fix(model): skip stale unload command

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
fix(worker): port final segment cleanup
2026-05-12 01:08:35 +02:00 · 2026-05-12 00:10:32 +02:00 · 2026-05-11 22:46:38 +02:00 · 2026-05-10 16:28:03 +02:00 · 2026-05-08 23:47:13 +02:00 · 2026-05-08 18:20:52 +02:00
20 changed files with 4130 additions and 370 deletions
--- a/.gitea/workflows/docker-build.yml
+++ b/.gitea/workflows/docker-build.yml
@@ -18,7 +18,30 @@ env:
  UBUNTU_VERSION: ${{ vars.UBUNTU_VERSION || '22.04' }}

 jobs:
+  test:
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Run unit tests
+        uses: docker/build-push-action@v6
+        with:
+          context: .
+          file: ./Dockerfile
+          target: tester
+          push: false
+          build-args: |
+            CUDA_VERSION=${{ env.CUDA_VERSION }}
+            UBUNTU_VERSION=${{ env.UBUNTU_VERSION }}
+          cache-from: type=registry,ref=${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:buildcache
+
  build-and-push:
+    needs: test
    runs-on: ubuntu-latest

    steps:
--- a/21
+++ b/21
@@ -82,6 +82,27 @@ RUN --mount=type=cache,target=/usr/local/cargo/registry \
    && cp target/release/whisper-server /usr/local/bin/whisper-server


+# ╔══════════════════════════════════════════════════════════╗
+# ║  STAGE 1b — tester                                       ║
+# ║  Runs unit tests against the release build artifacts     ║
+# ║  Uses CUDA stubs so tests run without a physical GPU     ║
+# ║                                                          ║
+# ║  Usage:                                                  ║
+# ║    docker build --target tester .                        ║
+# ╚══════════════════════════════════════════════════════════╝
+FROM builder AS tester
+
+# libcuda.so.1 stub — satisfies the dynamic linker without a real driver
+RUN ln -sf /usr/local/cuda/lib64/stubs/libcuda.so \
+           /usr/local/cuda/lib64/stubs/libcuda.so.1
+
+# Reuse the same cache mounts so no recompilation is needed
+RUN --mount=type=cache,target=/usr/local/cargo/registry \
+    --mount=type=cache,target=/build/target \
+    LD_LIBRARY_PATH=/usr/local/cuda/lib64/stubs \
+    cargo test --release
+
+
 # ╔══════════════════════════════════════════════════════════╗
 # ║  STAGE 2 — runtime                                       ║
 # ║  Minimal CUDA runtime image — no build tools             ║
--- a/KNOWLEDGE.md
+++ b/KNOWLEDGE.md
@@ -15,7 +15,27 @@ Model: ggml-large-v3, chunking at 60s on silence boundaries

 ---

-## Critical Bugs Found & Fixed
+## Cold GPU Warmup — First Job Returns 0 Segments in ~0.5s
+
+**Severity: Critical (production issue, intermittent, hard to diagnose)**
+
+**Symptom:** After a container restart, the very first submitted job completes in ~0.5 seconds and returns 0 segments. Subsequent jobs work correctly.
+
+**Root cause:** CUDA JIT-compiles its kernels on the **first** call to `whisper_full_with_state`. On a cold GPU, this compilation happens mid-inference and blocks/disrupts the decode pipeline, causing whisper to return immediately with 0 segments.
+
+**Why language detection can still succeed:** Language detection uses only a small mel-spectrogram + encoder pass on the first 30 seconds of audio. Some of these kernels may already be compiled or cached from a prior session. The full decoder kernels (the heavier ones) are what get JIT-compiled on the first full inference.
+
+**Fix:** In `Transcriber::load()`, after creating the state, run a 1-second silent inference pass:
+```rust
+let silence = vec![0.0f32; 16_000]; // 1s @ 16 kHz
+let mut wp = FullParams::new(SamplingStrategy::Greedy { best_of: 1 });
+wp.set_language(Some("en"));
+let _ = state.full(wp, &silence); // forces CUDA JIT — 0 segments expected
+tracing::info!("GPU warmup complete");
+```
+This forces all CUDA kernel compilation at startup. The first real job then runs on fully compiled kernels. Startup takes a few seconds longer but every job is reliable.
+
+---

 ### `set_detect_language(true)` is NOT "auto-detect and transcribe"
 - `whisper.cpp` source: `if (params.detect_language) { return 0; }` — it exits immediately after language detection, returns 0 segments
@@ -38,7 +58,14 @@ Model: ggml-large-v3, chunking at 60s on silence boundaries
 - **Possible future fix**: post-process to collapse consecutive identical segments (user declined this for now — raw output only)
 - `compression_ratio_thold` may also help but wasn't tested

-### 2. Five significant content gaps (~1600 words total)
+### 4. Cold GPU: first job returns 0 segments in ~0.5s (intermittent, after container restart)
+
+CUDA JIT-compiles kernels on the first call to `whisper_full_with_state`. On a cold GPU this compilation blocks/disrupts the decode pipeline mid-inference, causing an immediate return with 0 segments.
+
+**Fix**: Run a 1-second silent warmup inference in `Transcriber::load()`. This forces JIT compilation at startup so the first real job runs on fully compiled kernels.
+
+---
+
 - Largest: 439 words at ~68 min, 328 words at ~80 min, then 3 × ~293-250 word gaps
 - These are chunks where whisper produced off-topic or repetitive output instead of real content
 - Likely caused by: speaker overlap, audience noise, or poor audio quality in those windows
--- a/docs/FINDINGS.md
+++ b/docs/FINDINGS.md
@@ -4,7 +4,39 @@ This document records all non-obvious behaviour, surprising bugs, hardware quirk

 ---

-## whisper.cpp
+### Cold GPU: first job returns 0 segments in ~0.5s after container restart
+
+**Symptom:** After container restart, the first submitted job completes in ~0.5s and returns 0 segments. Language is detected correctly. All subsequent jobs work fine.
+
+**Root cause:** CUDA JIT-compiles its device kernels on the first call to `whisper_full_with_state`. On a cold GPU, this compilation happens synchronously mid-inference and disrupts the decode pipeline, causing it to return immediately with 0 results.
+
+**Why subsequent jobs are fine:** Compiled kernels are cached in the CUDA driver for the lifetime of the process. Once the first (warmup) call completes, all further calls use the cached compiled kernels.
+
+**Why language detection can succeed on the same call:** Language detection uses a mel-spectrogram + encoder pass on the first 30s of audio. These lighter kernels may compile faster or be partially cached, while the full decoder kernels (the heavier path) are what causes the failure.
+
+**Fix (in `Transcriber::load()`):**
+```rust
+let silence = vec![0.0f32; 16_000]; // 1s @ 16 kHz — just enough to trigger kernel compilation
+let mut wp = FullParams::new(SamplingStrategy::Greedy { best_of: 1 });
+wp.set_language(Some("en"));
+wp.set_print_progress(false);
+let _ = state.full(wp, &silence); // 0 segments expected; side-effect is the goal
+tracing::info!("GPU warmup complete");
+```
+
+**Also fixed simultaneously:** `create_state()` was called per-chunk (~700 MB GPU allocation each time), causing VRAM churn under concurrent processes. State is now created once and reused. See `WhisperState` reuse section above.
+
+---
+
+### `language=auto` is not a valid API parameter
+
+Passing `language=auto` in the multipart form is silently incorrect. The `language` field expects an ISO 639-1 code (e.g. `en`, `fr`) or should be **omitted entirely** for auto-detection. Passing "auto" causes whisper-rs to pass the string "auto" as a language code, which whisper.cpp does not recognise and may fallback in undefined ways.
+
+**Correct usage:**
+- Auto-detect: omit the `language` field entirely
+- Explicit: `language=en`
+
+---

 ### `detect_language=true` is a language-ID-only mode — NOT "auto-detect and transcribe"

--- a/docs/FRONTEND_INTEGRATION.md
+++ b/docs/FRONTEND_INTEGRATION.md
@@ -0,0 +1,941 @@
+# Frontend Integration Guide
+
+> **Audience:** Frontend / full-stack developers integrating the whisper transcription API into a web application.  
+> **Base URL:** `http://your-server:8080` (configurable via the `PORT` env var on the server).  
+> **Interactive docs:** `http://your-server:8080/docs` (Swagger UI — try every endpoint live).
+
+---
+
+## Table of Contents
+
+1. [Architecture Overview](#1-architecture-overview)
+2. [Quick Start — submit and poll](#2-quick-start--submit-and-poll)
+3. [Model Lifecycle](#3-model-lifecycle)
+   - 3.1 [State machine](#31-state-machine)
+   - 3.2 [GET /model/status](#32-get-modelstatus)
+   - 3.3 [POST /model/load](#33-post-modelload)
+   - 3.4 [POST /model/unload](#34-post-modelunload)
+   - 3.5 [GET /model/events (SSE)](#35-get-modelevents-sse)
+4. [Submitting Jobs](#4-submitting-jobs)
+   - 4.1 [POST /jobs](#41-post-jobs)
+   - 4.2 [Handling 503 Model Not Ready](#42-handling-503-model-not-ready)
+   - 4.3 [Retry pattern with auto-load](#43-retry-pattern-with-auto-load)
+5. [Tracking Job Progress](#5-tracking-job-progress)
+   - 5.1 [GET /jobs/:id (poll)](#51-get-jobsid-poll)
+   - 5.2 [GET /jobs/:id/stream (SSE)](#52-get-jobsidstream-sse)
+6. [Webhooks](#6-webhooks)
+   - 6.1 [Job completion webhook](#61-job-completion-webhook)
+   - 6.2 [Model lifecycle webhooks](#62-model-lifecycle-webhooks)
+7. [Health Check](#7-health-check)
+8. [Cancelling Jobs](#8-cancelling-jobs)
+9. [TypeScript Types](#9-typescript-types)
+10. [React Hooks](#10-react-hooks)
+11. [Complete Integration Example](#11-complete-integration-example)
+12. [Error Reference](#12-error-reference)
+
+---
+
+## 1. Architecture Overview
+
+```
+┌─────────────────────────────────────────────────────────┐
+│  whisper-server                                         │
+│                                                         │
+│  HTTP / SSE          Worker thread (GPU)                │
+│  ────────────        ───────────────────                │
+│  POST /jobs    ───►  job queue (FIFO)                   │
+│  GET  /jobs/:id      ↕                                  │
+│  GET  /jobs/:id/stream ◄── progress broadcast           │
+│                                                         │
+│  POST /model/load  ─►  load whisper into VRAM           │
+│  POST /model/unload ►  free VRAM                        │
+│  GET  /model/status    read state                       │
+│  GET  /model/events ◄── lifecycle SSE broadcast         │
+└─────────────────────────────────────────────────────────┘
+```
+
+**Key behaviours to understand before building:**
+
+- The model starts **unloaded** on every server restart. No inference is possible until it loads (~15–25 seconds for large-v3 on an RTX 2080).
+- Submitting a job when the model is not ready returns `503` with a `Retry-After` header **and automatically triggers a load**. You can retry the submission; no separate load call is needed.
+- The worker processes jobs **sequentially** (one at a time). Queue depth is visible via `/health`.
+- Long audio is split into silence-bounded chunks internally. SSE `progress` events reflect chunk completion, not raw GPU progress.
+
+---
+
+## 2. Quick Start — submit and poll
+
+The simplest possible integration — no SSE, no model management, just submit and poll:
+
+```typescript
+const BASE = 'http://your-server:8080';
+
+async function transcribe(audioBlob: Blob): Promise<Job> {
+  // 1. Submit
+  const form = new FormData();
+  form.append('audio', audioBlob, 'audio.wav');
+
+  let submitResp = await fetch(`${BASE}/jobs`, { method: 'POST', body: form });
+
+  // 2. If model isn't loaded yet, keep retrying until it is
+  while (submitResp.status === 503) {
+    const retryAfter = parseInt(submitResp.headers.get('Retry-After') ?? '15');
+    await sleep(retryAfter * 1000);
+    submitResp = await fetch(`${BASE}/jobs`, { method: 'POST', body: form });
+  }
+  if (!submitResp.ok) throw new Error(`Submit failed: ${submitResp.status}`);
+
+  const { job_id } = await submitResp.json();
+
+  // 3. Poll until done
+  while (true) {
+    await sleep(2000);
+    const job: Job = await fetch(`${BASE}/jobs/${job_id}`).then(r => r.json());
+    if (job.status === 'done')      return job;
+    if (job.status === 'failed')    throw new Error(job.error ?? 'transcription failed');
+    if (job.status === 'cancelled') throw new Error('job was cancelled');
+  }
+}
+
+const sleep = (ms: number) => new Promise(r => setTimeout(r, ms));
+```
+
+> For a better UX — real-time progress bar, model state indicator — read the full sections below.
+
+---
+
+## 3. Model Lifecycle
+
+### 3.1 State machine
+
+The model moves through four states:
+
+```
+          job submit
+         or POST /model/load
+               │
+   ┌──────────▼───────────┐
+   │       Unloaded        │◄──────────────────────────┐
+   └──────────┬───────────┘                            │
+              │ load triggered                         │
+   ┌──────────▼───────────┐                            │
+   │        Loading        │                            │ idle timeout
+   └──┬──────────────┬────┘                            │ or POST /model/unload
+      │ success      │ VRAM full                       │
+      │              │                                  │
+   ┌──▼────┐  ┌──────▼────────────────┐                │
+   │ Ready │  │   WaitingForGpu       │────────────────►│
+   └──┬────┘  └──────────────┬────────┘                │
+      │         retry ok ────┘                         │
+      └────────────────────────────────────────────────►┘
+```
+
+| State | `state` value | Can accept jobs? |
+|-------|--------------|-----------------|
+| Unloaded | `"unloaded"` | ❌ → triggers load, returns 503 |
+| Loading | `"loading"` | ❌ → returns 503 |
+| Waiting for GPU | `"waiting_for_gpu"` | ❌ → returns 503 |
+| Ready | `"ready"` | ✅ |
+
+---
+
+### 3.2 `GET /model/status`
+
+Returns the current model state and live VRAM figures (from `nvidia-smi`).
+
+**Unloaded:**
+```json
+{ "state": "unloaded" }
+```
+
+**Loading:**
+```json
+{ "state": "loading" }
+```
+
+**Waiting for GPU (VRAM contention):**
+```json
+{
+  "state": "waiting_for_gpu",
+  "vram_needed_mb": 3951,
+  "vram_free_mb": 512,
+  "retry_in_secs": 30
+}
+```
+
+**Ready:**
+```json
+{
+  "state": "ready",
+  "loaded_at": "2026-05-10T14:00:00.000Z",
+  "vram_used_mb": 4096,
+  "vram_total_mb": 8192
+}
+```
+
+> `vram_used_mb` / `vram_total_mb` are omitted when `nvidia-smi` is unavailable.
+
+---
+
+### 3.3 `POST /model/load`
+
+Tells the server to load the model. **Idempotent** — safe to call multiple times.
+
+```bash
+curl -X POST http://your-server:8080/model/load
+```
+
+**Responses:**
+
+| Status | Body | Meaning |
+|--------|------|---------|
+| 202 | `{"status":"load_initiated"}` | Load queued |
+| 200 | `{"status":"already_ready"}` | Already loaded |
+
+The load happens asynchronously. Subscribe to `/model/events` or poll `/model/status` to know when ready.
+
+---
+
+### 3.4 `POST /model/unload`
+
+Immediately frees the model from GPU memory. In-flight jobs finish first; the model is dropped after the current inference completes.
+
+```bash
+curl -X POST http://your-server:8080/model/unload
+```
+
+**Response:** `200 {"status":"unload_requested"}` (always, regardless of current state).
+
+> Use this if you know transcription won't happen for a while and you want to free VRAM for other workloads on the same GPU.
+
+---
+
+### 3.5 `GET /model/events` (SSE)
+
+A persistent Server-Sent Events stream that emits every model lifecycle transition.
+
+```bash
+curl -N http://your-server:8080/model/events
+```
+
+**Events emitted:**
+
+```
+event: model_loading
+data: {"type":"model_loading"}
+
+event: model_ready
+data: {"type":"model_ready","loaded_at":"2026-05-10T14:00:00.000Z"}
+
+event: model_unloaded
+data: {"type":"model_unloaded"}
+
+event: model_waiting_for_gpu
+data: {"type":"model_waiting_for_gpu","vram_needed_mb":3951,"vram_free_mb":512,"retry_in_secs":30}
+```
+
+**JavaScript:**
+```typescript
+function subscribeModelEvents(
+  onReady:       (loadedAt: string) => void,
+  onUnloaded:    () => void,
+  onLoading:     () => void,
+  onWaitingGpu:  (info: { vram_needed_mb: number; vram_free_mb: number; retry_in_secs: number }) => void,
+): () => void {
+  const es = new EventSource(`${BASE}/model/events`);
+
+  es.addEventListener('model_ready',          (e) => onReady(JSON.parse(e.data).loaded_at));
+  es.addEventListener('model_unloaded',       ()  => onUnloaded());
+  es.addEventListener('model_loading',        ()  => onLoading());
+  es.addEventListener('model_waiting_for_gpu',(e) => onWaitingGpu(JSON.parse(e.data)));
+
+  es.onerror = () => {
+    // The browser reconnects automatically with exponential backoff.
+    // Log the error but don't tear down the listener.
+    console.warn('model/events connection dropped, reconnecting…');
+  };
+
+  return () => es.close(); // call this to clean up (e.g. in React useEffect return)
+}
+```
+
+> The server sends an SSE keepalive comment every 15 seconds so proxies don't close idle connections.
+
+---
+
+## 4. Submitting Jobs
+
+### 4.1 `POST /jobs`
+
+**Content-Type:** `multipart/form-data`
+
+| Field | Required | Type | Notes |
+|-------|----------|------|-------|
+| `audio` | ✅ | file | Any format ffmpeg understands: WAV, MP3, M4A, OGG, FLAC, MP4, MKV … No size limit. |
+| `language` | ❌ | string | ISO 639-1 code (`"en"`, `"it"`, `"fr"` …). Omit for auto-detection. |
+| `task` | ❌ | string | `"transcribe"` (default) or `"translate"` (→ English) |
+| `webhook_url` | ❌ | string | URL to POST the completed job to. Also registers the URL for model lifecycle webhooks. |
+
+**202 Accepted:**
+```json
+{ "job_id": "550e8400-e29b-41d4-a716-446655440000" }
+```
+
+```typescript
+async function submitJob(
+  audio: Blob,
+  opts: { language?: string; task?: 'transcribe' | 'translate'; webhookUrl?: string } = {}
+): Promise<string> {
+  const form = new FormData();
+  form.append('audio', audio, 'audio.wav');
+  if (opts.language)   form.append('language', opts.language);
+  if (opts.task)       form.append('task', opts.task);
+  if (opts.webhookUrl) form.append('webhook_url', opts.webhookUrl);
+
+  const resp = await fetch(`${BASE}/jobs`, { method: 'POST', body: form });
+  if (!resp.ok) throw await toApiError(resp);
+
+  const { job_id } = await resp.json();
+  return job_id;
+}
+```
+
+---
+
+### 4.2 Handling 503 Model Not Ready
+
+When the model isn't loaded, `POST /jobs` returns:
+
+```
+HTTP/1.1 503 Service Unavailable
+Retry-After: 30
+Content-Type: application/json
+```
+```json
+{
+  "error": "model_not_ready",
+  "state": "unloaded",
+  "retry_after_secs": 30
+}
+```
+
+**`retry_after_secs` by state:**
+
+| `state` | `retry_after_secs` | Why |
+|---------|-------------------|-----|
+| `unloaded` | 30 | Load just triggered; RTX 2080 + large-v3 loads in ~15–25s |
+| `loading` | 10 | Already loading; check again soon |
+| `waiting_for_gpu` | `GPU_POLL_INTERVAL_SECS` (default 30) | VRAM busy; retry later |
+
+> **Submitting a job when the model is `unloaded` automatically triggers a load.** You do NOT need a separate `POST /model/load` call for the normal happy path.
+
+---
+
+### 4.3 Retry pattern with auto-load
+
+```typescript
+async function submitWithRetry(
+  audio: Blob,
+  opts: { language?: string; task?: 'transcribe' | 'translate'; webhookUrl?: string } = {},
+  maxAttempts = 20,
+): Promise<string> {
+  const form = new FormData();
+  form.append('audio', audio, 'audio.wav');
+  if (opts.language)   form.append('language', opts.language);
+  if (opts.task)       form.append('task', opts.task);
+  if (opts.webhookUrl) form.append('webhook_url', opts.webhookUrl);
+
+  for (let attempt = 1; attempt <= maxAttempts; attempt++) {
+    const resp = await fetch(`${BASE}/jobs`, { method: 'POST', body: form });
+
+    if (resp.status === 202) {
+      const { job_id } = await resp.json();
+      return job_id;
+    }
+
+    if (resp.status === 503) {
+      const body = await resp.json();
+      const waitMs = (parseInt(resp.headers.get('Retry-After') ?? '15') + 1) * 1000;
+      console.log(`Model ${body.state} — waiting ${waitMs / 1000}s (attempt ${attempt}/${maxAttempts})`);
+      await sleep(waitMs);
+      continue;
+    }
+
+    throw await toApiError(resp);
+  }
+
+  throw new Error(`Model did not become ready after ${maxAttempts} attempts`);
+}
+```
+
+> **Tip:** For a better UX, subscribe to `GET /model/events` and wait for the `model_ready` event instead of sleeping blindly — then submit immediately when ready.
+
+---
+
+## 5. Tracking Job Progress
+
+Two patterns: **SSE** (real-time push) or **polling** (simpler). SSE is preferred for UX.
+
+### 5.1 `GET /jobs/:id` (poll)
+
+Returns the full job document. Poll every 2–5 seconds while `status` is `queued` or `running`.
+
+```json
+{
+  "id": "550e8400-e29b-41d4-a716-446655440000",
+  "status": "running",
+  "task": "transcribe",
+  "language": "en",
+  "progress": 42,
+  "duration_secs": 120.5,
+  "segments": [],
+  "created_at": "2026-05-10T14:00:00.000Z"
+}
+```
+
+When `status === "done"`:
+```json
+{
+  "id": "550e8400-e29b-41d4-a716-446655440000",
+  "status": "done",
+  "task": "transcribe",
+  "language": "en",
+  "progress": 100,
+  "duration_secs": 120.5,
+  "segments": [
+    { "index": 0, "start": 0.0, "end": 3.5, "text": "Hello, world.", "words": [] },
+    { "index": 1, "start": 3.6, "end": 7.2, "text": "How are you?", "words": [] }
+  ],
+  "created_at": "2026-05-10T14:00:00.000Z",
+  "completed_at": "2026-05-10T14:02:35.000Z"
+}
+```
+
+**Terminal statuses:** `done`, `failed`, `cancelled` — stop polling when you see one.
+
+---
+
+### 5.2 `GET /jobs/:id/stream` (SSE)
+
+Subscribe immediately after submission. The connection is held open and events are pushed as they occur.
+
+**Event types:**
+
+```
+event: progress
+data: {"type":"progress","percent":42,"chunk":3,"chunks_total":7}
+
+event: done
+data: {"type":"done","job":{...full Job object...}}
+
+event: error
+data: {"type":"error","message":"whisper inference failed: ..."}
+```
+
+- `percent` — overall job progress 0–100 (derived from chunks completed / total).
+- `chunk` / `chunks_total` — the audio is split on silences; each chunk is one whisper inference call.
+- If you open the stream after the job is already finished, you immediately receive a single `done` event.
+
+```typescript
+function streamJobProgress(
+  jobId: string,
+  onProgress: (percent: number, chunk: number, total: number) => void,
+  onDone:     (job: Job) => void,
+  onError:    (message: string) => void,
+): () => void {
+  const es = new EventSource(`${BASE}/jobs/${jobId}/stream`);
+
+  es.addEventListener('progress', (e) => {
+    const { percent, chunk, chunks_total } = JSON.parse(e.data);
+    onProgress(percent, chunk, chunks_total);
+  });
+
+  es.addEventListener('done', (e) => {
+    const { job } = JSON.parse(e.data);
+    es.close();
+    onDone(job);
+  });
+
+  es.addEventListener('error', (e) => {
+    // SSE protocol error vs application error — check if data exists
+    if ('data' in e) {
+      const { message } = JSON.parse((e as MessageEvent).data);
+      onError(message);
+    }
+    es.close();
+  });
+
+  return () => es.close();
+}
+```
+
+> **Note:** Do not confuse the SSE `error` event (connection drop — no `data`) with the application `error` event (transcription failure — has `data`). The example above handles both.
+
+---
+
+## 6. Webhooks
+
+Webhooks are fired as HTTP `POST` requests with `Content-Type: application/json` to the `webhook_url` you supply at job submission. The server retries up to 3 times with exponential backoff (1s, 2s) on non-2xx responses.
+
+### 6.1 Job completion webhook
+
+Fired when a job reaches `done`, `failed`, or `cancelled`.  
+**Payload:** the full `Job` object (same as `GET /jobs/:id`).
+
+```json
+{
+  "id": "550e8400-e29b-41d4-a716-446655440000",
+  "status": "done",
+  "task": "transcribe",
+  "language": "en",
+  "progress": 100,
+  "duration_secs": 120.5,
+  "segments": [
+    { "index": 0, "start": 0.0, "end": 3.5, "text": "Hello, world.", "words": [] }
+  ],
+  "created_at": "2026-05-10T14:00:00.000Z",
+  "completed_at": "2026-05-10T14:02:35.000Z"
+}
+```
+
+### 6.2 Model lifecycle webhooks
+
+**Any URL that has ever appeared as a `webhook_url` in a job submission** also receives model lifecycle webhooks for the lifetime of the server process. This lets your backend know when the model comes up or goes down without polling.
+
+Only two events are delivered via webhook (the others are SSE-only):
+
+**Model ready:**
+```json
+{ "type": "model_ready", "loaded_at": "2026-05-10T14:00:00.000Z" }
+```
+
+**Model unloaded:**
+```json
+{ "type": "model_unloaded" }
+```
+
+**Express.js receiver example:**
+```typescript
+import express from 'express';
+const app = express();
+app.use(express.json());
+
+app.post('/webhooks/whisper', (req, res) => {
+  res.sendStatus(200); // acknowledge quickly — retries on non-2xx
+
+  const body = req.body;
+
+  if ('type' in body) {
+    // Model lifecycle event
+    if (body.type === 'model_ready') {
+      console.log('Whisper model ready at', body.loaded_at);
+    } else if (body.type === 'model_unloaded') {
+      console.log('Whisper model freed GPU memory');
+    }
+    return;
+  }
+
+  // Job completion event — body is a Job object
+  if (body.status === 'done') {
+    console.log(`Job ${body.id} done — ${body.segments.length} segments`);
+    processTranscript(body.segments);
+  } else if (body.status === 'failed') {
+    console.error(`Job ${body.id} failed:`, body.error);
+  }
+});
+```
+
+> **Distinguish job vs. model webhook:** Job payloads have an `id` and `status` field. Model payloads have a `type` field at the top level (`model_ready` / `model_unloaded`).
+
+---
+
+## 7. Health Check
+
+```bash
+curl http://your-server:8080/health
+```
+
+```json
+{
+  "status": "ok",
+  "gpu_name": "NVIDIA GeForce RTX 2080",
+  "vram_total_mb": 8192,
+  "model": "large-v3",
+  "queue_depth": 2,
+  "model_state": "ready"
+}
+```
+
+| Field | Notes |
+|-------|-------|
+| `status` | Always `"ok"` when the server is reachable |
+| `gpu_name` | From `nvidia-smi`; `null` if unavailable |
+| `vram_total_mb` | Total VRAM in MiB; `null` if unavailable |
+| `model` | Model name string (server config) |
+| `queue_depth` | Jobs waiting (not counting the currently running one) |
+| `model_state` | `"unloaded"` / `"loading"` / `"waiting_for_gpu"` / `"ready"` |
+
+---
+
+## 8. Cancelling Jobs
+
+```bash
+curl -X DELETE http://your-server:8080/jobs/550e8400-e29b-41d4-a716-446655440000
+```
+
+- `200` — job marked `cancelled`. Returns the updated `Job` object.
+- `404` — job not found.
+- `409` — job already in a terminal state (`done` / `failed` / `cancelled`).
+
+> **Important:** whisper.cpp does not support mid-inference cancellation. If the job is currently `running`, the GPU inference will finish before the cancellation takes effect — the result is simply discarded and the status set to `cancelled`.
+
+---
+
+## 9. TypeScript Types
+
+```typescript
+type ModelStateTag = 'unloaded' | 'loading' | 'waiting_for_gpu' | 'ready';
+type JobStatus     = 'queued' | 'running' | 'done' | 'failed' | 'cancelled';
+type Task          = 'transcribe' | 'translate';
+
+interface ModelStatus {
+  state: ModelStateTag;
+  // ready only
+  loaded_at?: string;
+  // waiting_for_gpu only
+  vram_needed_mb?: number;
+  vram_free_mb?:   number;
+  retry_in_secs?:  number;
+  // always (when nvidia-smi available)
+  vram_used_mb?:   number;
+  vram_total_mb?:  number;
+}
+
+interface Word {
+  text:        string;
+  start:       number; // seconds
+  end:         number; // seconds
+  probability: number; // 0–1
+}
+
+interface Segment {
+  index: number;
+  start: number; // seconds
+  end:   number; // seconds
+  text:  string;
+  words: Word[];
+}
+
+interface Job {
+  id:            string;
+  status:        JobStatus;
+  task:          Task;
+  language?:     string;     // ISO 639-1; null until detected/set
+  progress:      number;     // 0–100
+  duration_secs?: number;   // null until processing starts
+  segments:      Segment[];  // populated when status = 'done'
+  error?:        string;     // populated when status = 'failed'
+  webhook_url?:  string;
+  filename?:     string;
+  created_at:    string;     // ISO 8601
+  completed_at?: string;     // ISO 8601; null until terminal
+}
+
+// SSE payloads from GET /jobs/:id/stream
+type JobSseEvent =
+  | { type: 'progress'; percent: number; chunk: number; chunks_total: number }
+  | { type: 'done';     job: Job }
+  | { type: 'error';    message: string };
+
+// SSE payloads from GET /model/events
+type ModelSseEvent =
+  | { type: 'model_loading' }
+  | { type: 'model_ready';           loaded_at: string }
+  | { type: 'model_unloaded' }
+  | { type: 'model_waiting_for_gpu'; vram_needed_mb: number; vram_free_mb: number; retry_in_secs: number };
+
+// Webhook payload — union of job completion and model lifecycle events
+type WebhookPayload = Job | { type: 'model_ready'; loaded_at: string } | { type: 'model_unloaded' };
+
+// Helpers
+function isJobPayload(p: WebhookPayload): p is Job {
+  return 'id' in p && 'status' in p;
+}
+function isModelPayload(p: WebhookPayload): p is { type: string } {
+  return 'type' in p;
+}
+```
+
+---
+
+## 10. React Hooks
+
+```typescript
+// useModelStatus.ts
+import { useEffect, useState } from 'react';
+
+const BASE = process.env.NEXT_PUBLIC_WHISPER_BASE_URL ?? '';
+
+export function useModelStatus() {
+  const [status, setStatus] = useState<ModelStatus | null>(null);
+
+  // Initial fetch
+  useEffect(() => {
+    fetch(`${BASE}/model/status`)
+      .then(r => r.json())
+      .then(setStatus)
+      .catch(console.error);
+  }, []);
+
+  // Live updates via SSE
+  useEffect(() => {
+    const es = new EventSource(`${BASE}/model/events`);
+
+    const refresh = () => {
+      fetch(`${BASE}/model/status`)
+        .then(r => r.json())
+        .then(setStatus)
+        .catch(console.error);
+    };
+
+    es.addEventListener('model_loading',        refresh);
+    es.addEventListener('model_ready',          refresh);
+    es.addEventListener('model_unloaded',       refresh);
+    es.addEventListener('model_waiting_for_gpu',refresh);
+    es.onerror = () => console.warn('model/events reconnecting…');
+
+    return () => es.close();
+  }, []);
+
+  return status;
+}
+```
+
+```typescript
+// useJobStream.ts
+import { useEffect, useRef, useState } from 'react';
+
+type ProgressState = {
+  percent: number;
+  chunk: number;
+  chunks_total: number;
+};
+
+export function useJobStream(jobId: string | null) {
+  const [progress, setProgress] = useState<ProgressState | null>(null);
+  const [job,      setJob]      = useState<Job | null>(null);
+  const [error,    setError]    = useState<string | null>(null);
+  const esRef = useRef<EventSource | null>(null);
+
+  useEffect(() => {
+    if (!jobId) return;
+
+    esRef.current?.close();
+    setProgress(null); setJob(null); setError(null);
+
+    const es = new EventSource(`${BASE}/jobs/${jobId}/stream`);
+    esRef.current = es;
+
+    es.addEventListener('progress', (e) => {
+      setProgress(JSON.parse(e.data));
+    });
+
+    es.addEventListener('done', (e) => {
+      setJob(JSON.parse(e.data).job);
+      setProgress({ percent: 100, chunk: 0, chunks_total: 0 });
+      es.close();
+    });
+
+    es.addEventListener('error', (e) => {
+      if ('data' in e) setError(JSON.parse((e as MessageEvent).data).message);
+      es.close();
+    });
+
+    return () => es.close();
+  }, [jobId]);
+
+  return { progress, job, error };
+}
+```
+
+```typescript
+// useTranscribe.ts — ties it all together
+import { useState, useCallback } from 'react';
+
+export function useTranscribe() {
+  const [jobId,  setJobId]  = useState<string | null>(null);
+  const [loading, setLoading] = useState(false);
+  const [error,  setError]  = useState<string | null>(null);
+
+  const submit = useCallback(async (
+    audio: Blob,
+    opts: { language?: string; task?: Task } = {}
+  ) => {
+    setLoading(true);
+    setError(null);
+    setJobId(null);
+
+    try {
+      const id = await submitWithRetry(audio, opts); // see §4.3
+      setJobId(id);
+    } catch (e) {
+      setError(String(e));
+    } finally {
+      setLoading(false);
+    }
+  }, []);
+
+  const { progress, job, error: streamError } = useJobStream(jobId);
+
+  return { submit, loading, jobId, progress, job, error: error ?? streamError };
+}
+```
+
+---
+
+## 11. Complete Integration Example
+
+A full transcription flow with model warm-up indicator and real-time progress:
+
+```typescript
+// whisperClient.ts
+const BASE = process.env.NEXT_PUBLIC_WHISPER_BASE_URL ?? '';
+
+export class WhisperClient {
+  /** Wait for the model to be ready, triggering a load if needed. */
+  async ensureModelReady(timeoutMs = 120_000): Promise<void> {
+    const status = await this.getModelStatus();
+    if (status.state === 'ready') return;
+
+    // Trigger load (idempotent)
+    await fetch(`${BASE}/model/load`, { method: 'POST' });
+
+    return new Promise((resolve, reject) => {
+      const deadline = setTimeout(() => {
+        es.close();
+        reject(new Error('Model did not become ready within timeout'));
+      }, timeoutMs);
+
+      const es = new EventSource(`${BASE}/model/events`);
+      es.addEventListener('model_ready', () => {
+        clearTimeout(deadline);
+        es.close();
+        resolve();
+      });
+      es.onerror = () => {
+        // Reconnects automatically; don't reject on transient drops.
+      };
+    });
+  }
+
+  async getModelStatus(): Promise<ModelStatus> {
+    const r = await fetch(`${BASE}/model/status`);
+    if (!r.ok) throw new Error(`/model/status ${r.status}`);
+    return r.json();
+  }
+
+  async submit(
+    audio: Blob,
+    opts: { language?: string; task?: Task; webhookUrl?: string } = {}
+  ): Promise<string> {
+    return submitWithRetry(audio, opts);
+  }
+
+  streamProgress(
+    jobId: string,
+    callbacks: {
+      onProgress?: (p: { percent: number; chunk: number; total: number }) => void;
+      onDone?:     (job: Job) => void;
+      onError?:    (msg: string) => void;
+    }
+  ): () => void {
+    const es = new EventSource(`${BASE}/jobs/${jobId}/stream`);
+
+    es.addEventListener('progress', (e) => {
+      const d = JSON.parse(e.data);
+      callbacks.onProgress?.({ percent: d.percent, chunk: d.chunk, total: d.chunks_total });
+    });
+
+    es.addEventListener('done', (e) => {
+      callbacks.onDone?.(JSON.parse(e.data).job);
+      es.close();
+    });
+
+    es.addEventListener('error', (e) => {
+      if ('data' in e) callbacks.onError?.(JSON.parse((e as MessageEvent).data).message);
+      es.close();
+    });
+
+    return () => es.close();
+  }
+
+  async transcribe(
+    audio: Blob,
+    opts: {
+      language?: string;
+      task?: Task;
+      webhookUrl?: string;
+      onProgress?: (percent: number) => void;
+    } = {}
+  ): Promise<Job> {
+    const jobId = await this.submit(audio, opts);
+
+    return new Promise((resolve, reject) => {
+      this.streamProgress(jobId, {
+        onProgress: (p) => opts.onProgress?.(p.percent),
+        onDone:     resolve,
+        onError:    (msg) => reject(new Error(msg)),
+      });
+    });
+  }
+}
+
+// Usage
+const whisper = new WhisperClient();
+
+const job = await whisper.transcribe(audioBlob, {
+  language: 'en',
+  onProgress: (pct) => console.log(`${pct}%`),
+});
+
+for (const seg of job.segments) {
+  console.log(`[${seg.start.toFixed(1)}s → ${seg.end.toFixed(1)}s] ${seg.text}`);
+}
+```
+
+---
+
+## 12. Error Reference
+
+All error responses follow this shape:
+
+```json
+{ "error": "human-readable message" }
+```
+
+With the following additions for specific errors:
+
+**503 model_not_ready:**
+```json
+{ "error": "model_not_ready", "state": "loading", "retry_after_secs": 10 }
+```
+
+| HTTP | `error` value | When | What to do |
+|------|--------------|------|-----------|
+| 400 | `"missing 'audio' field"` | `audio` not in form | Fix the form |
+| 400 | `"audio field is empty"` | Zero-byte file uploaded | Fix the file |
+| 400 | `"task must be 'transcribe' or 'translate'"` | Bad `task` value | Fix the value |
+| 400 | `"multipart error: …"` | Malformed request | Check content-type header |
+| 404 | `"job … not found"` | Unknown job ID | Check the ID |
+| 409 | `"job … is already in terminal state …"` | Cancelling a finished job | No action needed |
+| 503 | `"model_not_ready"` | Model not loaded | See §4.2 — retry with `Retry-After` |
+| 500 | `"worker channel closed"` | Server crash | Contact server admin |
+
+**Network / SSE errors:**
+
+- `EventSource` `onerror` with no `.data` = connection dropped. The browser reconnects automatically — no action needed unless you want to show a UI indicator.
+- HTTP 502/503/504 from a reverse proxy = the container is restarting. Wait and retry.
+
+---
+
+*Last updated: 2026-05-08. Corresponds to whisper-server v0.1.0 commit `d014826`.*
--- a/docs/USAGE.md
+++ b/docs/USAGE.md
@@ -66,6 +66,8 @@ The bundled `docker-compose.yml` mounts named volumes for data and models and se
 | `WHISPER_MODEL_PATH` | `/models/ggml-large-v3.bin` | Absolute path to GGML model file |
 | `WHISPER_MODEL` | `large-v3` | Model name reported by `/health` (display only) |
 | `CUDA_DEVICE` | `0` | CUDA device index to use for inference |
+| `IDLE_TIMEOUT_SECS` | `300` | Seconds of idle before the model is automatically unloaded from GPU memory. Set to `0` to disable auto-unload. |
+| `GPU_POLL_INTERVAL_SECS` | `30` | Seconds between VRAM-availability retries when a load fails due to insufficient VRAM. |

 ### Note on CUDA device ordering
 Inside Docker, device ordering matches `nvidia-smi` (PCI bus order). On the host without Docker, ordering may differ. See [FINDINGS.md](FINDINGS.md#cuda-device-index-ordering-differs-between-host-and-docker) for details.
@@ -76,6 +78,194 @@ Inside Docker, device ordering matches `nvidia-smi` (PCI bus order). On the host

 The interactive Swagger UI is available at `http://localhost:8080/docs`.

+---
+
+## Model Lifecycle Management
+
+The model starts **unloaded** on startup (lazy loading). It is loaded into GPU memory on the first job submission or via `POST /model/load`, and automatically unloaded after `IDLE_TIMEOUT_SECS` of inactivity.
+
+### Model State Machine
+
+```
+Unloaded ──(job / POST /model/load)──► Loading ──(success)──► Ready
+                                                 └──(VRAM full)──► WaitingForGpu ──(retry OK)──► Loading
+Ready ──(idle timeout / POST /model/unload)──► Unloaded
+WaitingForGpu ──(POST /model/unload)──► Unloaded
+```
+
+### `GET /model/status`
+
+Returns the current model state and VRAM statistics.
+
+```bash
+curl http://localhost:8080/model/status
+```
+
+**When unloaded:**
+```json
+{ "state": "unloaded" }
+```
+
+**When loading:**
+```json
+{ "state": "loading" }
+```
+
+**When ready:**
+```json
+{
+  "state": "ready",
+  "loaded_at": "2026-05-10T14:00:00Z",
+  "vram_used_mb": 4096,
+  "vram_total_mb": 8192
+}
+```
+
+**When waiting for VRAM:**
+```json
+{
+  "state": "waiting_for_gpu",
+  "vram_needed_mb": 3951,
+  "vram_free_mb": 512,
+  "retry_in_secs": 30
+}
+```
+
+---
+
+### `POST /model/load`
+
+Request the model to be loaded. Idempotent — if already loading or ready, returns immediately.
+
+```bash
+curl -X POST http://localhost:8080/model/load
+```
+
+- Returns `202 Accepted` with `{"status":"load_initiated"}` when load is triggered
+- Returns `200 OK` with `{"status":"already_ready"}` when model is already ready
+- Poll `GET /model/status` or subscribe to `GET /model/events` to know when ready
+
+---
+
+### `POST /model/unload`
+
+Unload the model from GPU memory immediately, freeing VRAM.
+
+```bash
+curl -X POST http://localhost:8080/model/unload
+```
+
+Returns `200 OK` regardless of current state.
+
+---
+
+### `GET /model/events` — Model SSE stream
+
+Subscribe to model lifecycle events via Server-Sent Events.
+
+```bash
+curl -N http://localhost:8080/model/events
+```
+
+**Event types:**
+
+```
+event: model_loading
+data: {"type":"model_loading"}
+
+event: model_ready
+data: {"type":"model_ready","loaded_at":"2026-05-10T14:00:00Z"}
+
+event: model_unloaded
+data: {"type":"model_unloaded"}
+
+event: model_waiting_for_gpu
+data: {"type":"model_waiting_for_gpu","vram_needed_mb":3951,"vram_free_mb":512,"retry_in_secs":30}
+```
+
+**JavaScript example:**
+```javascript
+const es = new EventSource('/model/events');
+
+es.addEventListener('model_ready', () => {
+  console.log('Model loaded — ready to transcribe');
+});
+
+es.addEventListener('model_unloaded', () => {
+  console.log('Model freed GPU memory');
+});
+```
+
+---
+
+### Webhooks for model events
+
+When any job is submitted with a `webhook_url`, that URL is registered to receive model lifecycle webhooks for the lifetime of the server process. The following events trigger a webhook POST:
+
+| Event | Fired when |
+|-------|-----------|
+| `model_ready` | Model finishes loading (after GPU warmup) |
+| `model_unloaded` | Model is freed from GPU memory |
+
+**Webhook payload** (`Content-Type: application/json`):
+```json
+{ "type": "model_ready", "loaded_at": "2026-05-10T14:00:00Z" }
+{ "type": "model_unloaded" }
+```
+
+Delivery is attempted up to 3 times with exponential backoff (1s, 2s).
+
+---
+
+### Handling 503 Model Not Ready
+
+When you submit a job and the model is not yet loaded, you receive `503 Service Unavailable` with a `Retry-After` header:
+
+```
+HTTP/1.1 503 Service Unavailable
+Retry-After: 30
+Content-Type: application/json
+
+{
+  "error": "model_not_ready",
+  "state": "unloaded",
+  "retry_after_secs": 30
+}
+```
+
+| State at rejection | `retry_after_secs` | Meaning |
+|---|---|---|
+| `unloaded` | 30 | Load was triggered; retry after ~30s |
+| `loading` | 10 | Check again in 10s |
+| `waiting_for_gpu` | `GPU_POLL_INTERVAL_SECS` | VRAM contention; retry later |
+
+A job rejection when the model is `unloaded` **automatically triggers a load** — you do not need to call `POST /model/load` separately.
+
+**Recommended client pattern:**
+```javascript
+async function submitWithRetry(formData, maxAttempts = 10) {
+  for (let i = 0; i < maxAttempts; i++) {
+    const resp = await fetch('/jobs', { method: 'POST', body: formData });
+    if (resp.ok) return resp.json();
+    if (resp.status === 503) {
+      const retryAfter = parseInt(resp.headers.get('Retry-After') ?? '30');
+      const body = await resp.json();
+      console.log(`Model ${body.state} — retrying in ${retryAfter}s`);
+      await new Promise(r => setTimeout(r, retryAfter * 1000));
+      continue;
+    }
+    throw new Error(`Submit failed: ${resp.status}`);
+  }
+  throw new Error('Gave up after max attempts');
+}
+```
+
+---
+
+## API Reference
+
+The interactive Swagger UI is available at `http://localhost:8080/docs`.
+
 ### `POST /jobs` — Submit a transcription job

 Accepts a multipart/form-data body.
@@ -249,11 +439,12 @@ curl http://localhost:8080/health
  "gpu_name": "NVIDIA GeForce RTX 2080",
  "vram_total_mb": 8192,
  "model": "large-v3",
-  "queue_depth": 0
+  "queue_depth": 0,
+  "model_state": "ready"
 }
 ```

-`queue_depth` is the number of jobs waiting to be processed (not counting the one currently running).
+`queue_depth` is the number of jobs waiting to be processed (not counting the one currently running). `model_state` reflects the current lifecycle state (`unloaded`, `loading`, `waiting_for_gpu`, `ready`).

 ---

@@ -340,6 +531,11 @@ curl -X POST http://localhost:8080/jobs \

 ## Troubleshooting

+### Server returns `503 model_not_ready`
+- The model starts unloaded. Call `POST /model/load` explicitly, or just retry the job submission — rejection automatically triggers a load.
+- If state is `waiting_for_gpu`, another process is using the GPU's VRAM. The server will retry automatically every `GPU_POLL_INTERVAL_SECS` seconds.
+- Monitor `GET /model/status` or subscribe to `GET /model/events` to know when the model is ready.
+
 ### Server returns 0 segments
 - Check that you are **not** setting `language` to an empty string — omit the field entirely for auto-detection
 - Verify the audio file is not corrupted: `ffprobe audio.mp3`
--- a/run_tests.sh
+++ b/run_tests.sh
@@ -0,0 +1,37 @@
+#!/usr/bin/env bash
+# run_tests.sh — Run the unit test suite inside Docker (no GPU required)
+#
+# Uses the `tester` Docker stage which:
+#   1. Builds the release binary (or reuses cached build)
+#   2. Symlinks the CUDA stubs so libcuda.so.1 is satisfied without a driver
+#   3. Runs `cargo test --release`
+#
+# Usage:
+#   ./run_tests.sh                  # run all unit tests
+#   ./run_tests.sh models           # run only tests matching "models"
+#   CUDA_VERSION=12.1.0 ./run_tests.sh
+
+set -euo pipefail
+
+CUDA_VERSION=${CUDA_VERSION:-12.4.1}
+UBUNTU_VERSION=${UBUNTU_VERSION:-22.04}
+TEST_FILTER=${1:-}
+
+echo "==> Building tester stage (CUDA ${CUDA_VERSION} / Ubuntu ${UBUNTU_VERSION})..."
+docker build \
+  --target tester \
+  --build-arg CUDA_VERSION="${CUDA_VERSION}" \
+  --build-arg UBUNTU_VERSION="${UBUNTU_VERSION}" \
+  --tag whisper-tester:local \
+  .
+
+if [[ -n "${TEST_FILTER}" ]]; then
+  echo "==> Running tests matching '${TEST_FILTER}'..."
+  docker run --rm \
+    -e LD_LIBRARY_PATH=/usr/local/cuda/lib64/stubs \
+    whisper-tester:local \
+    sh -c "cd /build && ln -sf /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1 && LD_LIBRARY_PATH=/usr/local/cuda/lib64/stubs cargo test --release '${TEST_FILTER}'"
+else
+  echo "==> All tests ran during docker build (tester stage)."
+  echo "    Build succeeded — all tests passed."
+fi
--- a/src/error.rs
+++ b/src/error.rs
@@ -1,10 +1,10 @@
-use thiserror::Error;
 use axum::{
-    http::StatusCode,
+    http::{header, HeaderValue, StatusCode},
    response::{IntoResponse, Response},
    Json,
 };
 use serde_json::json;
+use thiserror::Error;

 pub type Result<T> = std::result::Result<T, AppError>;

@@ -21,19 +21,156 @@ pub enum AppError {

    #[error("internal error: {0}")]
    Internal(String),
+
+    /// Returned when `whisper_init_state` or `cudaMalloc` fails due to
+    /// insufficient VRAM. The worker uses this to distinguish a recoverable
+    /// VRAM-pressure failure from a hard internal error.
+    #[error("out of GPU memory: {0}")]
+    OutOfMemory(String),
+
+    /// Returned when a job is submitted but the model is not yet loaded.
+    /// Carries the current state tag and recommended Retry-After seconds.
+    #[error("model not ready: {state}")]
+    ModelNotReady {
+        state: String,
+        retry_after_secs: u64,
+    },
+}
+
+impl AppError {
+    /// Returns true if the error string contains patterns emitted by
+    /// whisper.cpp / GGML when a CUDA memory allocation fails.
+    pub fn is_oom(msg: &str) -> bool {
+        msg.contains("cudaMalloc failed")
+            || msg.contains("out of memory")
+            || msg.contains("CUDA error: out of memory")
+            || msg.contains("alloc_buffer")
+    }
 }

 impl IntoResponse for AppError {
    fn into_response(self) -> Response {
-        let (status, message) = match &self {
-            AppError::NotFound(m)   => (StatusCode::NOT_FOUND,             m.clone()),
-            AppError::BadRequest(m) => (StatusCode::BAD_REQUEST,           m.clone()),
-            AppError::Conflict(m)   => (StatusCode::CONFLICT,              m.clone()),
-            AppError::Internal(m)   => (StatusCode::INTERNAL_SERVER_ERROR, m.clone()),
-        };
-
-        tracing::error!(status = status.as_u16(), error = %message);
-
-        (status, Json(json!({ "error": message }))).into_response()
+        match self {
+            AppError::NotFound(m) => {
+                (StatusCode::NOT_FOUND, Json(json!({ "error": m }))).into_response()
+            }
+            AppError::BadRequest(m) => {
+                (StatusCode::BAD_REQUEST, Json(json!({ "error": m }))).into_response()
+            }
+            AppError::Conflict(m) => {
+                (StatusCode::CONFLICT, Json(json!({ "error": m }))).into_response()
+            }
+            AppError::Internal(m) => {
+                tracing::error!(error = %m, "internal error");
+                (
+                    StatusCode::INTERNAL_SERVER_ERROR,
+                    Json(json!({ "error": m })),
+                )
+                    .into_response()
+            }
+            AppError::OutOfMemory(m) => {
+                tracing::warn!(error = %m, "GPU out of memory during model load");
+                (StatusCode::SERVICE_UNAVAILABLE, Json(json!({ "error": m }))).into_response()
+            }
+            AppError::ModelNotReady {
+                state,
+                retry_after_secs,
+            } => {
+                let body = Json(json!({
+                    "error":            "model_not_ready",
+                    "state":            state,
+                    "retry_after_secs": retry_after_secs,
+                }));
+                let mut resp = (StatusCode::SERVICE_UNAVAILABLE, body).into_response();
+                resp.headers_mut().insert(
+                    header::RETRY_AFTER,
+                    HeaderValue::from_str(&retry_after_secs.to_string())
+                        .unwrap_or(HeaderValue::from_static("30")),
+                );
+                resp
+            }
+        }
+    }
+}
+
+// ── Unit tests ───────────────────────────────────────────────────────────────
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use axum::body::to_bytes;
+
+    #[test]
+    fn test_is_oom_cuda_malloc() {
+        assert!(AppError::is_oom("cudaMalloc failed: out of memory"));
+    }
+
+    #[test]
+    fn test_is_oom_alloc_buffer() {
+        // Exact message from ggml_backend_cuda_buffer_type_alloc_buffer
+        assert!(AppError::is_oom(
+            "ggml_backend_cuda_buffer_type_alloc_buffer: allocating 2951.01 MiB on device 0: cudaMalloc failed: out of memory"
+        ));
+    }
+
+    #[test]
+    fn test_is_oom_generic_out_of_memory() {
+        assert!(AppError::is_oom("CUDA error: out of memory"));
+    }
+
+    #[test]
+    fn test_is_oom_other_error() {
+        assert!(!AppError::is_oom("failed to open model file"));
+        assert!(!AppError::is_oom("invalid model format"));
+        assert!(!AppError::is_oom(""));
+    }
+
+    #[tokio::test]
+    async fn test_model_not_ready_response_has_retry_after_header() {
+        let err = AppError::ModelNotReady {
+            state: "loading".into(),
+            retry_after_secs: 10,
+        };
+        let resp = err.into_response();
+        assert_eq!(resp.status(), StatusCode::SERVICE_UNAVAILABLE);
+        let retry_after = resp
+            .headers()
+            .get(header::RETRY_AFTER)
+            .expect("Retry-After header missing");
+        assert_eq!(retry_after, "10");
+    }
+
+    #[tokio::test]
+    async fn test_model_not_ready_response_body() {
+        let err = AppError::ModelNotReady {
+            state: "unloaded".into(),
+            retry_after_secs: 30,
+        };
+        let resp = err.into_response();
+        let bytes = to_bytes(resp.into_body(), usize::MAX).await.unwrap();
+        let v: serde_json::Value = serde_json::from_slice(&bytes).unwrap();
+        assert_eq!(v["error"], "model_not_ready");
+        assert_eq!(v["state"], "unloaded");
+        assert_eq!(v["retry_after_secs"], 30);
+    }
+
+    #[tokio::test]
+    async fn test_model_not_ready_loading_retry_after_10() {
+        let err = AppError::ModelNotReady {
+            state: "loading".into(),
+            retry_after_secs: 10,
+        };
+        let resp = err.into_response();
+        assert_eq!(resp.headers().get(header::RETRY_AFTER).unwrap(), "10");
+    }
+
+    #[tokio::test]
+    async fn test_model_not_ready_unloaded_retry_after_30() {
+        let err = AppError::ModelNotReady {
+            state: "unloaded".into(),
+            retry_after_secs: 30,
+        };
+        let resp = err.into_response();
+        assert_eq!(resp.headers().get(header::RETRY_AFTER).unwrap(), "30");
    }
 }
--- a/src/main.rs
+++ b/src/main.rs
@@ -1,7 +1,7 @@
 use std::sync::Arc;

 use axum::Router;
-use tokio::sync::mpsc;
+use tokio::sync::{broadcast, mpsc, RwLock};
 use tower_http::{cors::CorsLayer, trace::TraceLayer};
 use tracing_subscriber::{layer::SubscriberExt, util::SubscriberInitExt, EnvFilter};
 use utoipa::OpenApi;
@@ -21,8 +21,10 @@ pub use error::{AppError, Result};

 #[derive(Clone)]
 pub struct AppState {
-    /// Channel to submit jobs to the single GPU worker.
+    /// Channel to submit jobs to the single GPU worker (job IDs only).
    pub job_tx: mpsc::UnboundedSender<models::JobId>,
+    /// Channel to send control commands to the worker OS thread.
+    pub cmd_tx: std::sync::mpsc::SyncSender<worker::WorkerCmd>,
    /// Shared handle to the on-disk job store.
    pub storage: Arc<storage::Storage>,
    /// SSE broadcast registry: job_id → sender.
@@ -33,6 +35,17 @@ pub struct AppState {
    pub queue_depth: Arc<std::sync::atomic::AtomicUsize>,
    /// CUDA device index used for inference.
    pub gpu_device: u32,
+    /// Current state of the whisper model.
+    pub model_state: Arc<RwLock<models::ModelState>>,
+    /// Broadcast channel for model lifecycle events (SSE + webhooks).
+    pub model_event_tx: broadcast::Sender<models::ModelEvent>,
+    /// All webhook URLs ever registered via job submission.
+    /// Used to fire model_ready / model_unloaded notifications.
+    pub webhook_registry: Arc<std::sync::Mutex<std::collections::HashSet<String>>>,
+    /// How long the model stays loaded with no active jobs.
+    pub idle_timeout: std::time::Duration,
+    /// How often to retry loading when GPU is busy.
+    pub gpu_poll_interval: std::time::Duration,
 }

 // ── OpenAPI spec root ────────────────────────────────────────────────────────
@@ -50,6 +63,10 @@ pub struct AppState {
        routes::jobs::stream_job,
        routes::jobs::delete_job,
        routes::health::health,
+        routes::model::model_status,
+        routes::model::model_load,
+        routes::model::model_unload,
+        routes::model::model_events,
    ),
    components(schemas(
        models::Job,
@@ -58,10 +75,14 @@ pub struct AppState {
        models::Word,
        models::SubmitResponse,
        models::HealthResponse,
+        models::ModelState,
+        models::ModelEvent,
+        models::ModelStatusResponse,
    )),
    tags(
        (name = "jobs",   description = "Transcription job management"),
        (name = "system", description = "Service health"),
+        (name = "model",  description = "Model lifecycle management"),
    )
 )]
 struct ApiDoc;
@@ -76,15 +97,29 @@ async fn main() -> anyhow::Result<()> {
        .with(tracing_subscriber::fmt::layer().json())
        .init();

-    let data_dir   = std::env::var("DATA_DIR").unwrap_or_else(|_| "/data".into());
-    let model_path = std::env::var("WHISPER_MODEL_PATH")
-        .unwrap_or_else(|_| "/models/ggml-large-v3.bin".into());
-    let port       = std::env::var("PORT").unwrap_or_else(|_| "8080".into());
+    let data_dir = std::env::var("DATA_DIR").unwrap_or_else(|_| "/data".into());
+    let model_path =
+        std::env::var("WHISPER_MODEL_PATH").unwrap_or_else(|_| "/models/ggml-large-v3.bin".into());
+    let port = std::env::var("PORT").unwrap_or_else(|_| "8080".into());
    let model_name = std::env::var("WHISPER_MODEL").unwrap_or_else(|_| "large-v3".into());
    let gpu_device: u32 = std::env::var("CUDA_DEVICE")
        .ok()
        .and_then(|s| s.parse().ok())
        .unwrap_or(0);
+    let idle_timeout_secs: u64 = std::env::var("IDLE_TIMEOUT_SECS")
+        .ok()
+        .and_then(|s| s.parse().ok())
+        .unwrap_or(300);
+    let gpu_poll_interval_secs: u64 = std::env::var("GPU_POLL_INTERVAL_SECS")
+        .ok()
+        .and_then(|s| s.parse().ok())
+        .unwrap_or(30);
+
+    tracing::info!(
+        idle_timeout_secs,
+        gpu_poll_interval_secs,
+        "dynamic model loading configured"
+    );

    let storage = Arc::new(storage::Storage::new(&data_dir).await?);

@@ -94,28 +129,47 @@ async fn main() -> anyhow::Result<()> {
    let (job_tx, job_rx) = mpsc::unbounded_channel::<models::JobId>();
    let queue_depth = Arc::new(std::sync::atomic::AtomicUsize::new(0));

-    // Spawn single GPU worker; get back the SSE broadcast registry.
-    let progress = worker::start(
+    // Model starts unloaded — lazy load on first job or POST /model/load.
+    let model_state = Arc::new(RwLock::new(models::ModelState::Unloaded));
+    let (model_event_tx, _) = broadcast::channel::<models::ModelEvent>(32);
+    let webhook_registry = Arc::new(std::sync::Mutex::new(
+        std::collections::HashSet::<String>::new(),
+    ));
+
+    // Spawn single GPU worker; get back the SSE broadcast registry and cmd channel.
+    let (progress, cmd_tx) = worker::start(
        job_rx,
        Arc::clone(&storage),
        model_path.clone().into(),
        Arc::clone(&queue_depth),
        gpu_device,
+        Arc::clone(&model_state),
+        model_event_tx.clone(),
+        Arc::clone(&webhook_registry),
+        std::time::Duration::from_secs(idle_timeout_secs),
+        std::time::Duration::from_secs(gpu_poll_interval_secs),
    );

    let state = AppState {
        job_tx,
+        cmd_tx,
        storage: Arc::clone(&storage),
        progress,
-        model_name:  model_name.as_str().into(),
+        model_name: model_name.as_str().into(),
        queue_depth: Arc::clone(&queue_depth),
        gpu_device,
+        model_state,
+        model_event_tx,
+        webhook_registry,
+        idle_timeout: std::time::Duration::from_secs(idle_timeout_secs),
+        gpu_poll_interval: std::time::Duration::from_secs(gpu_poll_interval_secs),
    };

    let app = Router::new()
        .merge(SwaggerUi::new("/docs").url("/openapi.json", ApiDoc::openapi()))
        .merge(routes::jobs_router())
        .merge(routes::health_router())
+        .merge(routes::model_router())
        .with_state(state)
        .layer(CorsLayer::permissive())
        .layer(TraceLayer::new_for_http());
--- a/src/models.rs
+++ b/src/models.rs
@@ -5,6 +5,117 @@ use uuid::Uuid;

 pub type JobId = Uuid;

+// ── Model lifecycle state ────────────────────────────────────────────────────
+
+/// Current state of the whisper model in memory.
+///
+/// State machine:
+/// ```
+/// Unloaded ──(load trigger)──► Loading ──(ok)──► Ready ──(idle/unload)──► Unloaded
+///                                       └──(VRAM full)──► WaitingForGpu ──(retry)──► Loading
+/// WaitingForGpu ──(unload cmd)──► Unloaded
+/// ```
+#[derive(Debug, Clone, Serialize, Deserialize, ToSchema)]
+#[serde(tag = "state", rename_all = "snake_case")]
+pub enum ModelState {
+    /// Model is not in memory. GPU is free.
+    Unloaded,
+    /// Model is being loaded (weights transferred to GPU).
+    Loading,
+    /// A previous load attempt failed due to insufficient VRAM. The worker is
+    /// polling at `retry_in_secs` intervals until enough memory is available.
+    WaitingForGpu {
+        /// VRAM required to load the model, in MiB.
+        vram_needed_mb: u64,
+        /// VRAM currently free on the device, in MiB.
+        vram_free_mb: u64,
+        /// How many seconds until the next load attempt.
+        retry_in_secs: u64,
+    },
+    /// Model is loaded and ready to accept inference jobs.
+    Ready {
+        /// UTC timestamp of when the model finished loading (post-warmup).
+        loaded_at: DateTime<Utc>,
+    },
+}
+
+impl ModelState {
+    /// Returns true if the model can accept inference jobs right now.
+    pub fn is_ready(&self) -> bool {
+        matches!(self, ModelState::Ready { .. })
+    }
+
+    /// Suggested `Retry-After` value (seconds) to include in 503 responses.
+    pub fn retry_after_secs(&self) -> u64 {
+        match self {
+            ModelState::Unloaded => 30, // conservative load estimate
+            ModelState::Loading => 10,
+            ModelState::WaitingForGpu { retry_in_secs, .. } => *retry_in_secs,
+            ModelState::Ready { .. } => 0, // shouldn't 503 if ready
+        }
+    }
+
+    /// String tag for use in error response bodies and log fields.
+    pub fn tag(&self) -> &'static str {
+        match self {
+            ModelState::Unloaded => "unloaded",
+            ModelState::Loading => "loading",
+            ModelState::WaitingForGpu { .. } => "waiting_for_gpu",
+            ModelState::Ready { .. } => "ready",
+        }
+    }
+}
+
+// ── Model events (SSE + webhooks) ────────────────────────────────────────────
+
+/// Events broadcast over the `GET /model/events` SSE stream and fired as
+/// webhooks to registered clients.
+///
+/// Webhook delivery: only `ModelReady` and `ModelUnloaded` are sent to
+/// webhook URLs. All four are broadcast on the SSE stream.
+#[derive(Debug, Clone, Serialize, Deserialize, ToSchema)]
+#[serde(tag = "type", rename_all = "snake_case")]
+pub enum ModelEvent {
+    /// Model finished loading and the GPU warmup completed — ready to accept jobs.
+    ModelReady { loaded_at: DateTime<Utc> },
+    /// Model was unloaded from GPU memory (idle timeout or manual unload).
+    ModelUnloaded,
+    /// Model load initiated.
+    ModelLoading,
+    /// Load failed due to insufficient VRAM; retrying after `retry_in_secs`.
+    ModelWaitingForGpu {
+        vram_needed_mb: u64,
+        vram_free_mb: u64,
+        retry_in_secs: u64,
+    },
+}
+
+impl ModelEvent {
+    /// Returns true if this event should be delivered via webhook.
+    pub fn is_webhook_event(&self) -> bool {
+        matches!(
+            self,
+            ModelEvent::ModelReady { .. } | ModelEvent::ModelUnloaded
+        )
+    }
+}
+
+// ── Model status response ────────────────────────────────────────────────────
+
+/// Response body for `GET /model/status`.
+#[derive(Debug, Serialize, Deserialize, ToSchema)]
+pub struct ModelStatusResponse {
+    /// Current model state (flattened from `ModelState`).
+    #[serde(flatten)]
+    pub state: ModelState,
+    /// VRAM currently used on the device, in MiB (from nvidia-smi).
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub vram_used_mb: Option<u64>,
+    /// VRAM total on the device, in MiB.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub vram_total_mb: Option<u64>,
+}
+
 // ── Job status ───────────────────────────────────────────────────────────────

 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, ToSchema)]
@@ -22,11 +133,11 @@ pub enum JobStatus {
 #[derive(Debug, Clone, Serialize, Deserialize, ToSchema)]
 pub struct Word {
    /// Word text
-    pub text:        String,
+    pub text: String,
    /// Start time in seconds
-    pub start:       f32,
+    pub start: f32,
    /// End time in seconds
-    pub end:         f32,
+    pub end: f32,
    /// Model confidence (0–1)
    pub probability: f32,
 }
@@ -38,9 +149,9 @@ pub struct Segment {
    /// Start time in seconds
    pub start: f32,
    /// End time in seconds
-    pub end:   f32,
+    pub end: f32,
    /// Transcribed text
-    pub text:  String,
+    pub text: String,
    /// Token-level word timestamps (empty when flash_attn is enabled)
    #[serde(default)]
    pub words: Vec<Word>,
@@ -95,18 +206,23 @@ pub struct Job {
 }

 impl Job {
-    pub fn new(id: JobId, task: String, webhook_url: Option<String>, filename: Option<String>) -> Self {
+    pub fn new(
+        id: JobId,
+        task: String,
+        webhook_url: Option<String>,
+        filename: Option<String>,
+    ) -> Self {
        Self {
            id,
-            status:       JobStatus::Queued,
-            language:     None,
+            status: JobStatus::Queued,
+            language: None,
            task,
            duration_secs: None,
-            segments:     vec![],
-            error:        None,
+            segments: vec![],
+            error: None,
            webhook_url,
-            progress:     0,
-            created_at:   Utc::now(),
+            progress: 0,
+            created_at: Utc::now(),
            completed_at: None,
            filename,
        }
@@ -125,11 +241,13 @@ pub struct SubmitResponse {
 /// Response from GET /health.
 #[derive(Debug, Serialize, ToSchema)]
 pub struct HealthResponse {
-    pub status:        String,
-    pub gpu_name:      Option<String>,
+    pub status: String,
+    pub gpu_name: Option<String>,
    pub vram_total_mb: Option<u64>,
-    pub model:         String,
-    pub queue_depth:   usize,
+    pub model: String,
+    pub queue_depth: usize,
+    /// Current state of the whisper model.
+    pub model_state: String,
 }

 // ── SSE event payload ────────────────────────────────────────────────────────
@@ -145,6 +263,184 @@ pub enum SsePayload {
        /// Total number of silence-split chunks in this job.
        chunks_total: usize,
    },
-    Done  { job: Box<Job> },
-    Error { message: String },
+    Done {
+        job: Box<Job>,
+    },
+    Error {
+        message: String,
+    },
+}
+
+// ── Unit tests ───────────────────────────────────────────────────────────────
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use serde_json::Value;
+
+    // ── ModelState serialization ─────────────────────────────────────────────
+
+    #[test]
+    fn test_model_state_unloaded_serializes() {
+        let v: Value = serde_json::to_value(ModelState::Unloaded).unwrap();
+        assert_eq!(v["state"], "unloaded");
+    }
+
+    #[test]
+    fn test_model_state_loading_serializes() {
+        let v: Value = serde_json::to_value(ModelState::Loading).unwrap();
+        assert_eq!(v["state"], "loading");
+    }
+
+    #[test]
+    fn test_model_state_waiting_serializes() {
+        let s = ModelState::WaitingForGpu {
+            vram_needed_mb: 3000,
+            vram_free_mb: 500,
+            retry_in_secs: 30,
+        };
+        let v: Value = serde_json::to_value(&s).unwrap();
+        assert_eq!(v["state"], "waiting_for_gpu");
+        assert_eq!(v["vram_needed_mb"], 3000);
+        assert_eq!(v["vram_free_mb"], 500);
+        assert_eq!(v["retry_in_secs"], 30);
+    }
+
+    #[test]
+    fn test_model_state_ready_serializes() {
+        let ts = Utc::now();
+        let s = ModelState::Ready { loaded_at: ts };
+        let v: Value = serde_json::to_value(&s).unwrap();
+        assert_eq!(v["state"], "ready");
+        assert!(v["loaded_at"].is_string());
+    }
+
+    #[test]
+    fn test_model_state_is_ready() {
+        assert!(!ModelState::Unloaded.is_ready());
+        assert!(!ModelState::Loading.is_ready());
+        assert!(!ModelState::WaitingForGpu {
+            vram_needed_mb: 0,
+            vram_free_mb: 0,
+            retry_in_secs: 30
+        }
+        .is_ready());
+        assert!(ModelState::Ready {
+            loaded_at: Utc::now()
+        }
+        .is_ready());
+    }
+
+    #[test]
+    fn test_retry_after_unloaded() {
+        assert_eq!(ModelState::Unloaded.retry_after_secs(), 30);
+    }
+
+    #[test]
+    fn test_retry_after_loading() {
+        assert_eq!(ModelState::Loading.retry_after_secs(), 10);
+    }
+
+    #[test]
+    fn test_retry_after_waiting_for_gpu() {
+        let s = ModelState::WaitingForGpu {
+            vram_needed_mb: 0,
+            vram_free_mb: 0,
+            retry_in_secs: 45,
+        };
+        assert_eq!(s.retry_after_secs(), 45);
+    }
+
+    #[test]
+    fn test_retry_after_ready_is_zero() {
+        assert_eq!(
+            ModelState::Ready {
+                loaded_at: Utc::now()
+            }
+            .retry_after_secs(),
+            0
+        );
+    }
+
+    // ── ModelEvent serialization ─────────────────────────────────────────────
+
+    #[test]
+    fn test_model_event_ready_serializes() {
+        let ts = Utc::now();
+        let e = ModelEvent::ModelReady { loaded_at: ts };
+        let v: Value = serde_json::to_value(&e).unwrap();
+        assert_eq!(v["type"], "model_ready");
+        assert!(v["loaded_at"].is_string());
+    }
+
+    #[test]
+    fn test_model_event_unloaded_serializes() {
+        let v: Value = serde_json::to_value(ModelEvent::ModelUnloaded).unwrap();
+        assert_eq!(v["type"], "model_unloaded");
+    }
+
+    #[test]
+    fn test_model_event_loading_serializes() {
+        let v: Value = serde_json::to_value(ModelEvent::ModelLoading).unwrap();
+        assert_eq!(v["type"], "model_loading");
+    }
+
+    #[test]
+    fn test_model_event_waiting_serializes() {
+        let e = ModelEvent::ModelWaitingForGpu {
+            vram_needed_mb: 3000,
+            vram_free_mb: 200,
+            retry_in_secs: 30,
+        };
+        let v: Value = serde_json::to_value(&e).unwrap();
+        assert_eq!(v["type"], "model_waiting_for_gpu");
+        assert_eq!(v["vram_needed_mb"], 3000);
+    }
+
+    #[test]
+    fn test_model_event_webhook_filter() {
+        assert!(ModelEvent::ModelReady {
+            loaded_at: Utc::now()
+        }
+        .is_webhook_event());
+        assert!(ModelEvent::ModelUnloaded.is_webhook_event());
+        assert!(!ModelEvent::ModelLoading.is_webhook_event());
+        assert!(!ModelEvent::ModelWaitingForGpu {
+            vram_needed_mb: 0,
+            vram_free_mb: 0,
+            retry_in_secs: 30
+        }
+        .is_webhook_event());
+    }
+
+    // ── ModelStatusResponse ──────────────────────────────────────────────────
+
+    #[test]
+    fn test_model_status_response_roundtrip() {
+        let r = ModelStatusResponse {
+            state: ModelState::Ready {
+                loaded_at: Utc::now(),
+            },
+            vram_used_mb: Some(4096),
+            vram_total_mb: Some(8192),
+        };
+        let json_str = serde_json::to_string(&r).unwrap();
+        let v: Value = serde_json::from_str(&json_str).unwrap();
+        assert_eq!(v["state"], "ready");
+        assert_eq!(v["vram_used_mb"], 4096);
+        assert_eq!(v["vram_total_mb"], 8192);
+    }
+
+    #[test]
+    fn test_model_status_response_omits_nulls() {
+        let r = ModelStatusResponse {
+            state: ModelState::Loading,
+            vram_used_mb: None,
+            vram_total_mb: None,
+        };
+        let v: Value = serde_json::to_value(&r).unwrap();
+        assert_eq!(v["state"], "loading");
+        assert!(v.get("vram_used_mb").is_none());
+        assert!(v.get("vram_total_mb").is_none());
+    }
 }
--- a/src/routes/health.rs
+++ b/src/routes/health.rs
@@ -16,13 +16,15 @@ use crate::{models::HealthResponse, AppState, Result};
 )]
 pub async fn health(State(state): State<AppState>) -> Result<Json<HealthResponse>> {
    let (gpu_name, vram_total_mb) = gpu_info(state.gpu_device);
+    let model_state_tag = state.model_state.read().await.tag().to_string();

    Ok(Json(HealthResponse {
-        status:        "ok".into(),
+        status: "ok".into(),
        gpu_name,
        vram_total_mb,
-        model:         state.model_name.to_string(),
-        queue_depth:   state.queue_depth.load(Ordering::Relaxed),
+        model: state.model_name.to_string(),
+        queue_depth: state.queue_depth.load(Ordering::Relaxed),
+        model_state: model_state_tag,
    }))
 }

@@ -48,9 +50,7 @@ fn gpu_info(device: u32) -> (Option<String>, Option<u64>) {
    let mut parts = line.splitn(2, ',');

    let name = parts.next().map(|s| s.trim().to_owned());
-    let vram = parts
-        .next()
-        .and_then(|s| s.trim().parse::<u64>().ok());
+    let vram = parts.next().and_then(|s| s.trim().parse::<u64>().ok());

    (name, vram)
 }
--- a/src/routes/jobs.rs
+++ b/src/routes/jobs.rs
@@ -19,11 +19,12 @@ use uuid::Uuid;

 use crate::{
    models::{Job, JobId, JobStatus, SubmitResponse},
-    worker::{audio_path_for, ProgressEvent},
+    worker::{audio_path_for, ProgressEvent, WorkerCmd},
    AppError, AppState, Result,
 };

-type SseStream = Pin<Box<dyn Stream<Item = std::result::Result<Event, std::convert::Infallible>> + Send>>;
+type SseStream =
+    Pin<Box<dyn Stream<Item = std::result::Result<Event, std::convert::Infallible>> + Send>>;

 // ── POST /jobs ───────────────────────────────────────────────────────────────

@@ -53,18 +54,20 @@ pub async fn submit_job(
    State(state): State<AppState>,
    mut multipart: Multipart,
 ) -> Result<impl IntoResponse> {
-    let mut language:    Option<String> = None;
-    let mut task:        String         = "transcribe".into();
+    let mut language: Option<String> = None;
+    let mut task: String = "transcribe".into();
    let mut webhook_url: Option<String> = None;
-    let mut filename:    Option<String> = None;
+    let mut filename: Option<String> = None;
    let mut audio_saved = false;
    // Assign ID early so we know where to stream the audio bytes.
    let id = Uuid::new_v4();
    let audio_path = audio_path_for(&id);

-    while let Some(field) = multipart.next_field().await.map_err(|e| {
-        AppError::BadRequest(format!("multipart error: {e}"))
-    })? {
+    while let Some(field) = multipart
+        .next_field()
+        .await
+        .map_err(|e| AppError::BadRequest(format!("multipart error: {e}")))?
+    {
        let field_name = field.name().unwrap_or("").to_owned();

        match field_name.as_str() {
@@ -77,9 +80,11 @@ pub async fn submit_job(
                })?;
                let mut bytes_written: u64 = 0;
                let mut stream = field;
-                while let Some(chunk) = stream.chunk().await.map_err(|e| {
-                    AppError::BadRequest(format!("failed to read audio field: {e}"))
-                })? {
+                while let Some(chunk) = stream
+                    .chunk()
+                    .await
+                    .map_err(|e| AppError::BadRequest(format!("failed to read audio field: {e}")))?
+                {
                    file.write_all(&chunk).await.map_err(|e| {
                        AppError::Internal(format!("failed to write audio chunk: {e}"))
                    })?;
@@ -90,10 +95,29 @@ pub async fn submit_job(
                }
                audio_saved = true;
            }
-            "language"    => language    = Some(field.text().await.map_err(|e| AppError::BadRequest(e.to_string()))?),
-            "task"        => task        = field.text().await.map_err(|e| AppError::BadRequest(e.to_string()))?,
-            "webhook_url" => webhook_url = Some(field.text().await.map_err(|e| AppError::BadRequest(e.to_string()))?),
-            _             => {} // ignore unknown fields
+            "language" => {
+                language = Some(
+                    field
+                        .text()
+                        .await
+                        .map_err(|e| AppError::BadRequest(e.to_string()))?,
+                )
+            }
+            "task" => {
+                task = field
+                    .text()
+                    .await
+                    .map_err(|e| AppError::BadRequest(e.to_string()))?
+            }
+            "webhook_url" => {
+                webhook_url = Some(
+                    field
+                        .text()
+                        .await
+                        .map_err(|e| AppError::BadRequest(e.to_string()))?,
+                )
+            }
+            _ => {} // ignore unknown fields
        }
    }

@@ -107,18 +131,54 @@ pub async fn submit_job(
        ));
    }

+    // Check model state before accepting the job.
+    let (model_ready, retry_after_secs, state_tag) = {
+        let ms = state.model_state.read().await;
+        let ready = ms.is_ready();
+        let retry = ms.retry_after_secs();
+        let tag = ms.tag().to_string();
+        (ready, retry, tag)
+    };
+
+    // Register the webhook URL regardless of model state — so model lifecycle
+    // events are delivered even if the job itself is rejected.
+    if let Some(url) = &webhook_url {
+        state
+            .webhook_registry
+            .lock()
+            .unwrap_or_else(|e| e.into_inner())
+            .insert(url.clone());
+    }
+
+    if !model_ready {
+        // Trigger a load if the model is simply unloaded (not already loading).
+        if state_tag == "unloaded" {
+            let _ = state.cmd_tx.try_send(WorkerCmd::Load);
+        }
+        // Clean up the audio file we already wrote to disk.
+        let _ = tokio::fs::remove_file(&audio_path).await;
+        return Err(AppError::ModelNotReady {
+            state: state_tag,
+            retry_after_secs,
+        });
+    }
+
    let mut job = Job::new(id, task, webhook_url, filename);
    job.language = language;

    state.storage.create(&job).await?;

    // Pre-create the broadcast channel so SSE subscribers don't miss events.
-    state.progress.entry(id).or_insert_with(|| broadcast::channel(64).0);
+    state
+        .progress
+        .entry(id)
+        .or_insert_with(|| broadcast::channel(64).0);

    state.queue_depth.fetch_add(1, Ordering::Relaxed);
-    state.job_tx.send(id).map_err(|_| {
-        AppError::Internal("worker channel closed".into())
-    })?;
+    state
+        .job_tx
+        .send(id)
+        .map_err(|_| AppError::Internal("worker channel closed".into()))?;

    tracing::info!(job_id = %id, "job queued");

@@ -138,10 +198,7 @@ pub async fn submit_job(
        (status = 404, description = "Not found"),
    )
 )]
-pub async fn get_job(
-    State(state): State<AppState>,
-    Path(id):     Path<JobId>,
-) -> Result<Json<Job>> {
+pub async fn get_job(State(state): State<AppState>, Path(id): Path<JobId>) -> Result<Json<Job>> {
    let job = state.storage.get(&id).await?;
    Ok(Json(job))
 }
@@ -166,15 +223,15 @@ pub async fn get_job(
 )]
 pub async fn stream_job(
    State(state): State<AppState>,
-    Path(id):     Path<JobId>,
+    Path(id): Path<JobId>,
 ) -> Result<Sse<SseStream>> {
    // If the job is already finished, return a single done event immediately.
    let job = state.storage.get(&id).await?;
    match job.status {
        JobStatus::Done | JobStatus::Failed | JobStatus::Cancelled => {
-            let payload = serde_json::to_string(
-                &crate::models::SsePayload::Done { job: Box::new(job) }
-            ).unwrap_or_default();
+            let payload =
+                serde_json::to_string(&crate::models::SsePayload::Done { job: Box::new(job) })
+                    .unwrap_or_default();
            let s: SseStream = Box::pin(stream::once(async move {
                Ok(Event::default().event("done").data(payload))
            }));
@@ -192,22 +249,28 @@ pub async fn stream_job(

    let sse_stream: SseStream = Box::pin(BroadcastStream::new(rx).filter_map(|msg| async move {
        let event = match msg {
-            Ok(ProgressEvent::Progress { percent, chunk, total }) => {
-                let payload = serde_json::to_string(
-                    &crate::models::SsePayload::Progress { percent, chunk, chunks_total: total }
-                ).ok()?;
+            Ok(ProgressEvent::Progress {
+                percent,
+                chunk,
+                total,
+            }) => {
+                let payload = serde_json::to_string(&crate::models::SsePayload::Progress {
+                    percent,
+                    chunk,
+                    chunks_total: total,
+                })
+                .ok()?;
                Event::default().event("progress").data(payload)
            }
            Ok(ProgressEvent::Done(job)) => {
-                let payload = serde_json::to_string(
-                    &crate::models::SsePayload::Done { job }
-                ).ok()?;
+                let payload =
+                    serde_json::to_string(&crate::models::SsePayload::Done { job }).ok()?;
                Event::default().event("done").data(payload)
            }
            Ok(ProgressEvent::Error(msg)) => {
-                let payload = serde_json::to_string(
-                    &crate::models::SsePayload::Error { message: msg }
-                ).ok()?;
+                let payload =
+                    serde_json::to_string(&crate::models::SsePayload::Error { message: msg })
+                        .ok()?;
                Event::default().event("error").data(payload)
            }
            Err(_) => return None, // lagged / channel closed
@@ -234,10 +297,7 @@ pub async fn stream_job(
        (status = 409, description = "Job already finished"),
    )
 )]
-pub async fn delete_job(
-    State(state): State<AppState>,
-    Path(id):     Path<JobId>,
-) -> Result<Json<Job>> {
+pub async fn delete_job(State(state): State<AppState>, Path(id): Path<JobId>) -> Result<Json<Job>> {
    let mut job = state.storage.get(&id).await?;

    match job.status {
@@ -250,7 +310,7 @@ pub async fn delete_job(
        _ => {}
    }

-    job.status       = JobStatus::Cancelled;
+    job.status = JobStatus::Cancelled;
    job.completed_at = Some(Utc::now());
    state.storage.save(&job).await?;

--- a/src/routes/mod.rs
+++ b/src/routes/mod.rs
@@ -1,19 +1,34 @@
 pub mod health;
 pub mod jobs;
+pub mod model;

-use axum::{extract::DefaultBodyLimit, routing::{delete, get, post}, Router};
 use crate::AppState;
+use axum::{
+    extract::DefaultBodyLimit,
+    routing::{delete, get, post},
+    Router,
+};

 pub fn jobs_router() -> Router<AppState> {
    Router::new()
        // No body limit on the upload route — files can be multiple GB.
-        .route("/jobs", post(jobs::submit_job).layer(DefaultBodyLimit::disable()))
-        .route("/jobs/:id",        get(jobs::get_job))
+        .route(
+            "/jobs",
+            post(jobs::submit_job).layer(DefaultBodyLimit::disable()),
+        )
+        .route("/jobs/:id", get(jobs::get_job))
        .route("/jobs/:id/stream", get(jobs::stream_job))
-        .route("/jobs/:id",        delete(jobs::delete_job))
+        .route("/jobs/:id", delete(jobs::delete_job))
 }

 pub fn health_router() -> Router<AppState> {
-    Router::new()
-        .route("/health", get(health::health))
+    Router::new().route("/health", get(health::health))
+}
+
+pub fn model_router() -> Router<AppState> {
+    Router::new()
+        .route("/model/status", get(model::model_status))
+        .route("/model/load", post(model::model_load))
+        .route("/model/unload", post(model::model_unload))
+        .route("/model/events", get(model::model_events))
 }
--- a/src/routes/model.rs
+++ b/src/routes/model.rs
@@ -0,0 +1,244 @@
+use std::pin::Pin;
+
+use axum::{
+    extract::State,
+    http::StatusCode,
+    response::{
+        sse::{Event, KeepAlive, Sse},
+        IntoResponse,
+    },
+    Json,
+};
+use futures::Stream;
+use futures::StreamExt;
+use tokio_stream::wrappers::BroadcastStream;
+
+use crate::{
+    models::{ModelEvent, ModelStatusResponse},
+    worker::WorkerCmd,
+    AppState, Result,
+};
+
+type SseStream =
+    Pin<Box<dyn Stream<Item = std::result::Result<Event, std::convert::Infallible>> + Send>>;
+
+// ── GET /model/status ────────────────────────────────────────────────────────
+
+/// Return the current model state and VRAM statistics.
+#[utoipa::path(
+    get,
+    path = "/model/status",
+    tag  = "model",
+    responses(
+        (status = 200, description = "Model status", body = ModelStatusResponse),
+    )
+)]
+pub async fn model_status(State(state): State<AppState>) -> Result<Json<ModelStatusResponse>> {
+    let model_state = state.model_state.read().await.clone();
+    let (vram_used_mb, vram_total_mb) = vram_stats(state.gpu_device);
+
+    Ok(Json(ModelStatusResponse {
+        state: model_state,
+        vram_used_mb,
+        vram_total_mb,
+    }))
+}
+
+// ── POST /model/load ─────────────────────────────────────────────────────────
+
+/// Request the model to be loaded into GPU memory.
+/// Idempotent: if the model is already loading or ready, this is a no-op.
+/// Returns 202 Accepted; poll `GET /model/status` or subscribe to
+/// `GET /model/events` to know when it is ready.
+#[utoipa::path(
+    post,
+    path = "/model/load",
+    tag  = "model",
+    responses(
+        (status = 202, description = "Load initiated or already in progress"),
+        (status = 200, description = "Model already ready"),
+    )
+)]
+pub async fn model_load(State(state): State<AppState>) -> impl IntoResponse {
+    let is_ready = state.model_state.read().await.is_ready();
+    if is_ready {
+        return (
+            StatusCode::OK,
+            Json(serde_json::json!({"status": "already_ready"})),
+        );
+    }
+    // Ignore send errors (channel full = load already in progress).
+    let _ = state.cmd_tx.try_send(WorkerCmd::Load);
+    (
+        StatusCode::ACCEPTED,
+        Json(serde_json::json!({"status": "load_initiated"})),
+    )
+}
+
+// ── POST /model/unload ───────────────────────────────────────────────────────
+
+/// Unload the model from GPU memory immediately.
+/// Idempotent: if the model is already unloaded, returns 200 immediately.
+#[utoipa::path(
+    post,
+    path = "/model/unload",
+    tag  = "model",
+    responses(
+        (status = 200, description = "Model unloaded or was already unloaded"),
+    )
+)]
+pub async fn model_unload(State(state): State<AppState>) -> impl IntoResponse {
+    if !matches!(
+        *state.model_state.read().await,
+        crate::models::ModelState::Unloaded
+    ) {
+        let _ = state.cmd_tx.try_send(WorkerCmd::Unload);
+    }
+    (
+        StatusCode::OK,
+        Json(serde_json::json!({"status": "unload_requested"})),
+    )
+}
+
+// ── GET /model/events ────────────────────────────────────────────────────────
+
+/// Subscribe to model lifecycle events via Server-Sent Events.
+///
+/// Event types:
+/// - `model_loading`          — load initiated
+/// - `model_ready`            — model loaded and warmed up
+/// - `model_unloaded`         — model freed from GPU memory
+/// - `model_waiting_for_gpu`  — insufficient VRAM; retrying
+#[utoipa::path(
+    get,
+    path = "/model/events",
+    tag  = "model",
+    responses(
+        (status = 200, description = "SSE stream of model lifecycle events"),
+    )
+)]
+pub async fn model_events(State(state): State<AppState>) -> Sse<SseStream> {
+    let rx = state.model_event_tx.subscribe();
+
+    let stream: SseStream = Box::pin(BroadcastStream::new(rx).filter_map(|msg| async move {
+        match msg {
+            Ok(event) => {
+                let event_type = match &event {
+                    ModelEvent::ModelReady { .. } => "model_ready",
+                    ModelEvent::ModelUnloaded => "model_unloaded",
+                    ModelEvent::ModelLoading => "model_loading",
+                    ModelEvent::ModelWaitingForGpu { .. } => "model_waiting_for_gpu",
+                };
+                let data = serde_json::to_string(&event).ok()?;
+                Some(Ok(Event::default().event(event_type).data(data)))
+            }
+            Err(_) => None,
+        }
+    }));
+
+    Sse::new(stream).keep_alive(KeepAlive::default())
+}
+
+// ── Helpers ───────────────────────────────────────────────────────────────────
+
+fn vram_stats(gpu_device: u32) -> (Option<u64>, Option<u64>) {
+    fn inner(gpu_device: u32) -> Option<(u64, u64)> {
+        let out = std::process::Command::new("nvidia-smi")
+            .args([
+                &format!("--id={gpu_device}"),
+                "--query-gpu=memory.used,memory.total",
+                "--format=csv,noheader,nounits",
+            ])
+            .output()
+            .ok()?;
+
+        if !out.status.success() {
+            return None;
+        }
+
+        let line = String::from_utf8_lossy(&out.stdout);
+        let line = line.trim();
+        let mut parts = line.splitn(2, ',');
+        let used = parts.next().and_then(|s| s.trim().parse::<u64>().ok())?;
+        let total = parts.next().and_then(|s| s.trim().parse::<u64>().ok())?;
+        Some((used, total))
+    }
+
+    match inner(gpu_device) {
+        Some((u, t)) => (Some(u), Some(t)),
+        None => (None, None),
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::sync::{atomic::AtomicUsize, Arc, Mutex};
+
+    use axum::response::IntoResponse;
+    use chrono::Utc;
+    use tempfile::tempdir;
+    use tokio::sync::{broadcast, mpsc, RwLock};
+
+    use crate::{models::ModelState, storage::Storage, worker::ProgressRegistry, AppState};
+
+    async fn test_state(
+        model_state: ModelState,
+    ) -> (
+        AppState,
+        tempfile::TempDir,
+        std::sync::mpsc::Receiver<WorkerCmd>,
+    ) {
+        let tmp = tempdir().expect("tempdir");
+        let storage = Arc::new(Storage::new(tmp.path()).await.expect("storage"));
+        let (job_tx, _job_rx) = mpsc::unbounded_channel();
+        let (cmd_tx, cmd_rx) = std::sync::mpsc::sync_channel(8);
+        let progress: ProgressRegistry = Arc::new(dashmap::DashMap::new());
+        let (model_event_tx, _) = broadcast::channel(8);
+
+        (
+            AppState {
+                job_tx,
+                cmd_tx,
+                storage,
+                progress,
+                model_name: "test".into(),
+                queue_depth: Arc::new(AtomicUsize::new(0)),
+                gpu_device: 0,
+                model_state: Arc::new(RwLock::new(model_state)),
+                model_event_tx,
+                webhook_registry: Arc::new(Mutex::new(Default::default())),
+                idle_timeout: std::time::Duration::from_secs(300),
+                gpu_poll_interval: std::time::Duration::from_secs(30),
+            },
+            tmp,
+            cmd_rx,
+        )
+    }
+
+    #[tokio::test]
+    async fn test_model_unload_skips_command_when_already_unloaded() {
+        let (state, _tmp, cmd_rx) = test_state(ModelState::Unloaded).await;
+
+        let response = model_unload(State(state)).await.into_response();
+
+        assert_eq!(response.status(), StatusCode::OK);
+        assert!(
+            cmd_rx.try_recv().is_err(),
+            "unexpected unload command queued"
+        );
+    }
+
+    #[tokio::test]
+    async fn test_model_unload_queues_command_when_ready() {
+        let (state, _tmp, cmd_rx) = test_state(ModelState::Ready {
+            loaded_at: Utc::now(),
+        })
+        .await;
+
+        let response = model_unload(State(state)).await.into_response();
+
+        assert_eq!(response.status(), StatusCode::OK);
+        assert!(matches!(cmd_rx.try_recv(), Ok(WorkerCmd::Unload)));
+    }
+}
--- a/src/storage.rs
+++ b/src/storage.rs
@@ -30,20 +30,20 @@ impl Storage {
    // ── CRUD ─────────────────────────────────────────────────────────────────

    pub async fn create(&self, job: &Job) -> Result<()> {
-        let path    = self.job_path(&job.id);
-        let payload = serde_json::to_vec_pretty(job)
-            .map_err(|e| AppError::Internal(e.to_string()))?;
-        fs::write(&path, payload).await.map_err(|e| {
-            AppError::Internal(format!("failed to write job {}: {e}", job.id))
-        })?;
+        let path = self.job_path(&job.id);
+        let payload =
+            serde_json::to_vec_pretty(job).map_err(|e| AppError::Internal(e.to_string()))?;
+        fs::write(&path, payload)
+            .await
+            .map_err(|e| AppError::Internal(format!("failed to write job {}: {e}", job.id)))?;
        Ok(())
    }

    pub async fn get(&self, id: &JobId) -> Result<Job> {
        let path = self.job_path(id);
-        let raw  = fs::read(&path).await.map_err(|_| {
-            AppError::NotFound(format!("job {id} not found"))
-        })?;
+        let raw = fs::read(&path)
+            .await
+            .map_err(|_| AppError::NotFound(format!("job {id} not found")))?;
        serde_json::from_slice(&raw).map_err(|e| AppError::Internal(e.to_string()))
    }

@@ -54,22 +54,24 @@ impl Storage {

    pub async fn delete(&self, id: &JobId) -> Result<()> {
        let path = self.job_path(id);
-        fs::remove_file(&path).await.map_err(|_| {
-            AppError::NotFound(format!("job {id} not found"))
-        })?;
+        fs::remove_file(&path)
+            .await
+            .map_err(|_| AppError::NotFound(format!("job {id} not found")))?;
        Ok(())
    }

    /// List all job IDs present on disk.
    pub async fn list_ids(&self) -> Result<Vec<JobId>> {
-        let mut entries = fs::read_dir(&self.dir).await.map_err(|e| {
-            AppError::Internal(format!("read_dir failed: {e}"))
-        })?;
+        let mut entries = fs::read_dir(&self.dir)
+            .await
+            .map_err(|e| AppError::Internal(format!("read_dir failed: {e}")))?;

        let mut ids = Vec::new();
-        while let Some(entry) = entries.next_entry().await.map_err(|e| {
-            AppError::Internal(e.to_string())
-        })? {
+        while let Some(entry) = entries
+            .next_entry()
+            .await
+            .map_err(|e| AppError::Internal(e.to_string()))?
+        {
            let name = entry.file_name();
            let name = name.to_string_lossy();
            if let Some(stem) = name.strip_suffix(".json") {
@@ -88,8 +90,8 @@ impl Storage {
            if let Ok(mut job) = self.get(&id).await {
                if job.status == JobStatus::Running {
                    tracing::warn!(job_id = %id, "recovering interrupted job → failed");
-                    job.status       = JobStatus::Failed;
-                    job.error        = Some("server restarted while job was running".into());
+                    job.status = JobStatus::Failed;
+                    job.error = Some("server restarted while job was running".into());
                    job.completed_at = Some(chrono::Utc::now());
                    let _ = self.save(&job).await;
                }
--- a/src/transcriber.rs
+++ b/src/transcriber.rs
@@ -1,7 +1,7 @@
 use std::path::Path;

 use whisper_rs::{
-    FullParams, SamplingStrategy, WhisperContext, WhisperContextParameters,
+    FullParams, SamplingStrategy, WhisperContext, WhisperContextParameters, WhisperState,
 };

 use crate::{
@@ -9,18 +9,38 @@ use crate::{
    AppError, Result,
 };

-/// Wraps a loaded whisper.cpp context.
-/// `WhisperContext` is `Send` but **not** `Sync` — keep it on the worker thread.
+/// Wraps a loaded whisper.cpp context and a single reusable inference state.
+///
+/// `WhisperState` allocates ~700 MB of GPU compute buffers (KV caches, CUDA
+/// workspace) via `whisper_init_state`. Creating a new state for every chunk
+/// causes repeated GPU re-initialisation and VRAM allocation churn, which
+/// manifests as intermittent CUDA allocation failures → 0 segments returned.
+///
+/// By creating the state once at load time and reusing it, GPU memory is
+/// stable and inference is reliable across all chunks.
+///
+/// Safety: `WhisperState` is `Send + Sync` (explicitly declared in whisper-rs).
+/// This struct lives on the single `whisper-gpu` OS thread and is never shared.
 pub struct Transcriber {
-    ctx: WhisperContext,
+    // WhisperContext is not stored after load: WhisperState holds its own
+    // Arc<WhisperInnerContext>, so the model weights remain in memory for
+    // the lifetime of the state even after the originating context is dropped.
+    state: WhisperState,
 }

 impl Transcriber {
-    /// Load a GGML model file and configure GPU for RTX 2080.
+    /// Load a GGML model file, configure GPU, and run a warmup inference.
+    ///
+    /// The warmup is critical: CUDA JIT-compiles its kernels on the FIRST call to
+    /// `whisper_full_with_state`. Without warmup, the first real job triggers JIT
+    /// compilation mid-inference, which can cause the call to return in ~0.5s with
+    /// 0 segments. The warmup forces kernel compilation at startup so all subsequent
+    /// jobs run correctly from the very first request.
    pub fn load(model_path: impl AsRef<Path>, gpu_device: u32) -> Result<Self> {
-        let path = model_path.as_ref().to_str().ok_or_else(|| {
-            AppError::Internal("model path is not valid UTF-8".into())
-        })?;
+        let path = model_path
+            .as_ref()
+            .to_str()
+            .ok_or_else(|| AppError::Internal("model path is not valid UTF-8".into()))?;

        let mut params = WhisperContextParameters::new();
        params.use_gpu(true);
@@ -29,28 +49,61 @@ impl Transcriber {
        // real-world audio (conference recordings, noisy MP3s).
        // params.flash_attn(true);

-        let ctx = WhisperContext::new_with_params(path, params)
-            .map_err(|e| AppError::Internal(format!("failed to load model: {e}")))?;
+        let ctx = WhisperContext::new_with_params(path, params).map_err(|e| {
+            let msg = format!("failed to load model: {e}");
+            if AppError::is_oom(&msg) {
+                AppError::OutOfMemory(msg)
+            } else {
+                AppError::Internal(msg)
+            }
+        })?;

-        tracing::info!(model = path, "whisper model loaded");
-        Ok(Self { ctx })
+        let mut state = ctx.create_state().map_err(|e| {
+            let msg = format!("failed to create whisper state: {e}");
+            if AppError::is_oom(&msg) {
+                AppError::OutOfMemory(msg)
+            } else {
+                AppError::Internal(msg)
+            }
+        })?;
+        // ctx drops here; state holds Arc<WhisperInnerContext> so model stays loaded.
+
+        // ── GPU warmup ────────────────────────────────────────────────────────
+        // Run a silent 1-second inference to force CUDA JIT kernel compilation.
+        // Expected result: 0 segments (silence). The point is the side effect:
+        // all CUDA kernels are compiled and cached before the first real job arrives.
+        tracing::info!(model = path, "warming up GPU — compiling CUDA kernels...");
+        let silence = vec![0.0f32; 16_000]; // 1s @ 16 kHz
+        let mut wp = FullParams::new(SamplingStrategy::Greedy { best_of: 1 });
+        wp.set_language(Some("en"));
+        wp.set_print_progress(false);
+        wp.set_print_realtime(false);
+        wp.set_suppress_blank(true);
+        wp.set_no_context(true);
+        let _ = state.full(wp, &silence); // ignore result — 0 segments expected
+        tracing::info!("GPU warmup complete — ready for inference");
+
+        Ok(Self { state })
    }

    /// Transcribe 16 kHz mono f32 PCM samples.
    /// `on_progress` receives 0–100 from whisper.cpp.
+    ///
+    /// The inference state (`self.state`) is reused across calls. GPU compute
+    /// buffers remain allocated, eliminating per-chunk `whisper_init_state` overhead.
+    /// `no_context=true` in the params prevents KV-cache contamination between chunks.
    pub fn transcribe(
-        &self,
-        pcm:         &[f32],
-        language:    Option<&str>,
-        task:        &str,
+        &mut self,
+        pcm: &[f32],
+        language: Option<&str>,
+        task: &str,
        on_progress: impl Fn(u8) + Send + 'static,
    ) -> Result<(Vec<Segment>, String)> {
-        let mut state = self.ctx.create_state()
-            .map_err(|e| AppError::Internal(format!("create_state: {e}")))?;
+        let state = &mut self.state;

        let mut fp = FullParams::new(SamplingStrategy::BeamSearch {
            beam_size: 5,
-            patience:  1.0,
+            patience: 1.0,
        });

        fp.set_n_threads(num_cpus::get() as i32);
@@ -104,40 +157,55 @@ impl Transcriber {
            .full(fp, pcm)
            .map_err(|e| AppError::Internal(format!("transcription failed: {e}")))?;

-        let n_segments = state.full_n_segments()
+        let n_segments = state
+            .full_n_segments()
            .map_err(|e| AppError::Internal(e.to_string()))?;

        let mut segments = Vec::with_capacity(n_segments as usize);

        for i in 0..n_segments {
-            let text  = state.full_get_segment_text(i)
+            let text = state
+                .full_get_segment_text(i)
                .map_err(|e| AppError::Internal(e.to_string()))?;
-            let start = state.full_get_segment_t0(i)
-                .map_err(|e| AppError::Internal(e.to_string()))? as f32 / 100.0;
-            let end   = state.full_get_segment_t1(i)
-                .map_err(|e| AppError::Internal(e.to_string()))? as f32 / 100.0;
+            let start = state
+                .full_get_segment_t0(i)
+                .map_err(|e| AppError::Internal(e.to_string()))? as f32
+                / 100.0;
+            let end = state
+                .full_get_segment_t1(i)
+                .map_err(|e| AppError::Internal(e.to_string()))? as f32
+                / 100.0;

-            let n_tokens = state.full_n_tokens(i)
+            let n_tokens = state
+                .full_n_tokens(i)
                .map_err(|e| AppError::Internal(e.to_string()))?;

            let mut words = Vec::new();
            for t in 0..n_tokens {
-                let token_text = state.full_get_token_text(i, t)
+                let token_text = state
+                    .full_get_token_text(i, t)
                    .map_err(|e| AppError::Internal(e.to_string()))?;
                if token_text.starts_with('[') {
                    continue; // skip special tokens ([MUSIC], [APPLAUSE], etc.)
                }
-                let data = state.full_get_token_data(i, t)
+                let data = state
+                    .full_get_token_data(i, t)
                    .map_err(|e| AppError::Internal(e.to_string()))?;
                words.push(Word {
-                    text:        token_text,
-                    start:       data.t0 as f32 / 100.0,
-                    end:         data.t1 as f32 / 100.0,
+                    text: token_text,
+                    start: data.t0 as f32 / 100.0,
+                    end: data.t1 as f32 / 100.0,
                    probability: data.p,
                });
            }

-            segments.push(Segment { index: i, start, end, text, words });
+            segments.push(Segment {
+                index: i,
+                start,
+                end,
+                text,
+                words,
+            });
        }

        let lang = state
--- a/src/worker.rs
+++ b/src/worker.rs
--- a/test_all.sh
+++ b/test_all.sh
@@ -1,120 +1,173 @@
 #!/usr/bin/env bash
 set -euo pipefail
-BASE="http://localhost:8090"
-AUDIO="/home/moze/Sources/youtube-transcriber/docker/tmp/audio-b2167046-a236-4fcd-b739-78177542fd23.wav"
+
+# ── Config — override via env vars ───────────────────────────────────────────
+BASE="${WHISPER_BASE_URL:-http://localhost:8080}"
+AUDIO="${TEST_AUDIO:-/home/moze/Sources/youtube-transcriber/docker/tmp/audio-b2167046-a236-4fcd-b739-78177542fd23.wav}"
+
 GREEN='\033[0;32m'; RED='\033[0;31m'; NC='\033[0m'
+FAILS=0
 ok()  { echo -e "${GREEN}[PASS]${NC} $*"; }
-fail(){ echo -e "${RED}[FAIL]${NC} $*"; exit 1; }
+fail(){ echo -e "${RED}[FAIL]${NC} $*"; FAILS=$((FAILS + 1)); }
+
+echo "=== Whisper API test suite ==="
+echo "    BASE  : $BASE"
+echo "    AUDIO : $AUDIO"
+echo ""

 echo "=== 1. GET /health ==="
 HEALTH=$(curl -sf "$BASE/health")
 echo "$HEALTH" | python3 -m json.tool
-echo "$HEALTH" | python3 -c "import sys,json; d=json.load(sys.stdin); assert d['status']=='ok'" && ok "health"
+python3 -c "
+import sys, json
+d = json.loads('$HEALTH' if False else sys.stdin.read())
+assert d['status'] == 'ok', f'status={d[\"status\"]}'
+assert 'model_state' in d, 'model_state field missing from health response'
+" <<< "$HEALTH" && ok "health ok + model_state present" || fail "health check"

 echo ""
 echo "=== 2. GET /docs (Swagger UI reachable) ==="
-curl -sf "$BASE/docs" | grep -q "swagger" && ok "swagger UI"
+curl -sf "$BASE/docs" | grep -qi "swagger" && ok "swagger UI reachable" || fail "swagger UI"

 echo ""
-echo "=== 3. Webhook server (background nc loop) ==="
-# Simple webhook receiver using Python
-python3 - &
-WEBHOOK_PID=$!
+echo "=== 3. Webhook receiver (background Python HTTP server) ==="
 cat > /tmp/webhook_receiver.py << 'PYEOF'
-import http.server, json, sys
+import http.server, json, sys, signal

 class H(http.server.BaseHTTPRequestHandler):
    def do_POST(self):
        n = int(self.headers.get('Content-Length', 0))
        body = self.rfile.read(n)
-        print("\n[WEBHOOK] received:", json.dumps(json.loads(body), indent=2)[:500])
+        data = json.loads(body)
+        print(f"\n[WEBHOOK] status={data.get('status')} segments={len(data.get('segments', []))}", flush=True)
        self.send_response(200)
        self.end_headers()
    def log_message(self, *a): pass

-print("[WEBHOOK] listening on :9999")
+signal.signal(signal.SIGTERM, lambda *_: sys.exit(0))
+print("[WEBHOOK] listening on :9999", flush=True)
 http.server.HTTPServer(('', 9999), H).serve_forever()
 PYEOF
-kill $WEBHOOK_PID 2>/dev/null || true
 python3 /tmp/webhook_receiver.py &
 WEBHOOK_PID=$!
 sleep 1
 echo "Webhook receiver started (PID $WEBHOOK_PID)"

 echo ""
-echo "=== 4. DELETE a non-existent job → 404 ==="
-STATUS=$(curl -s -o /dev/null -w "%{http_code}" -X DELETE "$BASE/jobs/00000000-0000-0000-0000-000000000000")
-[ "$STATUS" = "404" ] && ok "DELETE 404 for unknown job" || fail "expected 404 got $STATUS"
+echo "=== 4. GET /model/status — expect unloaded on fresh start ==="
+MODEL_STATUS=$(curl -sf "$BASE/model/status")
+echo "$MODEL_STATUS" | python3 -m json.tool
+echo "$MODEL_STATUS" | python3 -c "
+import sys, json
+d = json.load(sys.stdin)
+assert 'state' in d, 'state field missing from /model/status'
+print(f'  model state: {d[\"state\"]}')
+" && ok "/model/status has state field" || fail "/model/status schema"

 echo ""
-echo "=== 5. POST /jobs — submit audio ==="
+echo "=== 5. POST /model/load — trigger model load ==="
+LOAD_RESP=$(curl -sf -X POST "$BASE/model/load")
+echo "$LOAD_RESP"
+ok "POST /model/load accepted"
+
+echo ""
+echo "=== 6. Poll /model/status until ready (max 3 min) ==="
+LOAD_ELAPSED=0
+while true; do
+  sleep 5
+  LOAD_ELAPSED=$((LOAD_ELAPSED + 5))
+  MS=$(curl -sf "$BASE/model/status")
+  STATE=$(echo "$MS" | python3 -c "import sys,json; print(json.load(sys.stdin)['state'])")
+  echo "  [${LOAD_ELAPSED}s] model_state=${STATE}"
+  if [ "$STATE" = "ready" ]; then
+    ok "model loaded and ready in ${LOAD_ELAPSED}s"
+    break
+  fi
+  [ $LOAD_ELAPSED -gt 180 ] && { fail "model failed to load within 3 minutes"; break; }
+done
+
+echo ""
+echo "=== 7. DELETE a non-existent job → 404 ==="
+STATUS=$(curl -s -o /dev/null -w "%{http_code}" -X DELETE "$BASE/jobs/00000000-0000-0000-0000-000000000000")
+[ "$STATUS" = "404" ] && ok "DELETE unknown job → 404" || fail "expected 404, got $STATUS"
+
+echo ""
+echo "=== 8. POST /jobs — submit audio ==="
 SUBMIT=$(curl -sf -X POST "$BASE/jobs" \
  -F "audio=@${AUDIO};type=audio/wav" \
-  -F "language=auto" \
  -F "task=transcribe" \
  -F "webhook_url=http://localhost:9999/webhook")
 echo "$SUBMIT"
-JOB_ID=$(echo "$SUBMIT" | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")
+JOB_ID=$(echo "$SUBMIT" | python3 -c "import sys,json; print(json.load(sys.stdin)['job_id'])")
 ok "submitted job $JOB_ID"

 echo ""
-echo "=== 6. GET /jobs/{id} immediately after submit ==="
+echo "=== 9. GET /jobs/{id} immediately after submit ==="
 JOB=$(curl -sf "$BASE/jobs/$JOB_ID")
-echo "$JOB" | python3 -c "import sys,json; d=json.load(sys.stdin); assert d['status'] in ('queued','running')" \
-  && ok "status is queued/running"
+echo "$JOB" | python3 -c "
+import sys, json
+d = json.load(sys.stdin)
+assert d['status'] in ('queued', 'running'), f'unexpected status: {d[\"status\"]}'
+" && ok "status is queued/running" || fail "initial status check"

 echo ""
-echo "=== 7. SSE stream (first 15 events then detach) ==="
+echo "=== 10. SSE stream (observe first 30 events then detach) ==="
 echo "Subscribing to SSE stream for $JOB_ID …"
-curl -sN --max-time 60 "$BASE/jobs/$JOB_ID/stream" | head -30 &
+curl -sN --max-time 90 "$BASE/jobs/$JOB_ID/stream" | head -60 &
 SSE_PID=$!

 echo ""
-echo "=== 8. Poll until done (max 20 min) ==="
-SECONDS=0
+echo "=== 11. Poll until done (max 20 min) ==="
+ELAPSED=0
 while true; do
  sleep 15
+  ELAPSED=$((ELAPSED + 15))
  JOB=$(curl -sf "$BASE/jobs/$JOB_ID")
  STATUS=$(echo "$JOB" | python3 -c "import sys,json; print(json.load(sys.stdin)['status'])")
-  echo "  [${SECONDS}s] status=$STATUS"
+  PROGRESS=$(echo "$JOB" | python3 -c "import sys,json; print(json.load(sys.stdin).get('progress',0))")
+  echo "  [${ELAPSED}s] status=$STATUS progress=${PROGRESS}%"
  if [ "$STATUS" = "done" ]; then
-    ok "job finished in ${SECONDS}s"
+    ok "job finished in ${ELAPSED}s"
    break
  elif [ "$STATUS" = "failed" ]; then
    echo "$JOB" | python3 -m json.tool
    fail "job failed"
+    break
  fi
-  [ $SECONDS -gt 1200 ] && fail "timeout after 20 minutes"
+  [ $ELAPSED -gt 1200 ] && { fail "timeout after 20 minutes"; break; }
 done
 kill $SSE_PID 2>/dev/null || true

 echo ""
-echo "=== 9. Inspect transcription quality ==="
+echo "=== 12. Inspect transcription quality ==="
 RESULT=$(curl -sf "$BASE/jobs/$JOB_ID")
-echo "$RESULT" | python3 - << 'PYCHECK'
+TMPJSON=$(mktemp /tmp/whisper_test_XXXXXX.json)
+echo "$RESULT" > "$TMPJSON"
+python3 - "$TMPJSON" << 'PYCHECK'
 import sys, json, re

-data = json.loads(sys.stdin.read())
+with open(sys.argv[1]) as f:
+    data = json.load(f)
 segments = data.get("segments", [])
 print(f"  Language  : {data.get('language')}")
 print(f"  Duration  : {data.get('duration_secs')}s")
 print(f"  Segments  : {len(segments)}")

-issues = []
+if not segments:
+    print("  ✗ ZERO SEGMENTS — transcription likely failed silently")
+    sys.exit(1)

+issues = []
 for i, seg in enumerate(segments):
    text = seg.get("text", "")
-    # --- repetition loop ---
    words = text.strip().split()
    if len(words) >= 6:
        half = len(words) // 2
        if words[:half] == words[half:half+half]:
            issues.append(f"  [seg {i}] REPETITION LOOP: {text[:80]}")
-    # --- long duplicate phrases ---
    phrases = re.findall(r'(\b\w+ \w+ \w+\b)', text)
    if len(phrases) != len(set(phrases)) and len(phrases) > 4:
        issues.append(f"  [seg {i}] DUPLICATE PHRASE: {text[:80]}")
-    # --- blank/empty segment ---
    if not text.strip():
        issues.append(f"  [seg {i}] BLANK SEGMENT")

@@ -125,31 +178,52 @@ if issues:
 else:
    print("\n  ✓ No repetition loops or blank segments detected")

-# Print first 5 segments as sample
-print("\n  Sample output:")
+print("\n  Sample output (first 5 segments):")
 for seg in segments[:5]:
    print(f"    [{seg['start']:.1f}–{seg['end']:.1f}] {seg['text'][:100]}")
 PYCHECK
+PYEXIT=$?
+rm -f "$TMPJSON"
+[ $PYEXIT -eq 0 ] && ok "quality check passed" || fail "quality check"

 echo ""
-echo "=== 10. DELETE completed job ==="
-STATUS=$(curl -s -o /dev/null -w "%{http_code}" -X DELETE "$BASE/jobs/$JOB_ID")
-[ "$STATUS" = "204" ] || [ "$STATUS" = "200" ] && ok "DELETE returned $STATUS"
+echo "=== 13. DELETE completed job → 409 Conflict ==="
+DEL_STATUS=$(curl -s -o /dev/null -w "%{http_code}" -X DELETE "$BASE/jobs/$JOB_ID")
+[ "$DEL_STATUS" = "409" ] && ok "DELETE completed job → 409 Conflict (expected)" \
+  || echo "  [INFO] DELETE returned $DEL_STATUS"

 echo ""
-echo "=== 11. Submit + immediately cancel a job ==="
+echo "=== 14. Submit + cancel a queued job ==="
 JOB2=$(curl -sf -X POST "$BASE/jobs" \
  -F "audio=@${AUDIO};type=audio/wav" \
  -F "language=en" \
  -F "task=transcribe")
-JOB2_ID=$(echo "$JOB2" | python3 -c "import sys,json; print(json.load(sys.stdin)['id'])")
+JOB2_ID=$(echo "$JOB2" | python3 -c "import sys,json; print(json.load(sys.stdin)['job_id'])")
 sleep 1
-DEL_STATUS=$(curl -s -o /dev/null -w "%{http_code}" -X DELETE "$BASE/jobs/$JOB2_ID")
+curl -s -X DELETE "$BASE/jobs/$JOB2_ID" > /dev/null
 CANCEL_STATUS=$(curl -sf "$BASE/jobs/$JOB2_ID" | python3 -c "import sys,json; print(json.load(sys.stdin)['status'])")
-[ "$CANCEL_STATUS" = "cancelled" ] && ok "cancel works ($DEL_STATUS → cancelled)"
+[ "$CANCEL_STATUS" = "cancelled" ] && ok "cancel works → status=cancelled" \
+  || echo "  [INFO] cancel status: $CANCEL_STATUS (may be running — worker ignores cancel mid-chunk)"

 echo ""
-echo "=== 12. Verify webhook was fired ==="
+echo "=== 15. POST /model/unload ==="
+UNLOAD_RESP=$(curl -sf -X POST "$BASE/model/unload")
+echo "$UNLOAD_RESP"
+sleep 2
+UNLOAD_STATE=$(curl -sf "$BASE/model/status" | python3 -c "import sys,json; print(json.load(sys.stdin)['state'])")
+[ "$UNLOAD_STATE" = "unloaded" ] && ok "model unloaded → state=unloaded" \
+  || echo "  [INFO] state after unload: $UNLOAD_STATE"
+
+echo ""
+echo "=== 16. Verify webhook fired ==="
 sleep 3
 kill $WEBHOOK_PID 2>/dev/null || true
-ok "all tests done"
+ok "webhook server stopped"
+
+echo ""
+if [ $FAILS -eq 0 ]; then
+  echo -e "${GREEN}=== ALL TESTS PASSED ===${NC}"
+else
+  echo -e "${RED}=== $FAILS TEST(S) FAILED ===${NC}"
+  exit 1
+fi
--- a/tests/test_idle_timeout.sh
+++ b/tests/test_idle_timeout.sh
@@ -0,0 +1,246 @@
+#!/usr/bin/env bash
+# tests/test_idle_timeout.sh
+#
+# Integration tests for the idle-timeout auto-unload feature.
+# REQUIRES the server to be started with a short idle timeout:
+#
+#   IDLE_TIMEOUT_SECS=5 ./whisper-server
+#   # or via Docker:
+#   docker run -e IDLE_TIMEOUT_SECS=5 ...
+#
+# The default idle timeout is 5 minutes; these tests use a 5-second window
+# to keep the suite fast.
+
+set -euo pipefail
+
+BASE="${WHISPER_BASE_URL:-http://localhost:8080}"
+IDLE_TIMEOUT="${EXPECTED_IDLE_TIMEOUT_SECS:-5}"
+AUDIO="${TEST_AUDIO:-}"
+
+GREEN='\033[0;32m'; RED='\033[0;31m'; YELLOW='\033[0;33m'; NC='\033[0m'
+
+PASS=0; FAIL=0
+
+ok()   { echo -e "${GREEN}[PASS]${NC} $1"; PASS=$((PASS+1)); }
+fail() { echo -e "${RED}[FAIL]${NC} $1"; FAIL=$((FAIL+1)); }
+skip() { echo -e "${YELLOW}[SKIP]${NC} $1"; }
+info() { echo "       $1"; }
+
+echo "=== Idle Timeout Tests ==="
+echo "    BASE:              $BASE"
+echo "    IDLE_TIMEOUT_SECS: $IDLE_TIMEOUT  (must be configured on the server)"
+echo ""
+echo "NOTE: These tests require the server to be running with IDLE_TIMEOUT_SECS=$IDLE_TIMEOUT"
+echo ""
+
+# ── Helpers ──────────────────────────────────────────────────────────────────
+
+get_state() {
+  curl -sf "$BASE/model/status" | python3 -c "import sys,json; print(json.load(sys.stdin)['state'])"
+}
+
+ensure_ready() {
+  local state
+  state=$(get_state)
+  if [ "$state" = "ready" ]; then return 0; fi
+  curl -sf -X POST "$BASE/model/load" > /dev/null
+  local elapsed=0
+  while true; do
+    sleep 3; elapsed=$((elapsed+3))
+    state=$(get_state)
+    [ "$state" = "ready" ] && return 0
+    [ $elapsed -gt 180 ] && return 1
+  done
+}
+
+ensure_unloaded() {
+  curl -sf -X POST "$BASE/model/unload" > /dev/null || true
+  sleep 2
+}
+
+# ── TEST 1: Load model, complete a job, then wait for idle unload ─────────────
+echo "--- Test 1: Idle timeout triggers auto-unload ---"
+
+ensure_unloaded
+ensure_ready || { fail "T1: model load failed"; }
+
+WAIT_SECS=$((IDLE_TIMEOUT + 3))
+info "Model is ready. Waiting $WAIT_SECS seconds (idle timeout=$IDLE_TIMEOUT + 3s buffer)..."
+sleep $WAIT_SECS
+
+STATE=$(get_state)
+if [ "$STATE" = "unloaded" ]; then
+  ok "T1: model auto-unloaded after ${IDLE_TIMEOUT}s idle"
+else
+  fail "T1: expected unloaded after idle timeout, got $STATE"
+  info "Is the server running with IDLE_TIMEOUT_SECS=$IDLE_TIMEOUT?"
+fi
+
+# ── TEST 2: model_unloaded webhook fires on idle timeout ─────────────────────
+echo ""
+echo "--- Test 2: model_unloaded webhook fires on idle timeout ---"
+
+ensure_unloaded
+
+# Start webhook receiver
+python3 - <<'PYEOF' &
+import http.server, json, sys, signal
+
+class H(http.server.BaseHTTPRequestHandler):
+    def do_POST(self):
+        n = int(self.headers.get('Content-Length', 0))
+        body = json.loads(self.rfile.read(n))
+        with open('/tmp/idle_wh_event.json', 'w') as f:
+            json.dump(body, f)
+        self.send_response(200); self.end_headers()
+    def log_message(self, *a): pass
+
+signal.signal(signal.SIGTERM, lambda *_: sys.exit(0))
+http.server.HTTPServer(('', 9995), H).serve_forever()
+PYEOF
+WH_PID=$!
+sleep 1
+
+# Register webhook via a job submission (will 503 since unloaded)
+curl -sf -X POST "$BASE/jobs" \
+  -F "audio=@/dev/urandom;type=audio/wav" \
+  -F "webhook_url=http://localhost:9995/wh" \
+  --max-time 5 > /dev/null 2>&1 || true
+
+# Load model
+ensure_ready || { fail "T2: model load failed"; kill $WH_PID 2>/dev/null; }
+
+# Wait for idle timeout
+WAIT_SECS=$((IDLE_TIMEOUT + 5))
+info "Waiting ${WAIT_SECS}s for idle timeout..."
+sleep $WAIT_SECS
+
+kill $WH_PID 2>/dev/null || true
+wait $WH_PID 2>/dev/null || true
+
+if [ -f /tmp/idle_wh_event.json ]; then
+  EVENT_TYPE=$(python3 -c "import json; print(json.load(open('/tmp/idle_wh_event.json')).get('type','?'))")
+  rm -f /tmp/idle_wh_event.json
+  [ "$EVENT_TYPE" = "model_unloaded" ] && ok "T2: model_unloaded webhook fired on idle timeout" \
+    || fail "T2: webhook type=$EVENT_TYPE (expected model_unloaded)"
+else
+  fail "T2: no webhook received within timeout"
+fi
+
+# ── TEST 3: Job submission after idle timeout → 503 → triggers reload ─────────
+echo ""
+echo "--- Test 3: Job triggers reload after idle unload ---"
+
+ensure_unloaded
+ensure_ready || { fail "T3: initial load failed"; }
+
+# Wait for auto-unload
+WAIT_SECS=$((IDLE_TIMEOUT + 3))
+info "Waiting ${WAIT_SECS}s for idle unload..."
+sleep $WAIT_SECS
+
+STATE=$(get_state)
+[ "$STATE" = "unloaded" ] || info "Note: state=$STATE (expected unloaded)"
+
+# Submit job → 503, triggers reload
+HTTP=$(curl -s -o /tmp/t3_body.json -w "%{http_code}" -X POST "$BASE/jobs" \
+  -F "audio=@/dev/urandom;type=audio/wav" \
+  --max-time 5 2>/dev/null || echo "000")
+
+if [ "$HTTP" = "503" ]; then
+  ok "T3a: POST /jobs → 503 after idle unload"
+else
+  skip "T3a: POST /jobs returned $HTTP (model may have reloaded)"
+fi
+
+# State should be loading or ready (reload triggered by job submission)
+sleep 2
+STATE=$(get_state)
+if [ "$STATE" = "loading" ] || [ "$STATE" = "ready" ]; then
+  ok "T3b: reload triggered by job submission ($STATE)"
+else
+  fail "T3b: expected loading/ready, got $STATE"
+fi
+
+rm -f /tmp/t3_body.json
+
+# ── TEST 4: Idle timer resets per job (wait 60% of timeout → still ready) ─────
+echo ""
+echo "--- Test 4: Idle timer resets with each completed job ---"
+
+ensure_unloaded
+ensure_ready || { fail "T4: model load failed"; }
+
+HALF_WAIT=$((IDLE_TIMEOUT - 1))
+info "Waiting ${HALF_WAIT}s (less than idle timeout)..."
+sleep $HALF_WAIT
+
+STATE=$(get_state)
+if [ "$STATE" = "ready" ]; then
+  ok "T4a: model still ready after ${HALF_WAIT}s (less than ${IDLE_TIMEOUT}s timeout)"
+else
+  fail "T4a: model unexpectedly $STATE after only ${HALF_WAIT}s"
+fi
+
+# Wait for full unload
+REMAINING=$((IDLE_TIMEOUT - HALF_WAIT + 3))
+info "Waiting another ${REMAINING}s for full idle unload..."
+sleep $REMAINING
+STATE=$(get_state)
+[ "$STATE" = "unloaded" ] && ok "T4b: model unloaded after total > ${IDLE_TIMEOUT}s idle" \
+  || fail "T4b: expected unloaded, got $STATE"
+
+# ── TEST 5: Job resets idle timer ─────────────────────────────────────────────
+echo ""
+echo "--- Test 5: Completing a job resets the idle timer ---"
+
+if [ -z "$AUDIO" ]; then
+  skip "T5: TEST_AUDIO not set — skipping timer-reset test"
+else
+  ensure_unloaded
+  ensure_ready || { fail "T5: model load failed"; }
+
+  # Submit a job
+  SUBMIT=$(curl -sf -X POST "$BASE/jobs" \
+    -F "audio=@${AUDIO};type=audio/wav" \
+    -F "task=transcribe" 2>&1)
+  JOB_ID=$(echo "$SUBMIT" | python3 -c "import sys,json; print(json.load(sys.stdin)['job_id'])" 2>/dev/null || echo "")
+
+  if [ -z "$JOB_ID" ]; then
+    fail "T5: job submission failed"
+  else
+    # Wait for job to finish
+    elapsed=0
+    while true; do
+      sleep 5; elapsed=$((elapsed+5))
+      STATUS=$(curl -sf "$BASE/jobs/$JOB_ID" | python3 -c "import sys,json; print(json.load(sys.stdin)['status'])")
+      [ "$STATUS" = "done" ] || [ "$STATUS" = "failed" ] && break
+      [ $elapsed -gt 300 ] && break
+    done
+    info "Job finished in ${elapsed}s with status=$STATUS"
+
+    # Now wait IDLE_TIMEOUT - 2 seconds — should still be ready
+    SAFE_WAIT=$((IDLE_TIMEOUT - 2))
+    [ $SAFE_WAIT -lt 1 ] && SAFE_WAIT=1
+    info "Waiting ${SAFE_WAIT}s after job completion (less than idle timeout)..."
+    sleep $SAFE_WAIT
+    STATE=$(get_state)
+    [ "$STATE" = "ready" ] && ok "T5a: model still ready ${SAFE_WAIT}s after job completion" \
+      || fail "T5a: model unexpectedly $STATE after job"
+
+    # Wait for idle timeout
+    REMAINING=$((IDLE_TIMEOUT - SAFE_WAIT + 3))
+    info "Waiting ${REMAINING}s more for idle unload..."
+    sleep $REMAINING
+    STATE=$(get_state)
+    [ "$STATE" = "unloaded" ] && ok "T5b: model auto-unloaded after idle period post-job" \
+      || fail "T5b: expected unloaded, got $STATE"
+  fi
+fi
+
+# ── Summary ────────────────────────────────────────────────────────────────────
+echo ""
+echo "=========================================="
+echo "  Results: ${PASS} passed, ${FAIL} failed"
+echo "=========================================="
+[ $FAIL -eq 0 ] && echo -e "${GREEN}ALL PASSED${NC}" || { echo -e "${RED}FAILURES: $FAIL${NC}"; exit 1; }
--- a/tests/test_model_lifecycle.sh
+++ b/tests/test_model_lifecycle.sh
@@ -0,0 +1,470 @@
+#!/usr/bin/env bash
+# tests/test_model_lifecycle.sh
+#
+# Integration tests for dynamic model loading/unloading.
+# Requires a running whisper-server with GPU access.
+#
+# Usage:
+#   WHISPER_BASE_URL=http://localhost:8080 bash tests/test_model_lifecycle.sh
+#
+# Tests are designed to be independent; each section that needs a specific
+# state resets it explicitly at the start.
+
+set -euo pipefail
+
+BASE="${WHISPER_BASE_URL:-http://localhost:8080}"
+AUDIO="${TEST_AUDIO:-}"
+
+GREEN='\033[0;32m'; RED='\033[0;31m'; YELLOW='\033[0;33m'; NC='\033[0m'
+
+PASS=0; FAIL=0
+
+ok()   { echo -e "${GREEN}[PASS]${NC} $1"; PASS=$((PASS+1)); }
+fail() { echo -e "${RED}[FAIL]${NC} $1"; FAIL=$((FAIL+1)); }
+skip() { echo -e "${YELLOW}[SKIP]${NC} $1"; }
+info() { echo "       $1"; }
+
+echo "=== Model Lifecycle Integration Tests ==="
+echo "    BASE: $BASE"
+echo ""
+
+# ── Helpers ──────────────────────────────────────────────────────────────────
+
+get_state() {
+  curl -sf "$BASE/model/status" | python3 -c "import sys,json; print(json.load(sys.stdin)['state'])"
+}
+
+ensure_unloaded() {
+  curl -sf -X POST "$BASE/model/unload" > /dev/null
+  sleep 2
+  local s
+  s=$(get_state)
+  if [ "$s" != "unloaded" ]; then
+    echo "  WARNING: expected unloaded, got $s — waiting 5s"
+    sleep 5
+  fi
+}
+
+ensure_ready() {
+  local state
+  state=$(get_state)
+  if [ "$state" = "ready" ]; then return 0; fi
+  curl -sf -X POST "$BASE/model/load" > /dev/null
+  local elapsed=0
+  while true; do
+    sleep 5; elapsed=$((elapsed+5))
+    state=$(get_state)
+    [ "$state" = "ready" ] && return 0
+    [ $elapsed -gt 180 ] && echo "  TIMEOUT: model did not become ready" && return 1
+  done
+}
+
+poll_state_transition() {
+  local target="$1" max_secs="${2:-120}"
+  local elapsed=0
+  while true; do
+    sleep 2; elapsed=$((elapsed+2))
+    local s
+    s=$(get_state)
+    [ "$s" = "$target" ] && return 0
+    [ $elapsed -ge $max_secs ] && return 1
+  done
+}
+
+# ── TEST 1: Startup state is unloaded ────────────────────────────────────────
+echo "--- Test 1: Startup state is unloaded (or after explicit unload) ---"
+ensure_unloaded
+STATE=$(get_state)
+if [ "$STATE" = "unloaded" ]; then
+  ok "T1: state=unloaded after explicit unload"
+else
+  fail "T1: expected unloaded, got $STATE"
+fi
+
+# ── TEST 2: POST /model/load returns 202 ─────────────────────────────────────
+echo ""
+echo "--- Test 2: POST /model/load returns 202 ---"
+ensure_unloaded
+HTTP=$(curl -s -o /dev/null -w "%{http_code}" -X POST "$BASE/model/load")
+if [ "$HTTP" = "202" ]; then
+  ok "T2: POST /model/load → 202 Accepted"
+else
+  fail "T2: expected 202, got $HTTP"
+fi
+# Cancel the in-progress load to clean up
+curl -sf -X POST "$BASE/model/unload" > /dev/null || true
+sleep 2
+
+# ── TEST 3: State transitions to loading/ready after load trigger ─────────────
+echo ""
+echo "--- Test 3: State transitions to loading (not stuck at unloaded) ---"
+ensure_unloaded
+curl -sf -X POST "$BASE/model/load" > /dev/null
+sleep 1
+STATE=$(get_state)
+if [ "$STATE" = "loading" ] || [ "$STATE" = "ready" ]; then
+  ok "T3: state transitioned to $STATE (not stuck at unloaded)"
+else
+  fail "T3: expected loading or ready, got $STATE"
+fi
+
+# ── TEST 4: Model reaches ready state and loaded_at is set ───────────────────
+echo ""
+echo "--- Test 4: Model reaches ready state with loaded_at timestamp ---"
+# Already loading from T3 — wait for ready
+if ! poll_state_transition "ready" 180; then
+  fail "T4: model did not become ready within 3 minutes"
+else
+  STATUS_JSON=$(curl -sf "$BASE/model/status")
+  LOADED_AT=$(echo "$STATUS_JSON" | python3 -c "import sys,json; d=json.load(sys.stdin); print(d.get('loaded_at','MISSING'))" 2>/dev/null || echo "MISSING")
+  if [ "$LOADED_AT" != "MISSING" ] && [ "$LOADED_AT" != "null" ] && [ -n "$LOADED_AT" ]; then
+    ok "T4: model=ready, loaded_at=$LOADED_AT"
+  else
+    fail "T4: model ready but loaded_at is missing or null"
+  fi
+fi
+
+# ── TEST 5: Idempotent load — POST /model/load when ready returns 200 ─────────
+echo ""
+echo "--- Test 5: POST /model/load when already ready → 200 ---"
+ensure_ready || { fail "T5: could not load model"; }
+HTTP=$(curl -s -o /dev/null -w "%{http_code}" -X POST "$BASE/model/load")
+STATE=$(get_state)
+if [ "$HTTP" = "200" ] && [ "$STATE" = "ready" ]; then
+  ok "T5: idempotent load → 200, state stays ready"
+elif [ "$HTTP" = "202" ] && [ "$STATE" = "ready" ]; then
+  ok "T5: idempotent load → 202, state stays ready"
+else
+  fail "T5: expected 200 and ready, got HTTP=$HTTP state=$STATE"
+fi
+
+# ── TEST 6: Job accepted when ready (segments > 0) ────────────────────────────
+echo ""
+echo "--- Test 6: Job accepted when model is ready ---"
+if [ -z "$AUDIO" ]; then
+  skip "T6: TEST_AUDIO not set — skipping job submission test"
+else
+  ensure_ready || { fail "T6: model load failed"; }
+  SUBMIT=$(curl -sf -X POST "$BASE/jobs" -F "audio=@${AUDIO};type=audio/wav" -F "task=transcribe" 2>&1)
+  JOB_ID=$(echo "$SUBMIT" | python3 -c "import sys,json; print(json.load(sys.stdin)['job_id'])" 2>/dev/null || echo "")
+  if [ -n "$JOB_ID" ]; then
+    ok "T6: job accepted, id=$JOB_ID"
+    # Poll to done
+    elapsed=0
+    while true; do
+      sleep 10; elapsed=$((elapsed+10))
+      STATUS=$(curl -sf "$BASE/jobs/$JOB_ID" | python3 -c "import sys,json; print(json.load(sys.stdin)['status'])")
+      [ "$STATUS" = "done" ] && break
+      [ "$STATUS" = "failed" ] && break
+      [ $elapsed -gt 600 ] && break
+    done
+    SEGS=$(curl -sf "$BASE/jobs/$JOB_ID" | python3 -c "import sys,json; d=json.load(sys.stdin); print(len(d.get('segments',[])))")
+    [ "$SEGS" -gt 0 ] && ok "T6b: job done with $SEGS segments" || fail "T6b: job done but 0 segments"
+  else
+    fail "T6: job submission failed: $SUBMIT"
+  fi
+fi
+
+# ── TEST 7: POST /model/unload → state=unloaded ───────────────────────────────
+echo ""
+echo "--- Test 7: POST /model/unload ---"
+ensure_ready || { fail "T7: model load failed"; }
+curl -sf -X POST "$BASE/model/unload" > /dev/null
+sleep 3
+STATE=$(get_state)
+if [ "$STATE" = "unloaded" ]; then
+  ok "T7: POST /model/unload → state=unloaded"
+else
+  fail "T7: expected unloaded after unload, got $STATE"
+fi
+
+# ── TEST 8: POST /jobs when unloaded → 503 + Retry-After ─────────────────────
+echo ""
+echo "--- Test 8: POST /jobs when unloaded → 503 + Retry-After ---"
+ensure_unloaded
+# Submit a tiny dummy payload (won't be valid audio but that's ok for this test)
+HTTP=$(curl -s -o /tmp/t8_body.json -w "%{http_code}" -X POST "$BASE/jobs" \
+  -F "audio=@/dev/urandom;type=audio/wav" \
+  --max-time 5 2>/dev/null || echo "000")
+# If the model auto-loads it might start processing; check for 503 first
+if [ "$HTTP" = "503" ]; then
+  RETRY_AFTER=$(curl -sI -X POST "$BASE/jobs" \
+    -F "audio=@/dev/urandom;type=audio/wav" \
+    --max-time 5 2>/dev/null | grep -i "retry-after" | awk '{print $2}' | tr -d '\r' || echo "")
+  BODY=$(cat /tmp/t8_body.json 2>/dev/null || echo "{}")
+  HAS_STATE=$(echo "$BODY" | python3 -c "import sys,json; d=json.load(sys.stdin); print('state' in d)" 2>/dev/null || echo "False")
+  HAS_RETRY=$(echo "$BODY" | python3 -c "import sys,json; d=json.load(sys.stdin); print('retry_after_secs' in d)" 2>/dev/null || echo "False")
+  if [ "$HAS_STATE" = "True" ] && [ "$HAS_RETRY" = "True" ]; then
+    ok "T8: 503 with state + retry_after_secs in body"
+  else
+    fail "T8: 503 but body missing state/retry_after_secs. body=$BODY"
+  fi
+  if [ -n "$RETRY_AFTER" ]; then
+    ok "T8b: Retry-After header present: $RETRY_AFTER"
+  else
+    fail "T8b: Retry-After header missing from 503 response"
+  fi
+else
+  skip "T8: got HTTP $HTTP (model may have loaded before check) — skipping"
+fi
+
+# ── TEST 9: Rejected job triggers load ────────────────────────────────────────
+echo ""
+echo "--- Test 9: Job rejection triggers model load ---"
+ensure_unloaded
+# Send a job (we expect 503)
+curl -sf -X POST "$BASE/jobs" \
+  -F "audio=@/dev/urandom;type=audio/wav" \
+  --max-time 5 > /dev/null 2>&1 || true
+sleep 2
+STATE=$(get_state)
+if [ "$STATE" = "loading" ] || [ "$STATE" = "ready" ]; then
+  ok "T9: model started loading after job rejection ($STATE)"
+else
+  fail "T9: expected loading/ready after job rejection, got $STATE"
+fi
+# Stop the load to clean up
+curl -sf -X POST "$BASE/model/unload" > /dev/null || true
+sleep 2
+
+# ── TEST 10: Retry-After values ───────────────────────────────────────────────
+echo ""
+echo "--- Test 10: Retry-After values match state ---"
+ensure_unloaded
+# Unloaded → Retry-After: 30
+RESP_UNLOADED=$(curl -si -X POST "$BASE/jobs" -F "audio=@/dev/urandom;type=audio/wav" --max-time 5 2>/dev/null || echo "")
+RA_UNLOADED=$(echo "$RESP_UNLOADED" | grep -i "retry-after" | awk '{print $2}' | tr -d '\r' || echo "")
+[ "$RA_UNLOADED" = "30" ] && ok "T10a: Retry-After=30 when unloaded" \
+  || skip "T10a: Retry-After=$RA_UNLOADED (expected 30) — model may have started loading"
+
+# ── TEST 11: Retry-After=10 during loading ────────────────────────────────────
+echo ""
+echo "--- Test 11: Retry-After=10 when loading ---"
+ensure_unloaded
+curl -sf -X POST "$BASE/model/load" > /dev/null
+sleep 1  # In loading state
+STATE=$(get_state)
+if [ "$STATE" = "loading" ]; then
+  RESP_LOADING=$(curl -si -X POST "$BASE/jobs" -F "audio=@/dev/urandom;type=audio/wav" --max-time 5 2>/dev/null || echo "")
+  RA_LOADING=$(echo "$RESP_LOADING" | grep -i "retry-after" | awk '{print $2}' | tr -d '\r' || echo "")
+  [ "$RA_LOADING" = "10" ] && ok "T11: Retry-After=10 when loading" \
+    || fail "T11: expected Retry-After=10, got '$RA_LOADING' (state=$STATE)"
+else
+  skip "T11: model already $STATE — can't test loading state Retry-After"
+fi
+
+# ── TEST 12: 503 body schema validation ──────────────────────────────────────
+echo ""
+echo "--- Test 12: 503 body schema validation ---"
+ensure_unloaded
+BODY=$(curl -sf -X POST "$BASE/jobs" -F "audio=@/dev/urandom;type=audio/wav" --max-time 5 2>/dev/null || echo "{}")
+python3 - <<PYCHECK
+import json
+body = json.loads('$BODY')
+required = {'error', 'state', 'retry_after_secs'}
+missing = required - set(body.keys())
+if missing:
+    print(f"MISSING: {missing}")
+    exit(1)
+assert body['error'] == 'model_not_ready', f"error={body['error']}"
+assert isinstance(body['retry_after_secs'], int), f"retry_after_secs not int: {body['retry_after_secs']}"
+print("schema ok")
+PYCHECK
+[ $? -eq 0 ] && ok "T12: 503 body has correct schema" || fail "T12: 503 body schema invalid"
+
+# ── TEST 13: GET /health has model_state field ────────────────────────────────
+echo ""
+echo "--- Test 13: GET /health has model_state ---"
+HEALTH=$(curl -sf "$BASE/health")
+HAS_MODEL_STATE=$(echo "$HEALTH" | python3 -c "import sys,json; d=json.load(sys.stdin); print('model_state' in d)")
+[ "$HAS_MODEL_STATE" = "True" ] && ok "T13: /health has model_state" || fail "T13: /health missing model_state"
+
+# ── TEST 14: SSE /model/events delivers model_ready event ─────────────────────
+echo ""
+echo "--- Test 14: GET /model/events SSE delivers model_ready ---"
+ensure_unloaded
+
+# Collect SSE events for up to 3 minutes
+SSE_LOG=$(mktemp /tmp/sse_events_XXXXXX.txt)
+curl -sN --max-time 180 "$BASE/model/events" > "$SSE_LOG" &
+SSE_PID=$!
+sleep 1
+
+# Trigger load
+curl -sf -X POST "$BASE/model/load" > /dev/null
+poll_state_transition "ready" 180 || true
+
+sleep 2
+kill $SSE_PID 2>/dev/null || true
+wait $SSE_PID 2>/dev/null || true
+
+if grep -q "model_loading" "$SSE_LOG" 2>/dev/null; then
+  ok "T14a: SSE received model_loading event"
+else
+  fail "T14a: SSE did not receive model_loading event"
+fi
+
+if grep -q "model_ready" "$SSE_LOG" 2>/dev/null; then
+  ok "T14b: SSE received model_ready event"
+else
+  fail "T14b: SSE did not receive model_ready event"
+fi
+
+# Now unload to get model_unloaded event
+curl -sf -X POST "$BASE/model/unload" > /dev/null
+sleep 1
+
+SSE_LOG2=$(mktemp /tmp/sse_events_XXXXXX.txt)
+curl -sN --max-time 10 "$BASE/model/events" > "$SSE_LOG2" &
+SSE_PID2=$!
+sleep 2
+kill $SSE_PID2 2>/dev/null || true
+wait $SSE_PID2 2>/dev/null || true
+
+# model_unloaded fires immediately on unload command
+if grep -q "model_unloaded" "$SSE_LOG" 2>/dev/null || grep -q "model_unloaded" "$SSE_LOG2" 2>/dev/null; then
+  ok "T14c: SSE received model_unloaded event"
+else
+  fail "T14c: SSE did not receive model_unloaded event"
+fi
+
+rm -f "$SSE_LOG" "$SSE_LOG2"
+
+# ── TEST 15: model_ready webhook fires after load ──────────────────────────────
+echo ""
+echo "--- Test 15: model_ready webhook ---"
+ensure_unloaded
+
+# Start webhook receiver
+WEBHOOK_LOG=$(mktemp /tmp/webhook_log_XXXXXX.txt)
+python3 - <<'PYEOF' &
+import http.server, json, sys, signal, os
+
+class H(http.server.BaseHTTPRequestHandler):
+    def do_POST(self):
+        n = int(self.headers.get('Content-Length', 0))
+        body = json.loads(self.rfile.read(n))
+        with open('/tmp/t15_webhook.json', 'w') as f:
+            json.dump(body, f)
+        self.send_response(200); self.end_headers()
+    def log_message(self, *a): pass
+
+signal.signal(signal.SIGTERM, lambda *_: sys.exit(0))
+http.server.HTTPServer(('', 9998), H).serve_forever()
+PYEOF
+WBOOK_PID=$!
+sleep 1
+
+# Register a webhook via a (doomed) job submission — this registers the URL
+# even though the model is unloaded (and the job will 503)
+curl -sf -X POST "$BASE/jobs" \
+  -F "audio=@/dev/urandom;type=audio/wav" \
+  -F "webhook_url=http://localhost:9998/wh" \
+  --max-time 5 > /dev/null 2>&1 || true
+
+# Now load the model
+curl -sf -X POST "$BASE/model/load" > /dev/null
+poll_state_transition "ready" 180 || true
+sleep 3
+
+kill $WBOOK_PID 2>/dev/null || true
+wait $WBOOK_PID 2>/dev/null || true
+
+if [ -f /tmp/t15_webhook.json ]; then
+  EVENT_TYPE=$(python3 -c "import json; d=json.load(open('/tmp/t15_webhook.json')); print(d.get('type','?'))")
+  [ "$EVENT_TYPE" = "model_ready" ] && ok "T15: model_ready webhook fired" \
+    || fail "T15: webhook fired but type=$EVENT_TYPE (expected model_ready)"
+  rm -f /tmp/t15_webhook.json
+else
+  fail "T15: model_ready webhook not received within timeout"
+fi
+
+# ── TEST 16: model_unloaded webhook fires ─────────────────────────────────────
+echo ""
+echo "--- Test 16: model_unloaded webhook ---"
+
+python3 - <<'PYEOF' &
+import http.server, json, sys, signal
+
+class H(http.server.BaseHTTPRequestHandler):
+    def do_POST(self):
+        n = int(self.headers.get('Content-Length', 0))
+        body = json.loads(self.rfile.read(n))
+        with open('/tmp/t16_webhook.json', 'w') as f:
+            json.dump(body, f)
+        self.send_response(200); self.end_headers()
+    def log_message(self, *a): pass
+
+signal.signal(signal.SIGTERM, lambda *_: sys.exit(0))
+http.server.HTTPServer(('', 9997), H).serve_forever()
+PYEOF
+WBOOK2_PID=$!
+sleep 1
+
+# Register webhook URL
+curl -sf -X POST "$BASE/jobs" \
+  -F "audio=@/dev/urandom;type=audio/wav" \
+  -F "webhook_url=http://localhost:9997/wh" \
+  --max-time 5 > /dev/null 2>&1 || true
+
+ensure_ready
+
+# Unload
+curl -sf -X POST "$BASE/model/unload" > /dev/null
+sleep 5
+
+kill $WBOOK2_PID 2>/dev/null || true
+wait $WBOOK2_PID 2>/dev/null || true
+
+if [ -f /tmp/t16_webhook.json ]; then
+  EVENT_TYPE=$(python3 -c "import json; d=json.load(open('/tmp/t16_webhook.json')); print(d.get('type','?'))")
+  [ "$EVENT_TYPE" = "model_unloaded" ] && ok "T16: model_unloaded webhook fired" \
+    || fail "T16: webhook type=$EVENT_TYPE (expected model_unloaded)"
+  rm -f /tmp/t16_webhook.json
+else
+  fail "T16: model_unloaded webhook not received"
+fi
+
+# ── TEST 17: Concurrent load requests — single load, stable ready ─────────────
+echo ""
+echo "--- Test 17: Concurrent POST /model/load requests ---"
+ensure_unloaded
+# Send 3 concurrent load requests
+curl -sf -X POST "$BASE/model/load" > /dev/null &
+curl -sf -X POST "$BASE/model/load" > /dev/null &
+curl -sf -X POST "$BASE/model/load" > /dev/null &
+wait
+poll_state_transition "ready" 180 || true
+STATE=$(get_state)
+[ "$STATE" = "ready" ] && ok "T17: concurrent loads handled cleanly, state=ready" \
+  || fail "T17: expected ready after concurrent loads, got $STATE"
+
+# ── TEST 18: POST /model/unload during loading → clean unloaded ───────────────
+echo ""
+echo "--- Test 18: POST /model/unload during loading ---"
+ensure_unloaded
+curl -sf -X POST "$BASE/model/load" > /dev/null
+sleep 1  # Hopefully still in loading state
+curl -sf -X POST "$BASE/model/unload" > /dev/null
+# Allow time for the unload to propagate
+sleep 5
+STATE=$(get_state)
+if [ "$STATE" = "unloaded" ]; then
+  ok "T18: unload during loading → clean unloaded"
+elif [ "$STATE" = "ready" ]; then
+  # Load completed before unload arrived — immediately unload
+  curl -sf -X POST "$BASE/model/unload" > /dev/null
+  sleep 3
+  STATE=$(get_state)
+  [ "$STATE" = "unloaded" ] && ok "T18: load completed then unloaded (race condition OK)" \
+    || fail "T18: state=$STATE after load+unload"
+else
+  fail "T18: unexpected state after unload-during-load: $STATE"
+fi
+
+# ── Summary ────────────────────────────────────────────────────────────────────
+echo ""
+echo "=========================================="
+echo "  Results: ${PASS} passed, ${FAIL} failed"
+echo "=========================================="
+[ $FAIL -eq 0 ] && echo -e "${GREEN}ALL PASSED${NC}" || { echo -e "${RED}FAILURES: $FAIL${NC}"; exit 1; }
Author	SHA1	Message	Date
Giancarmine Salucci	7c01d7f77f	fix(model): skip stale unload command All checks were successful Build & Push Docker Image / test (push) Successful in 6m9s Details Build & Push Docker Image / build-and-push (push) Successful in 6m33s Details Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>	2026-05-12 01:08:35 +02:00
Giancarmine Salucci	d8a73e150a	fix(worker): port final segment cleanup All checks were successful Build & Push Docker Image / test (push) Successful in 6m2s Details Build & Push Docker Image / build-and-push (push) Successful in 6m31s Details Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>	2026-05-12 00:10:32 +02:00
Giancarmine Salucci	cb0b07b2ff	fix(worker): collapse incremental segments All checks were successful Build & Push Docker Image / test (push) Successful in 6m20s Details Build & Push Docker Image / build-and-push (push) Successful in 6m29s Details Normalize rolling partial-hypothesis chains before final job persistence so downstream clients receive stable transcript segments instead of echoed continuations. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>	2026-05-11 22:46:38 +02:00
mozempk	d3a67f11b3	fix: four worker.rs bugs found during E2E testing All checks were successful Build & Push Docker Image / test (push) Successful in 6m21s Details Build & Push Docker Image / build-and-push (push) Successful in 6m32s Details	2026-05-10 16:28:03 +02:00
mozempk	bcaf8680db	docs: add FRONTEND_INTEGRATION.md developer guide All checks were successful Build & Push Docker Image / test (push) Successful in 5m54s Details Build & Push Docker Image / build-and-push (push) Successful in 17s Details Comprehensive integration guide for frontend/full-stack developers: - Architecture overview diagram - Quick start (submit + poll in ~20 lines) - Model lifecycle: state machine diagram, all 4 /model/* endpoints, SSE event subscription with JS examples - Job submission: multipart fields, 503 model_not_ready handling, retry-with-auto-load pattern - Job progress: polling vs SSE, all event types with payloads - Webhooks: job completion + model lifecycle, Express receiver example, how to distinguish job vs model payloads - Health check field reference - Cancellation semantics (GPU inference not interruptible) - Full TypeScript type definitions for all API shapes - React hooks: useModelStatus, useJobStream, useTranscribe - Complete WhisperClient class example with ensureModelReady, streamProgress, and end-to-end transcribe() - Error reference table with all 400/404/409/503/500 shapes Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>	2026-05-08 23:47:13 +02:00
mozempk	d0148260e3	test: add unit test infrastructure (Docker tester stage + CI) All checks were successful Build & Push Docker Image / test (push) Successful in 5m42s Details Build & Push Docker Image / build-and-push (push) Successful in 21s Details - Add Dockerfile 'tester' stage (FROM builder): - Symlinks /usr/local/cuda/lib64/stubs/libcuda.so → libcuda.so.1 so the test binary can satisfy the dynamic linker without a real GPU - Runs `cargo test --release` reusing the cached release build artifacts (no recompilation — tests complete in ~6s) - docker build --target tester . to run all 30 unit tests - Add 'test' job to .gitea/workflows/docker-build.yml: - Runs before build-and-push (build-and-push needs: test) - Builds --target tester with registry build cache - Gate: build-and-push only runs when all tests pass - Add run_tests.sh convenience script for local use: - Accepts optional test name filter as first argument - Respects CUDA_VERSION / UBUNTU_VERSION env overrides All 30 unit tests pass: error::tests — 7 tests (OOM detection, ModelNotReady HTTP shape) models::tests — 17 tests (state machine, serialization, retry-after) worker::tests — 6 tests (chunk ranges, silence snap/trim) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>	2026-05-08 18:20:52 +02:00
mozempk	b191fbe200	feat: dynamic model loading/unloading with GPU polling All checks were successful Build & Push Docker Image / build-and-push (push) Successful in 8m41s Details - Model starts unloaded (lazy); loads on first job or POST /model/load - Auto-unloads after IDLE_TIMEOUT_SECS (default 300) of inactivity - POST /model/unload for immediate manual release - GPU-busy detection: on VRAM OOM, enters WaitingForGpu and retries every GPU_POLL_INTERVAL_SECS (default 30) indefinitely - POST /jobs when unloaded → 503 + Retry-After header, triggers load - AppError::OutOfMemory and AppError::ModelNotReady variants - WorkerCmd channel (SyncSender<WorkerCmd>) replaces bare tx_req channel - Idle timer via recv_timeout(1s) tick inside OS thread (no extra thread) - Model lifecycle events broadcast via tokio broadcast channel (SSE + webhooks) - webhook_registry: all clients that ever submitted a webhook_url receive model_ready and model_unloaded webhooks - GPU warmup retained on every (re)load New routes: GET /model/status — current state + VRAM stats POST /model/load — trigger load (idempotent) POST /model/unload — immediate unload GET /model/events — SSE stream of model lifecycle events New env vars: IDLE_TIMEOUT_SECS (default 300) GPU_POLL_INTERVAL_SECS (default 30) Tests: tests/test_model_lifecycle.sh — 18 integration tests (full state machine, SSE events, webhooks, concurrency, unload-during-load) tests/test_idle_timeout.sh — 5 tests with short IDLE_TIMEOUT_SECS=5 test_all.sh updated: loads model before job submission, asserts model_state in /health, adds POST /model/unload at end Docs: docs/USAGE.md: model lifecycle section, new env vars, 503 retry pattern, updated /health response shape Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>	2026-05-08 17:57:20 +02:00
mozempk	78c6fab81b	fix: remove duplicate old test suite and fix step 9 pipe/heredoc bug All checks were successful Build & Push Docker Image / build-and-push (push) Successful in 16s Details Step 9 used 'echo $RESULT \| python3 - << HEREDOC' which is a bash gotcha: the heredoc takes over stdin (as the script source), so the pipe is silently ignored and sys.stdin.read() returns empty string → JSONDecodeError. Fix: write RESULT to a temp file and pass it as sys.argv[1] to the script. Also removed the old buggy test suite that was accidentally left appended at lines 181-327 (had language=auto, ['id'] field, wrong DELETE assertion). Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>	2026-05-06 12:13:15 +02:00
mozempk	fd8d4deefb	fix: GPU warmup on startup + fix test_all.sh + document cold-GPU finding All checks were successful Build & Push Docker Image / build-and-push (push) Successful in 6m39s Details GPU warmup (src/transcriber.rs): After creating WhisperState, run a 1s silent inference pass in load(). CUDA JIT-compiles device kernels on the first whisper_full_with_state call. On a cold GPU this compilation disrupts the decode pipeline mid-inference, returning 0 segments in ~0.5s. The warmup forces all kernel compilation at startup so the first real job runs on fully compiled kernels. test_all.sh: - Fix submit response field: 'id' → 'job_id' (was breaking all downstream steps) - Remove language=auto: not a valid ISO 639-1 code; omit field for auto-detect - Make BASE and AUDIO configurable via env vars (WHISPER_BASE_URL, TEST_AUDIO) - Fix DELETE assertion: completed jobs return 409 Conflict, not 204 - Add explicit zero-segments failure check in quality inspection (step 9) - Add progress reporting to poll loop docs/FINDINGS.md + KNOWLEDGE.md: Document cold GPU warmup issue, root cause, and fix. Document language=auto as invalid API usage. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>	2026-05-06 11:57:30 +02:00
mozempk	d5a88d1866	fix: create WhisperState once at load time, reuse across all chunks Some checks failed Build & Push Docker Image / build-and-push (push) Has been cancelled Details Previously create_state() was called for every 60s audio chunk, triggering whisper_init_state() each time. This allocates ~700 MB of GPU compute buffers (KV caches, CUDA workspace) and re-initialises the CUDA backend per chunk. For a 101-minute audio (102 chunks), this caused 102 GPU re-initialisations and VRAM allocation cycles. Under VRAM pressure from concurrent processes, CUDA allocation failures occurred silently — whisper returned language detection results but 0 segments. Fix: create WhisperState once in Transcriber::load() and reuse it for every transcription call. GPU memory is stable; no_context=true prevents KV-cache contamination between chunks. WhisperState is Send+Sync (explicitly declared in whisper-rs) and holds its own Arc<WhisperInnerContext>, so the model weights stay alive even after WhisperContext is dropped. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>	2026-05-06 11:51:33 +02:00