feat(services): add llmlingua-server supervisor installer and HTTP compression integration

.env.example

@@ -2506,15 +2506,16 @@ APP_LOG_TO_FILE=true
 # DARIO_HOST=127.0.0.1
 # DARIO_PORT=3456
 
-# ── Dario embedded service ──
-# Override the host/port the embedded Dario (Claude Code subscription proxy)
-# daemon binds to and is reached at. Always bound to 127.0.0.1 — never
-# configurable to 0.0.0.0. Rarely needed — defaults to 127.0.0.1:3456.
-# Used by: src/lib/services/installers/dario.ts, src/lib/services/bootstrap.ts,
-# src/app/api/services/dario/_lib.ts, src/app/api/services/dario/admin/_lib.ts,
-# open-sse/executors/dario.ts
-# DARIO_HOST=127.0.0.1
-# DARIO_PORT=3456
+# ── LLMLingua embedded service ──
+# Override the port and base URL for the embedded LLMLingua-2 prompt compression
+# sidecar daemon. Always bound to 127.0.0.1 — never configurable to 0.0.0.0.
+# Defaults to port 20135 and http://127.0.0.1:20135.
+# Used by: src/lib/services/bootstrap.ts, src/lib/services/installers/llmlingua.ts,
+# open-sse/services/compression/engines/llmlingua/index.ts
+# LLMLINGUA_PORT=20135
+# LLMLINGUA_BASE_URL=http://127.0.0.1:20135
+# Hugging Face model id the sidecar loads (LLMLingua-2 JS port).
+# LLMLINGUA_MODEL=atjsh/llmlingua-2-js-tinybert-meetingbank
 
 # ── Local hostnames (Docker networking) ──
 # Comma-separated additional hostnames treated as "local" for provider routing.

changelog.d/features/12967-llmlingua-server.md

@@ -0,0 +1 @@
+- **feat(services):** Added an LLMLingua embedded-service supervisor + HTTP sidecar wrapping the real `@atjsh/llmlingua-2` prompt-compression package (ONNX TinyBERT model, downloaded from Hugging Face on first use). The compression engine dispatches `/compress` calls to the sidecar over HTTP, falling back to the existing in-process worker-thread backend when the sidecar is down.

config/quality/gate-manifest.json

@@ -613,6 +613,11 @@
       "command": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --import tsx/esm --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test tests/unit/security-fase01.test.ts",
       "disposition": "separately-invoked"
     },
+    {
+      "name": "test:services:int",
+      "command": "cross-env RUN_SERVICES_INT=1 DISABLE_SQLITE_AUTO_BACKUP=true node --import tsx/esm --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit --test-concurrency=1 \"tests/integration/services/*.test.ts\"",
+      "disposition": "separately-invoked"
+    },
     {
       "name": "test:system",
       "command": "cross-env DISABLE_SQLITE_AUTO_BACKUP=true node --import tsx/esm --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit --test-concurrency=1 tests/e2e/system-failover.test.ts",

config/quality/test-discovery-baseline.json

@@ -1,15 +1,12 @@
 {
   "_comment": "Catraca de test-discovery (check-test-discovery.mjs). Cada entrada e um arquivo de teste que NENHUM runner coleta (ele nunca roda) — divida congelada na auditoria 6A.1 (2026-06-09; 195 originais, 135 religados no node runner em 6A.1c). So pode DIMINUIR: religue o teste (ajustando o glob do runner ou movendo o arquivo) e remova a entrada via --update. NAO adicione novos orfaos — corrija o runner.",
-  "_remaining_10": "10 orfaos restantes: 4 golden-set + 1 benchmark + 1 teste live + 1 stress (deliberadamente manuais — decidir runner/gating); 3 integration/services (gated RUN_SERVICES_INT=1, sem runner CI).",
+  "_remaining_6": "6 orfaos restantes: 4 golden-set + 1 benchmark + 1 teste live (deliberadamente manuais — decidir runner/gating). Os 3 integration/services foram religados por test:services:int (#12967).",
   "orphans": [
     "tests/benchmarks/pipeline-accuracy.test.ts",
     "tests/golden-set/compression-caveman-v2.test.ts",
     "tests/golden-set/compression-quality.test.ts",
     "tests/golden-set/compression-savings.test.ts",
     "tests/golden-set/compression-upstream-parity.test.ts",
-    "tests/integration/services/cliproxy-coexistence.test.ts",
-    "tests/integration/services/full-lifecycle.int.test.ts",
-    "tests/integration/services/route-guard-services.int.test.ts",
     "tests/live/deepseek-web-live.test.ts"
   ]
 }

docs/frameworks/EMBEDDED-SERVICES.md

@@ -1,13 +1,13 @@
 ---
 title: "Embedded Services"
-description: "Reference for 9Router, CLIProxyAPI, Mux, Bifrost, and open-wa"
+description: "Reference for 9Router, CLIProxyAPI, Mux, Bifrost, open-wa, and LLMLingua"
 ---
 
 # Embedded Services
 
 > **Version:** v3.8.44
-> **Last updated:** 2026-09-09
-> **Audience:** Engineers adding, maintaining, or debugging embedded services (9Router, CLIProxyAPI, Mux, Bifrost, open-wa).
+> **Last updated:** 2026-09-16
+> **Audience:** Engineers adding, maintaining, or debugging embedded services (9Router, CLIProxyAPI, Mux, Bifrost, open-wa, LLMLingua).
 
 Embedded services are locally-installed process sidecar tools that OmniRoute installs, supervises, and
 exposes as first-class routing targets. Unlike external providers (which are reached over the internet
@@ -32,7 +32,7 @@ via API keys), embedded services run on the same machine as OmniRoute and commun
 
 ### Why embedded services?
 
-Six services are embedded:
+Seven services are embedded:
 
 | Service         | npm package                        | Default port | Purpose                                                                                                                                                                                |
 | --------------- | ---------------------------------- | :----------: | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
@@ -42,8 +42,9 @@ Six services are embedded:
 | **Bifrost**     | `@maximhq/bifrost`                 |     8080     | Go AI-gateway relay backend. When running, auto-selected by the relay route (`/v1/relay/`)                                                                                             |
 | **Dario**       | `@askalf/dario`                    |     3456     | Claude-subscription proxy — alternative/failover to CLIProxyAPI for Claude-Code-shaped traffic; the injected key becomes `DARIO_ADMIN_TOKEN` gating its `/admin/*` OAuth control plane |
 | **open-wa**     | `@open-wa/wa-automate`             |     8323     | WhatsApp Web automation (headless Chromium via Puppeteer). Lifecycle-managed only — not a routing target.                                                                              |
+| **LLMLingua**   | `@atjsh/llmlingua-2`               |    20135     | Prompt-compression sidecar — real LLMLingua-2 ONNX model (JS/TS port of Microsoft's algorithm). `open-sse/services/compression/engines/llmlingua/index.ts` dispatches `/compress` over HTTP to it, falling back to the in-process worker-thread backend when the sidecar is down. Lifecycle-managed only — not a routing target. |
 
-All six follow the same supervisory model:
+All seven follow the same supervisory model:
 
 - OmniRoute installs them under `DATA_DIR/services/{name}/` (isolated from OmniRoute's own `package.json`)
 - OmniRoute spawns and monitors them as child processes
@@ -153,6 +154,7 @@ All six follow the same supervisory model:
 | `src/lib/services/installers/cliproxy.ts`   | npm install/update/uninstall for CLIProxyAPI     |
 | `src/lib/services/installers/mux.ts`        | npm install/update/uninstall for Mux             |
 | `src/lib/services/installers/openwa.ts`     | npm install/update/uninstall for open-wa         |
+| `src/lib/services/installers/llmlingua.ts`   | npm install/update/uninstall for LLMLingua       |
 | `src/app/api/services/9router/_lib.ts`      | `getOrInitSupervisor()` helper                   |
 | `src/app/api/services/[name]/logs/route.ts` | Shared SSE logs endpoint                         |
 | `open-sse/executors/ninerouter.ts`          | Provider executor (Layer 4)                      |
@@ -540,7 +542,39 @@ in this integration yet.
 
 ---
 
-### 4.7 Reverse proxy (9Router dashboard embed)
+### 4.7 LLMLingua endpoints (8 routes)
+
+LLMLingua is a prompt-compression sidecar wrapping `@atjsh/llmlingua-2` (real
+ONNX token-classification model, downloaded from Hugging Face on first
+`/compress` call). It uses the same endpoint shape as Bifrost (no API key —
+`needsApiKey: false`, it never handles credentials).
+
+| Method | Path                                           | Description                                                               |
+| ------ | ---------------------------------------------- | ------------------------------------------------------------------------- |
+| `POST` | `/api/services/llmlingua/install`              | npm-install `@atjsh/llmlingua-2` + peers, write the sidecar server script |
+| `POST` | `/api/services/llmlingua/start`                | Start the sidecar on port 20135 (default)                                 |
+| `POST` | `/api/services/llmlingua/stop`                 | Stop the sidecar                                                          |
+| `POST` | `/api/services/llmlingua/restart`              | Restart the sidecar                                                       |
+| `POST` | `/api/services/llmlingua/update`               | Update to the newer package version                                       |
+| `GET`  | `/api/services/llmlingua/status`               | Live + DB status                                                          |
+| `POST` | `/api/services/llmlingua/auto-start`           | Toggle auto-start                                                         |
+| `POST` | `/api/services/llmlingua/auto-restart-adopted` | Toggle auto-restart of an adopted (pre-existing) instance                 |
+| `GET`  | `/api/services/llmlingua/logs`                 | SSE log tail (via shared `[name]/logs` dynamic route)                     |
+
+**Sidecar contract:** the server script exposes `GET /health` (instant — does
+not wait on the model) and `POST /compress` (`{ text, rate }` →
+`{ text, compressed, ratio }`). The model loads lazily on the first
+`/compress` call.
+
+**Compression wiring:** `open-sse/services/compression/engines/llmlingua/index.ts`'s
+`httpSidecarBackend` calls `LLMLINGUA_BASE_URL` (default
+`http://127.0.0.1:20135`) and only accepts the sidecar's response when it is
+strictly shorter than the input; any failure (not running, timeout, no-op
+response) falls back to the in-process worker-thread backend (`./worker.ts`).
+
+---
+
+### 4.8 Reverse proxy (9Router dashboard embed)
 
 The dashboard embeds the 9Router web UI inside an iframe via an internal reverse
 proxy at:

docs/openapi.yaml

@@ -5930,6 +5930,145 @@ paths:
         "500":
           description: Import failed
 
+  /api/services/llmlingua/install:
+    post:
+      tags: [Embedded Services]
+      summary: Install LLMLingua
+      description: >-
+        Installs `@atjsh/llmlingua-2` (real LLMLingua-2 ONNX prompt-compression
+        package) plus its `@huggingface/transformers`/`js-tiktoken` peers under
+        DATA_DIR/services/llmlingua/, and writes the sidecar server script.
+        Accepts an optional `version` field (semver or `latest`).
+        **LOCAL_ONLY** — loopback only.
+      requestBody:
+        required: false
+        content:
+          application/json:
+            schema:
+              type: object
+              properties:
+                version:
+                  type: string
+                  default: latest
+      responses:
+        "200":
+          description: Installation result
+          content:
+            application/json:
+              schema:
+                type: object
+                properties:
+                  ok:
+                    type: boolean
+                  installedVersion:
+                    type: string
+                  installPath:
+                    type: string
+                  durationMs:
+                    type: number
+
+  /api/services/llmlingua/start:
+    post:
+      tags: [Embedded Services]
+      summary: Start LLMLingua
+      description: Starts the supervised LLMLingua sidecar process. **LOCAL_ONLY** — loopback only.
+      responses:
+        "200":
+          description: Service status after start
+        "409":
+          description: LLMLingua is not installed
+
+  /api/services/llmlingua/stop:
+    post:
+      tags: [Embedded Services]
+      summary: Stop LLMLingua
+      description: Stops the supervised LLMLingua sidecar process. **LOCAL_ONLY** — loopback only.
+      responses:
+        "200":
+          description: Service status after stop
+
+  /api/services/llmlingua/restart:
+    post:
+      tags: [Embedded Services]
+      summary: Restart LLMLingua
+      description: Restarts the supervised LLMLingua sidecar process. **LOCAL_ONLY** — loopback only.
+      responses:
+        "200":
+          description: Service status after restart
+        "409":
+          description: LLMLingua is not installed
+
+  /api/services/llmlingua/update:
+    post:
+      tags: [Embedded Services]
+      summary: Update LLMLingua
+      description: >-
+        Updates LLMLingua to the latest npm version. Stops the running process,
+        installs the new version, and restarts if it was previously running.
+        **LOCAL_ONLY** — loopback only.
+      responses:
+        "200":
+          description: Update result
+
+  /api/services/llmlingua/status:
+    get:
+      tags: [Embedded Services]
+      summary: Get LLMLingua status
+      description: Returns live and DB status for the supervised LLMLingua service. **LOCAL_ONLY** — loopback only.
+      responses:
+        "200":
+          description: LLMLingua service status
+
+  /api/services/llmlingua/auto-start:
+    post:
+      tags: [Embedded Services]
+      summary: Toggle LLMLingua auto-start
+      description: >-
+        When enabled, LLMLingua starts automatically on the next OmniRoute boot.
+        **LOCAL_ONLY** — loopback only.
+      requestBody:
+        required: true
+        content:
+          application/json:
+            schema:
+              type: object
+              required: [enabled]
+              properties:
+                enabled:
+                  type: boolean
+      responses:
+        "204":
+          description: Auto-start flag updated
+        "400":
+          description: Invalid request body
+
+  /api/services/llmlingua/auto-restart-adopted:
+    post:
+      tags: [Embedded Services]
+      summary: Toggle LLMLingua auto-restart-when-adopted
+      description: >-
+        When enabled, an externally-adopted (not OmniRoute-spawned) LLMLingua
+        process is restarted under OmniRoute's own supervisor on the next
+        health-check cycle instead of being left as adopted-only.
+        **LOCAL_ONLY** — loopback only.
+      requestBody:
+        required: true
+        content:
+          application/json:
+            schema:
+              type: object
+              required: [enabled]
+              properties:
+                enabled:
+                  type: boolean
+      responses:
+        "204":
+          description: Flag updated
+        "400":
+          description: Invalid request body
+        "500":
+          description: Update failed
+
   /api/services/{name}/logs:
     get:
       tags: [Embedded Services]
@@ -5945,7 +6084,7 @@ paths:
           required: true
           schema:
             type: string
-            enum: [9router, cliproxy]
+            enum: [9router, cliproxy, mux, bifrost, dario, openwa, llmlingua]
         - name: tail
           in: query
           schema:

docs/reference/ENVIRONMENT.md

@@ -1525,6 +1525,9 @@ Provider quota endpoints, network tunnels (Tailscale, Ngrok, MITM debug proxy),
 | `BIFROST_ENABLED`                           | `1`                                                                         | `src/app/api/v1/relay/chat/completions/bifrost/route.ts`                  | Master kill switch for the bifrost sidecar proxy. When set to `0`, the route returns 503 with the `X-Bifrost-Killswitch` header and the operator is bounced to the TS path. Use to disable the sidecar without redeploying (tier-1 router incident, key rotation).                                                                                                    |
 | `BIFROST_BASE_URL`                          | _(unset)_                                                                   | `src/app/api/v1/relay/chat/completions/bifrost/route.ts`                  | When set, the Bifrost sidecar proxy route forwards `/v1/chat/completions` traffic to this Go gateway instead of the TS relay handler. Unset → 503-with-fallback. Trailing slash is stripped.                                                                                                                                                                          |
 | `BIFROST_PORT`                              | `8080`                                                                      | `src/lib/services/bootstrap.ts`                                          | Port the supervised Bifrost embedded service binds to (`127.0.0.1:<port>`) when OmniRoute manages the Bifrost sidecar lifecycle. Defaults to `8080`.                                                                                                                                                                                                                  |
+| `LLMLINGUA_PORT`                            | `20135`                                                                     | `src/lib/services/bootstrap.ts`                                          | Port the supervised LLMLingua compression service binds to (`127.0.0.1:<port>`). Defaults to `20135`.                                                                                                                                                                                                                          |
+| `LLMLINGUA_BASE_URL`                        | `http://127.0.0.1:20135`                                                    | `open-sse/services/compression/engines/llmlingua/index.ts`                | Base URL for the local LLMLingua HTTP compression microservice daemon. Defaults to `http://127.0.0.1:20135`.                                                                                                                                                                                                                     |
+| `LLMLINGUA_MODEL` | `atjsh/llmlingua-2-js-tinybert-meetingbank` | `src/lib/services/installers/llmlingua.ts` | Hugging Face model id the supervised LLMLingua sidecar loads (passed to the child process env). |
 | `BIFROST_API_KEY`                           | _(unset)_                                                                   | `src/app/api/v1/relay/chat/completions/bifrost/route.ts`                  | API key for the Bifrost gateway (sent as `Authorization: Bearer ...`). If unset, the route expects the request to carry a valid OmniRoute API key; this key is for gateway-side auth only.                                                                                                                                                                            |
 | `BIFROST_STREAMING_ENABLED`                 | `true`                                                                      | `src/app/api/v1/relay/chat/completions/bifrost/route.ts`                  | When true, the Bifrost sidecar route streams responses back via SSE through the gateway rather than the TS streaming executor. Set to `0` to force non-streaming JSON responses through the gateway.                                                                                                                                                                  |
 | `BIFROST_TIMEOUT_MS`                        | `30000`                                                                     | `src/app/api/v1/relay/chat/completions/bifrost/route.ts`                  | Per-request timeout when proxying to the Bifrost gateway (ms). On timeout the route returns the TS relay path via the `X-Bifrost-Fallback` header.                                                                                                                                                                                                                    |