26" height="11" rx="2.5" fill="#1c2330" stroke="#7ee787" stroke-opacity="0.8" stroke-width="1"/><text x="147" y="70.5" font-family="Consolas, 'Courier New', monospace" font-size="8.5" fill="#71717a">91</text><rect x="116" y="86.0" width="26" height="11" rx="2.5" fill="#1c2330" stroke="#ffffff" stroke-opacity="0.10" stroke-width="1"/><text x="147" y="94.5" font-family="Consolas, 'Courier New', monospace" font-size="8.5" fill="#71717a">64</text><rect x="116" y="110.0" width="26" height="11" rx="2.5" fill="#1c2330" stroke="#ffffff" stroke-opacity="0.10" stroke-width="1"/><text x="147" y="118.5" font-family="Consolas, 'Courier New', monospace" font-size="8.5" fill="#71717a">55</text><circle r="2.8" fill="#a78bfa" opacity="0"><animateMotion path="M 26,79 C 62,79 84,67.5 110,67.5" begin="3.3s" dur="3.6s" repeatCount="indefinite"/><animate attributeName="opacity" values="0;1;1;0;0" keyTimes="0;0.02;0.3;0.33999999999999997;1" begin="3.3s" dur="3.6s" repeatCount="indefinite"/></circle>
-<text x="12" y="148" font-family="Inter, 'Segoe UI', Arial, Helvetica, system-ui, sans-serif" font-size="9.5" fill="#71717a">live 15-factor scoring</text>
+<text x="12" y="148" font-family="Inter, 'Segoe UI', Arial, Helvetica, system-ui, sans-serif" font-size="9.5" fill="#71717a">live 16-factor scoring</text>
 </g><g transform="translate(796,456)">
 <rect width="173" height="158" rx="10" fill="#161b22" stroke="#ffffff" stroke-opacity="0.07" stroke-width="1"/>
 <text x="12" y="21" font-family="Consolas, 'Courier New', monospace" font-size="11" fill="#a78bfa">fusion</text>

docs/frameworks/OPEN_SSE_ARCHITECTURE.md

@@ -105,7 +105,7 @@ Resolves the request to a concrete `(provider, model, account, credentials)` tup
 
 For `auto/*` models, this stage also:
 
-- Runs the **15-factor scoring** algorithm (`services/autoCombo/`)
+- Runs the **16-factor scoring** algorithm (`services/autoCombo/`)
 - Selects a `provider+model` pair based on health, cost, latency, etc.
 
 ### Stage 2: Translate (translator/)
@@ -245,7 +245,7 @@ Supports **19 routing strategies** (see `src/shared/constants/routingStrategies.
 | `reset-window`      | Reset window-based routing                                                |
 | `headroom`          | Most remaining quota headroom first                                       |
 | `strict-random`     | Truly uniform (no quality weighting)                                      |
-| `auto`              | Use 15-factor scoring (`autoCombo/`)                                      |
+| `auto`              | Use 16-factor scoring (`autoCombo/`)                                      |
 | `lkgp`              | Last known good provider first                                            |
 | `context-optimized` | Best for long-context requests                                            |
 | `fusion`            | Fan out to a panel in parallel, then synthesize via a judge (`fusion.ts`) |
@@ -280,7 +280,7 @@ Services are **focused, single-purpose modules** that handlers compose. The big
 ### Routing & Combo
 
 - `combo.ts` — entry point for combo-routed requests
-- `services/autoCombo/` — 15-factor scoring, 8 auto routing strategies
+- `services/autoCombo/` — 16-factor scoring, 8 auto routing strategies
 - `wildcardRouter.ts` — matches wildcard routes (`gpt-*`)
 - `modelFamilyFallback.ts` — T5 intra-family fallback
 
@@ -570,7 +570,7 @@ The routing engine has strict performance budgets:
 - [ARCHITECTURE.md](../architecture/ARCHITECTURE.md) — high-level architecture
 - [CODEBASE_DOCUMENTATION.md](../architecture/CODEBASE_DOCUMENTATION.md) — engineering reference
 - [REPOSITORY_MAP.md](../architecture/REPOSITORY_MAP.md) — directory-by-directory
-- [AUTO-COMBO.md](../routing/AUTO-COMBO.md) — 15-factor scoring
+- [AUTO-COMBO.md](../routing/AUTO-COMBO.md) — 16-factor scoring
 - [MCP-SERVER.md](./MCP-SERVER.md) — MCP server
 - [A2A-SERVER.md](./A2A-SERVER.md) — A2A server
 - Source: `open-sse/` (400+ files, ~143K LOC)

docs/getting-started/AUTO-COMBO-GUIDE.md

@@ -46,14 +46,14 @@ model: "auto/cheap"     # Cheapest option
 
 ## Which "auto" Should I Use?
 
-| If you want... | Use this | Best for | How it works |
-|----------------|----------|----------|--------------|
-| **Best overall** | `auto` | General questions, chat | Balances speed, cost, and quality |
-| **Best code** | `auto/coding` | Writing code, debugging | Picks models good at coding tasks |
-| **Fastest response** | `auto/fast` | Quick answers, low latency | Prioritizes speed over everything |
-| **Cheapest option** | `auto/cheap` | Saving money | Picks the cheapest provider |
-| **Smartest model** | `auto/smart` | Complex tasks | Quality-first + explores new models |
-| **Most available** | `auto/offline` | When providers are busy | Picks providers with most capacity |
+| If you want...       | Use this       | Best for                   | How it works                        |
+| -------------------- | -------------- | -------------------------- | ----------------------------------- |
+| **Best overall**     | `auto`         | General questions, chat    | Balances speed, cost, and quality   |
+| **Best code**        | `auto/coding`  | Writing code, debugging    | Picks models good at coding tasks   |
+| **Fastest response** | `auto/fast`    | Quick answers, low latency | Prioritizes speed over everything   |
+| **Cheapest option**  | `auto/cheap`   | Saving money               | Picks the cheapest provider         |
+| **Smartest model**   | `auto/smart`   | Complex tasks              | Quality-first + explores new models |
+| **Most available**   | `auto/offline` | When providers are busy    | Picks providers with most capacity  |
 
 ### Examples
 
@@ -81,7 +81,7 @@ curl http://localhost:20128/v1/chat/completions \
 When you send a request with `model: "auto"`, OmniRoute:
 
 1. **Looks at all your connected providers** — Every provider you've added (OpenAI, Anthropic, Google, etc.)
-2. **Scores each one** on 5 factors:
+2. **Scores each one**, weighing among other things:
    - Is it working? (health)
    - Does it have capacity? (quota)
    - How much does it cost? (price)
@@ -94,29 +94,29 @@ When you send a request with `model: "auto"`, OmniRoute:
 
 Each provider gets a score from 0 to 1. The higher the score, the better the fit.
 
-| Factor | Weight | What it means |
-|--------|--------|---------------|
-| Health | 20% | Is the provider working? (circuit breaker state) |
-| Quota | 15% | Does it have capacity remaining? |
-| Cost | 15% | How expensive is it? (cheaper = higher score) |
-| Speed | 12% | How fast is it? (lower latency = higher score) |
-| Task Fit | 8% | Is it good at this type of task? |
-| Stability | 5% | Is it consistent? (low error rate) |
-| Tier | 5% | Account tier (Ultra > Pro > Free) |
-| Other | 20% | Context affinity, connection density, etc. |
+| Factor    | Weight | What it means                                    |
+| --------- | ------ | ------------------------------------------------ |
+| Health    | 20%    | Is the provider working? (circuit breaker state) |
+| Quota     | 15%    | Does it have capacity remaining?                 |
+| Cost      | 15%    | How expensive is it? (cheaper = higher score)    |
+| Speed     | 12%    | How fast is it? (lower latency = higher score)   |
+| Task Fit  | 8%     | Is it good at this type of task?                 |
+| Stability | 5%     | Is it consistent? (low error rate)               |
+| Tier      | 5%     | Account tier (Ultra > Pro > Free)                |
+| Other     | 20%    | Context affinity, connection density, etc.       |
 
 ### How Variants Change the Scoring
 
 Each variant uses different weights:
 
-| Variant | Prioritizes | Key Weights |
-|---------|-------------|-------------|
-| `auto` | Balanced | health=20%, quota=15%, cost=15% |
-| `auto/coding` | Quality | taskFit=37%, stability=15% |
-| `auto/fast` | Speed | latency=32%, health=28% |
-| `auto/cheap` | Cost | cost=37% |
-| `auto/smart` | Quality + Explore | taskFit=37%, exploration=10% |
-| `auto/offline` | Capacity | quota=37%, health=28% |
+| Variant        | Prioritizes       | Key Weights                     |
+| -------------- | ----------------- | ------------------------------- |
+| `auto`         | Balanced          | health=20%, quota=15%, cost=15% |
+| `auto/coding`  | Quality           | taskFit=37%, stability=15%      |
+| `auto/fast`    | Speed             | latency=32%, health=28%         |
+| `auto/cheap`   | Cost              | cost=37%                        |
+| `auto/smart`   | Quality + Explore | taskFit=37%, exploration=10%    |
+| `auto/offline` | Capacity          | quota=37%, health=28%           |
 
 ---
 
@@ -125,15 +125,19 @@ Each variant uses different weights:
 OmniRoute has **three layers of protection**:
 
 ### 1. Auto-Fallback
+
 If the best provider fails, OmniRoute automatically tries the next one. You don't need to do anything.
 
 ### 2. Self-Healing
+
 If a provider keeps failing:
+
 - **Score < 0.2** → Excluded for 5 minutes
 - **Circuit breaker open** → Auto-excluded
 - **More than 50% providers down** → Incident mode (no exploration)
 
 ### 3. Emergency Fallback
+
 If all providers fail, OmniRoute routes to stable free providers (like Kiro or Qoder) as a last resort.
 
 ---
@@ -209,7 +213,8 @@ Round-robin cycles through providers in order. Auto-combo **scores each provider
 ## Learn More
 
 For developers and contributors, see the [Auto-Combo Technical Reference](../routing/AUTO-COMBO.md) for:
-- Full 13-factor scoring algorithm
+
+- Full 16-factor scoring algorithm
 - Mode pack weight tables
 - Implementation file paths
 - API endpoints

docs/guides/FEATURES.md

@@ -18,7 +18,7 @@ Visual guide to every section of the OmniRoute dashboard.
 
 The v3.7.x → v3.8.0 cycle added zero-config auto routing, new providers, OAuth flows, deeper resilience, and a much richer CLI experience. Headline features below — full details further in the document and in linked specs.
 
-- 🤖 **Auto Combo / Zero-config auto-routing** — use prefixes `auto/coding`, `auto/fast`, `auto/cheap`, `auto/offline`, `auto/smart`, `auto/lkgp`, `auto/chaos`. Backed by a 15-factor scoring engine and 6 curated **mode packs** (ship-fast, cost-saver, quality-first, offline-friendly)
+- 🤖 **Auto Combo / Zero-config auto-routing** — use prefixes `auto/coding`, `auto/fast`, `auto/cheap`, `auto/offline`, `auto/smart`, `auto/lkgp`, `auto/chaos`. Backed by a 16-factor scoring engine and 6 curated **mode packs** (ship-fast, cost-saver, quality-first, offline-friendly, reliability-first, chaos-mode)
 - 🆕 **Command Code provider** (#2199) — first-class registration with model catalog and quota tracking
 - 🆕 **Z.AI provider** — new free-tier provider with quota labels
 - 🎬 **KIE media expansion** — extended catalog including video generation models

docs/guides/FREE_PROVIDER_RANKINGS.md

@@ -238,7 +238,7 @@ provider that ships at least one strong free model.
    `user_override → arena_elo → models_dev_tier → static table`). So after you connect
    the top free providers, routing with `model: "auto"` (e.g. `auto/coding`) will
    automatically prefer the higher-quality free models per request. See
-   [Auto-Combo](../routing/AUTO-COMBO.md) for the full 15-factor scoring.
+   [Auto-Combo](../routing/AUTO-COMBO.md) for the full 16-factor scoring.
 
 ---
 
@@ -248,7 +248,7 @@ provider that ships at least one strong free model.
   providers, no credit card required.
 - [Free Tiers directory](../reference/FREE_TIERS.md) — full catalog of free providers
   and their limits.
-- [Auto-Combo](../routing/AUTO-COMBO.md) — the 15-factor routing engine that consumes the
+- [Auto-Combo](../routing/AUTO-COMBO.md) — the 16-factor routing engine that consumes the
   same Arena ELO task-fitness data.
 - [Environment variables](../reference/ENVIRONMENT.md) — `ARENA_ELO_SYNC_ENABLED` /
   `ARENA_ELO_SYNC_INTERVAL` reference.

docs/guides/TIERS.md

@@ -69,7 +69,7 @@ rate limits — circuit breaker recovers them on backoff.
 Dashboard → **Tiers** → assign your providers. Defaults (from `tierDefaults.json`) are
 sensible; edit when you have specific subscriptions to prioritize or providers to exclude.
 
-Auto-Combo's 15-factor scoring also considers tier. See
+Auto-Combo's 16-factor scoring also considers tier. See
 [`docs/routing/AUTO-COMBO.md`](../routing/AUTO-COMBO.md).
 
 ## Telemetry

docs/i18n/ar/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/az/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/bg/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/bn/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/cs/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/da/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/de/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/es/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/fa/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/fi/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/fr/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/gu/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/he/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/hi/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/hu/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/id/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/in/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/it/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/ja/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/ko/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/mr/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/ms/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/nl/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/no/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/phi/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/pl/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/pt-BR/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/pt/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/ro/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/ru/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/sk/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/sv/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/sw/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/ta/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/te/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/th/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/tr/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/uk-UA/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/ur/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/vi/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/zh-CN/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/i18n/zh-TW/llm.txt

@@ -227,7 +227,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -267,7 +267,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -286,7 +286,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -445,7 +445,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -480,7 +480,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

docs/reference/FREE_TIERS.md

@@ -54,10 +54,10 @@ A 50-agent web-research pass (official docs + last-7-days news, adversarially ve
 OmniRoute answers "is it free?" through two regimes that intentionally read
 different sources:
 
-| Regime | Source of truth | Surfaces |
-|---|---|---|
-| **Counting / displaying** | Resolved catalog — the shipped baseline overlaid by the Radar feed (`getRadarCatalog`) | Free-tier totals, budget card, dashboards |
-| **Deciding** | Shipped catalog only (`FREE_MODEL_BUDGETS` in `open-sse/config/freeModelCatalog.data.ts`) plus the local heuristics (`:free` suffix, zero pricing, `grantsFreeAccess`) | Every consumer of `src/shared/utils/freeModels.ts`: model import, `auto/*` routing, `GET /v1/models`, and the browser previews |
+| Regime                    | Source of truth                                                                                                                                                        | Surfaces                                                                                                                       |
+| ------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------ |
+| **Counting / displaying** | Resolved catalog — the shipped baseline overlaid by the Radar feed (`getRadarCatalog`)                                                                                 | Free-tier totals, budget card, dashboards                                                                                      |
+| **Deciding**              | Shipped catalog only (`FREE_MODEL_BUDGETS` in `open-sse/config/freeModelCatalog.data.ts`) plus the local heuristics (`:free` suffix, zero pricing, `grantsFreeAccess`) | Every consumer of `src/shared/utils/freeModels.ts`: model import, `auto/*` routing, `GET /v1/models`, and the browser previews |
 
 Counting can improve whenever a feed is available. Deciding stays on the
 release artifact, so the answer is identical in the browser and on the server,
@@ -68,7 +68,8 @@ purpose.
 
 ## Methodology & caveats
 
-- Numbers are **upper-bound estimates** from each provider's documented free-tier limits as of **2026-06-17**, gathered by web research (confidence tagged per row). Free tiers change constantly — re-verify before relying on a figure.
+- Numbers are **upper-bound estimates** from each provider's documented free-tier limits as of **2026-06-17**, gathered by web research. Free tiers change constantly — re-verify before relying on a figure.
+- **What an entry actually vouches for.** No entry carries a per-row confidence rating, and the API serves none — treat every figure above as an estimate of the same, unstated quality. Two facts are different, because they are curated by hand rather than inferred: 7 entries carry an independently documented hard stop, and 13 entries carry a prompt-training disclosure. `hardStopGuaranteed` is set only when the provider's own terms say that exceeding the free allowance refuses the request rather than silently starting to bill you, with the source in a comment next to the entry; it is never defaulted to `true`, and an entry nobody has verified stays unset. So a missing hard-stop flag means "not established", not "known to bill you".
 - `estMonthlyFreeTokens` = recurring monthly tokens only. **One-time signup credits do not recur** and count as 0. Discontinued tiers are also 0.
 - Daily token cap → `monthly = daily × 30`. Only RPD documented → `RPD × ~800 output tokens × 30`. Only RPM/TPM (no daily cap) → **uncapped** (see below).
 - **Permanently free, but no published token cap** (`siliconflow`, `glm-cn`, `tencent`, `baidu`, `kilo-gateway`, `opencode-zen`): these are real recurring free access, rate/concurrency-limited. We classify them `recurring-uncapped` and **never sum them** — multiplying `RPM × 24/7 × 30d` would produce a fantasy ceiling (the inflation we reject). They are listed so you know they exist.

docs/routing/AUTO-COMBO.md

@@ -184,54 +184,64 @@ See [#7992](https://github.com/diegosouzapw/OmniRoute/issues/7992) and [#7111](h
 
 ## How It Works (Persisted Auto-Combos)
 
-The Auto-Combo Engine dynamically selects the best provider/model for each request using a **15-factor scoring function** (defined in `open-sse/services/autoCombo/scoring.ts` → `DEFAULT_WEIGHTS`). The default weights sum to `1.0`; custom weights are renormalized by `normalizeScoringWeights()`. Two of the fifteen — `cacheAffinity` and `resetWindowAffinity` — carry a default weight of `0`: they are still computed for every candidate, and `cacheAffinity` gates prompt-cache deduplication outside the score, so they are declared factors that simply do not vote by default.
+The Auto-Combo Engine dynamically selects the best provider/model for each request using a **16-factor scoring function** (defined in `open-sse/services/autoCombo/scoring.ts` → `DEFAULT_WEIGHTS`). The default weights sum to `1.0`; custom weights are renormalized by `normalizeScoringWeights()`. Three of the sixteen — `cacheAffinity`, `resetWindowAffinity` and `reliability` — carry a default weight of `0`: they are still computed for every candidate, and `cacheAffinity` gates prompt-cache deduplication outside the score, so they are declared factors that simply do not vote by default.
 
-![Auto-Combo 15-factor scoring](../diagrams/exported/auto-combo-scoring.svg)
+![Auto-Combo 16-factor scoring](../diagrams/exported/auto-combo-scoring.svg)
 
-> Source: [diagrams/auto-combo-scoring.mmd](../diagrams/auto-combo-scoring.mmd) (regenerate via `npm run docs:render-diagrams`). The filename is historical; the source and rendered diagram show all 15 factors declared in `DEFAULT_WEIGHTS`.
+> Source: [diagrams/auto-combo-scoring.mmd](../diagrams/auto-combo-scoring.mmd) (regenerate via `npm run docs:render-diagrams`). The filename is historical; the source and rendered diagram show all 16 factors declared in `DEFAULT_WEIGHTS`.
 
-| Factor                | Default Weight | Description                                                                                                                                                                                 |
-| :-------------------- | :------------- | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| `quota`               | 0.1429         | Remaining quota / rate-limit headroom [0..1]                                                                                                                                                |
-| `health`              | 0.1605         | Health score from circuit breaker (CLOSED=1.0, HALF_OPEN=0.5, OPEN=0.0)                                                                                                                     |
-| `costInv`             | 0.1429         | Inverse **blended** cost (60% input + 40% output token price, normalized) — cheaper = higher score                                                                                          |
-| `latencyInv`          | 0.1143         | Inverse p95 latency normalized to pool — faster = higher score                                                                                                                              |
-| `taskFit`             | 0.0762         | Task-type fitness (coding, review, planning, analysis, debugging, docs)                                                                                                                     |
-| `stability`           | 0.0476         | Variance-based stability (low latency stdDev / error rate)                                                                                                                                  |
-| `tierPriority`        | 0.0476         | Account-tier priority — Ultra=1.0, Pro=0.67, Standard=0.33, Free=0.0                                                                                                                        |
-| `tierAffinity`        | 0.0476         | Affinity between the candidate's tier and the manifest-recommended tier                                                                                                                     |
-| `specificityMatch`    | 0.0476         | Match between request specificity (manifest hint) and model tier                                                                                                                            |
-| `contextAffinity`     | 0.0476         | Affinity between the request's context-window need and the model's context window                                                                                                           |
-| `sessionAvailability` | 0.0476         | OAuth session availability of the candidate connection for this session (`getOAuthSessionAvailability()`; non-OAuth connections score 1.0)                                                  |
-| `connectionDensity`   | 0.0476         | Spreads load across connections of the same provider (anti-concentration)                                                                                                                   |
-| `cacheAffinity`       | 0.00           | Rendezvous-hash affinity toward the connection likeliest to already hold this request's prompt-cache prefix (`open-sse/services/combo/promptCacheAffinity.ts`); disabled by default (#8008) |
-| `resetWindowAffinity` | 0.00           | Bias toward connections whose quota reset window is favorable (disabled by default)                                                                                                         |
-| `quality`             | 0.03           | Feedback-driven output-quality signal from the routing-event quality tracker; candidates without observations receive a neutral 0.5                                                         |
+| Factor                | Default Weight | Description                                                                                                                                                                                    |
+| :-------------------- | :------------- | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `quota`               | 0.1429         | Remaining quota / rate-limit headroom [0..1]                                                                                                                                                   |
+| `health`              | 0.1605         | Health score from circuit breaker (CLOSED=1.0, HALF_OPEN=0.5, OPEN=0.0)                                                                                                                        |
+| `costInv`             | 0.1429         | Inverse **blended** cost (60% input + 40% output token price, normalized) — cheaper = higher score                                                                                             |
+| `latencyInv`          | 0.1143         | Inverse p95 latency normalized to pool — faster = higher score                                                                                                                                 |
+| `taskFit`             | 0.0762         | Task-type fitness (coding, review, planning, analysis, debugging, docs)                                                                                                                        |
+| `stability`           | 0.0476         | Variance-based stability from latency standard deviation — a candidate whose response time swings scores lower                                                                                 |
+| `tierPriority`        | 0.0476         | Account-tier priority — Ultra=1.0, Pro=0.67, Standard=0.33, Free=0.0                                                                                                                           |
+| `tierAffinity`        | 0.0476         | Affinity between the candidate's tier and the manifest-recommended tier                                                                                                                        |
+| `specificityMatch`    | 0.0476         | Match between request specificity (manifest hint) and model tier                                                                                                                               |
+| `contextAffinity`     | 0.0476         | Affinity between the request's context-window need and the model's context window                                                                                                              |
+| `sessionAvailability` | 0.0476         | OAuth session availability of the candidate connection for this session (`getOAuthSessionAvailability()`; non-OAuth connections score 1.0)                                                     |
+| `connectionDensity`   | 0.0476         | Spreads load across connections of the same provider (anti-concentration)                                                                                                                      |
+| `cacheAffinity`       | 0.00           | Rendezvous-hash affinity toward the connection likeliest to already hold this request's prompt-cache prefix (`open-sse/services/combo/promptCacheAffinity.ts`); disabled by default (#8008)    |
+| `resetWindowAffinity` | 0.00           | Bias toward connections whose quota reset window is favorable (disabled by default)                                                                                                            |
+| `quality`             | 0.03           | Feedback-driven output-quality signal from the routing-event quality tracker; candidates without observations receive a neutral 0.5                                                            |
+| `reliability`         | 0.00           | Observed success share, `1 - failureRate`, from 24h of usage history behind a ten-sample floor (real-time metrics otherwise); candidates with no observations read as 1.0. Disabled by default |
 
-**Sum:** `0.1429 + 0.1605 + 0.1429 + 0.1143 + 0.0762 + (7 × 0.0476) + 0.00 + 0.00 + 0.03 = 1.0` as declared in `DEFAULT_WEIGHTS`; user-configured weights are renormalized into a distribution by `normalizeScoringWeights()` before scoring.
+**Sum:** `0.1429 + 0.1605 + 0.1429 + 0.1143 + 0.0762 + (7 × 0.0476) + 0.00 + 0.00 + 0.03 + 0.00 = 1.0` as declared in `DEFAULT_WEIGHTS`; user-configured weights are renormalized into a distribution by `normalizeScoringWeights()` before scoring.
 
 ## Mode Packs
 
-Six pre-defined weight profiles in `open-sse/services/autoCombo/modePacks.ts` — `ship-fast`, `cost-saver`, `quality-first`, `offline-friendly`, `reliability-first` and `chaos-mode` (fault-injection). Each pack overrides the default weights to bias selection toward a specific goal; the seed weights below are renormalized to sum 1.0 at runtime together with the session/context factors every pack also sets. The table shows the four original packs — see `modePacks.ts` for `reliability-first` and `chaos-mode`.
+6 pre-defined weight profiles in `open-sse/services/autoCombo/modePacks.ts`. Each pack replaces the default weights outright to bias selection toward one goal. Every pack already sums to `1.0` (`0.9999` as printed at four decimals), so `normalizeScoringWeights()` has nothing meaningful to correct when a pack is active — the values below are, to rounding, the ones the scorer applies.
 
-| Factor       | ship-fast | cost-saver | quality-first | offline-friendly |
-| :----------- | :-------- | :--------- | :------------ | :--------------- |
-| quota        | 0.14      | 0.14       | 0.10          | **0.37**         |
-| health       | 0.28      | 0.19       | 0.18          | 0.28             |
-| costInv      | 0.05      | **0.37**   | 0.05          | 0.10             |
-| latencyInv   | **0.32**  | 0.05       | 0.05          | 0.05             |
-| taskFit      | 0.10      | 0.10       | **0.37**      | 0.00             |
-| stability    | 0.00      | 0.05       | 0.15          | 0.10             |
-| tierPriority | 0.05      | 0.05       | 0.05          | 0.05             |
+| Factor                | ship-fast  | cost-saver | quality-first | offline-friendly | reliability-first | chaos-mode |
+| :-------------------- | :--------- | :--------- | :------------ | :--------------- | :---------------- | :--------- |
+| `quota`               | 0.1333     | 0.1333     | 0.0952        | **0.3524**       | 0.1333            | 0.0476     |
+| `health`              | 0.2667     | 0.1810     | 0.1714        | 0.2667           | **0.3524**        | **0.4000** |
+| `costInv`             | 0.0476     | **0.3524** | 0.0476        | 0.0952           | 0.0381            | 0.0190     |
+| `latencyInv`          | **0.3048** | 0.0476     | 0.0476        | 0.0476           | 0.0476            | 0.0286     |
+| `taskFit`             | 0.0952     | 0.0952     | **0.3524**    | 0.0000           | 0.0952            | 0.1905     |
+| `stability`           | 0.0000     | 0.0476     | 0.1429        | 0.0952           | 0.1905            | 0.1714     |
+| `tierPriority`        | 0.0476     | 0.0476     | 0.0476        | 0.0476           | 0.0476            | 0.0190     |
+| `tierAffinity`        | 0.0000     | 0.0000     | 0.0000        | 0.0000           | 0.0000            | 0.0000     |
+| `specificityMatch`    | 0.0000     | 0.0000     | 0.0000        | 0.0000           | 0.0000            | 0.0000     |
+| `contextAffinity`     | 0.0095     | 0.0000     | 0.0000        | 0.0000           | 0.0000            | 0.0286     |
+| `sessionAvailability` | 0.0476     | 0.0476     | 0.0476        | 0.0476           | 0.0476            | 0.0476     |
+| `resetWindowAffinity` | 0.0000     | 0.0000     | 0.0000        | 0.0000           | 0.0000            | 0.0000     |
+| `connectionDensity`   | 0.0476     | 0.0476     | 0.0476        | 0.0476           | 0.0476            | 0.0476     |
 
 Notes:
 
-- `tierAffinity` and `specificityMatch` are explicitly set to `0` in every mode pack.
+- **No pack sets `quality`, and a pack replaces the weight map wholesale** (`weights = pack`, not a merge). `quality` carries `0.03` in `DEFAULT_WEIGHTS`, but under any mode pack it normalizes to `0` — selecting a pack silences the observed-quality signal completely. If you want quality feedback to influence routing, leave `modePack` unset and tune the weights directly. (`cacheAffinity` is also unset by every pack, but it defaults to `0` anyway, so nothing changes there.)
+- `tierAffinity`, `specificityMatch` and `resetWindowAffinity` are explicitly `0` in every pack.
 - Each pack's emphasis at a glance:
-  - **ship-fast** → latencyInv 0.32 + health 0.28 (low-latency, healthy connections)
-  - **cost-saver** → costInv 0.37 (cheapest tokens win)
-  - **quality-first** → taskFit 0.37 + stability 0.15 (best model for the task, consistent)
-  - **offline-friendly** → quota 0.37 + health 0.28 (max headroom regardless of speed/cost)
+  - **ship-fast** → latencyInv 0.3048 + health 0.2667 (low-latency, healthy connections)
+  - **cost-saver** → costInv 0.3524 (cheapest tokens win)
+  - **quality-first** → taskFit 0.3524 + stability 0.1429 (best model for the task, consistent)
+  - **offline-friendly** → quota 0.3524 + health 0.2667 (max headroom regardless of speed/cost)
+  - **reliability-first** → health 0.3524 + stability 0.1905 (fewest surprises)
+  - **chaos-mode** → health 0.4000 + taskFit 0.1905 (fault-injection profile)
 
 ### Per-Request Controls (headers) — #6023 / #6024 / #6025 / #3470
 
@@ -280,7 +290,7 @@ OmniRoute's combo engine supports **19 routing strategies** (declared in `src/sh
 | `reset-window`      | Prefer targets whose quota window resets soonest                                                                                                                                          |
 | `headroom`          | Pick the target with the most remaining quota headroom                                                                                                                                    |
 | `strict-random`     | Random without deduplication of repeats                                                                                                                                                   |
-| `auto`              | Use Auto Combo scoring (15-factor) — **recommended**                                                                                                                                      |
+| `auto`              | Use Auto Combo scoring (16-factor) — **recommended**                                                                                                                                      |
 | `lkgp`              | Last-Known-Good Path (pins to the last successful provider, then falls back to rules)                                                                                                     |
 | `context-optimized` | Pick target with best fit for current context size                                                                                                                                        |
 | `cache-optimized`   | Reorder targets by prompt-cache affinity — the connection likeliest to already hold this request's cached prefix is tried first (`open-sse/services/combo/promptCacheAffinity.ts`, #8008) |
@@ -389,7 +399,7 @@ The Auto Combo engine doesn't require pre-defined combos. Instead, `open-sse/ser
 3. Cross-references with `getProviderRegistry()` for model availability + pricing
 4. For each tuple `(provider, model, connection)`, builds a `VirtualAutoComboCandidate`
 5. Picks `connection.defaultModel` (or the registry's first model) as the dispatch target
-6. Scores each candidate using the 15-factor `scorePool()` and the variant's weight pack
+6. Scores each candidate using the 16-factor `scorePool()` and the variant's weight pack
 7. Returns the resulting in-memory `AutoComboConfig` for `handleComboChat()` — never persisted to DB
 
 This means **adding a new provider with `auto/*` enabled automatically expands the candidate pool** — no manual combo editing needed. The virtual combo is rebuilt per request, so newly-added or newly-healthy connections are picked up immediately.
@@ -450,7 +460,7 @@ Each strategy picks one provider from the candidate pool, given a `RoutingContex
 (task type, tool/vision hints, token estimate, optional SLA policy, optional
 last-known-good provider).
 
-#### 1. `rules` (default) — 15-factor weighted scoring
+#### 1. `rules` (default) — 16-factor weighted scoring
 
 Wraps the existing scoring engine. Filters out `OPEN` circuit-breaker
 candidates, then runs `scorePool()` with the current task type and `getTaskFitness()`,
@@ -459,7 +469,7 @@ picking the top-scoring provider.
 ```ts
 class RulesStrategyImpl implements RouterStrategy {
   readonly name = "rules";
-  readonly description = "15-factor weighted scoring (see DEFAULT_WEIGHTS)";
+  readonly description = "16-factor weighted scoring (see DEFAULT_WEIGHTS)";
 
   select(pool, context) {
     const eligible = pool.filter((c) => c.circuitBreakerState !== "OPEN");
@@ -700,7 +710,7 @@ Including the bare `auto` (default) plus the 6 `AutoVariant` values declared in
 
 ## How tiers fit Auto-Combo
 
-The 15-factor scoring function (`open-sse/services/autoCombo/scoring.ts`) treats tier
+The 16-factor scoring function (`open-sse/services/autoCombo/scoring.ts`) treats tier
 membership as two signals: `tierPriority` (0.0476) and `tierAffinity` (0.0476). See the
 canonical [scoring factor table](#how-it-works-persisted-auto-combos) above for the full
 `DEFAULT_WEIGHTS` set — the per-pack overrides (ship-fast/cost-saver/quality-first/
@@ -753,15 +763,15 @@ intentionally excluded from CI because they require live credentials and VPS acc
 
 ## Files
 
-| File                                                      | Purpose                                                                    |
-| :-------------------------------------------------------- | :------------------------------------------------------------------------- |
-| `open-sse/services/autoCombo/scoring.ts`                  | 15-factor scoring function, `DEFAULT_WEIGHTS`, pool norm                   |
-| `open-sse/services/autoCombo/taskFitness.ts`              | Model × task fitness lookup                                                |
-| `open-sse/services/autoCombo/engine.ts`                   | Selection logic, bandit, budget cap                                        |
-| `open-sse/services/autoCombo/selfHealing.ts`              | Exclusion, probes, incident mode                                           |
-| `open-sse/services/autoCombo/modePacks.ts`                | 4 weight profiles (ship-fast, cost-saver, quality-first, offline-friendly) |
-| `open-sse/services/autoCombo/autoPrefix.ts`               | `auto/` prefix parser + 6 variants                                         |
-| `open-sse/services/autoCombo/virtualFactory.ts`           | Builds in-memory `AutoComboConfig` from live connections                   |
-| `open-sse/services/autoCombo/providerRegistryAccessor.ts` | Test hook for mocking provider registry                                    |
-| `src/shared/constants/routingStrategies.ts`               | `ROUTING_STRATEGY_VALUES` (19 strategies)                                  |
-| `src/sse/handlers/chat.ts`                                | Integration: auto-prefix short-circuit                                     |
+| File                                                      | Purpose                                                                                                   |
+| :-------------------------------------------------------- | :-------------------------------------------------------------------------------------------------------- |
+| `open-sse/services/autoCombo/scoring.ts`                  | 16-factor scoring function, `DEFAULT_WEIGHTS`, pool norm                                                  |
+| `open-sse/services/autoCombo/taskFitness.ts`              | Model × task fitness lookup                                                                               |
+| `open-sse/services/autoCombo/engine.ts`                   | Selection logic, bandit, budget cap                                                                       |
+| `open-sse/services/autoCombo/selfHealing.ts`              | Exclusion, probes, incident mode                                                                          |
+| `open-sse/services/autoCombo/modePacks.ts`                | 6 weight profiles (ship-fast, cost-saver, quality-first, offline-friendly, reliability-first, chaos-mode) |
+| `open-sse/services/autoCombo/autoPrefix.ts`               | `auto/` prefix parser + 6 variants                                                                        |
+| `open-sse/services/autoCombo/virtualFactory.ts`           | Builds in-memory `AutoComboConfig` from live connections                                                  |
+| `open-sse/services/autoCombo/providerRegistryAccessor.ts` | Test hook for mocking provider registry                                                                   |
+| `src/shared/constants/routingStrategies.ts`               | `ROUTING_STRATEGY_VALUES` (19 strategies)                                                                 |
+| `src/sse/handlers/chat.ts`                                | Integration: auto-prefix short-circuit                                                                    |

docs/routing/STRICT_ZERO_COST.md

@@ -116,13 +116,37 @@ no waiting out the TTL after a 402/403/quota-exhausted response.
 `freeAccessPolicy`: a candidate can be economically `SAFE` and still excluded here for
 contractual reasons, or left in when this guard is off even with `freeAccessPolicy: "strict"` on.
 
-## What passes today
+## Seeing what the guard excludes
 
-Run `npx tsx scripts/ad-hoc/dry-run-strict-zero-cost.ts` against a live instance's
-`GET /v1/auto-combo/{channel}/candidates` output for a real before/after — the script reads each
-candidate's real `connectionId`, so it also proves the connection-safety fix live, not just in
-unit tests. Keyless candidates must arrive with the synthetic no-auth `connectionId`, never a
-real connection. The current built-in keyless auto path is OpenCode Free; exact candidate counts
+`GET /v1/auto-combo/{channel}/candidates` lists every candidate, including the ones this guard
+would keep out of dispatch, and each carries `freeAccessExclusion` — `null` when the guard is
+satisfied, otherwise the reason. The listing reports; it never enforces. Turning the policy off
+leaves the field `null` everywhere and costs nothing.
+
+| `freeAccessExclusion`  | What it means                                                                                                 | What to do about it                                                                                                                 |
+| :--------------------- | :------------------------------------------------------------------------------------------------------------ | :---------------------------------------------------------------------------------------------------------------------------------- |
+| `not-in-catalog`       | The provider/model pair is absent from `FREE_MODEL_BUDGETS`.                                                  | Add a curated entry, or accept that new pairs start excluded — that is the design.                                                  |
+| `regime-not-free`      | Catalogued, but its `freeType` is not one that grants free access (a discontinued tier, for instance).        | Nothing to fix. The model costs money.                                                                                              |
+| `no-hard-stop`         | Free regime, but `hardStopGuaranteed` is not `true`, so exceeding the allowance might silently start billing. | Verify the provider's terms and set the flag with the source in a comment — never to grow the catalog.                              |
+| `contradictory-noauth` | A no-auth candidate whose catalog entry is not `keyless`. Fail-closed on inconsistent metadata.               | Fix the catalog entry; the two facts disagree.                                                                                      |
+| `exhausted`            | A fresh reading says the allowance is used up.                                                                | Wait for the reset. This one resolves itself.                                                                                       |
+| `state-unknown`        | No quota reading, or one too old to trust.                                                                    | Go look: the provider may have no usage adapter registered, or the quota fetch is failing.                                          |
+| `no-connection`        | The candidate carries no account to check at all.                                                             | Not a quota problem: the candidate was built without a connection, so nothing was ever looked up. Check how the pool was assembled. |
+
+The last two are the pair worth separating. An exhausted allowance resets on its own; a reading
+that never arrives means the lookup itself is broken, and until now both looked identical from
+outside — the candidate simply vanished.
+
+**One gap remains, and it is deliberate.** `excludeTosAvoid` still removes candidates before the
+listing is built, so a model curated `tos: "avoid"` is absent with no reason given — the same
+invisibility this section just closed for the zero-cost guard. Closing it too means deciding what
+a ToS exclusion should report, which is a separate question from economic safety; this page names
+the gap rather than pretending it is not there.
+
+For an offline before/after, `npx tsx scripts/ad-hoc/dry-run-strict-zero-cost.ts` still works
+against a live instance's candidates output; it reads each candidate's real `connectionId`, so it
+also exercises the connection-safety path. Keyless candidates must arrive with the synthetic
+no-auth `connectionId`, never a real connection. The current built-in keyless auto path is OpenCode Free; exact candidate counts
 still depend on live model discovery and should be measured on the target deployment instead of
 copied from an older run. A `recurring-*` candidate passes only when it has both a registered
 usage adapter and `hardStopGuaranteed: true`; incomplete metadata remains fail-closed.

llm.txt

@@ -223,7 +223,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   │   ├── accountFallback.ts    # Multi-account fallback
 │   │   ├── sessionManager.ts     # Session management
 │   │   ├── wildcardRouter.ts     # Wildcard model routing
-│   │   ├── autoCombo/            # Auto-combo engine (15-factor scoring, bandit exploration)
+│   │   ├── autoCombo/            # Auto-combo engine (16-factor scoring, bandit exploration)
 │   │   ├── intentClassifier.ts   # Request intent classification
 │   │   ├── taskAwareRouter.ts    # Task-aware routing
 │   │   ├── thinkingBudget.ts     # Thinking budget management
@@ -263,7 +263,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 │   ├── architecture/             # ARCHITECTURE.md, CODEBASE_DOCUMENTATION.md, REPOSITORY_MAP.md, AUTHZ_GUIDE.md, RESILIENCE_GUIDE.md, QUALITY_GATES.md
 │   ├── reference/                # API_REFERENCE.md, PROVIDER_REFERENCE.md, CLI-TOOLS.md
 │   ├── frameworks/               # MCP-SERVER.md (110 tools), A2A-SERVER.md, SKILLS.md, MEMORY.md, CLOUD_AGENT.md, EVALS.md, WEBHOOKS.md
-│   ├── routing/                  # AUTO-COMBO.md (15-factor scoring), REASONING_REPLAY.md
+│   ├── routing/                  # AUTO-COMBO.md (16-factor scoring), REASONING_REPLAY.md
 │   ├── security/                 # GUARDRAILS.md, COMPLIANCE.md, STEALTH_GUIDE.md, PUBLIC_CREDS.md, ERROR_SANITIZATION.md
 │   ├── guides/                   # USER_GUIDE.md, TROUBLESHOOTING.md, ELECTRON_GUIDE.md, I18N.md
 │   ├── ops/                      # RELEASE_CHECKLIST.md, TUNNELS_GUIDE.md, VM deployment
@@ -282,7 +282,7 @@ OmniRoute solves the problem of managing multiple AI provider subscriptions, quo
 - **19 routing strategies**: priority, weighted, round-robin, fill-first, p2c, random, least-used, cost-optimized, reset-aware, reset-window, headroom, strict-random, auto, lkgp, context-optimized, cache-optimized, context-relay, fusion, pipeline
 - **4-tier fallback**: Subscription → API Key → Cheap → Free
 - **Context Relay strategy**: Session handoff summaries on account rotation for continuity
-- **Auto-combo engine**: Self-healing routing optimization with **15-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
+- **Auto-combo engine**: Self-healing routing optimization with **16-factor scoring** (see `docs/routing/AUTO-COMBO.md`), bandit exploration, progressive cooldown
 - **Semantic caching** with cache hit/miss headers
 - **Idempotency** with configurable dedup window
 - **3-layer resilience**: Provider Circuit Breaker / Connection Cooldown / Model Lockout
@@ -441,7 +441,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 8. **ACP agents** are in `src/lib/acp/registry.ts` with detection cache. Custom agents stored via settings DB.
 
-9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **15-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
+9. **Auto-combo engine** in `open-sse/services/autoCombo/` — **16-factor scoring** (weights and factors in `docs/routing/AUTO-COMBO.md`), 6 mode packs, bandit exploration, progressive cooldown.
 
 10. **Docker:** Dockerfile has two targets: `runner-base` and `runner-cli`. `docker-compose.yml` for dev (8 profiles), `docker-compose.prod.yml` for production (port 20130).
 
@@ -476,7 +476,7 @@ diagnostics) plus **memory**, **skill**, **agentSkill**, **githubSkill**, **pool
 
 - **352-provider catalog** with 150+ free tiers, one-click account imports, and bulk key add
 - **19 routing strategies** — including `fusion` (parallel panel + judge synthesis), `pipeline`, `reset-aware`, `reset-window`, `headroom`, and `context-relay`
-- **15-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
+- **16-factor Auto-Combo scoring** with bandit exploration and progressive cooldown
 - **MCP server expanded to 110 tools / 33 scopes** (canonical + memory/skill/agentSkill/githubSkill/pool/notion/obsidian/localCorpus/gamification/plugin modules)
 - **Cloud Agents** (Codex Cloud, Cursor, Devin, Jules), **Guardrails**, **Evals**, **Webhooks**, **Compliance** frameworks
 - **Embedded services** manager (install/start/stop bundled services from the dashboard)

open-sse/handlers/autoComboCandidates.ts

@@ -30,6 +30,7 @@ import { buildErrorBody } from "@omniroute/open-sse/utils/error.ts";
 import { getCircuitBreaker } from "@/shared/utils/circuitBreaker";
 import { isModelLocked } from "@omniroute/open-sse/services/accountFallback.ts";
 import { parseModel } from "@omniroute/open-sse/services/model.ts";
+import type { StrictZeroCostExclusionReason } from "@omniroute/open-sse/services/autoCombo/strictZeroCostFilter.ts";
 import { getProviderConnectionById } from "@/lib/db/providers";
 import { getExcludedConnectionIds } from "@/lib/db/autoCandidateOverrides";
 
@@ -51,6 +52,13 @@ export interface AutoComboCandidateView {
   breakerState: string;
   connectionCooldown: boolean;
   modelLocked: boolean;
+  /**
+   * Why STRICT_ZERO_COST would exclude this candidate from dispatch, or null
+   * when it would not — and null as well when the policy is off, which is the
+   * default. Reported, never enforced: this listing shows the candidate either
+   * way, the routing path is what acts on it.
+   */
+  freeAccessExclusion: StrictZeroCostExclusionReason | null;
 }
 
 export interface AutoComboCandidatesResult {
@@ -69,6 +77,7 @@ async function decorateCandidate(candidate: {
   connectionId: string;
   model: string;
   modelStr: string;
+  freeAccessExclusion?: StrictZeroCostExclusionReason | null;
 }): Promise<AutoComboCandidateView> {
   const breaker = getCircuitBreaker(candidate.provider);
   const breakerStatus = breaker.getStatus();
@@ -111,6 +120,7 @@ async function decorateCandidate(candidate: {
     breakerState: String(breakerStatus.state),
     connectionCooldown,
     modelLocked,
+    freeAccessExclusion: candidate.freeAccessExclusion ?? null,
   };
 }
 
@@ -160,6 +170,7 @@ export async function getAutoComboCandidates(
     connectionId: string | null;
     allowedConnectionIds?: string[];
     model: string;
+    freeAccessExclusion?: StrictZeroCostExclusionReason | null;
   }> = Array.isArray(virtualCombo?.models) ? virtualCombo.models : [];
   // Routing keeps one logical provider/model candidate, but the management API
   // remains account-oriented so operators can inspect and toggle each fallback.
@@ -178,6 +189,7 @@ export async function getAutoComboCandidates(
         connectionId: candidate.connectionId,
         model: candidate.model,
         modelStr: candidate.model,
+        freeAccessExclusion: candidate.freeAccessExclusion,
       });
       return { ...decorated, excluded: excludedConnectionIds.has(candidate.connectionId) };
     })

open-sse/handlers/chatCore/clientUsageBuffer.ts

@@ -22,6 +22,7 @@ import {
   addBufferToUsage as defaultAddBuffer,
   filterUsageForFormat as defaultFilterUsage,
   estimateUsage as defaultEstimateUsage,
+  isEmptyUsage,
   sanitizeProviderUsageForRequest,
   type UsageLike,
 } from "../../utils/usageTracking.ts";
@@ -46,35 +47,6 @@ const DEFAULT_DEPS: ClientUsageBufferDeps = {
   estimateUsage: defaultEstimateUsage,
 };
 
-/** True when a usage object is present but every token field is zero/absent.
- * Web/unofficial providers often emit `{prompt_tokens:0,completion_tokens:0,total_tokens:0}`
- * because the upstream has no metering. Treating that as "has usage" makes
- * `addBufferToUsage` turn zeros into a constant `USAGE_TOKEN_BUFFER` (default 2000),
- * so every request shows exactly 2000 tokens. Prefer estimating instead. */
-function isEmptyUsage(usage: unknown): boolean {
-  if (!usage || typeof usage !== "object" || Array.isArray(usage)) return true;
-  const u = usage as Record<string, unknown>;
-  const fields = [
-    "prompt_tokens",
-    "completion_tokens",
-    "total_tokens",
-    "input_tokens",
-    "output_tokens",
-    "promptTokenCount",
-    "candidatesTokenCount",
-    "totalTokenCount",
-  ];
-  let sawNumber = false;
-  for (const key of fields) {
-    const v = u[key];
-    if (typeof v !== "number" || !Number.isFinite(v)) continue;
-    sawNumber = true;
-    if (v > 0) return false;
-  }
-  // No positive counts (or no numeric fields at all) → treat as empty.
-  return true;
-}
-
 /** context_budget_* → visible-field mapping folded back in for Claude-Code-compatible
  * responses only (see module docstring above). */
 const CONTEXT_BUDGET_TO_VISIBLE_FIELD: Record<string, string> = {

open-sse/services/autoCombo/routerStrategy.ts

@@ -3,7 +3,7 @@
  *
  * Inspired by ClawRouter commit 14c83c258 "refactor: extract routing into pluggable RouterStrategy system".
  * Provides a RouterStrategy interface and built-in implementations:
- *   - RulesStrategy (default): wraps the existing 15-factor scoring engine
+ *   - RulesStrategy (default): wraps the existing 16-factor scoring engine
  *   - ScoreStrategy: highest configured weighted score, with explicit exploration
  *   - CostStrategy: always picks cheapest available model
  *   - LatencyStrategy: prioritizes low p95 latency with reliability weighting
@@ -53,7 +53,7 @@ export interface RouterStrategy {
   select(pool: ProviderCandidate[], context: RoutingContext): RoutingDecision;
 }
 
-// ── RulesStrategy: wraps 15-factor scoring engine ───────────────────────────
+// ── RulesStrategy: wraps 16-factor scoring engine ───────────────────────────
 
 function toSpeedCandidate(c: ProviderCandidate): SpeedCandidate {
   return {
@@ -87,7 +87,7 @@ function toSpeedCandidate(c: ProviderCandidate): SpeedCandidate {
 
 class RulesStrategyImpl implements RouterStrategy {
   readonly name = "rules";
-  readonly description = "15-factor weighted scoring (see DEFAULT_WEIGHTS)";
+  readonly description = "16-factor weighted scoring (see DEFAULT_WEIGHTS)";
 
   select(pool: ProviderCandidate[], context: RoutingContext): RoutingDecision {
     const eligible = pool.filter((c) => c.circuitBreakerState !== "OPEN");

open-sse/services/autoCombo/scoring.ts

@@ -29,6 +29,13 @@ export interface ScoringFactors {
    * observed events default to neutral (0.5) and are never penalized.
    */
   quality?: number;
+  /**
+   * Observed success share over the routing window: 1 - failure rate. Optional
+   * so a candidate nobody has called yet reads as 1 rather than 0 -- it has not
+   * failed anything. That differs from `quality` on purpose: a score with no
+   * observations is neutral at 0.5, a failure rate with no observations is 0.
+   */
+  reliability?: number;
 }
 
 export interface ScoringWeights {
@@ -48,6 +55,8 @@ export interface ScoringWeights {
   connectionDensity: number;
   /** Weight for the feedback-driven quality factor (#feedback-foundation). */
   quality?: number;
+  /** Weight for the observed failure-rate factor. 0 by default. */
+  reliability?: number;
 }
 
 export const DEFAULT_WEIGHTS: ScoringWeights = {
@@ -69,6 +78,12 @@ export const DEFAULT_WEIGHTS: ScoringWeights = {
   // the new quality signal (observed output quality over time) gets a real,
   // if smaller, vote. Sum remains exactly 1.0.
   quality: 0.03,
+  // Declared but silent, like `cacheAffinity` and `resetWindowAffinity`: every
+  // candidate already carries a measured failure rate (24h of usage history
+  // behind a ten-sample floor, real-time metrics otherwise) and the scorer had
+  // no way to read it. Which weight it deserves is a product call backed by
+  // measurement, so this ships at 0 and leaves the ranking exactly as it was.
+  reliability: 0,
 };
 
 /** Normalize independently configured UI weights into a scoring distribution. */
@@ -162,7 +177,10 @@ export function calculateScore(factors: ScoringFactors, weights: ScoringWeights)
       (weights.connectionDensity ?? 0) * factors.connectionDensity +
       // Missing quality factor → neutral 0.5: a cold candidate is neither boosted
       // (which would let optimistic initialization dominate) nor penalized.
-      (weights.quality ?? 0) * (factors.quality ?? 0.5)
+      (weights.quality ?? 0) * (factors.quality ?? 0.5) +
+      // Missing reliability factor -> neutral 1, not 0.5: a candidate with no
+      // observations has not failed anything. See the field doc on ScoringFactors.
+      (weights.reliability ?? 0) * (factors.reliability ?? 1)
   );
 }
 
@@ -255,6 +273,17 @@ export function computePoolMaxima(pool: ProviderCandidate[]): PoolMaxima {
   return { maxCost, maxLatency, maxStdDev };
 }
 
+/**
+ * Bound an observed failure rate to [0,1], treating anything missing or
+ * non-finite as 0 (nothing observed has failed). Mirrors `toBoundedRate` in
+ * `speedRanking.ts` so both consumers of the same signal agree, including on
+ * garbage input.
+ */
+function boundedRate(value: number | null | undefined): number {
+  if (typeof value !== "number" || !Number.isFinite(value) || value < 0) return 0;
+  return Math.min(1, value);
+}
+
 export function calculateFactors(
   candidate: ProviderCandidate,
   pool: ProviderCandidate[],
@@ -292,6 +321,13 @@ export function calculateFactors(
     // Feedback quality signal; neutral 0.5 when the tracker has no data yet
     // (cold providers are neither boosted nor unfairly penalized).
     quality: clamp01(candidate.quality ?? 0.5),
+    // Same formula and same precedence as `speedRanking.ts` uses for its own
+    // reliability factor: an explicit failure rate wins over the coarser error
+    // rate, and an unobserved candidate reads as fully reliable. The rate is
+    // bounded BEFORE the subtraction, exactly as `toBoundedRate` does there --
+    // `clamp01(1 - NaN)` would be 0, i.e. "fails every call", which is the
+    // opposite of what corrupt telemetry should mean.
+    reliability: clamp01(1 - boundedRate(candidate.failureRate ?? candidate.errorRate)),
   };
 }
 

open-sse/services/autoCombo/strictZeroCostFilter.ts

@@ -129,25 +129,52 @@ export function findBudgetEntry(
   return catalog.find((m) => m.provider === candidate.provider && m.modelId === candidate.model);
 }
 
-function isConnectionStateSafe(
+/** Why the guard cannot trust a candidate right now. */
+export type StrictZeroCostExclusionReason =
+  | "not-in-catalog"
+  | "regime-not-free"
+  | "no-hard-stop"
+  | "contradictory-noauth"
+  | "exhausted"
+  | "state-unknown"
+  | "no-connection";
+
+export type StrictZeroCostVerdict =
+  { outcome: "safe"; safeConnectionIds: string[] } | { outcome: StrictZeroCostExclusionReason };
+
+/**
+ * Why one connection cannot be trusted right now. Splitting "exhausted" from
+ * "state-unknown" is the whole point: an exhausted allowance resets on its own
+ * and the operator waits, while a missing or stale reading means the quota
+ * lookup itself is not working and the operator has to go fix something.
+ * Opposite actions, and until now the same silence.
+ *
+ * Freshness is checked before status, so a stale EXHAUSTED reading reports
+ * "state-unknown" rather than asserting an exhaustion nobody has confirmed
+ * lately. The exclusion verdict is identical either way -- only the reason
+ * shown to the operator differs.
+ */
+export function classifyConnectionState(
   provider: string,
   connectionId: string,
   resolveFreeAccessState: StrictZeroCostOptions["resolveFreeAccessState"],
   options: Pick<StrictZeroCostOptions, "minRemainingAllowance" | "maxStateAgeMs" | "now">
-): boolean {
+): "safe" | "exhausted" | "state-unknown" {
   const state = resolveFreeAccessState(provider, connectionId);
-  if (!state) return false; // no usage adapter for this provider, or lookup never ran/is stale
-  if (state.status !== "SAFE") return false;
+  if (!state) return "state-unknown"; // no usage adapter for this provider, or lookup never ran
 
   const now = (options.now ?? Date.now)();
   const checkedAtMs = Date.parse(state.checkedAt);
-  if (!Number.isFinite(checkedAtMs) || now - checkedAtMs > options.maxStateAgeMs) return false;
+  if (!Number.isFinite(checkedAtMs) || now - checkedAtMs > options.maxStateAgeMs)
+    return "state-unknown";
 
-  if (state.remainingFreeAllowance === null) return false;
+  if (state.status === "EXHAUSTED") return "exhausted";
+  if (state.status !== "SAFE") return "state-unknown";
+  if (state.remainingFreeAllowance === null) return "state-unknown";
   // A negative threshold would let a negative/garbage reading pass; a caller
   // that genuinely wants "any allowance greater than zero" should pass 0.
-  if (options.minRemainingAllowance < 0) return false;
-  return state.remainingFreeAllowance > options.minRemainingAllowance;
+  if (options.minRemainingAllowance < 0) return "state-unknown";
+  return state.remainingFreeAllowance > options.minRemainingAllowance ? "safe" : "exhausted";
 }
 
 /**
@@ -168,7 +195,28 @@ export function evaluateCandidateConnections(
   resolveFreeAccessState: StrictZeroCostOptions["resolveFreeAccessState"],
   options: Pick<StrictZeroCostOptions, "minRemainingAllowance" | "maxStateAgeMs" | "now">
 ): string[] {
-  if (!budgetEntry) return []; // not in the catalog at all → paid, or genuinely unknown
+  const verdict = classifyStrictZeroCostCandidate(
+    candidate,
+    budgetEntry,
+    resolveFreeAccessState,
+    options
+  );
+  return verdict.outcome === "safe" ? verdict.safeConnectionIds : [];
+}
+
+/**
+ * Same decision as `evaluateCandidateConnections`, but it says why instead of
+ * answering with an empty list. The read-only candidate listing needs the why;
+ * the pool filter only needs the list, so it reads this one and throws the
+ * reason away.
+ */
+export function classifyStrictZeroCostCandidate(
+  candidate: StrictZeroCostCandidate,
+  budgetEntry: FreeModelBudget | undefined,
+  resolveFreeAccessState: StrictZeroCostOptions["resolveFreeAccessState"],
+  options: Pick<StrictZeroCostOptions, "minRemainingAllowance" | "maxStateAgeMs" | "now">
+): StrictZeroCostVerdict {
+  if (!budgetEntry) return { outcome: "not-in-catalog" }; // paid, or genuinely unknown
 
   const isGenuineNoAuthCandidate = candidate.connectionId === SYNTHETIC_NOAUTH_CONNECTION_ID;
   if (allowsNoAuthShortcut(budgetEntry.freeType)) {
@@ -180,30 +228,47 @@ export function evaluateCandidateConnections(
     // any other freeType, and is excluded there unless hardStopGuaranteed is
     // also set for it (which the curated catalog does not do for keyless
     // entries today, so it will correctly exclude).
-    if (isGenuineNoAuthCandidate) return [SYNTHETIC_NOAUTH_CONNECTION_ID];
+    if (isGenuineNoAuthCandidate)
+      return { outcome: "safe", safeConnectionIds: [SYNTHETIC_NOAUTH_CONNECTION_ID] };
   }
-  if (!grantsFreeAccess(budgetEntry.freeType)) return [];
-  if (isGenuineNoAuthCandidate) return []; // no-auth path but a non-keyless catalog entry: contradictory metadata, fail closed
+  if (!grantsFreeAccess(budgetEntry.freeType)) return { outcome: "regime-not-free" };
+  // no-auth path but a non-keyless catalog entry: contradictory metadata, fail closed
+  if (isGenuineNoAuthCandidate) return { outcome: "contradictory-noauth" };
 
   // Every remaining freeType (recurring-*, one-time-initial, a keyless entry
   // reached via a real connection, and any future type this module doesn't
   // special-case) requires a documented hard stop before any live check even
   // runs — no point burning a quota lookup on a connection we could never
   // trust regardless of its answer.
-  if (budgetEntry.hardStopGuaranteed !== true) return [];
+  if (budgetEntry.hardStopGuaranteed !== true) return { outcome: "no-hard-stop" };
 
   const candidateConnectionIds = candidate.connectionId
     ? [candidate.connectionId]
     : (candidate.allowedConnectionIds ?? []);
 
+  // No account at all to check. Reporting `state-unknown` here would send the
+  // operator hunting a quota lookup that was never attempted; this is a wiring
+  // problem, not a quota one.
+  if (candidateConnectionIds.length === 0) return { outcome: "no-connection" };
+
   const safe: string[] = [];
+  // An observed exhaustion outranks a missing reading: one is a fact, the other
+  // is the absence of one, and the operator needs the fact. Without this rule the
+  // reason would depend on the order the connections happen to be listed in.
+  let sawExhausted = false;
   for (const connectionId of candidateConnectionIds) {
     if (connectionId === SYNTHETIC_NOAUTH_CONNECTION_ID) continue; // never reachable here, defensive
-    if (isConnectionStateSafe(candidate.provider, connectionId, resolveFreeAccessState, options)) {
-      safe.push(connectionId);
-    }
+    const state = classifyConnectionState(
+      candidate.provider,
+      connectionId,
+      resolveFreeAccessState,
+      options
+    );
+    if (state === "safe") safe.push(connectionId);
+    else if (state === "exhausted") sawExhausted = true;
   }
-  return safe;
+  if (safe.length > 0) return { outcome: "safe", safeConnectionIds: safe };
+  return { outcome: sawExhausted ? "exhausted" : "state-unknown" };
 }
 
 /**

open-sse/services/autoCombo/virtualFactory.ts

@@ -37,7 +37,12 @@ import {
   orderPoolByRung,
   type LadderOptions,
 } from "./subscriptionLadder";
-import { filterStrictZeroCostCandidates, filterTosAvoidCandidates } from "./strictZeroCostFilter";
+import {
+  classifyStrictZeroCostCandidate,
+  filterStrictZeroCostCandidates,
+  filterTosAvoidCandidates,
+  findBudgetEntry,
+} from "./strictZeroCostFilter";
 import { resolveFreeAccessState } from "./freeAccessQuota";
 import { isModelExcludedByConnection } from "@/domain/connectionModelRules";
 import { resolveProviderAlias } from "../model.ts";
@@ -106,6 +111,13 @@ export interface VirtualAutoComboCandidate {
   resolvedSupportsVision?: boolean;
   resolvedReasoning?: boolean;
   resolvedSupportsThinking?: boolean;
+  /**
+   * Why STRICT_ZERO_COST would exclude this candidate, or null when it would
+   * not. Only populated for the read-only inspector build (`skip`), where the
+   * guard is deliberately not applied — dispatch builds leave it undefined and
+   * do no extra work.
+   */
+  freeAccessExclusion?: import("./strictZeroCostFilter").StrictZeroCostExclusionReason | null;
 }
 
 type VirtualAutoCombo = AutoComboConfig & {
@@ -119,6 +131,9 @@ type VirtualAutoCombo = AutoComboConfig & {
     allowedConnectionIds?: string[];
     weight: number;
     label: string;
+    /** Carried through from the candidate for the read-only inspector; absent
+     * on every dispatch build. */
+    freeAccessExclusion?: import("./strictZeroCostFilter").StrictZeroCostExclusionReason | null;
   }>;
   /** MAX of candidates' context windows — safe to advertise because the
    * auto-combo context pre-filter routes oversized requests to large-window
@@ -749,17 +764,42 @@ export async function prepareVirtualAutoComboInputs(
     // per-candidate: `resolveFreeAccessState` here is a raw pass-through of the real
     // per-(provider,connectionId) resolver; the filter itself decides which connection(s)
     // on each candidate to check and rewrites `allowedConnectionIds` to the SAFE subset.
-    const strictFilteredPool = filterStrictZeroCostCandidates(pool, {
-      enabled: settings.freeAccessPolicy === "strict",
-      resolveFreeAccessState,
+    const strictZeroCostThresholds = {
       // 1 percentage point of headroom, not 0: `freeAccessQuota.ts` reports
       // 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: toNumber(settings.autoRefreshProviderQuotaInterval, 180) * 1000,
+    };
+    const strictZeroCostOn = settings.freeAccessPolicy === "strict";
+    const strictFilteredPool = filterStrictZeroCostCandidates(pool, {
+      // The read-only candidate inspector (#9133) must be able to see what the
+      // guard would exclude, and why — the same opt-out the resilience filter
+      // already honours through `skip`. Dispatch (`skip === false`) is unaffected.
+      enabled: strictZeroCostOn && !skip,
+      resolveFreeAccessState,
+      ...strictZeroCostThresholds,
     });
     if (strictFilteredPool !== pool) pool = strictFilteredPool;
 
+    // Annotate here rather than in the handler: this is where the thresholds and
+    // `resolveFreeAccessState` already live. Doing it downstream would mean a second
+    // copy of both, with nothing to keep them in agreement.
+    if (strictZeroCostOn && skip) {
+      pool = pool.map((candidate) => {
+        const verdict = classifyStrictZeroCostCandidate(
+          candidate,
+          findBudgetEntry(candidate),
+          resolveFreeAccessState,
+          strictZeroCostThresholds
+        );
+        return {
+          ...candidate,
+          freeAccessExclusion: verdict.outcome === "safe" ? null : verdict.outcome,
+        };
+      });
+    }
+
     // Separate, optional ToS guard — independent of economic safety on purpose.
     const tosFilteredPool = filterTosAvoidCandidates(pool, settings.excludeTosAvoid === true);
     if (tosFilteredPool !== pool) pool = tosFilteredPool;
@@ -1058,6 +1098,9 @@ export async function createVirtualAutoComboFromPrepared(
       : {}),
     weight: snapshotScores.get(candidate.modelStr) ?? 1,
     label: candidate.provider,
+    ...(candidate.freeAccessExclusion === undefined
+      ? {}
+      : { freeAccessExclusion: candidate.freeAccessExclusion }),
   }));
   const autoConfig = {
     candidatePool: providerPool,

open-sse/services/combo/comboStructure.ts

@@ -529,7 +529,10 @@ function hasKnownCompatibleContextLimit(
   requirements: RequestCompatibilityRequirements
 ): boolean {
   if (requirements.requiredContextTokens <= 0) return false;
-  const capabilities = getResolvedModelCapabilities(target.modelStr);
+  const capabilities = getResolvedModelCapabilities({
+    provider: target.providerId || target.provider || null,
+    model: target.modelStr,
+  });
   return evaluateContextLimit(capabilities, requirements, target.modelStr) === true;
 }
 
@@ -546,7 +549,10 @@ export function isVisionIncompatibleTarget(
   requirements: RequestCompatibilityRequirements
 ): boolean {
   if (!requirements.requiresVision) return false;
-  const capabilities = getResolvedModelCapabilities(target.modelStr);
+  const capabilities = getResolvedModelCapabilities({
+    provider: target.providerId || target.provider || null,
+    model: target.modelStr,
+  });
   return capabilities.supportsVision !== true;
 }
 
@@ -571,7 +577,10 @@ function getTargetCompatibilityFailures(
   target: ResolvedComboTarget,
   requirements: RequestCompatibilityRequirements
 ): string[] {
-  const capabilities = getResolvedModelCapabilities(target.modelStr);
+  const capabilities = getResolvedModelCapabilities({
+    provider: target.providerId || target.provider || null,
+    model: target.modelStr,
+  });
   const failures: string[] = [];
 
   if (

open-sse/services/compression/compressionWorkerPool.ts

@@ -1,6 +1,5 @@
 import { existsSync } from "node:fs";
-import { dirname, join } from "node:path";
-import { fileURLToPath, pathToFileURL } from "node:url";
+import { dirname, join, resolve } from "node:path";
 import { Worker } from "node:worker_threads";
 import type { CompressionResult } from "./types.ts";
 import type { StackedCompressionStep } from "./strategySelector.ts";
@@ -14,14 +13,70 @@ function positiveInteger(value: string | undefined, fallback: number): number {
   const parsed = Number(value);
   return Number.isSafeInteger(parsed) && parsed > 0 ? parsed : fallback;
 }
-function workerUrl(): URL {
-  const dir = dirname(fileURLToPath(import.meta.url));
-  for (const name of ["compressionWorker.js", "compressionWorker.ts"]) {
-    const candidate = join(dir, name);
-    if (existsSync(candidate)) return pathToFileURL(candidate);
+
+/** Relative path (from an install root) to the compression worker. */
+const WORKER_JS_REL = join("open-sse", "services", "compression", "compressionWorker.js");
+const WORKER_TS_REL = join("open-sse", "services", "compression", "compressionWorker.ts");
+
+const MAX_WALK_UP = 8;
+
+/**
+ * Walk up from each anchor directory (≤ MAX_WALK_UP levels) and return the first
+ * ancestor that actually contains `relPath`, or null. Pure + exported for tests.
+ *
+ * This deliberately avoids `import.meta.url`/`__dirname` (both dead in the standalone
+ * bundle) — see the LLMLingua worker comments in llmlingua/worker.ts.
+ */
+export function firstAncestorWith(anchors: string[], relPath: string): string | null {
+  for (const anchor of anchors) {
+    if (!anchor) continue;
+    let dir = resolve(anchor);
+    for (let i = 0; i <= MAX_WALK_UP; i++) {
+      if (existsSync(join(dir, relPath))) return dir;
+      const parent = dirname(dir);
+      if (parent === dir) break;
+      dir = parent;
+    }
   }
-  return pathToFileURL(join(dir, "compressionWorker.js"));
+  return null;
 }
+
+/**
+ * Runtime install-root anchors that SURVIVE the standalone bundle:
+ *  - `process.cwd()` — `dist/server.js` runs `process.chdir(__dirname)` → the dist root.
+ *  - `dirname(process.argv[1])` — the entry script (server.js / bin), walked up.
+ */
+function runtimeAnchors(): string[] {
+  const anchors = [process.cwd()];
+  const argv1 = process.argv[1];
+  if (typeof argv1 === "string" && argv1) anchors.push(dirname(argv1));
+  return anchors;
+}
+
+/**
+ * Resolve the worker entry file across dev and prod WITHOUT `import.meta.url`.
+ *
+ * Prod: the worker is likely a .js file under the install root
+ * Dev: the same relative path resolves to the `.ts` source under the project
+ * root (cwd) and runs via the default Node.js loader.
+ *
+ * First existing candidate wins. Exported for tests.
+ */
+export function resolveWorkerFile(): string {
+  const anchors = runtimeAnchors();
+
+  // Prod first: the .js under the install root.
+  const jsRoot = firstAncestorWith(anchors, WORKER_JS_REL);
+  if (jsRoot) return join(jsRoot, WORKER_JS_REL);
+
+  // Dev: the .ts source.
+  const tsRoot = firstAncestorWith(anchors, WORKER_TS_REL);
+  if (tsRoot) return join(tsRoot, WORKER_TS_REL);
+
+  // Nothing found — return a cwd-relative .js path; the spawn will fail-open.
+  return join(process.cwd(), WORKER_JS_REL);
+}
+
 function unchanged(body: Record<string, unknown>): CompressionResult {
   return { body, compressed: false, stats: null };
 }
@@ -80,7 +135,7 @@ export class CompressionWorkerPool {
   }
   private spawn(): PoolWorker {
     const slot: PoolWorker = {
-      worker: new Worker(workerUrl()),
+      worker: new Worker(resolveWorkerFile()),
       job: null,
       timeout: null,
       idle: null,

open-sse/services/taskAwareRouter.ts

@@ -174,7 +174,7 @@ const TASK_PATTERNS: Record<TaskType, TaskPattern> = {
  *  - The list rotted. Those ids aged out by a generation or two, and every model
  *    release made them staler. Naming an intent instead removes the maintenance.
  *  - It bypassed the router. `applyTaskAwareRouting` overwrites `body.model`, so a
- *    literal target skipped auto-combo's 15-factor scoring (quota, circuit-breaker
+ *    literal target skipped auto-combo's 16-factor scoring (quota, circuit-breaker
  *    health, cost, latency, stability), connection cooldown and model lockout — and
  *    hard-failed for any operator who simply had no connection for that provider.
  *

open-sse/services/usage/fetcherProviders.ts

@@ -59,6 +59,14 @@ export const USAGE_FETCHER_PROVIDERS = [
   // PromptQL playground credits (data.pro.ql.app getCreditSummary)
   "promptql",
   "pql",
+  // Adobe Firefly credit balance (GET firefly.adobe.io/v1/credits/balance).
+  // Dispatched since #8006; the registration list was missed at the time, so the
+  // manifest and the quota fetchers reported no usage support for either id.
+  // Declaring them here is what makes `registerGenericQuotaFetchers` wire a
+  // generic fetcher for them and `resolveFreeAccessState` stop returning early —
+  // i.e. the balance is now actually fetched, which is the point.
+  "adobe-firefly",
+  "firefly",
   // HyperAgent billing usage (creditBlocks USD)
   "hyperagent",
   "ha",

open-sse/services/usage/qwen-token-plan.ts

@@ -11,6 +11,7 @@
 
 import {
   fetchQwenTokenPlanQuota,
+  resolveConsoleSite,
   QWEN_TOKEN_PLAN_WINDOW_5H,
   QWEN_TOKEN_PLAN_WINDOW_WEEKLY,
   type QwenTokenPlanQuota,
@@ -54,13 +55,26 @@ export async function getQwenTokenPlanUsage(
     });
 
     if (!quota) {
+      // The same plan is sold through two consoles with different portals, gateway
+      // hosts and login tickets — instructions for the wrong console produce a cookie
+      // the gateway rejects (console mismatch → NotLogined). With no cookie stored the
+      // console is inferred from the provider id, same rule the fetcher applies.
+      const site = resolveConsoleSite("", provider);
+      const guide =
+        site.consoleSite === "ALIYUN"
+          ? "Get it at modelstudio.console.alibabacloud.com (logged in): F12 › Network, " +
+            "reload, filter by api.json, click a request to " +
+            "bailian-singapore-cs.alibabacloud.com and copy the whole Cookie value from " +
+            "Request Headers (it contains login_aliyunid_ticket)."
+          : "Get it at home.qwencloud.com › Billing › Subscription (logged in): F12 › " +
+            "Network, reload, filter by api.json, click a request to " +
+            "cs-data.qwencloud.com and copy the whole Cookie value from Request Headers " +
+            "(it contains login_qwencloud_ticket).";
+      const brand = site.consoleSite === "ALIYUN" ? "Alibaba" : "Qwen";
       return {
         message:
-          "Qwen Token Plan connected. Quota needs a console session cookie — the inference " +
-          "API key cannot read it. Get it at home.qwencloud.com › Billing › Subscription " +
-          "(logged in): F12 › Network, reload, filter by api.json, click a request to " +
-          "cs-data.qwencloud.com and copy the whole Cookie value from Request Headers " +
-          "(it contains login_qwencloud_ticket). Paste it into the connection's " +
+          `${brand} Token Plan connected. Quota needs a console session cookie — the ` +
+          `inference API key cannot read it. ${guide} Paste it into the connection's ` +
           "'Qwen / Model Studio console cookie' field, or set QWEN_CLOUD_COOKIE. " +
           "The cookie expires with the browser session — re-paste it when this message returns.",
       };

open-sse/utils/stream.ts

@@ -763,6 +763,8 @@ export function createSSEStream(options: StreamOptions = {}) {
   let passthroughAccumulatedContent = "";
   let passthroughAccumulatedReasoning = "";
   let passthroughBufferedTextualToolCallContent = "";
+  /** Passthrough: whether a usage block was already forwarded to the client (prevents double). */
+  let passthroughForwardedUsage = false;
   // Passthrough Responses SSE: snapshots of items seen via `response.output_item.done`,
   // used to backfill `response.completed.response.output` when upstream returns it
   // empty (which happens when `store: false` — see backfillResponsesCompletedOutput).
@@ -839,13 +841,6 @@ export function createSSEStream(options: StreamOptions = {}) {
   const clientPayloadCollector = createStructuredSSECollector({
     stage: "client_response",
   });
-  const requestRecord = asRecord(body);
-  const requestStreamOptions = asRecord(
-    requestRecord.stream_options ?? requestRecord.streamOptions
-  );
-  const expectsOpenAIUsageOnlyChunk =
-    requestStreamOptions.include_usage === true || requestStreamOptions.includeUsage === true;
-
   // Per-stream instances to avoid shared state with concurrent streams
   const decoder = new TextDecoder();
   const encoder = new TextEncoder();
@@ -1751,7 +1746,7 @@ export function createSSEStream(options: StreamOptions = {}) {
                         !parsed.choices[0]?.finish_reason))
                   ) {
                     const emptyChoicesUsage = extractUsage(parsed) ?? parsed.usage;
-                    if (hasValidUsage(emptyChoicesUsage)) {
+                    if (hasValidUsage(emptyChoicesUsage) && !passthroughForwardedUsage) {
                       // Some upstreams (e.g. Ollama Cloud) emit prompt_tokens: 0
                       // even when input was sent — they simply don't count input
                       // tokens.  When we have a non-zero output but zero input,
@@ -1764,7 +1759,7 @@ export function createSSEStream(options: StreamOptions = {}) {
                       ) {
                         const pt = emptyChoicesUsage.prompt_tokens ?? 0;
                         if (pt === 0) {
-                          const estimated = estimateUsage(body, totalContentLength, FORMATS.OPENAI);
+                          const estimated = estimateUsage(body, totalContentLength, sourceFormat || FORMATS.OPENAI);
                           if (estimated?.prompt_tokens > 0) {
                             emptyChoicesUsage.prompt_tokens = estimated.prompt_tokens;
                             emptyChoicesUsage.total_tokens =
@@ -1773,6 +1768,7 @@ export function createSSEStream(options: StreamOptions = {}) {
                         }
                       }
                       usage = emptyChoicesUsage;
+                      passthroughForwardedUsage = true;
                       output = `data: ${JSON.stringify(parsed)}\n\n`;
                       injectedUsage = true;
                       clientPayload = parsed;
@@ -1782,6 +1778,11 @@ export function createSSEStream(options: StreamOptions = {}) {
                       continue;
                     }
 
+                    // If we already forwarded usage, drop any trailing empty-choices valid usage
+                    if (passthroughForwardedUsage && hasValidUsage(emptyChoicesUsage)) {
+                      continue;
+                    }
+
                     console.warn(
                       `[STREAM] Upstream returned empty choices array (${provider || "provider"}:${model || "unknown"}) — dropping chunk`
                     );
@@ -1980,18 +1981,24 @@ export function createSSEStream(options: StreamOptions = {}) {
                   }
                   if (
                     isFinishChunk &&
+                    !passthroughForwardedUsage &&
                     !hasValidUsage(parsed.usage) &&
-                    !expectsOpenAIUsageOnlyChunk
+                    !hasValidUsage(usage) &&
+                    totalContentLength > 0
                   ) {
-                    const estimated = estimateUsage(body, totalContentLength, FORMATS.OPENAI);
-                    parsed.usage = filterUsageForFormat(estimated, FORMATS.OPENAI);
-                    output = `data: ${JSON.stringify(parsed)}\n\n`;
-                    usage = estimated;
-                    injectedUsage = true;
-                  } else if (isFinishChunk && usage) {
+                    const estimated = estimateUsage(body, totalContentLength, sourceFormat || FORMATS.OPENAI);
+                    if (hasValidUsage(estimated)) {
+                      parsed.usage = filterUsageForFormat(estimated, sourceFormat || FORMATS.OPENAI);
+                      output = `data: ${JSON.stringify(parsed)}\n\n`;
+                      usage = estimated;
+                      passthroughForwardedUsage = true;
+                      injectedUsage = true;
+                    }
+                  } else if (isFinishChunk && hasValidUsage(usage) && !passthroughForwardedUsage) {
                     const buffered = addBufferToUsage(usage);
-                    parsed.usage = filterUsageForFormat(buffered, FORMATS.OPENAI);
+                    parsed.usage = filterUsageForFormat(buffered, sourceFormat || FORMATS.OPENAI);
                     output = `data: ${JSON.stringify(parsed)}\n\n`;
+                    passthroughForwardedUsage = true;
                     injectedUsage = true;
                   } else if (textualToolCallConverted) {
                     output = `data: ${JSON.stringify(parsed)}\n\n`;

open-sse/utils/usageTracking.ts

@@ -655,6 +655,7 @@ export function hasValidUsage(usage: UsageLike | null | undefined) {
     "output_tokens", // Claude
     "promptTokenCount",
     "candidatesTokenCount", // Gemini
+    "totalTokenCount", // Gemini (was missing — caused !hasValid to misfire on {totalTokenCount:15})
   ];
 
   for (const field of tokenFields) {
@@ -666,6 +667,17 @@ export function hasValidUsage(usage: UsageLike | null | undefined) {
   return false;
 }
 
+/** True when present but every token field zero/absent — web relays emit `{prompt_tokens:0, ...}`. */
+export function isEmptyUsage(usage: unknown): boolean {
+  if (!usage || typeof usage !== "object" || Array.isArray(usage)) return true;
+  const u = usage as Record<string, unknown>;
+  for (const k of ["prompt_tokens","completion_tokens","total_tokens","input_tokens","output_tokens","promptTokenCount","candidatesTokenCount","totalTokenCount"]) {
+    const v = u[k];
+    if (typeof v === "number" && Number.isFinite(v)) { if (v > 0) return false; }
+  }
+  return true;
+}
+
 /**
  * Extract usage from supported formats (Claude, OpenAI, Gemini, Responses API)
  */

scripts/check/check-docs-counts-sync.mjs

@@ -85,6 +85,27 @@ function countScoringFactors() {
   return parseScoringFactors(fs.readFileSync(file, "utf8"));
 }
 
+// PURE: the reference document must NAME every shipped pack. Reporting which one is
+// missing is the point — "6 packs" tells a doc it is stale, "chaos-mode is missing"
+// tells it what to write.
+export function makeModePackNamesValidator(names) {
+  return (content) => {
+    if (!names.length) return { ok: true, detail: "no mode packs found in source — skipping" };
+    // Token boundary, not `includes`: "ship-fast" is a substring of
+    // "ship-fast-v2", so a doc could satisfy the gate while naming a pack that
+    // does not ship — and a future pack named as a prefix of another would be
+    // masked by it.
+    const missing = names.filter(
+      (name) =>
+        !new RegExp(`(^|[^\\w-])${name.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")}([^\\w-]|$)`).test(
+          content
+        )
+    );
+    if (!missing.length) return { ok: true, detail: `all ${names.length} mode packs are named` };
+    return { ok: false, detail: `mode pack(s) never named in this file: ${missing.join(", ")}` };
+  };
+}
+
 // PURE: parse the canonical provider total out of the auto-generated catalog text.
 export function parseProviderTotal(referenceText) {
   if (!referenceText) return 0;
@@ -163,7 +184,8 @@ export function tallyDrift(checks, getContent) {
 // Returns null when tsx is unavailable so the gate degrades to a skip, not a false red.
 function readCodeFacts() {
   const script = [
-    'import {computeFreeModelTotals} from "./open-sse/config/freeModelCatalog.ts";',
+    'import {computeFreeModelTotals,FREE_MODEL_BUDGETS} from "./open-sse/config/freeModelCatalog.ts";',
+    'import {MODE_PACKS} from "./open-sse/services/autoCombo/modePacks.ts";',
     'import {ENGINE_IDS} from "./open-sse/services/compression/engineCatalog.ts";',
     'import {CLI_TOOLS} from "./src/shared/constants/cliTools.ts";',
     'import {countUniqueMcpTools} from "./open-sse/mcp-server/toolCount.ts";',
@@ -203,7 +225,10 @@ function readCodeFacts() {
     'console.log("@@"+JSON.stringify({freeSteady:t.steadyRecurringTokens,entries:t.perModel.length,',
     "freeFirst:t.firstMonthRealisticTokens,freePools:t.poolCount,engines:ENGINE_IDS.length,",
     "cliTotal:cli.length,cliCode:by('code'),cliAgent:by('agent'),",
-    "mcpTools:countUniqueMcpTools(cols),mcpScopes:sc.size,providers:pids.size,freeForever:ff.size}));",
+    "mcpTools:countUniqueMcpTools(cols),mcpScopes:sc.size,providers:pids.size,freeForever:ff.size,",
+    "modePacks:Object.keys(MODE_PACKS),",
+    "hardStop:FREE_MODEL_BUDGETS.filter(e=>e.hardStopGuaranteed===true).length,",
+    "trainsOnPrompts:FREE_MODEL_BUDGETS.filter(e=>e.trainsOnPrompts===true).length}));",
   ].join("");
   const tmp = fs.mkdtempSync(path.join(os.tmpdir(), "docs-counts-"));
   try {
@@ -317,10 +342,31 @@ export function extractNumberClaims(content, { pattern, skipBefore, skipAfter })
   return claims;
 }
 
+// Three spellings of the same claim are in use across the docs, and all three
+// must be watched: "6 curated **mode packs**", "6 pre-defined weight profiles",
+// "4 weight profiles". Matching only the first left the other two unguarded.
+const MODE_PACK_CLAIM_PATTERN =
+  /(\d+)\s+(?:curated\s+|pre-defined\s+)?\*{0,2}(?:mode\s+packs?|weight\s+profiles?)\b/gi;
+
 export function makeNumberClaimValidator(expected, opts) {
   return (content) => {
     const claims = extractNumberClaims(content, opts);
-    if (!claims.length) return { ok: true, detail: `no ${opts.what} claim in this file` };
+    if (!claims.length) {
+      // Most files in a check's list legitimately never mention the number, so
+      // "no claim" is normally a pass. But for a reference document that is
+      // supposed to state it, silence is the failure mode that matters: reword
+      // the sentence past the pattern and the gate goes quiet while reporting
+      // green. `requireClaim` says this file must carry the claim.
+      if (opts.requireClaim)
+        return {
+          ok: false,
+          detail:
+            `no ${opts.what} claim found, and this file is required to state one — ` +
+            `either the sentence was reworded past the pattern, or it was deleted ` +
+            `(code has ${expected})`,
+        };
+      return { ok: true, detail: `no ${opts.what} claim in this file` };
+    }
     const stale = claims.filter((c) => c.value !== expected);
     if (!stale.length)
       return { ok: true, detail: `${claims.length} ${opts.what} claim(s) match the code` };
@@ -477,7 +523,56 @@ export function buildChecks() {
         files,
         validate: makeNumberClaimValidator(expected, { what, ...opts }),
       });
+      const packs = Array.isArray(f.modePacks) ? f.modePacks : [];
       return [
+        {
+          // Two packs shipped after the docs were written and nothing noticed.
+          // The count and the names are two different gates: a table can carry
+          // the right number and still describe the wrong four out of six.
+          label: "Auto-Combo mode packs",
+          actual: packs.length,
+          docKey: "mode packs",
+          strict: true,
+          files: [
+            "README.md",
+            "llm.txt",
+            "docs/guides/FEATURES.md",
+            "docs/architecture/ARCHITECTURE.md",
+            "docs/architecture/REPOSITORY_MAP.md",
+            "docs/routing/AUTO-COMBO.md",
+          ],
+          validate: makeNumberClaimValidator(packs.length, {
+            what: "mode packs",
+            // Three spellings are in use across the docs, and all three are the
+            // same claim: "6 curated **mode packs**", "6 pre-defined weight
+            // profiles", "4 weight profiles". Matching only the first left the
+            // other two unwatched.
+            pattern: MODE_PACK_CLAIM_PATTERN,
+          }),
+        },
+        {
+          // Same claim, but on the one document that MUST carry it. Without
+          // `requireClaim` the strongest gate in this file is also the easiest to
+          // silence: reword the sentence and "no claim in this file" reads as a pass.
+          label: "Auto-Combo mode packs (reference doc must state the count)",
+          actual: packs.length,
+          docKey: "mode packs",
+          strict: true,
+          files: ["docs/routing/AUTO-COMBO.md"],
+          validate: makeNumberClaimValidator(packs.length, {
+            what: "mode packs",
+            pattern: MODE_PACK_CLAIM_PATTERN,
+            requireClaim: true,
+          }),
+        },
+        {
+          label: "Auto-Combo mode packs (named in the reference doc)",
+          actual: packs.length,
+          docKey: "mode packs",
+          strict: true,
+          files: ["docs/routing/AUTO-COMBO.md"],
+          validate: makeModePackNamesValidator(packs),
+        },
         {
           label: "Provider reference total (doc vs live modules)",
           actual: f.providers,
@@ -570,6 +665,37 @@ export function buildChecks() {
           },
           ["README.md", "docs/diagrams/free-tier-budget.svg", "docs/reference/FREE_TIERS.md"]
         ),
+        // The reference page says what an entry vouches for. These two facts are
+        // curated by hand rather than inferred, so the page quotes their counts —
+        // and quoting a count is how a page goes stale. The patterns are deliberately
+        // narrow: FREE_TIERS.md is full of numbers, and a loose one would gate a
+        // token budget by accident.
+        claim(
+          f.hardStop,
+          "hard-stop-guaranteed entries",
+          {
+            // `requireClaim`: this page is the one place that states the number,
+            // so a reworded or deleted sentence must fail rather than pass as
+            // "no claim in this file" — otherwise the gate is one edit from silent.
+            requireClaim: true,
+            pattern:
+              /(\d+) entr(?:y|ies) (?:that )?(?:carry|carries) an? independently documented hard stop/gi,
+          },
+          ["docs/reference/FREE_TIERS.md"]
+        ),
+        claim(
+          f.trainsOnPrompts,
+          "training-disclosure entries",
+          {
+            // `requireClaim`: this page is the one place that states the number,
+            // so a reworded or deleted sentence must fail rather than pass as
+            // "no claim in this file" — otherwise the gate is one edit from silent.
+            requireClaim: true,
+            pattern:
+              /(\d+) entr(?:y|ies) (?:that )?(?:carry|carries) a (?:prompt-)?training disclosure/gi,
+          },
+          ["docs/reference/FREE_TIERS.md"]
+        ),
       ];
     })(),
     {
@@ -600,6 +726,12 @@ export function buildChecks() {
         "docs/diagrams/strategies-grid.svg",
         "docs/diagrams/auto-combo-scoring.mmd",
         "llm.txt",
+        "docs/architecture/ARCHITECTURE.md",
+        "docs/architecture/REPOSITORY_MAP.md",
+        "docs/architecture/RESILIENCE_GUIDE.md",
+        "docs/frameworks/OPEN_SSE_ARCHITECTURE.md",
+        "docs/getting-started/AUTO-COMBO-GUIDE.md",
+        "skills/omni-combos-routing/SKILL.md",
         "open-sse/services/autoCombo/routerStrategy.ts",
         "open-sse/services/taskAwareRouter.ts",
         "tests/unit/lkgp-enabled-context-11181.test.ts",

skills/omni-combos-routing/SKILL.md

@@ -228,7 +228,7 @@ curl -X POST $OMNIROUTE_URL/api/combos \
 | `reset-window` | Order targets by their configured reset window |
 | `headroom` | Prefer targets with more remaining quota headroom |
 | `strict-random` | Random without repeating until all targets have been used |
-| `auto` | Auto-Combo scoring across 13 factors |
+| `auto` | Auto-Combo scoring across 16 factors |
 | `lkgp` | Last-known-good-provider sticky routing |
 | `context-optimized` | Pick the best model for the request's context size |
 | `cache-optimized` | Prefer targets with stronger cache affinity |
@@ -237,7 +237,7 @@ curl -X POST $OMNIROUTE_URL/api/combos \
 
 ## Auto-combo (recommended for production)
 
-Auto-combo scores each candidate on 13 factors every request:
+Auto-combo scores each candidate on 16 factors every request:
 
 ```bash
 curl -X POST $OMNIROUTE_URL/api/combos \

src/app/(dashboard)/dashboard/combos/comboAgentFeatures.ts

@@ -0,0 +1,46 @@
+/**
+ * Agent-features clearing for the combos editor (#399 / #401 / #454, fixed in #12158).
+ *
+ * `PUT /api/combos/[id]` merges its body over the stored record, so an omitted field
+ * means "leave unchanged". Deleting a cleared field from the payload therefore left the
+ * previous value in the database: unchecking context cache protection, or emptying the
+ * system message or tool filter, never persisted. Only an explicit `null` reaches
+ * `updateCombo`'s null-means-delete pass.
+ *
+ * On create there is nothing to clear, so an empty field is simply absent — the same
+ * shape `description` and `context_length` already use in this editor.
+ */
+export interface AgentFeatureInput {
+  systemMessage: string;
+  toolFilter: string;
+  contextCache: boolean;
+  isEdit: boolean;
+}
+
+export interface AgentFeaturePatch {
+  system_message?: string | null;
+  tool_filter_regex?: string | null;
+  context_cache_protection?: true | null;
+}
+
+export function buildAgentFeaturePatch({
+  systemMessage,
+  toolFilter,
+  contextCache,
+  isEdit,
+}: AgentFeatureInput): AgentFeaturePatch {
+  const patch: AgentFeaturePatch = {};
+
+  const message = systemMessage.trim();
+  if (message) patch.system_message = message;
+  else if (isEdit) patch.system_message = null;
+
+  const filter = toolFilter.trim();
+  if (filter) patch.tool_filter_regex = filter;
+  else if (isEdit) patch.tool_filter_regex = null;
+
+  if (contextCache) patch.context_cache_protection = true;
+  else if (isEdit) patch.context_cache_protection = null;
+
+  return patch;
+}

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

@@ -18,6 +18,7 @@ import { useCopyToClipboard } from "@/shared/hooks/useCopyToClipboard";
 import { FieldLabelWithHelp, WeightTotalBar } from "./parts";
 import { ComboTargetOptions } from "./ComboQuotaOnlyFallbackToggle";
 import { applyQuotaOnlyFallbackConfig, setQuotaOnlyFallback } from "./comboQuotaOnlyFallback";
+import { buildAgentFeaturePatch } from "./comboAgentFeatures";
 import { useComboProxyAssignments } from "./useComboProxyAssignments";
 import { ResponseValidationEditor, type ResponseValidationValue } from "./ResponseValidationEditor";
 import ReasoningTokenBufferToggle from "./ReasoningTokenBufferToggle";
@@ -3011,13 +3012,20 @@ function ComboFormModal({ isOpen, combo, onClose, onSave, activeProviders, combo
       saveData.config = configToSave;
     }
 
-    // Agent features (#399 / #401 / #454)
-    if (agentSystemMessage.trim()) saveData.system_message = agentSystemMessage.trim();
-    else delete saveData.system_message;
-    if (agentToolFilter.trim()) saveData.tool_filter_regex = agentToolFilter.trim();
-    else delete saveData.tool_filter_regex;
-    if (agentContextCache) saveData.context_cache_protection = true;
-    else delete saveData.context_cache_protection;
+    // Agent features (#399 / #401 / #454). A cleared field is sent as null on edit
+    // rather than omitted, because PUT merges over the stored record (#12158).
+    delete saveData.system_message;
+    delete saveData.tool_filter_regex;
+    delete saveData.context_cache_protection;
+    Object.assign(
+      saveData,
+      buildAgentFeaturePatch({
+        systemMessage: agentSystemMessage,
+        toolFilter: agentToolFilter,
+        contextCache: agentContextCache,
+        isEdit,
+      })
+    );
 
     // Validate and save context_length
     if (contextLength !== undefined && contextLength !== null) {

src/app/(dashboard)/dashboard/free-provider-rankings/page.tsx

@@ -8,7 +8,11 @@ import {
   sortRankingsAuthTypeFirst,
   type ProviderAuthType,
 } from "@/lib/freeProviderRankingsAuthType";
-import { formatUsageReliability, usageToneClass } from "@/lib/freeProviderRankingsUsage";
+import {
+  formatUsageReliability,
+  sortRankingsByReliability,
+  usageToneClass,
+} from "@/lib/freeProviderRankingsUsage";
 // Type-only: `freeProviderRankings` wires DB modules at import time, so a
 // client component must never take a runtime value from it. The page used to
 // keep its own copy of this shape, which had already drifted past the API.
@@ -60,6 +64,7 @@ export default function FreeProviderRankingsPage() {
   const [availableOnly, setAvailableOnly] = useState(false);
   const [typeFilter, setTypeFilter] = useState<ProviderAuthType | "">("");
   const [groupByType, setGroupByType] = useState(false);
+  const [sortByReliability, setSortByReliability] = useState(false);
 
   const fetchRankings = useCallback(
     async (category?: string, opts?: { configuredOnly?: boolean; availableOnly?: boolean }) => {
@@ -97,8 +102,11 @@ export default function FreeProviderRankingsPage() {
   // from the already-fetched `rankings`, never trigger a refetch.
   const displayedRankings = useMemo(() => {
     const filtered = filterRankingsByAuthType(rankings, typeFilter);
-    return groupByType ? sortRankingsAuthTypeFirst(filtered) : filtered;
-  }, [rankings, typeFilter, groupByType]);
+    // Reliability first, then grouping: the type sort compares categories only,
+    // so a stable sort keeps the reliability order inside each group.
+    const ordered = sortByReliability ? sortRankingsByReliability(filtered) : filtered;
+    return groupByType ? sortRankingsAuthTypeFirst(ordered) : ordered;
+  }, [rankings, typeFilter, groupByType, sortByReliability]);
 
   return (
     <div className="flex flex-col gap-6">
@@ -182,6 +190,18 @@ export default function FreeProviderRankingsPage() {
         >
           {t("sortTypeFirst")}
         </button>
+        <button
+          onClick={() => setSortByReliability((v) => !v)}
+          aria-pressed={sortByReliability}
+          title={t("sortByReliabilityHelp")}
+          className={`px-4 py-2 text-sm font-medium rounded-lg border transition-colors ${
+            sortByReliability
+              ? "bg-emerald-500 border-emerald-500 text-white"
+              : "border-border text-text-muted hover:text-text-main hover:border-emerald-500/50"
+          }`}
+        >
+          {t("sortByReliability")}
+        </button>
       </div>
       <p className="text-xs text-text-muted">{t("typeLegend")}</p>
 

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

@@ -117,6 +117,8 @@ export default function ProviderDetailPageClient() {
     providerNode,
     loading,
     retestingId,
+    handleClearCooldown,
+    clearingCooldownId,
     batchTesting,
     batchTestResults,
     selectedIds,
@@ -671,7 +673,11 @@ export default function ProviderDetailPageClient() {
             />
           ) : (
             <>
-              <CoolingConnectionsPanel connections={connections} />
+              <CoolingConnectionsPanel
+                connections={connections}
+                onClearCooldown={handleClearCooldown}
+                clearingCooldownId={clearingCooldownId}
+              />
               <ConnectionsListPanel
                 connections={connections}
                 providerId={providerId}

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

@@ -12,6 +12,14 @@
  * indicator the user has no way to see "OmniRoute learned that this key is
  * exhausted — and for how long". This panel makes the lesson visible.
  *
+ * Each row also carries a manual "Clear cooldown" action. The cooldown is
+ * OmniRoute's local lesson, not upstream truth: when the quota has already
+ * refreshed upstream (daily/weekly resets, provider-side fix), the automatic
+ * clear paths (Test-button success, Edit-modal key re-validation) still
+ * require an upstream round-trip before the bench lifts. The button PUTs
+ * `rateLimitedUntil: null` directly so the connection rejoins routing
+ * immediately — the next request is the real test of whether the key works.
+ *
  * Acceptance criteria (Issue #1, fix scope D):
  *   1. Filters `connections` to those with a future `rateLimitedUntil`.
  *   2. Shows connection name + reset countdown.
@@ -19,6 +27,8 @@
  *   4. Renders nothing when no connection is cooling.
  *   5. Uses the same connection-shape type as ConnectionRow so the data flow
  *      stays consistent with the rest of the dashboard.
+ *   6. Offers a per-row manual clear (disabled while that row's request is
+ *      in flight) for the stale-bench case described above.
  */
 
 import { useEffect, useState } from "react";
@@ -34,6 +44,10 @@ import { providerText } from "../providerCredentialText";
 
 export interface CoolingConnectionsPanelProps {
   readonly connections: readonly ConnectionRowConnection[];
+  /** Clears a connection's persisted cooldown (PUT `rateLimitedUntil: null`). */
+  readonly onClearCooldown?: (connectionId: string) => void;
+  /** Connection id whose clear request is in flight — disables its button. */
+  readonly clearingCooldownId?: string | null;
 }
 
 function isCoolingNow(connection: ConnectionRowConnection, now: number): boolean {
@@ -42,8 +56,45 @@ function isCoolingNow(connection: ConnectionRowConnection, now: number): boolean
   return Number.isFinite(until) && until > now;
 }
 
+interface ClearCooldownButtonProps {
+  /** Row's connection id — without one there is nothing to PUT, so no button. */
+  readonly connectionId: string | undefined;
+  /** True while this row's clear request is in flight (disables the button). */
+  readonly clearing: boolean;
+  readonly onClearCooldown?: (connectionId: string) => void;
+}
+
+/**
+ * Per-row manual "Clear cooldown" action, split out of the panel body so the
+ * row map stays readable (and the panel inside the max-lines-per-function
+ * ratchet). Renders nothing when there is no handler or no connection id.
+ */
+function ClearCooldownButton(props: ClearCooldownButtonProps) {
+  const { connectionId, clearing, onClearCooldown } = props;
+  const t = useTranslations("providers");
+  if (!onClearCooldown || !connectionId) return null;
+  return (
+    <button
+      type="button"
+      data-testid={`clear-cooldown-${connectionId}`}
+      disabled={clearing}
+      onClick={() => onClearCooldown(connectionId)}
+      title={providerText(
+        t,
+        "clearConnectionCooldownTitle",
+        "Clear the cooldown now — use when the quota has already refreshed upstream"
+      )}
+      className="rounded border border-amber-500/50 px-2 py-0.5 text-xs text-amber-700 transition-colors hover:bg-amber-500/10 disabled:cursor-not-allowed disabled:opacity-50 dark:text-amber-300"
+    >
+      {clearing
+        ? providerText(t, "clearConnectionCooldownInProgress", "Clearing…")
+        : providerText(t, "clearConnectionCooldown", "Clear cooldown")}
+    </button>
+  );
+}
+
 export default function CoolingConnectionsPanel(props: CoolingConnectionsPanelProps) {
-  const { connections } = props;
+  const { connections, onClearCooldown, clearingCooldownId } = props;
   const t = useTranslations("providers");
   // Tick once per second so the human-readable countdown updates.
   const [now, setNow] = useState<number>(() => Date.now());
@@ -88,17 +139,25 @@ export default function CoolingConnectionsPanel(props: CoolingConnectionsPanelPr
             (c.id
               ? `${providerText(t, "connectionFallback", "connection")} ${c.id.slice(0, 8)}`
               : providerText(t, "connectionFallback", "connection"));
+          const clearing = clearingCooldownId != null && clearingCooldownId === c.id;
           return (
             <li
               key={c.id ?? label}
-              className="flex items-center justify-between rounded border border-amber-500/30 bg-background/40 px-3 py-2 text-sm"
+              className="flex items-center justify-between gap-2 rounded border border-amber-500/30 bg-background/40 px-3 py-2 text-sm"
             >
               <span className="font-medium">{label}</span>
-              <span
-                className="font-mono text-xs text-amber-700 dark:text-amber-300"
-                data-testid="cooling-countdown"
-              >
-                {formatResetCountdown(until)}
+              <span className="flex items-center gap-2">
+                <span
+                  className="font-mono text-xs text-amber-700 dark:text-amber-300"
+                  data-testid="cooling-countdown"
+                >
+                  {formatResetCountdown(until)}
+                </span>
+                <ClearCooldownButton
+                  connectionId={c.id}
+                  clearing={clearing}
+                  onClearCooldown={onClearCooldown}
+                />
               </span>
             </li>
           );

src/app/(dashboard)/dashboard/providers/[id]/components/__tests__/CoolingConnectionsPanel.test.tsx

@@ -0,0 +1,150 @@
+// @vitest-environment jsdom
+//
+// Biting tests for the manual "Clear cooldown" action in
+// CoolingConnectionsPanel (stale-bench case: quota already refreshed upstream,
+// OmniRoute's persisted rate_limited_until still benches the connection).
+//
+// What must NOT regress:
+//   1. Clicking the per-row button fires onClearCooldown with the row's
+//      connection id — and nothing else (no fetch here; the panel stays a
+//      dumb readout + action surface).
+//   2. The button is disabled (and stays silent) while that row's clear is
+//      in flight, so double-clicks can't stack PUTs.
+//   3. Rows without a connection id render no button (nothing to PUT).
+//   4. Omitting onClearCooldown renders the legacy read-only panel — the
+//      mount in ProviderDetailPageClient is the only consumer, but the prop
+//      is optional so the component stays independently mountable.
+//   5. The panel still renders nothing when no connection is cooling.
+import React, { act } from "react";
+import { createRoot } from "react-dom/client";
+import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
+import CoolingConnectionsPanel from "../CoolingConnectionsPanel";
+import type { ConnectionRowConnection } from "../ConnectionRow";
+
+vi.mock("next/navigation", () => ({
+  useParams: () => ({ id: "zai" }),
+  useRouter: () => ({ push: vi.fn(), replace: vi.fn() }),
+}));
+vi.mock("next-intl", () => ({
+  useTranslations: () => {
+    const t = ((key: string) => key) as ((key: string) => string) & {
+      has?: (key: string) => boolean;
+    };
+    t.has = () => false; // force providerText fallbacks, like phase1d tests
+    return t;
+  },
+}));
+
+function coolingConnection(overrides: Partial<ConnectionRowConnection> = {}) {
+  return {
+    id: "conn-cooling-1",
+    provider: "zai",
+    name: "main",
+    rateLimitedUntil: new Date(Date.now() + 60_000).toISOString(),
+    ...overrides,
+  } as ConnectionRowConnection;
+}
+
+const cleanups: Array<() => void> = [];
+
+function renderPanel(props: Partial<Parameters<typeof CoolingConnectionsPanel>[0]> = {}) {
+  const container = document.createElement("div");
+  document.body.appendChild(container);
+  const root = createRoot(container);
+  let panelRoot: HTMLDivElement | null = null;
+  act(() => {
+    root.render(
+      <CoolingConnectionsPanel
+        connections={[coolingConnection()]}
+        onClearCooldown={undefined}
+        {...props}
+      />
+    );
+  });
+  cleanups.push(() => {
+    act(() => root.unmount());
+    container.remove();
+  });
+  panelRoot = container.querySelector('[data-testid="cooling-connections-panel"]');
+  return { container, root, panelRoot };
+}
+
+describe("CoolingConnectionsPanel — manual clear cooldown", () => {
+  beforeEach(() => {
+    (
+      globalThis as typeof globalThis & { IS_REACT_ACT_ENVIRONMENT?: boolean }
+    ).IS_REACT_ACT_ENVIRONMENT = true;
+  });
+
+  afterEach(() => {
+    while (cleanups.length) cleanups.pop()?.();
+    document.body.innerHTML = "";
+  });
+
+  it("renders nothing when no connection is cooling", () => {
+    const { container } = renderPanel({
+      connections: [coolingConnection({ rateLimitedUntil: null })],
+    });
+    expect(container.querySelector('[data-testid="cooling-connections-panel"]')).toBeNull();
+  });
+
+  it("clicking the row button fires onClearCooldown with the connection id", () => {
+    const onClearCooldown = vi.fn();
+    const { panelRoot } = renderPanel({ onClearCooldown });
+    const button = panelRoot!.querySelector<HTMLButtonElement>(
+      '[data-testid="clear-cooldown-conn-cooling-1"]'
+    );
+    expect(button).not.toBeNull();
+    act(() => {
+      button!.dispatchEvent(new MouseEvent("click", { bubbles: true }));
+    });
+    expect(onClearCooldown).toHaveBeenCalledTimes(1);
+    expect(onClearCooldown).toHaveBeenCalledWith("conn-cooling-1");
+  });
+
+  it("disables the button and stays silent while the row's clear is in flight", () => {
+    const onClearCooldown = vi.fn();
+    const { panelRoot } = renderPanel({ onClearCooldown, clearingCooldownId: "conn-cooling-1" });
+    const button = panelRoot!.querySelector<HTMLButtonElement>(
+      '[data-testid="clear-cooldown-conn-cooling-1"]'
+    );
+    expect(button!.disabled).toBe(true);
+    act(() => {
+      button!.dispatchEvent(new MouseEvent("click", { bubbles: true }));
+    });
+    expect(onClearCooldown).not.toHaveBeenCalled();
+    expect(button!.textContent).toContain("Clearing…");
+  });
+
+  it("leaves another row's button enabled (per-row in-flight state)", () => {
+    const onClearCooldown = vi.fn();
+    const connA = coolingConnection({ id: "conn-a" });
+    const connB = coolingConnection({ id: "conn-b" });
+    const { panelRoot } = renderPanel({
+      connections: [connA, connB],
+      onClearCooldown,
+      clearingCooldownId: "conn-a",
+    });
+    const buttonB = panelRoot!.querySelector<HTMLButtonElement>(
+      '[data-testid="clear-cooldown-conn-b"]'
+    );
+    expect(buttonB!.disabled).toBe(false);
+    act(() => {
+      buttonB!.dispatchEvent(new MouseEvent("click", { bubbles: true }));
+    });
+    expect(onClearCooldown).toHaveBeenCalledWith("conn-b");
+  });
+
+  it("renders read-only (no buttons) when onClearCooldown is omitted", () => {
+    const { panelRoot } = renderPanel();
+    expect(panelRoot!.querySelector<HTMLButtonElement>("button")).toBeNull();
+  });
+
+  it("renders no button for a row without a connection id", () => {
+    const { panelRoot } = renderPanel({
+      connections: [coolingConnection({ id: undefined, name: "anon" })],
+      onClearCooldown: vi.fn(),
+    });
+    expect(panelRoot!.querySelectorAll("button").length).toBe(0);
+  });
+});

src/app/(dashboard)/dashboard/providers/[id]/hooks/useProviderConnections.ts

@@ -10,7 +10,7 @@
  *  - batch activate / deactivate / retest / delete (with MAX_BULK_IDS chunking)
  *  - single-connection handlers: delete, update status, proxy toggles,
  *    rate-limit, claude extra-usage, codex limit, cpa mode,
- *    retest, token refresh, swap priority
+ *    retest, clear-cooldown, token refresh, swap priority
  *  - selection state: selectedIds, handleToggleSelectOne/All, batchDeleteConfirmOpen
  *  - batch-test runner (runBatchTest / handleBatchTestAll / handleBatchRetest)
  *  - health/pagination filters (healthFilter, page)
@@ -155,6 +155,8 @@ export interface UseProviderConnectionsReturn {
   providerNode: any;
   loading: boolean;
   retestingId: string | null;
+  /** Connection id whose cooldown-clear PUT is in flight (drives button spinners). */
+  clearingCooldownId: string | null;
   batchTesting: boolean;
   batchTestResults: BatchTestResults;
   selectedIds: Set<string>;
@@ -208,6 +210,14 @@ export interface UseProviderConnectionsReturn {
     perKeyProxyEnabled: boolean
   ) => Promise<void>;
   handleRetestConnection: (connectionId: string) => Promise<void>;
+  /**
+   * Manually lifts a persisted 429 cooldown: PUTs `rateLimitedUntil: null`
+   * (plus backoff reset server-side) so the connection rejoins routing
+   * immediately. For the "quota already refreshed upstream but OmniRoute
+   * still benches the key" case — the cooldown timer is OmniRoute's own
+   * lesson, not upstream truth.
+   */
+  handleClearCooldown: (connectionId: string) => Promise<void>;
   handleRefreshToken: (connectionId: string) => Promise<void>;
   handleSwapPriority: (conn1: any, conn2: any) => Promise<void>;
   handleReorderByAvailability: () => Promise<void>;
@@ -253,6 +263,7 @@ export function useProviderConnections(
 
   // ── test state ──────────────────────────────────────────────────────────
   const [retestingId, setRetestingId] = useState<string | null>(null);
+  const [clearingCooldownId, setClearingCooldownId] = useState<string | null>(null);
   const [batchTesting, setBatchTesting] = useState(false);
   const [batchTestResults, setBatchTestResults] = useState<BatchTestResults>(null);
 
@@ -674,6 +685,43 @@ export function useProviderConnections(
     }
   };
 
+  // Manually lift a persisted 429 cooldown. Complements the automatic paths
+  // (Test-button success / Edit-modal key re-validation): those only clear the
+  // bench as a side effect of a successful upstream round-trip, so a user whose
+  // quota already refreshed upstream still waits out OmniRoute's local timer.
+  // PUT /api/providers/[id] applies updateProviderConnectionDefaults, which
+  // resets backoffLevel → 0 alongside rateLimitedUntil → null.
+  const handleClearCooldown = async (connectionId: string) => {
+    if (!connectionId || clearingCooldownId) return;
+    setClearingCooldownId(connectionId);
+    try {
+      const res = await fetch(`/api/providers/${connectionId}`, {
+        method: "PUT",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({ rateLimitedUntil: null }),
+      });
+      if (!res.ok) {
+        const data = await res.json().catch(() => ({}));
+        notify.error(data.error || t("failedClearConnectionCooldown"));
+        return;
+      }
+      // Optimistically drop the cooldown locally so the row leaves the cooling
+      // panel immediately; fetchConnections() reconciles with server truth.
+      setConnections((prev: any[]) =>
+        prev.map((c) =>
+          c.id === connectionId ? { ...c, rateLimitedUntil: null, backoffLevel: 0 } : c
+        )
+      );
+      notify.success(t("connectionCooldownCleared"));
+      await fetchConnections();
+    } catch (error) {
+      console.error("Error clearing cooldown:", error);
+      notify.error(t("failedClearConnectionCooldown"));
+    } finally {
+      setClearingCooldownId(null);
+    }
+  };
+
   const handleRefreshToken = async (connectionId: string) => {
     if (refreshingId) return;
     setRefreshingId(connectionId);
@@ -1072,6 +1120,8 @@ export function useProviderConnections(
     handleToggleProxyEnabled,
     handleTogglePerKeyProxyEnabled,
     handleRetestConnection,
+    handleClearCooldown,
+    clearingCooldownId,
     handleRefreshToken,
     handleSwapPriority,
     handleReorderByAvailability,

src/app/(dashboard)/dashboard/radar/RadarCatalogTable.tsx

@@ -45,11 +45,47 @@ interface RadarCatalogTableProps {
   onError: (message: string) => void;
 }
 
+/**
+ * Compact a plain count. Locale-independent on purpose: `toLocaleString()` would
+ * make the rendering — and any test asserting on it — depend on the machine's
+ * locale.
+ */
+export function compactCount(value: number): string {
+  // The feed is JSON off the network, so the declared `number` is a hope, not a
+  // guarantee: a string, a NaN or an Infinity would otherwise render as
+  // "InfinityM" or "NaN/min" in the operator's table.
+  const numeric = typeof value === "number" ? value : Number(value);
+  if (!Number.isFinite(numeric) || numeric < 0) return "?";
+  if (numeric >= 1_000_000) return `${(numeric / 1_000_000).toFixed(1)}M`;
+  if (numeric >= 1_000) return `${(numeric / 1_000).toFixed(0)}K`;
+  return String(numeric);
+}
+
 function formatTokens(value: number): string {
+  // A monthly budget of zero means the pool is rate-limited rather than
+  // token-limited. That reading is specific to budgets — see `formatLimits`.
   if (value === 0) return "rate-only";
-  if (value >= 1_000_000) return `${(value / 1_000_000).toFixed(1)}M`;
-  if (value >= 1_000) return `${(value / 1_000).toFixed(0)}K`;
-  return String(value);
+  return compactCount(value);
+}
+
+/**
+ * Render the rate limits the feed reported, and nothing else. Each of the four
+ * values is independently nullable, so an absent one is omitted rather than
+ * printed as zero — "no reported limit" and "a limit of zero" are opposite
+ * facts, and the second one is worth noticing. An entry with no limits at all,
+ * or with four nulls, reads "—".
+ *
+ * Deliberately not built on `formatTokens`: that one maps 0 to "rate-only",
+ * which would render a zero ceiling as "rate-only/min".
+ */
+export function formatLimits(entry: Pick<RadarMergedEntry, "limits">): string {
+  const { rpm, rpd, tpm, tpd } = entry.limits ?? {};
+  const parts: string[] = [];
+  if (rpm != null) parts.push(`${compactCount(rpm)}/min`);
+  if (rpd != null) parts.push(`${compactCount(rpd)}/day`);
+  if (tpm != null) parts.push(`${compactCount(tpm)} tok/min`);
+  if (tpd != null) parts.push(`${compactCount(tpd)} tok/day`);
+  return parts.length ? parts.join(" · ") : "—";
 }
 
 function budgetLabel(entry: RadarMergedEntry): string {
@@ -184,6 +220,7 @@ export function RadarCatalogTable({ entries, refreshCatalog, onError }: RadarCat
                 <th className="pb-3 font-medium">{t("colProvider")}</th>
                 <th className="pb-3 font-medium">{t("colModel")}</th>
                 <th className="pb-3 font-medium">{t("colQuota")}</th>
+                <th className="pb-3 font-medium">{t("colLimits")}</th>
                 <th className="pb-3 font-medium">{t("colContext")}</th>
                 <th className="pb-3 font-medium">{t("colCapabilities")}</th>
                 <th className="pb-3 font-medium">{t("colTos")}</th>
@@ -245,6 +282,7 @@ export function RadarCatalogTable({ entries, refreshCatalog, onError }: RadarCat
                       )}
                     </td>
                     <td className="py-3 text-sm">{budgetLabel(entry)}</td>
+                    <td className="py-3 text-sm text-text-muted">{formatLimits(entry)}</td>
                     <td className="py-3 text-sm text-text-muted">
                       {entry.contextWindow ? `${(entry.contextWindow / 1000).toFixed(0)}K` : "—"}
                     </td>
@@ -281,6 +319,14 @@ export function RadarCatalogTable({ entries, refreshCatalog, onError }: RadarCat
                       >
                         {entry.tos}
                       </span>
+                      {entry.trainsOnPrompts === true && (
+                        <span
+                          className="ml-1 text-xs px-2 py-1 rounded bg-orange-500/10 text-orange-400"
+                          title={t("trainsOnPromptsHelp")}
+                        >
+                          {t("trainsOnPrompts")}
+                        </span>
+                      )}
                     </td>
                     <td className="py-3 pl-3">
                       {editingKey === key ? (

src/app/api/free-provider-rankings/route.ts

@@ -27,6 +27,7 @@ const QuerySchema = z.object({
   // Rejected rather than silently coerced: a typo must not quietly return a
   // different window than the caller asked for.
   usageRange: z.enum(["1h", "24h", "7d", "30d"]).optional(),
+  sortBy: z.enum(["elo", "reliability"]).optional(),
 });
 
 export async function OPTIONS() {
@@ -42,6 +43,7 @@ export async function GET(request: NextRequest) {
     availableOnly: url.searchParams.get("availableOnly") || undefined,
     withUsage: url.searchParams.get("withUsage") || undefined,
     usageRange: url.searchParams.get("usageRange") || undefined,
+    sortBy: url.searchParams.get("sortBy") || undefined,
   });
 
   if (!parsed.success) {
@@ -51,12 +53,14 @@ export async function GET(request: NextRequest) {
     );
   }
 
-  const { category, limit, configuredOnly, availableOnly, withUsage, usageRange } = parsed.data;
+  const { category, limit, configuredOnly, availableOnly, withUsage, usageRange, sortBy } =
+    parsed.data;
   const rankings = await computeFreeProviderRankings(category, limit, {
     configuredOnly,
     availableOnly,
     withUsage,
     usageRange,
+    sortBy,
   });
 
   return NextResponse.json({ rankings }, { headers: CORS_HEADERS });

src/app/api/free-tier/summary/route.ts

@@ -7,7 +7,7 @@ import {
   FREE_MODEL_BUDGETS,
 } from "@omniroute/open-sse/config/freeModelCatalog.data.ts";
 import type { MergedEntry } from "@/lib/radar/applyFeed";
-import { getRadarCatalog } from "@/lib/radar";
+import { getCatalogWithoutOverlay, getRadarCatalog } from "@/lib/radar";
 import { sumUsageTokensThisMonth } from "@/lib/db/usageSummary";
 import { isAuthenticated } from "@/shared/utils/apiAuth";
 import { listNoCredentialProviders } from "@/shared/utils/providerCredentialRequirement";
@@ -69,11 +69,30 @@ export async function GET(req: Request): Promise<Response> {
   // live feed is supporter-key content, so it only reaches callers this
   // instance has authenticated — never anonymous visitors, or an exposed
   // instance would re-publish the paid feed for free.
-  const serveOverlay = meta !== null && (meta.tier !== "live" || (await isAuthenticated(req)));
+  // A feed built before the catalog this release ships is not an overlay, it is a
+  // regression: the totals would be recomputed from data older than the baseline
+  // the operator installed. An unknown build date — a cache row written before the
+  // column existed — counts as older: unknown never outranks known.
+  const overlayIsFresh =
+    meta !== null &&
+    meta.generatedAt !== null &&
+    meta.generatedAt.slice(0, 10) >= FREE_CATALOG_CURATED_AT;
+
+  const serveOverlay = overlayIsFresh && (meta.tier !== "live" || (await isAuthenticated(req)));
+
+  // Withheld only because it is stale: drop the feed, keep the operator's own
+  // local state. Falling back to the raw baseline here would resurrect models the
+  // operator disabled or tombstoned.
+  const overlayWithheldAsStale = meta !== null && !overlayIsFresh;
 
   const totals = serveOverlay
     ? computeFreeModelTotals({ excludeTosAvoid, entries: entries.map(toBudgetEntry) })
-    : computeFreeModelTotals({ excludeTosAvoid });
+    : overlayWithheldAsStale
+      ? computeFreeModelTotals({
+          excludeTosAvoid,
+          entries: getCatalogWithoutOverlay().map(toBudgetEntry),
+        })
+      : computeFreeModelTotals({ excludeTosAvoid });
 
   const usedThisMonth = sumUsageTokensThisMonth();
   const body = {

src/i18n/messages/ar.json

@@ -6448,7 +6448,12 @@
     "savedModelEndpointSettings": "إعدادات نقطة نهاية النموذج المحفوظ",
     "searchByModelAria": "البحث حسب الطراز",
     "selectSupportedEndpoint": "اختر نقطة نهاية مدعومة واحدة على الأقل",
-    "antigravityClientProfileHarness": "Harness / CLI"
+    "antigravityClientProfileHarness": "Harness / CLI",
+    "clearConnectionCooldown": "Clear cooldown",
+    "clearConnectionCooldownInProgress": "Clearing…",
+    "clearConnectionCooldownTitle": "Clear the cooldown now — use when the quota has already refreshed upstream",
+    "connectionCooldownCleared": "Cooldown cleared — connection rejoined routing",
+    "failedClearConnectionCooldown": "Failed to clear cooldown"
   },
   "settings": {
     "title": "الإعدادات",
@@ -12531,6 +12536,8 @@
     "typeApikey": "مفتاح واجهة برمجة التطبيقات",
     "sortTypeFirst": "الأسهل أولاً",
     "sortTypeFirstHelp": "التجميع حسب سهولة البدء (من دون تسجيل ← تسجيل دخول OAuth ← مفتاح API)، مع الحفاظ على ترتيب الجودة داخل كل مجموعة",
+    "sortByReliability": "__MISSING__:Most reliable first",
+    "sortByReliabilityHelp": "__MISSING__:Order providers by the share of calls that succeeded in the selected window. Providers with too few calls to state a rate keep their score order, below the measured ones.",
     "typeLegend": "من دون تسجيل = لا إعداد مطلوب · تسجيل دخول OAuth = سجّل الدخول بحسابك · مفتاح API = استخدم مفتاحك أو الخطة المجانية للمزود",
     "colReliability": "__MISSING__:Reliability",
     "colReliabilityHelp": "__MISSING__:Share of calls this provider actually answered over the last 24 hours. A dash means it served too little traffic to state a rate.",
@@ -13370,6 +13377,9 @@
     "colContext": "السياق",
     "colCapabilities": "القدرات",
     "colTos": "مخاطر شروط الخدمة",
+    "colLimits": "__MISSING__:Rate limits",
+    "trainsOnPrompts": "__MISSING__:Trains on prompts",
+    "trainsOnPromptsHelp": "__MISSING__:This provider's terms state it may train on the prompts you send. Models without this badge either state they do not, or do not document it — an absent statement is not a guarantee.",
     "newBadge": "جديد",
     "setupGuide": "دليل الإعداد",
     "disabledByFeed": "معطل بواسطة تغذية الرادار",

src/i18n/messages/az.json

@@ -6448,7 +6448,12 @@
     "savedModelEndpointSettings": "Saxlanmış model son nöqtəsi parametrləri",
     "searchByModelAria": "Model üzrə axtarış edin",
     "selectSupportedEndpoint": "Ən azı bir dəstəklənən son nöqtəni seçin",
-    "antigravityClientProfileHarness": "Harness / CLI"
+    "antigravityClientProfileHarness": "Harness / CLI",
+    "clearConnectionCooldown": "Clear cooldown",
+    "clearConnectionCooldownInProgress": "Clearing…",
+    "clearConnectionCooldownTitle": "Clear the cooldown now — use when the quota has already refreshed upstream",
+    "connectionCooldownCleared": "Cooldown cleared — connection rejoined routing",
+    "failedClearConnectionCooldown": "Failed to clear cooldown"
   },
   "settings": {
     "title": "Settings",
@@ -12531,6 +12536,8 @@
     "typeApikey": "API açarı",
     "sortTypeFirst": "Əvvəlcə ən asanlar",
     "sortTypeFirstHelp": "Qeydiyyat çətinliyinə görə qruplaşdırın (Qeydiyyatsız → OAuth ilə giriş → API açarı), hər qrup daxilində keyfiyyət ardıcıllığını qoruyun",
+    "sortByReliability": "__MISSING__:Most reliable first",
+    "sortByReliabilityHelp": "__MISSING__:Order providers by the share of calls that succeeded in the selected window. Providers with too few calls to state a rate keep their score order, below the measured ones.",
     "typeLegend": "Qeydiyyatsız = sıfır quraşdırma · OAuth ilə giriş = öz hesabınızla daxil olun · API açarı = öz açarınızı gətirin və ya həmin provayderin pulsuz paketindən istifadə edin",
     "colReliability": "__MISSING__:Reliability",
     "colReliabilityHelp": "__MISSING__:Share of calls this provider actually answered over the last 24 hours. A dash means it served too little traffic to state a rate.",
@@ -13370,6 +13377,9 @@
     "colContext": "Kontekst",
     "colCapabilities": "İmkanlar",
     "colTos": "ToS Riski",
+    "colLimits": "__MISSING__:Rate limits",
+    "trainsOnPrompts": "__MISSING__:Trains on prompts",
+    "trainsOnPromptsHelp": "__MISSING__:This provider's terms state it may train on the prompts you send. Models without this badge either state they do not, or do not document it — an absent statement is not a guarantee.",
     "newBadge": "yeni",
     "setupGuide": "Quraşdırma bələdçisi",
     "disabledByFeed": "Radar axını tərəfindən deaktiv edilmişdir",

src/i18n/messages/bg.json

@@ -6448,7 +6448,12 @@
     "savedModelEndpointSettings": "Настройки на крайна точка на запазен модел",
     "searchByModelAria": "Търсене по модел",
     "selectSupportedEndpoint": "Изберете поне една поддържана крайна точка",
-    "antigravityClientProfileHarness": "Harness / CLI"
+    "antigravityClientProfileHarness": "Harness / CLI",
+    "clearConnectionCooldown": "Clear cooldown",
+    "clearConnectionCooldownInProgress": "Clearing…",
+    "clearConnectionCooldownTitle": "Clear the cooldown now — use when the quota has already refreshed upstream",
+    "connectionCooldownCleared": "Cooldown cleared — connection rejoined routing",
+    "failedClearConnectionCooldown": "Failed to clear cooldown"
   },
   "settings": {
     "title": "Настройки",
@@ -12531,6 +12536,8 @@
     "typeApikey": "API ключ",
     "sortTypeFirst": "Най-лесните първо",
     "sortTypeFirstHelp": "Групиране по усилие за регистрация (Без регистрация → Вход с OAuth → API ключ), запазвайки подредбата по качество във всяка група",
+    "sortByReliability": "__MISSING__:Most reliable first",
+    "sortByReliabilityHelp": "__MISSING__:Order providers by the share of calls that succeeded in the selected window. Providers with too few calls to state a rate keep their score order, below the measured ones.",
     "typeLegend": "Без регистрация = нулева настройка · Вход с OAuth = влизане със собствен акаунт · API ключ = използване на собствен ключ или безплатния план на доставчика",
     "colReliability": "__MISSING__:Reliability",
     "colReliabilityHelp": "__MISSING__:Share of calls this provider actually answered over the last 24 hours. A dash means it served too little traffic to state a rate.",
@@ -13370,6 +13377,9 @@
     "colContext": "Контекст",
     "colCapabilities": "Възможности",
     "colTos": "Риск от ToS",
+    "colLimits": "__MISSING__:Rate limits",
+    "trainsOnPrompts": "__MISSING__:Trains on prompts",
+    "trainsOnPromptsHelp": "__MISSING__:This provider's terms state it may train on the prompts you send. Models without this badge either state they do not, or do not document it — an absent statement is not a guarantee.",
     "newBadge": "нов",
     "setupGuide": "Ръководство за настройка",
     "disabledByFeed": "Деактивирано от фийда на радара",