Fixes

.env

@@ -4,8 +4,13 @@
 # -----------------------------------------------------
 # Database
 # -----------------------------------------------------
-DB_ROOT_PASSWORD=changeme_root
-DB_PASSWORD=changeme_videodetect
+DB_ROOT_PASSWORD=videodetect
+DB_PASSWORD=videodetect
+DB_USER=videodetect
+DB_NAME=videodetect
+DB_HOST=mariadb
+
+API_HOST=https://api:3000
 
 # -----------------------------------------------------
 # Storage paths (host-side mounts)
@@ -14,7 +19,3 @@ NAS_OUTPUT_PATH=${PWD}/output
 MODELS_PATH=${PWD}/models
 TRAINING_PATH=${PWD}/training
 
-# -----------------------------------------------------
-# Grafana
-# -----------------------------------------------------
-GRAFANA_PASSWORD=changeme_grafana

REFACTOR.md

@@ -0,0 +1,486 @@
+# REFACTOR.md — Worker-to-API Refactor Plan
+
+## Context
+
+VideoDetect workers currently run as long-lived processes (`main.py` → `WorkerPool`) that query the database directly via `DBConnector`. The re-architecture moves to **independent, externally-invoked task processes** that communicate through a REST API provided by the Perl `api/app.pl` module. This enables finer-grained scaling, resilience to individual process crashes, and separation of concerns between scanning (discovery) and processing (AI inference).
+
+---
+
+## Target Architecture
+
+```
+┌───────────┐     /api/v1/nexttask/AISCAN     ┌─────────────┐
+│  Scanner   │ ──────────────────────────────> │  Perl API    │
+│  (long-    │ <────────────────────────────── │  (Dancer2)   │
+│   lived)   │  task reservation + assign_key  ├─────────────┤
+└───────────┘                                  │   DB         │
+                                                 │             │
+     ┌───────────────────────────────────────────┼──────────┐
+     │                                           ▼          │
+     │                                    ┌──────────────┐  │
+     │       POST /api/v1/task/:id/complete              │  │
+     │      ◄─────────────────────────────────────────►  │  │
+     │                                          │         │
+     │          ┌───────────────────────────────┘        │
+     ▼          ▼                                        ▼
+┌────────┐  ┌────────┐  ┌────────┐              ┌──────────────┐
+│worker-1│  │worker-2│  │worker-N│              │   MariaDB     │
+│(short- │  │(short- │  │(short- │              │              │
+│ lived) │  │ lived) │  │ lived) │              └──────────────┘
+└────────┘  └────────┘  └────────┘
+```
+
+Each worker process:
+1. Starts → accepts `--tasks N` (default 1) via CLI argument
+2. Loops up to N times: **claim task** → **fetch video data** → **process** → **submit results**
+3. Exits cleanly after completing the requested count
+
+---
+
+## API Reference (from `api/app.pl`)
+
+### GET `/api/v1/nexttask/:type`
+
+Claims the next pending task of the given type and reserves it.
+
+| Field     | Type    | Example                                      |
+|-----------|---------|----------------------------------------------|
+| task      | object  | Task record from DB                          |
+| assign_key| string  | `"worker_338"` (server-generated)            |
+
+**On success:** HTTP 200 with JSON body:
+```json
+{
+    "task": {
+        "id": 1,
+        "video_id": 42,
+        "task_type": "AISCAN",
+        "status": "PENDING",
+        "created_at": "2026-09-09T13:05:56",
+        "updated_at": "2026-09-09T14:53:57",
+        "assign_key": null,
+        "results": null,
+        "assigned_at": "2026-09-09T14:43:31"
+    },
+    "assign_key": "worker_338"
+}
+```
+
+**On no tasks available:** HTTP 404 with `{"message":"No task"}`.
+
+> **NOTE:** The Perl API marks the task `IN_PROGRESS` and sets `assign_key` + `assigned_at` atomically (`UPDATE ... WHERE id=? AND status='PENDING'`). This is safe for multiple workers racing to claim the same task.
+
+---
+
+### GET `/api/v1/video/:id`
+
+Returns full video metadata by ID.
+
+| Field         | Type    | Example                                              |
+|---------------|---------|------------------------------------------------------|
+| id            | integer | `1`                                                  |
+| file_path     | string  | `"/data/The Fappening/Sextape - Alyson Hannigan (American actress - American pie).wmv"` |
+| file_size     | integer | `19358016`                                           |
+| file_hash     | string  | `"1ffd178e9ee23039aebffd79ddcbc88e983633edcb75062e6bd4c269f4d7bf94"` |
+| resolution_w  | integer | `640`                                                |
+| resolution_h  | integer | `480`                                                |
+| codec         | string  | `"wmv1"`                                             |
+| duration      | float   | `102.499`                                            |
+| last_scan_time| datetime| `"2026-09-09T13:05:56"`                              |
+| created_at    | datetime| `"2026-09-09T13:05:56"`                              |
+| updated_at    | datetime| `"2026-09-09T13:05:56"`                              |
+
+**On not found:** HTTP 404 with `{"message":"No video"}`.
+
+---
+
+### POST `/api/v1/task/:task/complete`
+
+Submits processing results for a claimed task. The API verifies the task is `IN_PROGRESS` and assigned to the given `assign_key`.
+
+**Required parameters (form-encoded or JSON body):**
+
+| Field       | Type   | Description                                    |
+|-------------|--------|------------------------------------------------|
+| assign_key  | string | Must match the key returned by `/nexttask`     |
+| results     | object | JSON-serializable dict to store in DB `results` column |
+
+**On success:** HTTP 200 with `{"message":"Task completed successfully"}`.
+
+**On mismatch or task not IN_PROGRESS:** HTTP 403 with error message.
+
+> **NOTE:** The Perl API currently uses `body_parameters->get("results")`, which reads form-encoded data. For JSON body submission, the Dancer2 config may need `serializer: JSON` (which is already set in `api/config.yml`). However, `body_parameters` only parses form-encoded fields — JSON body content goes to `$app->request->body`. **This needs verification before implementation.**
+
+---
+
+## Refactor Steps
+
+### Phase 1 — Create the Task Worker Module
+
+**Goal:** A single entry-point script that can run as a standalone process.
+
+**File:** `src/task_worker.py` (new)
+
+**Acceptance Criteria:**
+- Starts with `python3 src/task_worker.py --tasks N` (default 1 task)
+- Prints startup log message showing number of tasks requested and model version
+- Exits cleanly after processing the requested count
+
+```bash
+# Default: process 1 task and exit
+python3 -m src.task_worker
+
+# Explicit: process 5 tasks and exit
+python3 -m src.task_worker --tasks 5
+
+# Also supported via direct script invocation
+python3 src/task_worker.py --tasks 10
+```
+
+**CLI Implementation:**
+
+```python
+import argparse
+
+parser = argparse.ArgumentParser(description="VideoDetect task worker")
+parser.add_argument("--tasks", type=int, default=1,
+                    help="Number of tasks to process before exiting (default: 1)")
+args = parser.parse_args()
+```
+
+---
+
+### Phase 2 — API Client Module
+
+**Goal:** A thin HTTP client layer for communicating with the Perl API.
+
+**File:** `src/api_client.py` (new)
+
+**Acceptance Criteria:**
+- Encapsulates all API calls in one class: `ApiClient(base_url)`
+- Handles JSON serialization/deserialization automatically
+- Raises a custom `ApiError` on HTTP errors (with status code and response body)
+- Returns `None` for 404 responses (no task remaining) from `/nexttask`
+
+**API Error handling:**
+
+```python
+class ApiError(Exception):
+    def __init__(self, status_code: int, message: str):
+        self.status_code = status_code
+        self.message = message
+        super().__init__(f"API error {status_code}: {message}")
+
+# Usage patterns:
+try:
+    response = client.get_next_task("AISCAN")
+except ApiError as e:
+    if e.status_code == 404:
+        logger.info("No more tasks available")
+        break  # exit processing loop
+    raise
+
+result = client.submit_results(task_id, assign_key, results_dict)
+```
+
+**Methods to implement:**
+
+| Method | Endpoint | Returns | Special handling |
+|--------|----------|---------|-----------------|
+| `get_next_task(task_type)` | `GET /api/v1/nexttask/{type}` | `{"task": {...}, "assign_key": str} \| None` | Returns None on 404 |
+| `get_video(video_id)` | `GET /api/v1/video/{id}` | dict with video metadata | Raises ApiError on 404/5xx |
+| `submit_results(task_id, assign_key, results)` | `POST /api/v1/task/{task}/complete` | dict with response | Passes assign_key + serialized results |
+
+**Notes for implementation:**
+- Use Python's standard library only (`urllib.request`) if possible to avoid adding dependencies. If JSON body submission is needed for the Perl API, check whether `body_parameters->get('results')` in Dancer2 handles raw JSON (it typically does not — it expects form-encoded data).
+- If form-encoding is required: `requests.post(url, data={'assign_key': key, 'results': json.dumps(results)})`
+- If JSON body works: `requests.post(url, json={'assign_key': key, 'results': results})`
+- **Recommendation:** Use the `requests` library (already in worker requirements.txt likely). Document both approaches and implement the one that matches your Dancer2 config.
+
+---
+
+### Phase 3 — Implement the AI Processing Pipeline
+
+**Goal:** Extract the core AI processing logic from `orchestrator.py` into a reusable function callable by the task worker.
+
+**File:** New or updated module in `src/` (tentatively `src/task_processor.py`)
+
+**Acceptance Criteria:**
+- Takes video metadata dict + file path as input
+- Returns a results dict matching what will be stored in the DB `results` column
+- Handles all error cases gracefully (returns errors instead of crashing)
+- Logs all significant decisions at INFO level
+
+**Input:**
+```python
+video = {
+    "id": 1,
+    "file_path": "/data/input/...",
+    "codec": "wmv1",
+    "duration": 102.499,
+    "resolution_w": 640,
+    "resolution_h": 480,
+    "file_size": 19358016,
+    "file_hash": "abc...",
+}
+```
+
+**Output (results dict stored in DB):**
+```python
+{
+    "status": "COMPLETED",           # or "FAILED" on error
+    "confidence": 0.87,              # video-level confidence score
+    "routing_decision": "MATCH",     # MATCH | REVIEW | SKIP
+    "face_count": 142,               # total faces detected across all frames
+    "frame_count": 23,               # frames extracted and processed
+    "model_version": "v0.0.0-placeholder",
+    "processing_time_seconds": 12.4,
+    "error": None,                   # error message if FAILED
+}
+```
+
+**Processing flow (extracted from orchestrator.py):**
+
+```python
+def process_video(video: dict) -> dict:
+    """Run the full AI scan pipeline on a single video.
+    
+    Returns results dict for submission via API.
+    """
+    start_time = time.time()
+    file_path = video["file_path"]
+    video_id = video["id"]
+    
+    try:
+        # 1. Probe (if metadata is incomplete)
+        prober = VideoProber(timeout=10)
+        metadata = prober.probe(file_path)
+        
+        # 2. Extract frames to scratch
+        scratch = ScratchManager(base_path="/scratch", video_id=str(video_id))
+        frame_dir = scratch.ensure_frame_dir()
+        sampler = FrameSampler(interval_seconds=30, quality=2)
+        frame_paths = sampler.extract_frames(
+            video_path=file_path, output_dir=str(frame_dir),
+            duration=metadata.duration, resolution=(metadata.resolution_w, metadata.resolution_h)
+        )
+        
+        # 3. Detect faces
+        detector = FaceDetector(engine_path="/models/face_detector/face_detector.trt")
+        detections_per_frame = detector.detect_faces(frame_paths)
+        total_faces = sum(len(dets) for dets in detections_per_frame)
+        
+        # 4. Crop and classify (aggregate confidences from face crops)
+        classifier = FaceClassifier(engine_path="/models/classifier/classifier.trt")
+        # ... crop paths → classify → get confidence list
+        
+        # 5. Aggregate
+        confidence = aggregate(confidences, strategy="max")
+        
+        # 6. Route
+        routing = router.route(confidence)
+        
+        processing_time = time.time() - start_time
+        
+        return {
+            "status": "COMPLETED",
+            "confidence": round(confidence, 4),
+            "routing_decision": routing,
+            "face_count": total_faces,
+            "frame_count": len(frame_paths),
+            "model_version": get_model_version(),
+            "processing_time_seconds": round(processing_time, 2),
+            "error": None,
+        }
+        
+    except Exception as exc:
+        logger.error("Processing failed for video %d: %s", video_id, exc)
+        return {
+            "status": "FAILED",
+            "confidence": 0.0,
+            "routing_decision": "REVIEW",
+            "face_count": 0,
+            "frame_count": 0,
+            "model_version": get_model_version(),
+            "processing_time_seconds": round(time.time() - start_time, 2),
+            "error": str(exc),
+        }
+```
+
+**Notes for implementation:**
+- Reuse existing classes from `orchestrator.py` (`FaceDetector`, `FaceClassifier`, `FrameSampler`, `ScratchManager`, `VideoProber`, `aggregate`, `router`) without refactoring them — just import and use them.
+- GPU setup (torch CUDA, GPUMemoryManager) can be done once at module level or lazily inside `process_video()` to avoid overhead per task invocation.
+- Model loading is expensive (~seconds). Consider lazy initialization or process-level caching via a singleton pattern.
+
+---
+
+### Phase 4 — Wire It Together in task_worker.py
+
+**Goal:** Combine the API client, processing pipeline, and CLI entry point into a working script.
+
+**Acceptance Criteria:**
+- Full end-to-end: claim → fetch → process → submit → repeat N times
+- Exits with code 0 on success, non-zero on unrecoverable errors
+- Logs startup/shutdown counts (tasks attempted, tasks succeeded, tasks failed)
+- Graceful shutdown on SIGTERM/SIGINT (finish current task, then exit)
+
+**Expected lifecycle log output:**
+
+```json
+{"timestamp": "...", "level": "INFO", "message": "Task worker starting. Will process 3 task(s)."}
+{"timestamp": "...", "level": "INFO", "message": "API client configured: base_url=http://localhost:8890"}
+{"timestamp": "...", "level": "INFO", "message": "Claiming task 1 (video_id=42) for AISCAN"}
+{"timestamp": "...", "level": "INFO", "message": "Processing video 42: /data/input/..."}
+{"timestamp": "...", "level": "INFO", "message": "Video 42 complete: C=0.87 routing=MATCH faces=142 frames=23 time=12.4s"}
+{"timestamp": "...", "level": "INFO", "message": "Submitting results for task 1 via API"}
+{"timestamp": "...", "level": "INFO", "message": "Claiming task 2 (video_id=99) for AISCAN"}
+{"timestamp": "...", "level": "INFO", "message": "Processing video 99: /data/input/..."}
+{"timestamp": "...", "level": "INFO", "message": "Video 99 complete: C=0.31 routing=SKIP faces=0 frames=18 time=8.7s"}
+{"timestamp": "...", "level": "INFO", "message": "Submitting results for task 2 via API"}
+{"timestamp": "...", "level": "INFO", "message": "Claiming task 3 (video_id=157) for AISCAN"}
+{"timestamp": "...", "level": "WARNING", "message": "Processing video 157 failed: ffprobe error: cannot decode stream"}
+{"timestamp": "...", "level": "INFO", "message": "Submitting FAILED results for task 3 via API"}
+{"timestamp": "...", "level": "INFO", "message": "No more tasks available (API returned 404)"}
+{"timestamp": "...", "level": "INFO", "message": "Task worker finished: 3 attempted, 2 succeeded, 1 failed"}
+```
+
+**Core loop:**
+
+```python
+for i in range(1, args.tasks + 1):
+    logger.info("=== Processing task %d/%d ===", i, args.tasks)
+    
+    # Claim a task
+    response = api.get_next_task("AISCAN")
+    if response is None:
+        logger.info("No more tasks available. Done.")
+        break
+    
+    task = response["task"]
+    assign_key = response["assign_key"]
+    video_id = task["video_id"]
+    
+    # Fetch video metadata
+    video = api.get_video(video_id)
+    
+    # Process
+    results = process_video(video)
+    
+    # Submit
+    api.submit_results(task["id"], assign_key, results)
+
+logger.info("Finished: %d attempted, %d succeeded, %d failed", total_attempted, total_succeeded, total_failed)
+```
+
+---
+
+### Phase 5 — Update docker-compose.yml
+
+**Goal:** Change the worker service from a long-lived `WorkerPool` to a short-lived task process.
+
+**Changes to `docker-compose.yml` worker section:**
+
+1. **Update command** to invoke the task worker instead of `main.py`:
+   ```yaml
+   command: python3 -m src.task_worker --tasks 50
+   ```
+
+2. **Add restart policy considerations** — since workers are now short-lived, you have two options:
+   - **Option A (simpler):** Run a single worker container with a large `--tasks` count (e.g., 100) so it processes many videos before exiting, then rely on Kubernetes/Cron/external scheduler to restart.
+   - **Option B (more flexible):** Make `--tasks` configurable via environment variable:
+     ```yaml
+     environment:
+       - TASK_COUNT=${TASK_COUNT:-50}
+     command: >
+       python3 -m src.task_worker --tasks ${TASK_COUNT}
+     ```
+
+3. **Remove the WorkerPool initialization** from `main.py` — it becomes unused (or is removed entirely in a later cleanup).
+
+---
+
+### Phase 6 — Verify API Compatibility
+
+**Goal:** Confirm the Dancer2 Perl API correctly handles JSON body submission from Python's `requests` library on `/api/v1/task/:task/complete`.
+
+**Acceptance Criteria:**
+- POST with `Content-Type: application/json` body containing `assign_key` + `results` (as a dict, not a string) returns 200
+- POST with form-encoded data also works as fallback
+- The Perl API correctly stores the results JSON in the database
+
+**Test procedure:**
+
+```bash
+# Step 1: Create a test task manually (via MySQL or via scanner)
+mysql videodetect -e "INSERT INTO tasks (task_type, video_id, status) VALUES ('AISCAN', 1, 'PENDING');"
+
+# Step 2: Claim the task via API
+curl -s http://localhost:8890/api/v1/nexttask/AISCAN | python3 -m json.tool
+
+# Step 3: Submit results (adjust video_id to a real one)
+curl -X POST http://localhost:8890/api/v1/task/1/complete \
+  -H "Content-Type: application/json" \
+  -d '{"assign_key":"worker_999","results":{"confidence":0.87,"routing_decision":"MATCH"}}' | python3 -m json.tool
+
+# Step 4: Verify results stored in DB
+mysql videodetect -e "SELECT id, status, results FROM tasks WHERE id=1;"
+```
+
+**If JSON body does NOT work**, update the Dancer2 API to handle JSON explicitly:
+
+```perl
+post '/api/v1/task/:task/complete' => sub {
+    my $task = route_parameters->get("task");
+    
+    # Handle both JSON body and form-encoded data
+    my $body;
+    if (request_content_type eq 'application/json') {
+        use Dancer2::Core::Request::Entity;
+        $body = decode_json(request_body);
+    } else {
+        $body = body_parameters->to_hash;
+    }
+    
+    my $assign_key = $body->{assign_key};
+    my $results    = $body->{results};
+    
+    # ... rest of the handler unchanged
+};
+```
+
+---
+
+## Files Modified / Created Summary
+
+| File | Action | Description |
+|------|--------|-------------|
+| `src/task_worker.py` | **NEW** | CLI entry point: `--tasks N`, loop claim→process→submit |
+| `src/api_client.py` | **NEW** | HTTP client for the Perl API (`ApiClient`) |
+| `src/task_processor.py` | **NEW** | `process_video(video_dict)` — core AI pipeline callable standalone |
+| `docker-compose.yml` | **MODIFY** | Change worker command to invoke task_worker; add TASK_COUNT env var |
+| `api/app.pl` | **MAYBE MODIFY** | Update `/task/:task/complete` to handle JSON body if Dancer2 doesn't support it natively |
+| `src/main.py` | **MODIFY (cleanup later)** | Remove or deprecate WorkerPool usage (long-lived process no longer needed) |
+| `REFACTOR.md` | **CURRENT FILE** | This plan document |
+
+---
+
+## Risks & Considerations
+
+1. **GPU model loading per invocation:** Each new Python process loads the ONNX/TensorRT engine into GPU memory. With `--tasks 1`, this is wasteful. Mitigation: use larger `--tasks` values (e.g., 20-100) so amortization is favorable, or implement a local cache server pattern later.
+
+2. **Permanence of scratch space:** The `ScratchManager` creates per-video temp files under `/scratch`. Since workers are now short-lived, ensure `cleanup=True` is always set (it is by default in the existing code).
+
+3. **API JSON body compatibility:** Dancer2's `body_parameters` may not parse JSON bodies — it expects form-encoded data. This is the highest-risk item. Test Phase 6 early.
+
+4. **Task idempotency:** If a worker crashes between processing and submitting results, the task remains `IN_PROGRESS`. The Perl API prevents re-claiming (the WHERE clause checks `status='PENDING'`). A manual SQL update or a `/api/v1/task/:task/reset` endpoint may be needed for recovery. This is outside the scope of this refactor but worth noting.
+
+---
+
+## Suggested Order of Execution
+
+1. **Phase 6 first** — verify API JSON compatibility (takes 5 minutes, unblocks everything)
+2. **Phase 2** — create `api_client.py` with minimal methods
+3. **Phase 3** — create `task_processor.py` by extracting from `orchestrator.py`
+4. **Phase 1** — create `task_worker.py` with CLI + loop
+5. **Phase 4** — wire together and test locally (`python3 -m src.task_worker --tasks 3`)
+6. **Phase 5** — update docker-compose.yml and deploy

STORY-01.md

@@ -1,170 +0,0 @@
-# STORY-01: Foundation & Infrastructure
-
-## Epic
-**E6: Infrastructure** — As a DevOps engineer, I can deploy the entire stack via Docker Compose.
-
-## Related Requirements
-| ID | Requirement |
-|----|-------------|
-| TC-01 | Hardware: 2× Tesla P40 24GB (compute capability 5.2, PCIe 3.0, no Tensor Cores) |
-| TC-02 | CUDA/Torch Compatibility: CUDA ≤ 11.8, PyTorch ≤ 2.1.0, FP32 inference only |
-| TC-03 | Storage I/O: Fast local NVMe/SSD for temp frame cache; shared NAS/SMB for video input/output |
-| TC-04 | Framework Stack: PyTorch → ONNX → TensorRT FP32; FFmpeg/OpenCV; MariaDB |
-| TC-05 | Deployment Model: Docker Compose orchestrates all services; GPUs passed via nvidia-container-toolkit |
-| TC-06 | Network Security: Internal LAN only; no reverse proxy, SSL, or auth |
-| NFR-04 | Determinism & Reproducibility: Config-seeded randomness, versioned models |
-
-## Description
-Establish the Docker environment, database schema, and basic connectivity. No video processing logic yet — just the operational skeleton that all subsequent stories depend on.
-
-## Scope
-
-### In Scope
-- Docker Compose multi-service orchestration
-- Custom Worker Dockerfile with CUDA 11.8 + PyTorch 2.1.0 + TensorRT FP32
-- MariaDB schema design and migration scripts
-- Volume mounts for NVMe scratch, NAS input/output, and model persistence
-- Configuration management (config.yaml + environment variables)
-- Structured logging setup (JSON format to stdout)
-- Database connection layer with connection pooling
-- Network connectivity verification between all services
-
-### Out of Scope
-- Video processing logic (covered in STORY-02 through STORY-05)
-- Model loading or inference (covered in STORY-04)
-- Review UI functionality (covered in STORY-06)
-- Monitoring dashboards (covered in STORY-08)
-- Active learning pipeline (covered in STORY-07)
-
-## Deliverables
-
-### 1.1 Docker Compose Structure
-**File:** `docker-compose.yml`
-
-Services defined:
-- `mariadb`: MariaDB 10.11+ with persistent volume
-- `worker`: PyTorch/TensorRT inference worker with GPU passthrough
-- `ui`: Placeholder service (nginx serving static page) for network verification
-
-Key configurations:
-- `nvidia-container-toolkit` runtime configuration for GPU passthrough
-- Volume mounts:
-  - NVMe → `/scratch` (tmpfs for speed)
-  - NAS/SMB → `/data/input` and `/data/output`
-  - Persistent → `/models` and `/data/training`
-- Network bridge for inter-service communication
-- Resource limits (GPU memory caps per NFR-03)
-
-### 1.2 Worker Dockerfile
-**File:** `worker/Dockerfile`
-
-Base image: `nvidia/cuda:11.8.0-runtime-ubuntu22.04`
-
-Installed packages:
-- PyTorch 2.1.0 (CUDA 11.8, FP32 only)
-- TensorRT 8.6+ (FP32)
-- OpenCV 4.8+
-- FFmpeg 5.x + ffprobe
-- ONNX Runtime
-- Python 3.10+
-- Required system libraries (libcudnn8, libglib2.0, etc.)
-
-### 1.3 Database Schema
-**File:** `db/schema.sql`
-
-Tables:
-- `videos`: `id` (BIGINT PK), `file_path` (VARCHAR), `file_hash` (CHAR(64)), `resolution_w` (INT), `resolution_h` (INT), `codec` (VARCHAR), `duration` (FLOAT), `status` (ENUM: NEW, PENDING, PROCESSING, COMPLETED, UNSCANNABLE, ERROR), `last_scan_time` (DATETIME), `last_processed_time` (DATETIME), `created_at` (DATETIME), `updated_at` (DATETIME)
-- `processing_logs`: `id` (BIGINT PK), `video_id` (BIGINT FK), `model_version` (VARCHAR), `frame_count` (INT), `confidence_score` (FLOAT), `routing_decision` (ENUM: MATCH, REVIEW, SKIP), `processed_at` (DATETIME), `error_message` (TEXT)
-- `models`: `version` (VARCHAR PK), `status` (ENUM: ACTIVE, CANDIDATE, ARCHIVED), `path` (VARCHAR), `calibration_temp` (FLOAT), `f1_score` (FLOAT), `ece_score` (FLOAT), `deployed_at` (DATETIME), `created_at` (DATETIME)
-- `review_queue`: `id` (BIGINT PK), `video_id` (BIGINT FK), `confidence_score` (FLOAT), `routing_decision` (ENUM: REVIEW), `annotated` (BOOLEAN), `ground_truth` (BOOLEAN), `annotated_at` (DATETIME), `created_at` (DATETIME)
-
-Indexes:
-- `idx_videos_status` on `videos(status)`
-- `idx_videos_file_hash` on `videos(file_hash)` (UNIQUE)
-- `idx_videos_last_scan` on `videos(last_scan_time)`
-- `idx_processing_logs_video` on `processing_logs(video_id)`
-- `idx_models_status` on `models(status)`
-- `idx_review_queue_annotated` on `review_queue(annotated)`
-
-### 1.4 Database Connection Layer
-**File:** `src/db_connector.py`
-
-Features:
-- Connection pooling (DBUtils + PyMySQL)
-- Configurable pool size (min=5, max=20)
-- Automatic reconnection on disconnect
-- Context manager support
-- Prepared statements for all queries
-- Transaction support for atomic state transitions
-
-### 1.5 Configuration Management
-**File:** `config.yaml`
-
-Contents:
-- `sampling`: `interval_seconds: 30`, `override_per_job: true`
-- `thresholds`: `T_high: 0.75`, `T_low: 0.45`
-- `gpu`: `max_memory_gb: 18`, `batch_size: auto`, `device: cuda`
-- `storage`: `scratch_path: /scratch`, `input_path: /data/input`, `output_path: /data/output`
-- `database`: `host: mariadb`, `port: 3306`, `pool_size: 20`
-- `logging`: `format: json`, `level: INFO`
-- `model`: `face_detector: yolo8n`, `classifier: mobilenetv3`, `input_size: 224`
-
-**File:** `src/config_loader.py`
-
-Features:
-- Load config.yaml with environment variable overrides
-- Validate all required fields
-- Provide typed accessors (e.g., `config.thresholds.T_high`)
-- Hot-reload support for config changes
-
-### 1.6 Logging Setup
-**File:** `src/logging_config.py`
-
-- JSON structured logging via `python-json-logger`
-- Log levels: DEBUG, INFO, WARNING, ERROR, CRITICAL
-- Fields: timestamp, level, service, video_id, message, metadata (key-value pairs)
-- Log rotation: 100MB per file, 10 files max
-- All logs to stdout for Docker capture
-
-## Acceptance Criteria
-
-### Functional
-- [ ] `docker-compose up` starts MariaDB, Worker, and UI containers successfully
-- [ ] Worker container can connect to MariaDB and execute schema migrations
-- [ ] GPU is visible inside Worker container (`nvidia-smi` shows Tesla P40)
-- [ ] Volume mounts are accessible and writable in all containers
-- [ ] Config.yaml loads correctly with all required fields validated
-- [ ] Structured logging produces valid JSON output in all services
-- [ ] Database connection pool handles concurrent connections (test with 20 simultaneous)
-
-### Non-Functional
-- [ ] Worker container starts within 30 seconds
-- [ ] MariaDB starts within 15 seconds
-- [ ] GPU memory usage in Worker is < 2GB at idle (before model loading)
-- [ ] All services communicate over Docker internal network (no host network exposure except UI port)
-- [ ] Schema migration is idempotent (running twice produces same result)
-
-### Technical Constraints
-- [ ] CUDA version in Worker is 11.8 (verified via `torch.version.cuda`)
-- [ ] PyTorch version ≤ 2.1.0 (verified via `torch.__version__`)
-- [ ] TensorRT runs in FP32 mode only
-- [ ] No Tensor Cores used (compute capability 5.2 constraint respected)
-- [ ] No SSL, auth, or reverse proxy configured (TC-06)
-
-## Dependencies
-- **Prerequisites:** NVIDIA Container Toolkit installed on host, Docker Compose v2+, NAS/SMB mounts configured
-- **Depends on:** None (this is the foundational story)
-- **Enables:** STORY-02 (Ingestion), STORY-03 (Orchestration), STORY-04 (Inference), STORY-05 (Results), STORY-06 (Review UI), STORY-07 (Active Learning), STORY-08 (Monitoring)
-
-## Risks & Mitigations
-| Risk | Mitigation |
-|------|-----------|
-| Tesla P40 (CC 5.2) incompatible with newer TensorRT | Use TensorRT 8.6 which supports CC 5.x; test early |
-| CUDA 11.8 + PyTorch 2.1.0 dependency conflicts | Pin all versions in Dockerfile; use nvidia base image |
-| NAS/SMB mount latency affects processing | Use local tmpfs for scratch; only read from NAS |
-| MariaDB connection pool exhaustion | Monitor pool metrics; tune pool_size based on worker count |
-
-## Estimated Effort
-- **Sprint:** 1-2
-- **Story Points:** 13
-- **Dependencies:** None

STORY-02.md

@@ -1,182 +0,0 @@
-# STORY-02: Core Ingestion & Codec Handling
-
-## Epic
-**E5: Data Management** — As a system, I can scan directories and sync file state to MariaDB.
-**E1: Core Pipeline** — As a system, I can detect video resolution and codec.
-
-## Related Requirements
-| ID | Requirement |
-|----|---------|
-| FR-09 | Directory Scanning & Sync: Process to scan input directories, detect new/removed files, and sync state to MariaDB |
-| FR-10 | Codec & Resolution Detection & Handling: Detect video properties; handle unsupported codecs by flagging files as UNSCANNABLE |
-| NFR-07 | Data Volume Handling: Efficient indexing for ~30TB dataset |
-| NFR-08 | Codec Agnosticism: Handle H.264, H.265, VP8, VP9, AV1, MJPEG, etc. |
-| NFR-05 | Fault Tolerance: Auto-retry on transient failures; skip & log on fatal errors |
-
-## Description
-Implement the directory scanner, video probing, and robust error handling for codecs and resolutions. This story enables the system to discover new files in the 30TB corpus, extract their metadata, validate codec support, and maintain accurate database state.
-
-## Scope
-
-### In Scope
-- Background directory scanner service that walks `/data/input`
-- File detection and deduplication via SHA-256 hashing
-- Video probing via ffprobe (codec, resolution, duration)
-- Codec whitelist/blacklist validation
-- UNSCANNABLE status for unsupported or corrupt files
-- Priority queue / DB-based locking for job assignment
-- Atomic state transitions (PENDING → PROCESSING)
-- Efficient handling of 30TB directory structure (incremental scanning)
-
-### Out of Scope
-- Frame extraction (covered in STORY-04)
-- Face detection and classification (covered in STORY-04)
-- Confidence scoring and routing (covered in STORY-05)
-- Review UI (covered in STORY-06)
-- Active learning pipeline (covered in STORY-07)
-- Monitoring dashboards (covered in STORY-08)
-
-## Deliverables
-
-### 2.1 Directory Scanner Service
-**File:** `src/scanner.py`
-
-Core components:
-- **Walker:** Recursive directory walker with configurable depth and path filters
-- **Incremental Sync:** Compare current filesystem state against DB `last_scan_time` — only process new/modified files
-- **File Detection:** Identify files not in DB or with status NEW/PENDING
-- **Hash Computation:** SHA-256 of first 1MB (or full file if < 1MB) for deduplication
-- **Scan Scheduler:** Configurable interval (default: every 60 seconds) via cron-like scheduler
-- **Concurrency:** Multi-threaded walker with configurable worker count (default: 8 threads)
-- **Error Handling:** Per-file error isolation — scanner continues on individual file failures
-
-Key behaviors:
-- On scan start: query DB for files with `last_scan_time < now()` or `status IN ('NEW', 'PENDING')`
-- For each file: compute hash → check DB for duplicate → if new, insert with status PENDING
-- For removed files: mark as REMOVED in DB (optional, configurable)
-- Update `last_scan_time` on `videos` table after successful scan
-
-### 2.2 Video Probing Module
-**File:** `src/prober.py`
-
-Features:
-- **ffprobe wrapper:** Execute ffprobe with optimized arguments for speed
-  ```bash
-  ffprobe -v error -select_streams v:0 -show_entries stream=codec_name,width,height,duration,r_frame_rate -show_entries format=duration -of json <file>
-  ```
-- **Metadata extraction:** Parse JSON output for:
-  - `codec_name`: Video codec identifier
-  - `width`, `height`: Resolution
-  - `duration`: Video duration in seconds
-  - `r_frame_rate`: Frame rate (for sampling calculations)
-- **Error handling:** Catch ffprobe failures (corrupt files, unsupported formats)
-- **Timeout:** ffprobe execution limited to 10 seconds per file
-
-### 2.3 Codec Validation
-**File:** `src/codec_validator.py`
-
-Codec whitelist (supported):
-- H.264 (avc1)
-- H.265 (hevc)
-- VP8 (vp8)
-- VP9 (vp9)
-- AV1 (av01)
-- MJPEG (mjpeg)
-- MPEG-4 (mp4v)
-
-Codec blacklist (unsupported):
-- Theora
-- DivX/Xvid (legacy)
-- ProRes (requires special handling)
-- Any codec not in whitelist
-
-Behavior:
-- If codec not in whitelist → set status UNSCANNABLE, log reason
-- If ffprobe fails → set status UNSCANNABLE, log error code
-- If file is not a valid video → set status UNSCANNABLE, log reason
-- UNSCANNABLE files are excluded from processing pipeline
-
-### 2.4 Batch Orchestration Skeleton
-**File:** `src/orchestrator.py`
-
-Components:
-- **Job Queue:** Priority queue based on file modification time (newest first)
-- **DB-based Locking:** Atomic state transition using UPDATE ... WHERE status = 'PENDING'
-- **Worker Pool:** Configurable number of worker processes (default: 2, one per GPU)
-- **State Machine:**
-  ```
-  NEW → PENDING → PROCESSING → COMPLETED
-                         → UNSCANNABLE
-                         → ERROR
-  ```
-- **Concurrency Control:** Max concurrent processing per GPU (default: 1 video at a time per worker)
-- **Idempotency:** Re-processing a file does not duplicate DB entries or outputs
-
-### 2.5 Configuration Updates
-**File:** `config.yaml` (updates to STORY-01)
-
-New fields:
-```yaml
-scanner:
-  scan_interval_seconds: 60
-  walker_threads: 8
-  ffprobe_timeout_seconds: 10
-  hash_algorithm: sha256
-  hash_chunk_size_mb: 1
-
-codec:
-  whitelist: [avc1, hevc, vp8, vp9, av01, mjpeg, mp4v]
-  default_status_on_error: UNSCANNABLE
-
-queue:
-  priority: modification_time  # newest first
-  max_concurrent_per_gpu: 1
-  lock_timeout_seconds: 300
-```
-
-## Acceptance Criteria
-
-### Functional
-- [ ] New files in `/data/input` appear in DB with correct metadata within 60 seconds of placement
-- [ ] File hash deduplication prevents re-processing identical files
-- [ ] ffprobe correctly extracts codec, resolution, and duration for all supported codecs
-- [ ] Unsupported codec files are marked UNSCANNABLE without crashing the scanner
-- [ ] Corrupt/unreadable files are marked UNSCANNABLE with appropriate error logged
-- [ ] Atomic state transitions: PENDING → PROCESSING succeeds only once (no duplicate processing)
-- [ ] Removed files are detected and marked REMOVED (if configured)
-- [ ] Scanner handles 30TB directory structure without OOM (memory < 500MB during scan)
-- [ ] Re-running scanner is idempotent — no duplicate entries or states
-
-### Non-Functional
-- [ ] Scanner completes full 30TB directory walk in < 4 hours (incremental: < 10 minutes for typical day)
-- [ ] ffprobe timeout (10s) is enforced — does not hang on corrupt files
-- [ ] Scanner memory usage stays < 500MB regardless of directory depth
-- [ ] Hash computation for 100MB file completes in < 5 seconds on NVMe
-- [ ] Worker pool respects GPU count (2 workers for 2 GPUs)
-
-### Technical Constraints
-- [ ] SHA-256 hash is deterministic and reproducible
-- [ ] Codec detection matches ffprobe output exactly
-- [ ] Resolution stored as integers (width, height)
-- [ ] Duration stored as float (seconds)
-- [ ] All scanner errors are logged with video_id, error_code, and file_path
-- [ ] Scanner does not modify video files (read-only operation)
-
-## Dependencies
-- **Prerequisites:** STORY-01 (Foundation & Infrastructure) — DB schema, connection layer, config
-- **Depends on:** None (runs in parallel with STORY-03 if needed)
-- **Enables:** STORY-04 (Inference Pipeline), STORY-05 (Results & Export)
-
-## Risks & Mitigations
-| Risk | Mitigation |
-|------|--|
-| 30TB directory walk is extremely slow | Incremental scanning using last_scan_time; only walk new directories |
-| ffprobe hangs on corrupt files | Enforce 10-second timeout via subprocess timeout |
-| Hash computation on slow NAS is bottleneck | Hash only first 1MB; full hash optional for verification |
-| Concurrent scanner + worker conflicts | DB-based locking; scanner only writes PENDING, workers read PENDING |
-| Network mount latency | Cache directory listings; batch DB operations |
-
-## Estimated Effort
-- **Sprint:** 3-4
-- **Story Points:** 21
-- **Dependencies:** STORY-01

STORY-03.md

@@ -1,152 +0,0 @@
-# STORY-03: Frame Sampling
-
-## Epic
-**E1: Core Pipeline** — As an engineer, I can configure frame sampling interval and extract frames uniformly.
-
-## Related Requirements
-| ID | Requirement |
-|----|---------|
-| FR-01 | Configurable frame sampling interval (default: 1 frame per 30 seconds); Must support override per job/batch; Uniform temporal sampling preferred |
-| NFR-02 | Latency per video: ≤ 45 seconds end-to-end (15-min avg video) |
-| NFR-03 | GPU Memory Safety: ≤ 18GB per GPU sustained |
-| NFR-05 | Fault Tolerance: Auto-retry on transient failures; skip & log on fatal errors |
-| TC-03 | Storage I/O: Fast local NVMe/SSD for temp frame cache |
-
-## Description
-Implement frame extraction from processed videos using FFmpeg/OpenCV. Extract one frame per configurable interval (default: 30 seconds) uniformly across the video duration. Handle variable FPS, high-resolution frames, and manage scratch space efficiently.
-
-## Scope
-
-### In Scope
-- Uniform temporal frame sampling via FFmpeg `-ss` timestamp extraction
-- Configurable sampling interval (default: 30 seconds, override per job)
-- Variable FPS handling (calculate correct timestamps)
-- Frame extraction to JPEG format for storage efficiency
-- Scratch space management (NVMe tmpfs with auto-cleanup)
-- High-resolution frame handling (downscale 4K to fit VRAM constraints)
-- Frame naming convention: `{video_id}_{timestamp}.jpg`
-- Memory management (delete frames after processing)
-
-### Out of Scope
-- Face detection (covered in STORY-04)
-- Classification (covered in STORY-05)
-- Confidence scoring (covered in STORY-05)
-- Results export (covered in STORY-06)
-- Review UI (covered in STORY-07)
-
-## Deliverables
-
-### 3.1 Frame Sampler Module
-**File:** `src/frame_sampler.py`
-
-Core functionality:
-- **Timestamp Calculator:** Compute uniform sampling timestamps
-  ```python
-  def calculate_timestamps(duration, interval):
-      """Return list of timestamps for frame extraction."""
-      count = max(1, int(duration / interval))
-      step = duration / count
-      return [i * step for i in range(count)]
-  ```
-- **FFmpeg Extraction:** Extract frames at computed timestamps
-  ```bash
-  ffmpeg -ss {timestamp} -i {video_path} -vframes 1 -q:v 2 -f jpeg {output_path}
-  ```
-- **Variable FPS Handling:** Adjust timestamps for videos with variable frame rates
-- **Quality Control:** JPEG quality factor 2 (high quality, reasonable size)
-- **Error Handling:** Per-frame error isolation — skip failed frames, log errors, continue
-
-### 3.2 Resolution Handling
-**File:** `src/frame_sampler.py` (resolution handling section)
-
-Features:
-- **Resolution Detection:** Use metadata from prober (STORY-02)
-- **Downscale Logic:** Auto-downscale frames > 1080p to fit VRAM
-  - 4K (3840x2160) → 1080p (1920x1080)
-  - 2K (2560x1440) → 720p (1280x720)
-  - ≤ 1080p → no change
-- **FFmpeg Scale Filter:** Apply during extraction
-  ```bash
-  ffmpeg -ss {ts} -i {video} -vf "scale=1920:1080:force_original_aspect_ratio=decrease,pad=1920:1080:(ow-iw)/2:(oh-ih)/2" -vframes 1 -q:v 2 -f jpeg {output}
-  ```
-- **Aspect Ratio Preservation:** Letterbox/pillarbox to maintain original aspect ratio
-
-### 3.3 Scratch Space Management
-**File:** `src/scratch_manager.py`
-
-Features:
-- **Directory Structure:** `/scratch/{video_id}/frames/` for each video
-- **Naming Convention:** `{video_id}_{timestamp_ms}.jpg`
-- **Auto-Cleanup:** Delete all frames after video processing completes
-- **Space Monitoring:** Alert if scratch usage exceeds 80% of tmpfs capacity
-- **Cleanup on Error:** Ensure frames are cleaned up even if processing fails
-- **Concurrent Access:** Per-video directory isolation (no conflicts between workers)
-
-### 3.4 Configuration Updates
-**File:** `config.yaml` (updates to STORY-01/02)
-
-New fields:
-```yaml
-sampling:
-  interval_seconds: 30
-  override_per_job: true
-  quality: 2  # JPEG quality (1-31, lower=better)
-  format: jpeg
-
-resolution:
-  max_height: 1080
-  preserve_aspect_ratio: true
-  pad_to_square: false  # Will be done in face detection if needed
-
-scratch:
-  path: /scratch
-  max_usage_percent: 80
-  auto_cleanup: true
-  cleanup_on_error: true
-```
-
-## Acceptance Criteria
-
-### Functional
-- [ ] Frames are extracted at uniform intervals (±1 frame tolerance) across video duration
-- [ ] Default interval of 30 seconds produces correct number of frames for test videos
-- [ ] Per-job interval override works (e.g., 10-second interval for specific videos)
-- [ ] Variable FPS videos produce evenly spaced timestamps (not evenly spaced frames)
-- [ ] 4K frames are downsampled to 1080p without crashing
-- [ ] Aspect ratio is preserved during downscaling (no distortion)
-- [ ] Frames are saved as high-quality JPEGs (quality factor 2)
-- [ ] Frame naming follows convention: `{video_id}_{timestamp_ms}.jpg`
-- [ ] Scratch space is cleaned up after video processing (no leftover frames)
-- [ ] Failed frame extraction does not stop processing of other frames
-
-### Non-Functional
-- [ ] Frame extraction for a 15-minute video completes in < 10 seconds on NVMe
-- [ ] Scratch space usage per video < 50MB (typical case)
-- [ ] Memory usage during extraction < 500MB per worker
-- [ ] FFmpeg process timeout enforced (30 seconds per frame)
-- [ ] No frames left in scratch after processing completes (verified by directory check)
-
-### Technical Constraints
-- [ ] FFmpeg uses `-ss` for accurate timestamp seeking (not frame-by-frame)
-- [ ] Downscaling uses FFmpeg scale filter (not post-processing resize)
-- [ ] JPEG quality is consistent across all extracted frames
-- [ ] Timestamps are in seconds with millisecond precision
-- [ ] Scratch directory is created per-video and isolated from other workers
-
-## Dependencies
-- **Prerequisites:** STORY-01 (Foundation), STORY-02 (Ingestion — provides metadata)
-- **Depends on:** None (runs after ingestion, before STORY-04)
-- **Enables:** STORY-04 (Face Detection — provides frames), STORY-05 (Classification)
-
-## Risks & Mitigations
-| Risk | Mitigation |
-|------|--|
-| FFmpeg seeking is inaccurate for some codecs | Use `-accurate_seek` flag; fall back to frame-by-frame if needed |
-| 4K downscaling increases processing time | Use FFmpeg scale filter (GPU-accelerated if available) |
-| Scratch space fills up with many concurrent videos | Monitor usage; implement cleanup queue; alert at 80% |
-| Variable FPS causes uneven frame distribution | Calculate timestamps based on duration, not frame count |
-
-## Estimated Effort
-- **Sprint:** 5 (first half)
-- **Story Points:** 13
-- **Dependencies:** STORY-01, STORY-02

STORY-04.md

@@ -1,168 +0,0 @@
-# STORY-04: Face Detection
-
-## Epic
-**E1: Core Pipeline** — As an engineer, I can run face detection on sampled frames in batch.
-
-## Related Requirements
-| ID | Requirement |
-|----|-------|
-| FR-02 | Face detection on all sampled frames: Lightweight detector only; no full-body or scene analysis |
-| NFR-03 | GPU Memory Safety: ≤ 18GB per GPU sustained |
-| NFR-01 | Throughput: ≥ 30 videos/hour/GPU |
-| TC-01 | Hardware: 2× Tesla P40 24GB (compute capability 5.2, PCIe 3.0, no Tensor Cores) |
-| TC-02 | CUDA/Torch Compatibility: CUDA ≤ 11.8, PyTorch ≤ 2.1.0, FP32 inference only |
-| TC-04 | Framework Stack: PyTorch → ONNX → TensorRT FP32 |
-
-## Description
-Implement face detection on all sampled frames using a lightweight detector (YOLOv8n or equivalent). Convert the model to ONNX and TensorRT FP32 for inference. Implement dynamic batching to maximize GPU utilization while staying within VRAM constraints. Generate face crops for classification.
-
-## Scope
-
-### In Scope
-- YOLOv8n face detection model (or equivalent lightweight detector)
-- ONNX export and TensorRT FP32 engine conversion
-- Dynamic batching for face detection inference
-- Face crop extraction and resizing to inference input size (224×224)
-- GPU memory management (stay within 18GB per GPU)
-- Batch size auto-tuning based on available VRAM
-- Face bounding box output with confidence scores
-- Handling frames with no detected faces
-
-### Out of Scope
-- Frame sampling (covered in STORY-03)
-- Demographic classification (covered in STORY-05)
-- Confidence aggregation (covered in STORY-05)
-- Results export (covered in STORY-06)
-- Review UI (covered in STORY-07)
-
-## Deliverables
-
-### 4.1 Face Detection Model
-**File:** `models/face_detector/`
-
-Components:
-- **Source Model:** YOLOv8n (Ultralytics) trained on face detection dataset (WIDER Face or equivalent)
-- **ONNX Export:** `face_detector.onnx` with proper input/output shapes
-  - Input: `[1, 3, 640, 640]` (RGB, BGR depending on model)
-  - Output: `[1, num_anchors, 4+1]` (bbox coordinates + confidence)
-- **TensorRT Engine:** `face_detector.trt` (FP32, built for Tesla P40 CC 5.2)
-  - Builder config: max_batch_size=32, max_workspace_size=4GB
-  - Serialization for runtime loading
-- **Model Metadata:** `model.json` with input shape, normalization params, calibration data
-
-### 4.2 Face Detection Runner
-**File:** `src/face_detector.py`
-
-Core components:
-- **Model Loader:** Load TensorRT engine at startup
-  ```python
-  class FaceDetector:
-      def __init__(self, engine_path, device='cuda'):
-          self.engine = load_trt_engine(engine_path)
-          self.context = self.engine.create_execution_context()
-          self.input_shape = (1, 3, 640, 640)
-          self.device = device
-  ```
-- **Preprocessing:** Convert frame to model input format
-  - Resize to 640×640
-  - Normalize (mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
-  - Convert to NHWC/NCHW as required by TensorRT
-- **Inference:** Run batched detection on multiple frames
-  - Dynamic batching: accumulate frames until batch_size reached or timeout
-  - Execute TensorRT engine
-  - Parse outputs (bounding boxes + confidence scores)
-- **Post-processing:** NMS (Non-Maximum Suppression) to remove duplicate detections
-  - IoU threshold: 0.45
-  - Confidence threshold: 0.25 (low to avoid missing faces)
-- **Crop Generation:** Extract face crops from original frames
-  - Resize crops to 224×224 for classification input
-  - Save crops to `/scratch/{video_id}/crops/`
-
-### 4.3 Dynamic Batching
-**File:** `src/batcher.py`
-
-Features:
-- **Batch Accumulator:** Collect face detections across frames
-  - Max batch size: auto-tuned based on available VRAM (start at 16, adjust dynamically)
-  - Timeout: 100ms (process batch even if not full)
-- **VRAM Monitoring:** Track GPU memory usage and adjust batch size
-  - If VRAM > 16GB: reduce batch size by 25%
-  - If VRAM < 10GB: increase batch size by 25%
-- **Frame Ordering:** Maintain frame-to-batch mapping for correct crop assignment
-
-### 4.4 GPU Memory Management
-**File:** `src/gpu_manager.py`
-
-Features:
-- **Memory Tracking:** Monitor GPU memory usage continuously
-- **Peak Allocation:** Ensure total GPU memory < 18GB (leaving 6GB headroom)
-- **Cleanup:** Explicitly free TensorRT buffers after each batch
-- **Per-Video Limits:** Max faces per video (e.g., 100) to prevent memory issues
-- **Fallback:** If VRAM is critically low, reduce batch size and retry
-
-### 4.5 Configuration Updates
-**File:** `config.yaml` (updates)
-
-New fields:
-```yaml
-face_detection:
-  model: yolo8n
-  model_path: /models/face_detector/face_detector.trt
-  input_size: 640
-  confidence_threshold: 0.25
-  iou_threshold: 0.45
-  max_faces_per_frame: 10
-  max_faces_per_video: 100
-
-batching:
-  max_batch_size: 16
-  batch_timeout_ms: 100
-  vram_target_gb: 16
-  vram_reduce_threshold_gb: 16
-  vram_increase_threshold_gb: 10
-```
-
-## Acceptance Criteria
-
-### Functional
-- [ ] YOLOv8n model loads and runs inference via TensorRT FP32
-- [ ] Face detection correctly identifies faces in test frames (mAP > 0.50 on WIDER Face val)
-- [ ] No faces detected in frames without faces (false positive rate < 5%)
-- [ ] Face crops are correctly extracted and resized to 224×224
-- [ ] Dynamic batching works correctly (multiple frames processed in single batch)
-- [ ] Batch size adjusts based on available VRAM
-- [ ] Frames with no detected faces are handled gracefully (no crashes, no crops generated)
-- [ ] NMS correctly removes duplicate detections (IoU > 0.45)
-- [ ] Face bounding boxes are accurate (center coordinates within 10 pixels of ground truth)
-
-### Non-Functional
-- [ ] GPU memory usage stays ≤ 18GB per GPU during face detection
-- [ ] Face detection for 10 frames (batched) completes in < 2 seconds on Tesla P40
-- [ ] Batch processing throughput: ≥ 50 faces/second/GPU
-- [ ] Model loading time < 5 seconds at startup
-- [ ] No GPU OOM errors during extended processing
-
-### Technical Constraints
-- [ ] TensorRT engine is FP32 only (no FP16, no INT8)
-- [ ] No Tensor Cores used (CC 5.2 constraint)
-- [ ] CUDA 11.8 compatible (verified via torch.version.cuda)
-- [ ] ONNX model exports without errors (verified via onnx.checker)
-- [ ] Face crops are saved in JPEG format at high quality
-
-## Dependencies
-- **Prerequisites:** STORY-01 (Foundation), STORY-03 (Frame Sampling — provides frames)
-- **Depends on:** None (runs after frame sampling)
-- **Enables:** STORY-05 (Classification — provides face crops)
-
-## Risks & Mitigations
-| Risk | Mitigation |
-|------|--|
-| Tesla P40 (CC 5.2) has no Tensor Cores — slower inference | Use TensorRT engine caching; maximize batch size to compensate |
-| YOLOv8n may be too large for VRAM with classification | Use YOLOv8n (smallest variant); monitor VRAM closely |
-| Many faces per frame could cause VRAM spike | Limit max faces per frame; reduce batch size if needed |
-| TensorRT engine build time is long | Pre-build engines; cache on disk; skip rebuild if model unchanged |
-
-## Estimated Effort
-- **Sprint:** 5-6 (second half)
-- **Story Points:** 21
-- **Dependencies:** STORY-01, STORY-03

STORY-05.md

@@ -1,187 +0,0 @@
-# STORY-05: Classification & Confidence Aggregation
-
-## Epic
-**E1: Core Pipeline** — I can classify face crops and aggregate video confidence.
-
-## Related Requirements
-| ID | Requirement |
-|----|------|
-| FR-03 | Binary demographic classification of detected face crops: Outputs probability p ∈ [0,1] for target class |
-| FR-04 | Video-level confidence aggregation & threshold routing: Aggregates frame-level scores → video confidence C; Routes to MATCH, REVIEW, or SKIP |
-| NFR-01 | Throughput: ≥ 30 videos/hour/GPU |
-| NFR-03 | GPU Memory Safety: ≤ 18GB per GPU sustained |
-| NFR-04 | Determinism & Reproducibility: Config-seeded randomness, versioned models |
-| TC-01 | Hardware: 2× Tesla P40 24GB (compute capability 5.2, PCIe 3.0, no Tensor Cores) |
-| TC-02 | CUDA/Torch Compatibility: CUDA ≤ 11.8, PyTorch ≤ 2.1.0, FP32 inference only |
-
-## Description
-Implement demographic classification of detected face crops using a lightweight classifier (MobileNetV3 or equivalent). Apply temperature scaling for calibrated confidence scores. Aggregate frame-level confidence to video-level confidence using configurable aggregation logic. Route videos to MATCH, REVIEW, or SKIP based on threshold comparison.
-
-## Scope
-
-### In Scope
-- MobileNetV3 (or equivalent) classification model via TensorRT FP32
-- Temperature scaling for confidence calibration
-- Frame-level confidence computation (softmax with temperature)
-- Video-level confidence aggregation (max or weighted mean, configurable)
-- Threshold-based routing (T_high, T_low)
-- Routing decision assignment (MATCH, REVIEW, SKIP)
-- Confidence score persistence to DB and output files
-
-### Out of Scope
-- Frame sampling (covered in STORY-03)
-- Face detection (covered in STORY-04)
-- Results export format (covered in STORY-06)
-- Review UI (covered in STORY-07)
-- Active learning / model retraining (covered in STORY-08)
-
-## Deliverables
-
-### 5.1 Classification Model
-**File:** `models/classifier/`
-
-Components:
-- **Source Model:** MobileNetV3-small (or equivalent lightweight classifier)
-  - Trained on demographic dataset (fairness considerations documented)
-  - Binary classification: Black male subject (class 1) vs. not (class 0)
-- **ONNX Export:** `classifier.onnx`
-  - Input: `[1, 3, 224, 224]` (RGB, resized face crop)
-  - Output: `[1, 2]` (logits for class 0 and class 1)
-- **TensorRT Engine:** `classifier.trt` (FP32)
-  - Builder config: max_batch_size=32, max_workspace_size=2GB
-- **Calibration Data:** Temperature parameter T stored with model
-- **Model Metadata:** `model.json` with architecture, training dataset, validation metrics
-
-### 5.2 Classification Runner
-**File:** `src/classifier.py`
-
-Core components:
-- **Model Loader:** Load TensorRT engine at startup
-  ```python
-  class FaceClassifier:
-      def __init__(self, engine_path, temperature=1.0, device='cuda'):
-          self.engine = load_trt_engine(engine_path)
-          self.context = self.engine.create_execution_context()
-          self.temperature = temperature
-          self.input_shape = (1, 3, 224, 224)
-  ```
-- **Preprocessing:** Convert face crop to model input
-  - Resize to 224×224
-  - Normalize (ImageNet statistics)
-  - Convert to tensor (NCHW format)
-- **Inference:** Run classification on face crops
-  - Dynamic batching (shared with face detection batcher)
-  - Execute TensorRT engine
-  - Get raw logits output
-- **Temperature Scaling:** Apply temperature to logits before softmax
-  ```python
-  def calibrated_softmax(logits, temperature):
-      scaled_logits = logits / temperature
-      return softmax(scaled_logits, axis=-1)
-  ```
-- **Frame-level Confidence:** Extract probability for target class (class 1)
-  ```python
-  p_i = calibrated_probs[:, 1]  # probability of target class
-  ```
-
-### 5.3 Confidence Aggregation
-**File:** `src/aggregator.py`
-
-Features:
-- **Aggregation Strategies (configurable):**
-  1. **Max:** `C = max(p_i)` — use highest confidence frame
-  2. **Weighted Mean:** `C = softmax(α·mean(p_i) + β·var(p_i))`
-  3. **Top-K Mean:** `C = mean(top_k(p_i))` — average of top K confidences
-- **Default Strategy:** Max (most conservative, aligns with FR-04)
-- **Parameter Configuration:** α, β, K configurable in config.yaml
-- **Variance Calculation:** Compute variance of frame-level confidences (for uncertainty estimation)
-
-### 5.4 Threshold Routing
-**File:** `src/router.py`
-
-Features:
-- **Threshold Comparison:**
-  ```python
-  if C >= T_high:  # default 0.75
-      routing = 'MATCH'
-  elif C >= T_low:  # default 0.45
-      routing = 'REVIEW'
-  else:
-      routing = 'SKIP'
-  ```
-- **Configurable Thresholds:** T_high and T_low in config.yaml
-- **Routing Decision Logging:** Log routing decision with confidence score
-- **Edge Case Handling:**
-  - No faces detected → routing = SKIP (with confidence = 0.0)
-  - All frames have same confidence → routing based on threshold comparison
-  - Confidence exactly at threshold → use >= comparison (inclusive)
-
-### 5.5 Configuration Updates
-**File:** `config.yaml` (updates)
-
-New fields:
-```yaml
-classifier:
-  model: mobilenetv3-small
-  model_path: /models/classifier/classifier.trt
-  input_size: 224
-  temperature: 1.0  # calibration temperature
-  default_strategy: max  # max, weighted_mean, top_k_mean
-
-aggregation:
-  strategy: max
-  alpha: 1.0   # for weighted_mean
-  beta: 0.1    # for weighted_mean
-  top_k: 3     # for top_k_mean
-
-routing:
-  T_high: 0.75
-  T_low: 0.45
-  no_faces_decision: SKIP
-```
-
-## Acceptance Criteria
-
-### Functional
-- [ ] MobileNetV3 model loads and runs inference via TensorRT FP32
-- [ ] Classification outputs calibrated probability p ∈ [0,1] for target class
-- [ ] Temperature scaling is applied correctly (verified on test set)
-- [ ] Frame-level confidence scores are deterministic (same input → same output)
-- [ ] Max aggregation produces C = max(p_i) correctly
-- [ ] Weighted mean aggregation produces correct result with configurable α, β
-- [ ] Threshold routing assigns exactly one of: MATCH, REVIEW, SKIP
-- [ ] Videos with no detected faces are routed to SKIP with confidence 0.0
-- [ ] Confidence exactly at threshold uses >= comparison (inclusive)
-
-### Non-Functional
-- [ ] Classification for 16 face crops (batched) completes in < 1 second on Tesla P40
-- [ ] GPU memory usage stays ≤ 18GB per GPU during classification
-- [ ] Aggregation computation is negligible (< 10ms per video)
-- [ ] Temperature parameter is stored with model checkpoint
-- [ ] All routing decisions are logged with confidence score and threshold values
-
-### Technical Constraints
-- [ ] TensorRT engine is FP32 only
-- [ ] No Tensor Cores used (CC 5.2 constraint)
-- [ ] CUDA 11.8 compatible
-- [ ] ONNX model exports without errors
-- [ ] Confidence scores are reproducible (deterministic inference)
-- [ ] Aggregation strategy is configurable without code changes
-
-## Dependencies
-- **Prerequisites:** STORY-01 (Foundation), STORY-03 (Frame Sampling), STORY-04 (Face Detection — provides crops)
-- **Depends on:** None (runs after face detection)
-- **Enables:** STORY-06 (Results Persistence), STORY-07 (Review UI — provides routing decisions)
-
-## Risks & Mitigations
-| Risk | Mitigation |
-|------|--|
-| Temperature scaling parameters need tuning | Evaluate on held-out calibration set during training cycle |
-| Max aggregation may be overly conservative | Offer weighted mean as alternative; allow per-job strategy selection |
-| Model bias concerns with demographic classification | Document training dataset; audit fairness metrics; legal review required |
-| TensorRT engine build time | Pre-build and cache engines; version control model files |
-
-## Estimated Effort
-- **Sprint:** 5-6 (second half)
-- **Story Points:** 21
-- **Dependencies:** STORY-01, STORY-03, STORY-04

STORY-06.md

@@ -1,181 +0,0 @@
-# STORY-06: Results Persistence & Export
-
-## Epic
-**E1: Core Pipeline** — As a system, I can persist processing results and export data for analytics.
-
-## Related Requirements
-| ID | Requirement |
-|----|------|
-| FR-07 | Metadata logging & audit trail: Stores video ID, timestamps, frame counts, confidence scores, routing decision, model version |
-| FR-04 | Video-level confidence aggregation & threshold routing: Results saved to Parquet |
-| NFR-05 | Fault Tolerance: Metadata persisted before cleanup |
-| NFR-07 | Data Volume Handling: Efficient indexing for ~30TB dataset |
-| TC-03 | Storage I/O: Shared NAS/SMB for video input/output |
-
-## Description
-Implement result persistence, metadata logging, and data export. Update the database with processing results, insert processing log entries, and export summary data in Parquet/JSONL format for analytics. Ensure all persistence happens before scratch space cleanup to prevent data loss.
-
-## Scope
-
-### In Scope
-- Database update for processed videos (status, confidence, routing, timestamps)
-- Processing log entry insertion (atomic transaction)
-- Parquet/JSONL export to `/data/output`
-- Confidence score persistence (frame-level and video-level)
-- Model version tracking in DB and export
-- Idempotent result persistence (re-processing doesn't duplicate entries)
-- Transaction-safe updates (all-or-nothing)
-
-### Out of Scope
-- Frame sampling (covered in STORY-03)
-- Face detection (covered in STORY-04)
-- Classification (covered in STORY-05)
-- Review UI (covered in STORY-07)
-- Active learning pipeline (covered in STORY-08)
-
-## Deliverables
-
-### 6.1 Result Updater
-**File:** `src/result_updater.py`
-
-Features:
-- **Video Record Update:** Update `videos` table with processing results
-  ```sql
-  UPDATE videos SET
-      status = 'COMPLETED',
-      last_processed_time = NOW(),
-      confidence_score = %s,
-      routing_decision = %s,
-      model_version = %s,
-      frame_count = %s,
-      updated_at = NOW()
-  WHERE id = %s AND status = 'PROCESSING'
-  ```
-- **Atomic Transaction:** Wrap video update + log insert in single transaction
-- **State Guard:** Only update if status is PROCESSING (prevents duplicate processing)
-- **Error Handling:** Rollback transaction on any error; re-queue video as PENDING
-
-### 6.2 Processing Logger
-**File:** `src/processing_logger.py`
-
-Features:
-- **Log Entry Insertion:** Insert record into `processing_logs` table
-  ```python
-  def insert_processing_log(video_id, model_version, frame_count,
-                            confidence_score, routing_decision):
-      cursor.execute("""
-          INSERT INTO processing_logs
-              (video_id, model_version, frame_count, confidence_score,
-               routing_decision, processed_at)
-          VALUES (%s, %s, %s, %s, %s, NOW())
-      """, (video_id, model_version, frame_count,
-            confidence_score, routing_decision))
-  ```
-- **Frame-Level Confidence Storage:** Store all frame-level confidence scores
-  - Option A: JSON array in processing_logs.confidence_scores (TEXT column)
-  - Option B: Separate `frame_confidences` table (for large datasets)
-  - Default: JSON array in processing_logs (simpler, sufficient for typical frame counts)
-- **Audit Trail:** All log entries include timestamp, model version, and routing decision
-
-### 6.3 Data Export
-**File:** `src/data_export.py`
-
-Features:
-- **Parquet Export:** Write summary data to Parquet format
-  ```python
-  # Schema:
-  # video_id (int64), file_path (string), file_hash (string),
-  # model_version (string), sample_count (int32),
-  # confidence_scores (list<float>), video_confidence (float),
-  # routing (string), processed_at (timestamp)
-  ```
-- **JSONL Export:** Alternative line-delimited JSON format for streaming analytics
-  ```json
-  {"video_id": 12345, "file_path": "/data/input/video.mp4",
-   "model_version": "v1.2.0", "sample_count": 8,
-   "confidence_scores": [0.82, 0.79, 0.85, 0.76, 0.81, 0.83, 0.78, 0.80],
-   "video_confidence": 0.85, "routing": "MATCH", "processed_at": "2026-08-03T10:30:00Z"}
-  ```
-- **Output Directory:** `/data/output/{model_version}/` organized by model version
-- **File Naming:** `{model_version}_{batch_id}.parquet` and `{model_version}_{batch_id}.jsonl`
-- **Batch ID:** Sequential counter or timestamp-based (e.g., `v1.2.0_20260803_103000`)
-- **Compression:** Parquet with Snappy compression
-- **Export Trigger:** Export after all videos in batch are processed (or periodically)
-
-### 6.4 Scratch Cleanup
-**File:** `src/scratch_manager.py` (cleanup section)
-
-Features:
-- **Post-Processing Cleanup:** Delete all frames and crops after results are persisted
-- **Cleanup Order:** Results → Logs → Export → Cleanup (critical ordering)
-- **Cleanup Verification:** Verify DB update before deleting scratch files
-- **Cleanup on Error:** If persistence fails, do NOT delete scratch (allow retry)
-- **Cleanup Logging:** Log cleanup completion with video_id and files removed
-
-### 6.5 Configuration Updates
-**File:** `config.yaml` (updates)
-
-New fields:
-```yaml
-results:
-  persist_before_cleanup: true
-  transaction_safe: true
-  state_guard: true  # only update PROCESSING → COMPLETED
-
-export:
-  format: parquet  # parquet, jsonl, or both
-  output_path: /data/output
-  compression: snappy
-  batch_size: 100  # export after N videos
-  include_frame_confidences: true
-```
-
-## Acceptance Criteria
-
-### Functional
-- [ ] Video status transitions from PROCESSING to COMPLETED after successful processing
-- [ ] Processing log entry is inserted for every processed video
-- [ ] Frame-level confidence scores are stored (all p_i values)
-- [ ] Video-level confidence score is stored correctly
-- [ ] Routing decision is persisted accurately
-- [ ] Model version is recorded in both videos and processing_logs tables
-- [ ] Parquet export produces valid, readable files with correct schema
-- [ ] JSONL export produces valid JSON lines with correct fields
-- [ ] Export files are organized by model version in /data/output/
-- [ ] Scratch space is cleaned up after persistence (verified by directory check)
-- [ ] Re-processing a video updates existing records (no duplicates)
-- [ ] Transaction rollback on error prevents partial state updates
-
-### Non-Functional
-- [ ] Result persistence completes in < 5 seconds per video
-- [ ] Parquet export for 100 videos completes in < 10 seconds
-- [ ] JSONL export for 100 videos completes in < 10 seconds
-- [ ] Export files are valid (verified by parquet.read_table and jsonl parsing)
-- [ ] Scratch cleanup frees > 95% of allocated space per video
-- [ ] No data loss if worker crashes between processing and persistence (PROCESSING state preserved)
-
-### Technical Constraints
-- [ ] All DB updates use parameterized queries (no SQL injection)
-- [ ] Transaction isolation level is READ COMMITTED or higher
-- [ ] Parquet files use Snappy compression (not gzip, for speed)
-- [ ] JSONL files are UTF-8 encoded
-- [ ] Confidence scores are stored with float64 precision
-- [ ] Timestamps are in UTC (ISO 8601 format)
-
-## Dependencies
-- **Prerequisites:** STORY-01 (Foundation — DB schema), STORY-05 (Classification — provides results)
-- **Depends on:** None (runs after classification)
-- **Enables:** STORY-07 (Review UI — provides routing decisions), STORY-08 (Active Learning — provides training data)
-
-## Risks & Mitigations
-| Risk | Mitigation |
-|------|--|
-| NAS write latency slows export | Batch writes; use local tmpfs for export staging |
-| Parquet library compatibility with CUDA container | Test pyarrow in Worker container early; pin version |
-| Transaction rollback leaves video in PROCESSING state | DB lock timeout (5 min) prevents permanent lock |
-| Scratch cleanup before persistence causes data loss | Enforce ordering: persist → verify → cleanup |
-
-## Estimated Effort
-- **Sprint:** 5-6 (second half)
-- **Story Points:** 13
-- **Dependencies:** STORY-01, STORY-05

STORY-07.md

@@ -1,216 +0,0 @@
-# STORY-07: Review Interface
-
-## Epic
-**E2: Routing & Review** — As an annotator, I can view low-confidence videos and label them.
-
-## Related Requirements
-| ID | Requirement |
-|----|------|
-| FR-05 | Manual review interface for low-confidence videos: Displays video + contributing frames/crops + model confidence; Supports binary labeling |
-| FR-07 | Metadata logging & audit trail: Stores annotated labels for active learning |
-| NFR-06 | Observability: Tracks review queue depth |
-| TC-06 | Network Security: Internal LAN only; no auth required |
-
-## Description
-Implement a lightweight web-based review interface for annotating low-confidence videos. Display the video player, top-k contributing frames, model confidence scores, and allow annotators to toggle the ground truth label. Support CSV/JSON export of annotated data for active learning.
-
-## Scope
-
-### In Scope
-- Lightweight web UI (Flask/FastAPI) serving on internal LAN
-- Query DB for videos with routing_decision = REVIEW
-- Video player with playback controls
-- Display top-k contributing frames (highest confidence frames)
-- Display model confidence scores per frame
-- Binary label toggle (True/False — target class present or not)
-- Label persistence to DB (review_queue table)
-- CSV/JSON export of annotated data with ground truth
-- Accessible via internal IP:Port (no auth, no SSL)
-
-### Out of Scope
-- Frame sampling (covered in STORY-03)
-- Face detection (covered in STORY-04)
-- Classification (covered in STORY-05)
-- Confidence aggregation (covered in STORY-05)
-- Active learning pipeline / model retraining (covered in STORY-08)
-- Monitoring dashboards (covered in STORY-09)
-
-## Deliverables
-
-### 7.1 Review Backend
-**File:** `src/review_api.py`
-
-API Endpoints:
-- `GET /api/review/queue` — List videos in review queue
-  - Query: `SELECT * FROM review_queue WHERE annotated = false ORDER BY created_at DESC`
-  - Pagination: 20 items per page
-  - Response: `{videos: [...], total: N, page: P, per_page: 20}`
-- `GET /api/review/{video_id}` — Get video details for annotation
-  - Response: `{video_id, file_path, confidence_score, routing_decision, model_version, frame_count, contributing_frames: [{timestamp, crop_path, confidence}], video_duration}`
-- `POST /api/review/{video_id}/label` — Submit annotation
-  - Body: `{ground_truth: true/false, notes: string (optional)}`
-  - Updates: `review_queue.annotated = true`, `review_queue.ground_truth = value`, `review_queue.annotated_at = NOW()`
-  - Response: `{status: 'annotated', video_id, ground_truth}`
-- `GET /api/review/export` — Export annotated data
-  - Query params: `format=csv|json`, `annotated=true/false`, `date_from`, `date_to`
-  - Response: File download with annotated data
-- `GET /api/review/stats` — Review queue statistics
-  - Response: `{total_in_queue: N, annotated_today: N, avg_confidence: F, confidence_distribution: {...}}`
-
-### 7.2 Review Frontend
-**File:** `ui/review/`
-
-Pages:
-- **Queue Page (`/`):** List of videos awaiting review
-  - Table columns: Video ID, File Path, Confidence Score, Model Version, Date Added, Actions (View)
-  - Sortable by confidence, date, file path
-  - Filter by confidence range, model version
-  - Pagination (20 items per page)
-- **Annotation Page (`/review/{video_id}`):** Video annotation interface
-  - Video player with playback controls (HTML5 `<video>` element)
-  - Top-k contributing frames displayed as thumbnails (k=5 default)
-  - Confidence scores displayed per frame
-  - Label toggle button (True/False) with confirmation
-  - Optional notes field
-  - Submit button (saves to DB via API)
-  - Navigation: Previous/Next video in queue
-
-### 7.3 Data Export
-**File:** `src/review_export.py`
-
-Features:
-- **CSV Export:**
-  ```csv
-  video_id,file_path,confidence_score,routing_decision,model_version,ground_truth,annotated_at,contributing_frames
-  12345,/data/input/video.mp4,0.62,REVIEW,v1.2.0,true,2026-08-03T10:30:00Z,"[{'timestamp': 30.0, 'crop_path': '/scratch/12345/crops/30000.jpg', 'confidence': 0.82}, ...]"
-  ```
-- **JSON Export:**
-  ```json
-  [
-    {
-      "video_id": 12345,
-      "file_path": "/data/input/video.mp4",
-      "confidence_score": 0.62,
-      "routing_decision": "REVIEW",
-      "model_version": "v1.2.0",
-      "ground_truth": true,
-      "annotated_at": "2026-08-03T10:30:00Z",
-      "contributing_frames": [
-        {"timestamp": 30.0, "crop_path": "/scratch/12345/crops/30000.jpg", "confidence": 0.82}
-      ]
-    }
-  ]
-  ```
-- **Export Options:**
-  - Filter by annotation status (annotated/unannotated)
-  - Filter by date range
-  - Filter by model version
-  - Filter by ground truth label
-- **Output Location:** `/data/output/reviews/`
-
-### 7.4 UI Container
-**File:** `ui/Dockerfile`
-
-Base image: `python:3.10-slim`
-
-Installed packages:
-- Flask 3.0+ or FastAPI 0.100+ (lightweight web framework)
-- Jinja2 3.1+ (template engine)
-- PyMySQL (database connection for API)
-- gunicorn (WSGI server)
-
-### 7.5 Docker Compose Update
-**File:** `docker-compose.yml` (update)
-
-Add UI service:
-```yaml
-ui:
-  build:
-    context: ./ui
-    dockerfile: Dockerfile
-  ports:
-    - "5000:5000"
-  volumes:
-    - ./ui:/app/ui
-  environment:
-    - DB_HOST=mariadb
-    - DB_PORT=3306
-    - DB_NAME=videodetect
-    - DB_USER=videodetect
-    - DB_PASSWORD=${DB_PASSWORD}
-  depends_on:
-    - mariadb
-  networks:
-    - videodetect-network
-```
-
-### 7.6 Database Update
-**File:** `db/schema.sql` (review_queue table — from STORY-01)
-
-The `review_queue` table was defined in STORY-01. This story populates and queries it.
-
-### 7.7 Configuration Updates
-**File:** `config.yaml` (updates)
-
-New fields:
-```yaml
-review_ui:
-  host: "0.0.0.0"
-  port: 5000
-  per_page: 20
-  top_k_frames: 5
-  export_path: /data/output/reviews
-  auth_enabled: false  # per TC-06
-  ssl_enabled: false   # per TC-06
-```
-
-## Acceptance Criteria
-
-### Functional
-- [ ] Review queue displays all videos with routing_decision = REVIEW and annotated = false
-- [ ] Video player loads and plays the video correctly
-- [ ] Top-k contributing frames are displayed as thumbnails with confidence scores
-- [ ] Annotator can toggle label (True/False) and submit
-- [ ] Submitted label is persisted to DB (review_queue table)
-- [ ] Annotated videos are removed from the default queue view
-- [ ] CSV export produces valid CSV with all required fields
-- [ ] JSON export produces valid JSON with all required fields
-- [ ] Export includes ground truth labels and contributing frame data
-- [ ] UI is accessible via http://<server-ip>:5000 (no auth, no SSL)
-- [ ] Pagination works correctly (20 items per page)
-- [ ] Sort and filter operations work on the queue page
-
-### Non-Functional
-- [ ] Queue page loads in < 2 seconds (with 1000+ videos in queue)
-- [ ] Video player loads in < 3 seconds
-- [ ] Label submission completes in < 1 second
-- [ ] Export of 1000 annotated videos completes in < 10 seconds
-- [ ] UI uses < 100MB RAM at idle
-- [ ] No authentication or SSL configured (per TC-06)
-
-### Technical Constraints
-- [ ] UI runs in Docker container (per TC-05)
-- [ ] No reverse proxy configured
-- [ ] No SSL certificates configured
-- [ ] No authentication mechanism configured
-- [ ] All API responses are JSON
-- [ ] Database queries use parameterized statements
-- [ ] Export files are UTF-8 encoded
-
-## Dependencies
-- **Prerequisites:** STORY-01 (Foundation — DB schema), STORY-05 (Classification — provides routing decisions)
-- **Depends on:** None (can be built in parallel with STORY-04, STORY-05)
-- **Enables:** STORY-08 (Active Learning — provides labeled training data)
-
-## Risks & Mitigations
-| Risk | Mitigation |
-|------|--|
-| Video playback in browser requires compatible format | Serve videos in web-compatible format (H.264 MP4); transcode if needed |
-| Crop paths may not be accessible from UI container | Store crop paths in DB; serve via API endpoint |
-| No auth means anyone on LAN can access | Acceptable per TC-06; document in security notes |
-| Large review queue slows page loads | Implement server-side pagination; lazy load thumbnails |
-
-## Estimated Effort
-- **Sprint:** 7
-- **Story Points:** 21
-- **Dependencies:** STORY-01, STORY-05

STORY-08.md

@@ -1,245 +0,0 @@
-# STORY-08: Active Learning Pipeline
-
-## Epic
-**E3: Active Learning** — As an ML engineer, I can fine-tune the classifier head with reviewed data.
-
-## Related Requirements
-| ID | Requirement |
-|----|------|
-| FR-06 | Active learning pipeline (label ingestion → fine-tuning → deployment): Batch retraining only. No online learning. Versioned model swaps. |
-| NFR-04 | Determinism & Reproducibility: Config-seeded randomness, versioned models |
-| NFR-05 | Fault Tolerance: Auto-retry on transient failures; skip & log on fatal errors |
-| TC-01 | Hardware: 2× Tesla P40 24GB (compute capability 5.2, PCIe 3.0, no Tensor Cores) |
-| TC-02 | CUDA/Torch Compatibility: CUDA ≤ 11.8, PyTorch ≤ 2.1.0, FP32 inference only |
-
-## Description
-Implement the active learning pipeline that ingests annotated review data, fine-tunes the classifier head, validates the candidate model, and deploys it if it meets quality gates. This enables incremental model improvement using human-labeled data from the review queue.
-
-## Scope
-
-### In Scope
-- Label ingestion from review queue (CSV/JSON export or direct DB query)
-- Dataset versioning and preparation for training
-- Head-only fine-tuning on face crops (freeze backbone, train classification head)
-- Training with AdamW, configurable learning rate, early stopping
-- Candidate model validation on held-out validation set
-- F1 score and ECE (Expected Calibration Error) calculation
-- Quality gate validation (ΔF1 > 0.02, ECE < 0.08)
-- Model registry updates (ACTIVE, CANDIDATE, ARCHIVED states)
-- Hot reload of new TensorRT engine
-- Rollback on regression
-
-### Out of Scope
-- Frame sampling (covered in STORY-03)
-- Face detection (covered in STORY-04)
-- Classification inference (covered in STORY-05)
-- Review UI (covered in STORY-07)
-- Monitoring dashboards (covered in STORY-09)
-
-## Deliverables
-
-### 8.1 Label Ingestion
-**File:** `src/active_learning/label_ingestor.py`
-
-Features:
-- **Data Source:** Query `review_queue` table for annotated labels
-  ```sql
-  SELECT video_id, ground_truth, contributing_frames
-  FROM review_queue
-  WHERE annotated = true AND ground_truth IS NOT NULL
-  ```
-- **Crop Extraction:** Extract face crops from stored paths or re-extract from video
-- **Dataset Versioning:** Create versioned dataset directory structure
-  ```
-  /data/training/v2.0.0/
-      crops/
-          class_0/  (negative samples)
-          class_1/  (positive samples)
-      labels.csv
-      metadata.json
-  ```
-- **Train/Val Split:** 80/20 split (stratified by class)
-- **Augmentation:** Apply standard augmentations at training time (not pre-computed)
-  - Random horizontal flip
-  - Random color jitter
-  - Random affine transform (±10 degrees rotation, ±10% scale)
-
-### 8.2 Training Pipeline
-**File:** `src/active_learning/trainer.py`
-
-Features:
-- **Model Loading:** Load pre-trained MobileNetV3 backbone (frozen)
-  ```python
-  model = load_pretrained_mobilenetv3()
-  for param in model.features.parameters():
-      param.requires_grad = False
-  # Replace classification head
-  model.classifier = nn.Sequential(
-      nn.Linear(1280, 256),
-      nn.ReLU(),
-      nn.Dropout(0.3),
-      nn.Linear(256, 2)  # binary classification
-  )
-  ```
-- **Head-Only Fine-Tuning:** Only train the classification head
-  - Backbone weights are frozen (no gradient updates)
-  - Head weights are trainable
-  - This is faster and requires less data than full fine-tuning
-- **Optimizer:** AdamW with configurable parameters
-  - Learning rate: 1e-3 (default), configurable
-  - Weight decay: 1e-2
-  - Betas: (0.9, 0.999)
-- **Loss Function:** Binary Cross-Entropy with class weights (if imbalanced)
-- **Training Configuration:**
-  - Epochs: 10-30 (configurable)
-  - Batch size: 32
-  - Early stopping: patience=5 epochs (no validation improvement)
-  - Learning rate scheduler: ReduceLROnPlateau (factor=0.5, patience=3)
-- **Checkpointing:** Save checkpoint to `/models/candidate/` after each epoch
-  ```
-  /models/candidate/
-      v2.0.0_epoch_01.pt
-      v2.0.0_epoch_02.pt
-      ...
-      v2.0.0_best.pt  (best validation F1)
-  ```
-
-### 8.3 Validation Pipeline
-**File:** `src/active_learning/validator.py`
-
-Features:
-- **Validation Set:** Held-out 20% of annotated data (never seen during training)
-- **Metrics Calculation:**
-  - **F1 Score:** Macro F1 on validation set
-  - **ECE (Expected Calibration Error):**
-    ```python
-    def compute_ece(predictions, labels, n_bins=15):
-        bin_boundaries = np.linspace(0, 1, n_bins + 1)
-        ece = 0.0
-        for i in range(n_bins):
-            mask = (predictions >= bin_boundaries[i]) & (predictions < bin_boundaries[i+1])
-            if mask.sum() > 0:
-                bin_confidence = predictions[mask].mean()
-                bin_accuracy = labels[mask].mean()
-                ece += (mask.sum() / len(predictions)) * abs(bin_confidence - bin_accuracy)
-        return ece
-    ```
-  - **Accuracy, Precision, Recall:** Standard classification metrics
-- **Quality Gates:**
-  - ΔF1 > 0.02 (improvement over current model)
-  - ECE < 0.08 (calibration acceptable)
-  - Both gates must pass for deployment
-
-### 8.4 Model Registry & Deployment
-**File:** `src/active_learning/registry.py`
-
-Features:
-- **Model Registry:** Update DB `models` table
-  ```sql
-  -- Promote candidate to active
-  UPDATE models SET status = 'ACTIVE' WHERE version = 'v2.0.0';
-  UPDATE models SET status = 'ARCHIVED' WHERE status = 'ACTIVE' AND version != 'v2.0.0';
-  ```
-- **TensorRT Engine Build:** Convert candidate model to TensorRT engine
-  ```bash
-  trtexec --onnx=models/candidate/v2.0.0.onnx \
-          --saveEngine=models/candidate/v2.0.0.trt \
-          --fp32 --maxBatch=32 --workspace=2048
-  ```
-- **Hot Reload:** Signal worker process to reload new engine
-  - Option A: Restart worker container (`docker-compose restart worker`)
-  - Option B: In-process reload (graceful, no downtime)
-  - Default: In-process reload via signal handling
-- **Rollback:** If new model causes issues, rollback to archived model
-  ```sql
-  UPDATE models SET status = 'ACTIVE' WHERE version = 'v1.2.0';
-  UPDATE models SET status = 'ARCHIVED' WHERE version = 'v2.0.0';
-  ```
-
-### 8.5 Training Configuration
-**File:** `config.yaml` (updates)
-
-New fields:
-```yaml
-active_learning:
-  enabled: true
-  min_annotated_samples: 100  # minimum labeled data to trigger training
-  training:
-    epochs: 20
-    batch_size: 32
-    learning_rate: 1e-3
-    weight_decay: 1e-2
-    early_stopping_patience: 5
-    lr_scheduler: ReduceLROnPlateau
-    lr_factor: 0.5
-    lr_patience: 3
-  validation:
-    val_split: 0.2
-    min_f1_improvement: 0.02
-    max_ece: 0.08
-  deployment:
-    auto_deploy: true  # deploy if quality gates pass
-    hot_reload: true
-    rollback_enabled: true
-  augmentation:
-    horizontal_flip: true
-    color_jitter: true
-    affine: true
-    affine_degrees: 10
-    affine_scale: 0.1
-```
-
-## Acceptance Criteria
-
-### Functional
-- [ ] Annotated labels are ingested from review_queue table correctly
-- [ ] Dataset is versioned with proper directory structure
-- [ ] Train/val split is stratified by class (80/20)
-- [ ] Data augmentation is applied at training time (not pre-computed)
-- [ ] Backbone weights are frozen during fine-tuning (verified by checking requires_grad)
-- [ ] Classification head is trainable and receives gradient updates
-- [ ] AdamW optimizer is used with correct parameters
-- [ ] Early stopping works (training stops if no improvement for patience epochs)
-- [ ] Candidate model is saved to /models/candidate/ with correct versioning
-- [ ] F1 score is calculated correctly on validation set
-- [ ] ECE is calculated correctly on validation set
-- [ ] Quality gates are enforced (ΔF1 > 0.02 AND ECE < 0.08)
-- [ ] Model registry is updated (ACTIVE, CANDIDATE, ARCHIVED states)
-- [ ] TensorRT engine is built from candidate model
-- [ ] Hot reload deploys new model without downtime
-- [ ] Rollback restores previous model if deployment fails
-
-### Non-Functional
-- [ ] Training completes in < 4 hours on Tesla P40 (typical dataset: 1000 samples)
-- [ ] Validation completes in < 30 minutes
-- [ ] TensorRT engine build completes in < 10 minutes
-- [ ] Hot reload completes in < 30 seconds
-- [ ] Training uses < 18GB GPU memory
-- [ ] Training is deterministic (same data + same config → same results)
-
-### Technical Constraints
-- [ ] CUDA 11.8 compatible (verified via torch.version.cuda)
-- [ ] PyTorch ≤ 2.1.0 (verified via torch.__version__)
-- [ ] FP32 training only (no mixed precision)
-- [ ] No Tensor Cores used (CC 5.2 constraint)
-- [ ] Model versioning follows semantic versioning (MAJOR.MINOR.PATCH)
-- [ ] All training hyperparameters are configurable via config.yaml
-
-## Dependencies
-- **Prerequisites:** STORY-01 (Foundation), STORY-05 (Classification — provides base model), STORY-07 (Review UI — provides labeled data)
-- **Depends on:** None (runs independently, triggered by labeled data threshold)
-- **Enables:** STORY-05 (provides new model for inference)
-
-## Risks & Mitigations
-| Risk | Mitigation |
-|------|--|
-| Insufficient labeled data for meaningful fine-tuning | Set min_annotated_samples threshold (e.g., 100); wait until reached |
-| Head-only fine-tuning may not be enough for domain shift | Offer full fine-tuning as option; document limitations |
-| Model regression in production | Strict quality gates; keep previous model in ARCHIVED state for quick rollback |
-| Long training times on Tesla P40 | Head-only training is faster; batch size tuning; early stopping |
-| ECE calculation sensitivity | Use proper binning; report ECI with confidence intervals |
-
-## Estimated Effort
-- **Sprint:** 7-8
-- **Story Points:** 34
-- **Dependencies:** STORY-01, STORY-05, STORY-07

STORY-09.md

@@ -1,360 +0,0 @@
-# STORY-09: Observability, Monitoring & Hardening
-
-## Epic
-**E4: Operations** — As a DevOps engineer, I can schedule, monitor, and resume batch jobs.
-
-## Related Requirements
-| ID | Requirement |
-|----|------|
-| NFR-05 | Fault Tolerance: Auto-retry on transient failures; skip & log on fatal errors |
-| NFR-06 | Observability: Prometheus/Grafana metrics + structured logging; Tracks FPS, queue depth, confidence distribution, drift alerts |
-| NFR-01 | Throughput: ≥ 30 videos/hour/GPU |
-| NFR-03 | GPU Memory Safety: ≤ 18GB per GPU sustained |
-| NFR-04 | Determinism & Reproducibility: Config-seeded randomness, versioned models |
-| NFR-07 | Data Volume Handling: Efficient indexing for ~30TB dataset |
-
-## Description
-Implement comprehensive monitoring, metrics collection, fault tolerance, and system hardening. Expose Prometheus metrics for throughput, GPU utilization, queue depth, and confidence distribution. Build Grafana dashboards for real-time system health. Implement crash recovery, idempotency, retry logic, and drift detection.
-
-## Scope
-
-### In Scope
-- Prometheus metrics exposure (videos processed, GPU utilization, queue depth, confidence distribution)
-- Grafana dashboard configuration (throughput, error rates, confidence drift)
-- Crash recovery with checkpointing
-- Idempotent processing guarantees
-- Retry logic for transient errors (up to 3 attempts)
-- Weekly drift detection job
-- Alerting on confidence distribution shifts and review queue growth
-- Worker health checks and auto-restart
-
-### Out of Scope
-- Frame sampling (covered in STORY-04)
-- Face detection (covered in STORY-05)
-- Classification (covered in STORY-06)
-- Review UI (covered in STORY-08)
-- Active learning pipeline (covered in STORY-09)
-
-## Deliverables
-
-### 9.1 Prometheus Metrics
-**File:** `src/metrics.py`
-
-Metrics to expose:
-- **Counter Metrics:**
-  - `videos_processed_total` (label: routing_decision=MATCH|REVIEW|SKIP|UNSCANNABLE|ERROR)
-  - `videos_processed_by_model_total` (label: model_version)
-  - `frames_extracted_total`
-  - `faces_detected_total`
-  - `inference_errors_total` (label: error_type)
-  - `retry_attempts_total` (label: step=probe|extract|detect|classify)
-- **Gauge Metrics:**
-  - `gpu_utilization_percent` (label: gpu_id)
-  - `gpu_memory_used_bytes` (label: gpu_id)
-  - `gpu_memory_free_bytes` (label: gpu_id)
-  - `queue_depth_pending`
-  - `queue_depth_processing`
-  - `queue_depth_review`
-  - `scratch_usage_bytes`
-  - `scratch_usage_percent`
-- **Histogram Metrics:**
-  - `video_processing_duration_seconds` (label: routing_decision)
-  - `confidence_score_distribution` (buckets: 0.0, 0.1, 0.2, ..., 0.9, 1.0)
-  - `frame_count_per_video`
-  - `face_count_per_video`
-- **Summary Metrics:**
-  - `throughput_videos_per_hour` (calculated from counter)
-  - `average_confidence` (calculated from histogram)
-
-Implementation:
-```python
-from prometheus_client import start_http_server, Counter, Gauge, Histogram
-
-# Start metrics HTTP server
-start_http_server(9090)  # /metrics endpoint
-
-# Define metrics
-videos_processed = Counter(
-    'videos_processed_total',
-    'Total videos processed',
-    ['routing_decision']
-)
-
-gpu_memory = Gauge(
-    'gpu_memory_used_bytes',
-    'GPU memory usage',
-    ['gpu_id']
-)
-
-confidence_hist = Histogram(
-    'confidence_score_distribution',
-    'Video confidence scores',
-    buckets=[0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]
-)
-```
-
-### 9.2 Grafana Dashboards
-**File:** `monitoring/grafana/dashboards/`
-
-Dashboard 1: **System Overview**
-- Panels:
-  - Throughput (videos/hour) — line chart, 1h window
-  - Queue depth (pending/processing/review) — stacked bar
-  - GPU utilization (both GPUs) — line chart
-  - GPU memory usage (both GPUs) — line chart
-  - Error rate (errors/100 videos) — bar chart
-  - Confidence distribution — histogram
-
-Dashboard 2: **Processing Details**
-- Panels:
-  - Processing duration per video — scatter plot
-  - Frame count per video — histogram
-  - Face count per video — histogram
-  - Confidence score by model version — box plot
-  - Routing decision distribution — pie chart
-
-Dashboard 3: **Active Learning**
-- Panels:
-  - Review queue size over time — line chart
-  - Annotation rate (labels/day) — bar chart
-  - Model version timeline — timeline panel
-  - F1 score by model version — line chart
-  - ECE by model version — line chart
-
-**File:** `monitoring/grafana/dashboards/system_overview.json`
-**File:** `monitoring/grafana/dashboards/processing_details.json`
-**File:** `monitoring/grafana/dashboards/active_learning.json`
-
-### 9.3 Crash Recovery
-**File:** `src/crash_recovery.py`
-
-Features:
-- **Checkpointing:** Periodically save processing state
-  ```python
-  def save_checkpoint(video_id, state, progress):
-      checkpoint = {
-          'video_id': video_id,
-          'state': state,  # 'processing', 'extracting', 'detecting', 'classifying'
-          'progress': progress,  # dict of step -> completed
-          'timestamp': datetime.utcnow().isoformat()
-      }
-      with open(f'/scratch/checkpoints/{video_id}.json', 'w') as f:
-          json.dump(checkpoint, f)
-  ```
-- **Recovery on Startup:** Scan for PROCESSING videos and re-queue them
-  ```sql
-  UPDATE videos SET status = 'PENDING', updated_at = NOW()
-  WHERE status = 'PROCESSING' AND updated_at < NOW() - INTERVAL 5 MINUTE
-  ```
-- **Lock Timeout:** 5 minutes (videos stuck in PROCESSING beyond this are re-queued)
-- **Re-queue Logic:** Only re-queue if worker is down (detected via health check)
-- **Idempotency:** Re-processing a video produces same results (no duplicates)
-
-### 9.4 Retry Logic
-**File:** `src/retry.py`
-
-Features:
-- **Retry Decorator:**
-  ```python
-  def retry(max_attempts=3, delay=1.0, backoff=2.0, exceptions=(Exception,)):
-      def decorator(func):
-          def wrapper(*args, **kwargs):
-              last_exception = None
-              for attempt in range(max_attempts):
-                  try:
-                      return func(*args, **kwargs)
-                  except exceptions as e:
-                      last_exception = e
-                      if attempt < max_attempts - 1:
-                          wait = delay * (backoff ** attempt)
-                          time.sleep(wait)
-              raise last_exception
-          return wrapper
-      return decorator
-  ```
-- **Applicable Steps:** ffprobe, frame extraction, face detection, classification, DB writes
-- **Transient Errors:** Network timeout, GPU OOM, file lock contention
-- **Non-Retryable Errors:** Codec unsupported, corrupt file, invalid path → skip and log
-- **Retry Logging:** Log each retry attempt with error type and delay
-
-### 9.5 Drift Detection
-**File:** `src/drift_detector.py`
-
-Features:
-- **Weekly Job:** Compare current confidence distribution to baseline
-  ```python
-  def detect_drift(current_confidences, baseline_confidences, threshold=0.10):
-      # Compare p(C > 0.5) shift
-      current_high = sum(1 for c in current_confidences if c > 0.5) / len(current_confidences)
-      baseline_high = sum(1 for c in baseline_confidences if c > 0.5) / len(baseline_confidences)
-      shift = abs(current_high - baseline_high)
-      
-      if shift > threshold:
-          alert(f"Confidence drift detected: {shift:.2%} shift in p(C > 0.5)")
-          return True
-      return False
-  ```
-- **Baseline:** Stored in DB or config (computed from last training cycle)
-- **Alert Conditions:**
-  - `p(C > 0.5)` shifts > 10% from baseline
-  - Review queue grows unbounded (> 1000 items for > 24 hours)
-  - Throughput drops below 20 videos/hour/GPU for > 1 hour
-  - Error rate exceeds 5% for any 1-hour window
-- **Alert Channels:** Email, Slack webhook, or log entry (configurable)
-
-### 9.6 Worker Health Checks
-**File:** `src/health_check.py`
-
-Features:
-- **Health Endpoint:** `/health` returns worker status
-  ```json
-  {
-      "status": "healthy",
-      "gpu_available": true,
-      "gpu_memory_used_gb": 12.5,
-      "queue_depth": 42,
-      "uptime_seconds": 86400,
-      "videos_processed_today": 156,
-      "last_error": null
-  }
-  ```
-- **Auto-Restart:** Docker restart policy for worker container
-  ```yaml
-  restart: unless-stopped
-  healthcheck:
-    test: ["CMD", "curl", "-f", "http://localhost:9090/health"]
-    interval: 30s
-    timeout: 10s
-    retries: 3
-  ```
-- **GPU Health:** Monitor GPU temperature and error counts
-- **Disk Health:** Monitor scratch space usage and NAS connectivity
-
-### 9.7 Docker Compose Update
-**File:** `docker-compose.yml` (update)
-
-Add monitoring services:
-```yaml
-prometheus:
-  image: prom/prometheus:v2.48.0
-  volumes:
-    - ./monitoring/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml
-    - prometheus_data:/prometheus
-  ports:
-    - "9090:9090"
-  networks:
-    - videodetect-network
-
-grafana:
-  image: grafana/grafana:10.2.0
-  volumes:
-    - grafana_data:/var/lib/grafana
-    - ./monitoring/grafana/dashboards:/etc/grafana/provisioning/dashboards
-  environment:
-    - GF_SECURITY_ADMIN_PASSWORD=${GRAFANA_PASSWORD}
-  ports:
-    - "3000:3000"
-  networks:
-    - videodetect-network
-
-volumes:
-  prometheus_data:
-  grafana_data:
-```
-
-### 9.8 Configuration Updates
-**File:** `config.yaml` (updates)
-
-New fields:
-```yaml
-monitoring:
-  prometheus:
-    enabled: true
-    port: 9090
-    metrics_path: /metrics
-  grafana:
-    enabled: true
-    port: 3000
-  alerts:
-    confidence_drift_threshold: 0.10
-    review_queue_max_size: 1000
-    review_queue_max_age_hours: 24
-    throughput_min_videos_per_hour: 20
-    throughput_min_duration_hours: 1
-    error_rate_threshold: 0.05
-    error_rate_window_hours: 1
-  drift_detection:
-    enabled: true
-    schedule: weekly  # cron: 0 2 * * 0 (Sundays at 2 AM)
-    baseline_source: db  # db or config
-  crash_recovery:
-    lock_timeout_minutes: 5
-    auto_requeue: true
-  retry:
-    max_attempts: 3
-    initial_delay: 1.0
-    backoff_factor: 2.0
-    retryable_errors:
-      - timeout
-      - gpu_oom
-      - file_lock
-    non_retryable_errors:
-      - codec_unsupported
-      - file_corrupt
-      - invalid_path
-```
-
-## Acceptance Criteria
-
-### Functional
-- [ ] Prometheus metrics are exposed at `/metrics` endpoint and queryable
-- [ ] All required metrics are present (counters, gauges, histograms, summaries)
-- [ ] Grafana dashboards load and display correct data
-- [ ] System Overview dashboard shows throughput, queue depth, GPU metrics
-- [ ] Processing Details dashboard shows duration, frame count, confidence distribution
-- [ ] Active Learning dashboard shows review queue, annotation rate, model versions
-- [ ] Crash recovery re-queues stuck PROCESSING videos on worker restart
-- [ ] Idempotent processing: re-processing produces same results (no duplicates)
-- [ ] Retry logic retries transient errors up to 3 times with exponential backoff
-- [ ] Non-retryable errors are skipped and logged (no infinite retry)
-- [ ] Drift detection runs weekly and alerts on > 10% confidence shift
-- [ ] Review queue growth alert triggers when queue > 1000 for > 24 hours
-- [ ] Worker health check returns correct status
-- [ ] Auto-restart triggers on health check failure
-
-### Non-Functional
-- [ ] Metrics collection overhead < 2% of CPU
-- [ ] Grafana dashboard loads in < 3 seconds
-- [ ] Crash recovery completes in < 30 seconds
-- [ ] Drift detection completes in < 5 minutes
-- [ ] Alert delivery completes in < 60 seconds
-- [ ] Prometheus data retention: 30 days (configurable)
-- [ ] Grafana data source refresh: 30 seconds
-
-### Technical Constraints
-- [ ] Prometheus metrics follow naming conventions (unit suffixes, proper labels)
-- [ ] Grafana dashboards are JSON-exportable and version-controlled
-- [ ] Crash recovery is idempotent (running twice produces same result)
-- [ ] Retry logic does not retry non-transient errors
-- [ ] Drift detection baseline is stored and versioned
-- [ ] Health check endpoint responds in < 1 second
-- [ ] All alerts are logged with timestamp and context
-
-## Dependencies
-- **Prerequisites:** STORY-01 (Foundation), STORY-05 (Classification — provides metrics data), STORY-08 (Active Learning — provides model metrics)
-- **Depends on:** None (can be built in parallel with other stories)
-- **Enables:** Production deployment and long-term operation
-
-## Risks & Mitigations
-| Risk | Mitigation |
-|------|--|
-| Prometheus metrics cardinality explosion | Limit label cardinality; use histograms instead of individual values |
-| Grafana dashboard load time with large datasets | Use Prometheus aggregations; pre-compute panels |
-| Crash recovery misses in-flight writes | Use DB transactions; lock timeout prevents permanent locks |
-| Drift detection baseline becomes stale | Update baseline with each training cycle |
-| Alert fatigue from too many alerts | Tune thresholds; implement alert grouping |
-
-## Estimated Effort
-- **Sprint:** 9+
-- **Story Points:** 34
-- **Dependencies:** STORY-01, STORY-05, STORY-08

aiworker/.env

@@ -0,0 +1,5 @@
+# VideoDetect — Docker Compose environment variables
+# Copy this file to .env and fill in real values before running docker-compose up.
+
+API_HOST=https://10.0.1.34:3000
+

aiworker/Dockerfile

@@ -0,0 +1,29 @@
+FROM ubuntu:22.04
+
+# Avoid interactive prompts during build
+ENV DEBIAN_FRONTEND=noninteractive
+ENV TZ=Etc/UTC
+
+# System dependencies
+# Install Perl, system deps, and ffmpeg (includes ffprobe)
+RUN apt update && apt install -y \
+    libdbd-mysql-perl \
+    libhttp-lite-perl \
+    libjson-perl \
+    libyaml-perl \
+    perl \
+    libdbi-perl \
+    build-essential \
+    ffmpeg \
+    libstring-shellquote-perl \
+    libwww-perl \
+    liblwp-protocol-https-perl
+
+
+RUN mkdir -p /app
+ADD run.pl /app/run.pl
+RUN chmod +x /app/run.pl
+WORKDIR /app
+
+# Default command: process tasks from the API until drained
+CMD ["perl", "/app/run.pl"]

aiworker/docker-compose.yml

@@ -0,0 +1,27 @@
+# VideoDetect — Worker service (standalone compose file)
+# Run: docker compose -f aiworker/docker-compose.yml up -d
+
+services:
+  worker:
+    build:
+      context: ./
+      dockerfile: Dockerfile
+    container_name: videodetect-aiworker
+    restart: "no"
+    environment:
+      - API_BASE_URL=${API_BASE_URL:-http://10.0.1.34:3000}
+      - OLLAMA_URL=${OLLAMA_URL:-http://10.0.1.7:11434}
+      - SECONDS_PER_FRAME=${SECONDS_PER_FRAME:-30}
+      - TASK_COUNT=${TASK_COUNT:-50}
+    volumes:
+      - nas_input:/data:ro
+      - ./run.pl:/app/run.pl:ro
+
+volumes:
+  nas_input:
+    driver: local
+    driver_opts:
+      type: nfs
+      o: addr=10.0.0.2,ro,nfsvers=4,hard,intr
+      device: ":/mnt/Bulk/Homes/ryan/Prawns"
+

aiworker/run.pl

@@ -0,0 +1,152 @@
+#!/usr/bin/perl
+use strict;
+use MIME::Base64 qw(encode_base64);
+use LWP::UserAgent;
+use JSON;
+use Data::Dumper;
+
+my $seconds = $ENV{'SECONDS_PER_FRAME'} || 30;
+my $ollama = $ENV{'OLLAMA_URL'} || die "OLLAMA_URL environment variable not set\n";
+my $api = $ENV{'API_BASE_URL'} || die "API_BASE_URL environment variable not set\n";
+my $task_count = $ENV{'TASK_COUNT'} || 1;
+my $sensitivity = 20;
+my $resolution = "672:672";
+
+chdir("/tmp") || die "Unable to change directory to /tmp: $!\n";
+my $ua = LWP::UserAgent->new;
+$ua->timeout(60);
+
+for(my $i=0; $i<$task_count; $i++) {
+    my $ref = GetNext();
+    my $task = $ref->{'task'};
+    my $video = $ref->{'video'};
+
+    my $task_id = $task->{'id'};
+    my $video_id = $video->{'id'};
+    my $path = $video->{'file_path'};
+    my $result = CheckPath($path);
+
+    next unless $result;
+    
+    my $response = PostResponse($task_id, $ref->{'assign_key'}, $result);
+}
+
+print "Done\n";
+exit(0);
+
+sub CheckPath {
+    my ($path) = @_;
+    die "File path not set\n" unless $path;
+    die "File does not exist: $path\n" unless -e $path;
+
+    unlink(<*.png>);
+    system("ffmpeg -i \"$path\" -vf fps=1/$seconds,scale=$resolution output_%04d.png");
+
+
+    my $total=0;
+    my $yes=0;
+    $/ = undef;
+    foreach my $file (<*.png>) {
+        print "$file...\n";
+        open(IN, '<:raw', $file) || die "Unable to read $file: $!\n";
+        my $data = <IN>;
+        close(IN);
+
+        my $base64 = encode_base64($data);
+        $data=undef;
+
+        my $ref = {
+            'model'     => 'srizon/pixie:latest',
+            'prompt'    => "Are there any african american people in this photo, whether they are shown completely or not?  Please answer 'yes' or 'no' only.",
+            'images'    => [$base64],
+        };
+
+        my $url = "$ollama/api/generate";
+        my $res = GetRest($url, $ref);
+        print "\t$res\n";
+        $total++;
+        if($res=~/yes/i) {
+            $yes++;
+        }
+        unlink($file);
+    }
+
+    if($total == 0) {
+        warn "No images were processed for $path\n";
+        return undef;
+    }
+
+    my $percentage = ($yes*100)/$total;
+    print "$yes of $total images gave yes - $percentage\n";
+    my $verdict = 0;
+    if($percentage >= $sensitivity) {
+        print "$path: Probably a black dude, $percentage\n";
+        $verdict = 1;
+    } else {
+        print "$path: Probably not, $percentage\n";
+        $verdict = 0;
+    }
+    return {
+        'images'    => $total,
+        'yes'       => $yes,
+        'percentage'=> $percentage,
+        'verdict'   => $verdict,
+
+    };
+}
+
+sub GetRest {
+    my($url, $ref)=@_;
+    my $payload = encode_json($ref);
+
+    my $response = $ua->post($url, 'Content_Type' => 'application/json', 'Content' => $payload);
+    if($response->is_success) {
+        my $content = $response->decoded_content;
+        my @parts = split(/\n/, $content);
+        my @refs = map { decode_json($_) } @parts;
+        return join(' ', map { $_->{'response'} } @refs);
+    } else {
+        die "Error: " . $response->status_line;
+    }
+}
+
+sub PostResponse {
+    my ($task_id, $assign_key, $results) = @_;
+    my $url = "$api/api/v1/task/$task_id/complete";
+    my $ref = {
+        'assign_key' => $assign_key,
+        'results'    => $results,
+    };
+    my $payload = encode_json($ref);
+    my $response = $ua->post($url, 'Content_Type' => 'application/json', 'Content' => $payload);
+    if($response->is_success) {
+        return $response->decoded_content;
+    } else {
+        die "Error posting response ($url): " . $response->status_line;
+    }
+}
+
+sub GetNext {
+    my $url = "$api/api/v1/nexttask/AISCAN";
+    my $task;
+    my $video;
+
+    my $response = $ua->get($url);
+    if($response->is_success) {
+        $task=decode_json($response->decoded_content);
+    } else {
+        die "Error fetching next task ($url): " . $response->status_line;
+    }
+
+    my $video_id = $task->{'task'}{'video_id'};
+    $url = "$api/api/v1/video/$video_id";
+    $response = $ua->get($url);
+    if($response->is_success) {
+        $video=decode_json($response->decoded_content);
+        $task->{'video'} = $video;
+    } else {
+        die "Error fetching video ($url): " . $response->status_line;
+    }
+
+    return $task;
+}

api/Dockerfile

@@ -0,0 +1,27 @@
+FROM ubuntu:22.04
+
+# Install Perl and system dependencies
+RUN apt update && apt install -y \
+    dnsutils \
+    libdancer2-perl \
+    libdancer2-plugin-database-perl \
+    libdancer-plugin-database-core-perl \
+    libdbd-mysql-perl \
+    libhttp-lite-perl \
+    libjson-perl \
+    libspreadsheet-parsexlsx-perl \
+    libsql-splitstatement-perl \
+    libyaml-perl \
+    perl \
+    libdbi-perl \
+    build-essential \
+    ffmpeg
+
+EXPOSE 3000
+
+RUN mkdir -p /app
+WORKDIR /app
+COPY . /app
+
+# Simple placeholder command
+CMD ["perl", "app.pl"]

api/app.pl

@@ -0,0 +1,249 @@
+#!/usr/bin/perl
+use strict;
+use Dancer2;
+use Dancer2::Plugin::Database;
+use Data::Dumper;
+use JSON ();
+use SQL::SplitStatement;
+use HTTP::Lite;
+
+$SIG{'INT'} = sub { exit; };
+
+
+hook before => sub {
+    my $origin = request_header('Origin') || 'http://localhost:8890';
+    my $method = request_header('Access-Control-Request-Method') || '';
+    my $headers = request_header('Access-Control-Request-Headers') || '';
+
+    response_header 'Access-Control-Allow-Origin' => $origin;
+    response_header 'Access-Control-Allow-Methods' => $method;
+    response_header 'Access-Control-Allow-Headers' => $headers;
+};
+
+
+options qr{/api/v1/.+} => sub {
+    return {};
+};
+
+get '/api/v1/hello' => sub {
+    my $user = session('user') || 'world';
+    return {'hello' => $user };
+};
+
+get '/api/v1/videos' => sub {
+    my $sth = database->prepare("SELECT id, file_path FROM videos ORDER BY id");
+    $sth->execute();
+    my $ref = $sth->fetchall_hashref('id');
+    $sth->finish();
+
+    return $ref;
+};
+
+get '/api/v1/video' => sub {
+    my $path = query_parameters->get("path");
+    my $sth = database->prepare("SELECT * FROM videos WHERE file_path=?");
+    $sth->execute($path);
+    my $ref = $sth->fetchrow_hashref();
+    $sth->finish();
+
+    unless(ref $ref) {
+        send_error("No video", 404);
+    }
+
+    return $ref;
+};
+
+get '/api/v1/video/:video' => sub {
+    my $video = route_parameters->get("video");
+    my $sth = database->prepare("SELECT * FROM videos WHERE id=?");
+    $sth->execute($video);
+    my $ref = $sth->fetchrow_hashref();
+    $sth->finish();
+
+    unless(ref $ref) {
+        send_error("No video", 404);
+    }
+
+    return $ref;
+};
+
+post '/api/v1/video' => sub {
+    my $file_path = body_parameters->get("file_path");
+    my $file_hash = body_parameters->get("file_hash") || undef;
+    my $resolution_w = body_parameters->get("resolution_w");
+    my $resolution_h = body_parameters->get("resolution_h");
+    my $codec = body_parameters->get("codec");
+
+    my $sth = database->prepare("INSERT INTO videos (file_path, file_hash, resolution_w, resolution_h, codec) VALUES (?, ?, ?, ?, ?)");
+    $sth->execute($file_path, $file_hash, $resolution_w, $resolution_h, $codec);
+    my $video_id = database->last_insert_id(undef, undef, 'videos', undef);
+    $sth->finish();
+
+    return { id => $video_id };
+};
+
+get '/api/v1/tasks' => sub {
+    my $status = query_parameters->get("status") || 'PENDING';
+    my $type = query_parameters->get("type");
+
+    unless ($type) {
+        send_error("Missing 'type' query parameter", 400);
+    }
+
+    # Optional pagination
+    my $page      = query_parameters->get("page")      || 1;
+    my $per_page  = query_parameters->get("per_page")  || 20;
+    my $offset    = ($page - 1) * $per_page;
+
+    my $sth = database->prepare(
+        "SELECT id, video_id, task_type AS type, status FROM tasks WHERE status=? AND task_type=? ORDER BY id LIMIT ? OFFSET ?"
+    );
+    $sth->execute($status, $type, $per_page, $offset);
+    my $ref = [];
+    while(my $row = $sth->fetchrow_hashref()) {
+        push @$ref, $row;
+    }
+    $sth->finish();
+
+    # Total count for pagination metadata
+    my $count_sth = database->prepare(
+        "SELECT COUNT(*) AS cnt FROM tasks WHERE status=? AND task_type=?"
+    );
+    $count_sth->execute($status, $type);
+    my $cnt = ($count_sth->fetchrow_hashref() || { cnt => 0 })->{cnt};
+    $count_sth->finish();
+
+    return {
+        tasks       => $ref,
+        page        => $page,
+        per_page    => $per_page,
+        total       => $cnt,
+        total_pages => max(1, int(($cnt + $per_page - 1) / $per_page)),
+    };
+};
+
+get '/api/v1/task/:task' => sub {
+    my $task = route_parameters->get("task");
+    my $sth = database->prepare("SELECT * FROM tasks WHERE id=?");
+    $sth->execute($task);
+    my $ref = $sth->fetchrow_hashref();
+    $sth->finish();
+
+    unless(ref $ref) {
+        send_error("No task", 404);
+    }
+
+    return $ref;
+};
+
+get '/api/v1/nexttask/:type' => sub {
+    my $type = route_parameters->get("type");
+    my $sth = database->prepare("SELECT * FROM tasks WHERE task_type=? AND status='PENDING' ORDER BY updated_at ASC LIMIT 1");
+    $sth->execute($type);
+    my $ref = $sth->fetchrow_hashref();
+    $sth->finish();
+
+    unless(ref $ref) {
+        send_error("No task", 404);
+    }
+
+    # Mark the task as in progress and assign it to a worker
+    my $assign_key = "worker_" . int(rand(1000)); # Example assign key, you can customize this
+    my $update_sth = database->prepare("UPDATE tasks SET status='IN_PROGRESS', assign_key=?, assigned_at=NOW() WHERE id=? AND status='PENDING'");
+    $update_sth->execute($assign_key, $ref->{id});
+    if($update_sth->rows == 0) {
+        send_error("Failed to claim task", 409);
+    }
+    $update_sth->finish();
+
+    return { task => $ref, assign_key => $assign_key };
+};
+
+post '/api/v1/task/:task/complete' => sub {
+    my $task_id = route_parameters->get("task");
+
+    # Accept both JSON body and form-encoded data for assign_key/results.
+    my ($assign_key, $results);
+    my $ct = request_header('Content-Type') || '';
+    if ($ct =~ m{application/json}i) {
+        my $body = JSON::decode_json(request->body());
+        $assign_key = $body->{assign_key};
+        $results    = $body->{results};
+    } else {
+        $assign_key = body_parameters->get("assign_key");
+        $results    = body_parameters->get("results");
+    }
+
+    # Optional worker verification: if assign_key is provided, require IN_PROGRESS + matching key.
+    # When omitted (e.g., UI-driven workflows), skip worker checks and just complete the task.
+    if ($assign_key) {
+        my $sth = database->prepare("SELECT * FROM tasks WHERE id=? AND assign_key=? AND status='IN_PROGRESS'");
+        $sth->execute($task_id, $assign_key);
+        my $ref = $sth->fetchrow_hashref();
+        $sth->finish();
+
+        unless(ref $ref) {
+            send_error("Task not assigned to this worker or not in progress", 403);
+        }
+    } else {
+        # No assign_key provided — caller may be an unauthenticated UI, etc.
+        my $sth = database->prepare("SELECT id FROM tasks WHERE id=?");
+        $sth->execute($task_id);
+        unless ($sth->fetchrow_hashref()) {
+            send_error("Task not found", 404);
+        }
+        $sth->finish();
+    }
+
+    # Update the task as completed (store results as JSON text for MariaDB)
+    my $results_json = defined $results ? JSON::encode_json($results) : 'null';
+    my $update_sth = database->prepare("UPDATE tasks SET status='COMPLETED', results=?, updated_at=NOW() WHERE id=?");
+    $update_sth->execute($results_json, $task_id);
+    if($update_sth->rows == 0) {
+        send_error("Failed to complete task", 500);
+    }
+    $update_sth->finish();
+
+    return { message => "Task completed successfully" };
+};
+
+post '/api/v1/task' => sub {
+    my $video_id = body_parameters->get("video_id");
+    my $task_type = body_parameters->get("task_type");
+
+    database->do("DELETE FROM tasks WHERE video_id=? AND task_type=?", undef, $video_id, $task_type);
+
+    my $sth = database->prepare("INSERT INTO tasks (video_id, task_type, status) VALUES (?, ?, 'PENDING')");
+    $sth->execute($video_id, $task_type);
+    my $task_id = database->last_insert_id(undef, undef, 'tasks', undef);
+    $sth->finish();
+
+    return { id => $task_id };
+};
+
+post '/api/v1/task/:task/unassign' => sub {
+    my $task_id = route_parameters->get("task");
+
+    my $update_sth = database->prepare(
+        "UPDATE tasks SET status='PENDING', assign_key=NULL, assigned_at=NULL, updated_at=NOW() WHERE id=?"
+    );
+    $update_sth->execute($task_id);
+
+    if ($update_sth->rows == 0) {
+        send_error("Task not found or already in PENDING state", 404);
+    }
+
+
+    return { message => 'Task unassigned' };
+};
+
+start();
+
+
+sub max {
+    return (sort { $b <=> $a } @_)[0];
+}
+
+sub min {
+    return (sort { $a <=> $b } @_)[0];
+}

api/config.yml

@@ -0,0 +1,39 @@
+serializer: JSON
+logger: console
+engines:
+    serializer:
+        JSON:
+            allow_nonref: 1
+            allow_blessed: 1
+            pretty: 1
+plugins:
+    Database:
+        driver: 'mysql'
+        database: 'videodetect'
+        host: 'mariadb'
+        port: 3306
+        username: 'videodetect'
+        password: 'videodetect'
+        connection_check_threshold: 10
+        dbi_params:
+            RaiseError: 1
+            AutoCommit: 1
+        on_connect_do: ["SET NAMES 'utf8mb4'", "SET CHARACTER SET 'utf8mb4'" ]
+        log_queries: 1
+
+server:
+  port: 5000
+  workers: 4
+  bind: 0.0.0.0
+
+logging:
+  level: info
+  logdir: /var/log/videodetect/api
+  logfile: videodetect-api.log
+  format: "[%d] %l [%P] %m"
+
+paths:
+  scratch: /scratch
+  input: /data/input
+  output: /data/output
+  models: /models

db.sh

@@ -0,0 +1,2 @@
+#!/bin/bash
+docker run --rm -it mariadb:10.11 mariadb --user=videodetect --password=videodetect --host=10.0.1.7 --port=3306 videodetect

db/schema.sql

@@ -18,119 +18,35 @@ USE videodetect;
 CREATE TABLE IF NOT EXISTS videos (
     id BIGINT AUTO_INCREMENT PRIMARY KEY,
     file_path VARCHAR(2048) NOT NULL,
-    file_hash CHAR(64) NOT NULL COMMENT 'SHA-256 hash of file',
+    file_size BIGINT NOT NULL COMMENT 'Size of the file in bytes',
+    file_hash CHAR(64) COMMENT 'SHA-256 hash of file',
     resolution_w INT DEFAULT NULL,
     resolution_h INT DEFAULT NULL,
     codec VARCHAR(50) DEFAULT NULL,
     duration FLOAT DEFAULT NULL COMMENT 'Duration in seconds',
-    status ENUM('NEW', 'PENDING', 'PROCESSING', 'COMPLETED', 'UNSCANNABLE', 'ERROR')
-        NOT NULL DEFAULT 'NEW' COMMENT 'Processing status',
     last_scan_time DATETIME DEFAULT NULL,
-    last_processed_time DATETIME DEFAULT NULL,
-    confidence_score FLOAT DEFAULT NULL COMMENT 'Video-level confidence score',
-    routing_decision ENUM('MATCH', 'REVIEW', 'SKIP') DEFAULT NULL COMMENT 'Routing decision',
-    model_version VARCHAR(50) DEFAULT NULL COMMENT 'Model version used for processing',
-    frame_count INT DEFAULT NULL COMMENT 'Number of frames sampled',
-    error_message TEXT DEFAULT NULL,
     created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
     updated_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
 
-    UNIQUE KEY uk_file_hash (file_hash),
-    INDEX idx_videos_status (status),
+    UNIQUE KEY uk_path (file_path),
     INDEX idx_videos_last_scan (last_scan_time),
-    INDEX idx_videos_status_last_scan (status, last_scan_time),
     INDEX idx_videos_file_path (file_path(255))
 ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
 
--- -----------------------------------------------------
--- Table: processing_logs
--- Audit trail for each video processing job
--- -----------------------------------------------------
-CREATE TABLE IF NOT EXISTS processing_logs (
+
+CREATE TABLE IF NOT EXISTS tasks (
     id BIGINT AUTO_INCREMENT PRIMARY KEY,
+    task_type VARCHAR(16) NOT NULL,
     video_id BIGINT NOT NULL,
-    model_version VARCHAR(50) NOT NULL,
-    frame_count INT NOT NULL DEFAULT 0,
-    confidence_score FLOAT DEFAULT NULL,
-    confidence_scores JSON DEFAULT NULL COMMENT 'Frame-level confidence scores',
-    routing_decision ENUM('MATCH', 'REVIEW', 'SKIP') NOT NULL,
-    processed_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    error_message TEXT DEFAULT NULL,
-    FOREIGN KEY (video_id) REFERENCES videos(id) ON DELETE CASCADE,
-    INDEX idx_processing_logs_video (video_id),
-    INDEX idx_processing_logs_processed_at (processed_at),
-    INDEX idx_processing_logs_routing (routing_decision)
-) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
-
--- -----------------------------------------------------
--- Table: models
--- Model registry tracking all deployed and archived models
--- -----------------------------------------------------
-CREATE TABLE IF NOT EXISTS models (
-    version VARCHAR(50) PRIMARY KEY,
-    status ENUM('ACTIVE', 'CANDIDATE', 'ARCHIVED') NOT NULL DEFAULT 'CANDIDATE',
-    path VARCHAR(2048) NOT NULL,
-    calibration_temp FLOAT DEFAULT NULL COMMENT 'Temperature scaling parameter',
-    f1_score FLOAT DEFAULT NULL COMMENT 'F1 score on validation set',
-    ece_score FLOAT DEFAULT NULL COMMENT 'Expected Calibration Error',
-    deployed_at DATETIME DEFAULT NULL,
-    created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    INDEX idx_models_status (status)
-) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
-
--- -----------------------------------------------------
--- Table: review_queue
--- Videos awaiting manual annotation
--- -----------------------------------------------------
-CREATE TABLE IF NOT EXISTS review_queue (
-    id BIGINT AUTO_INCREMENT PRIMARY KEY,
-    video_id BIGINT NOT NULL,
-    confidence_score FLOAT NOT NULL,
-    routing_decision ENUM('REVIEW') NOT NULL DEFAULT 'REVIEW',
-    annotated BOOLEAN NOT NULL DEFAULT FALSE,
-    ground_truth BOOLEAN DEFAULT NULL COMMENT 'True label from annotator',
-    annotated_at DATETIME DEFAULT NULL,
-    notes TEXT DEFAULT NULL,
+    status ENUM('PENDING', 'IN_PROGRESS', 'COMPLETED', 'FAILED') NOT NULL DEFAULT 'PENDING',
     created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    updated_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
+    assign_key VARCHAR(64) DEFAULT NULL COMMENT 'Key to identify the worker assigned to this task',
+    assigned_at DATETIME DEFAULT NULL COMMENT 'Timestamp when the task was assigned to a worker',
+    results JSON DEFAULT NULL COMMENT 'Task-specific results or metadata',
     FOREIGN KEY (video_id) REFERENCES videos(id) ON DELETE CASCADE,
-    INDEX idx_review_queue_annotated (annotated),
-    INDEX idx_review_queue_created (created_at)
+    INDEX idx_tasks_video (video_id),
+    INDEX idx_tasks_status (status),
+    INDEX idx_tasks_type_status (task_type, status),
+    UNIQUE KEY uk_tasks_video_type (video_id, task_type)
 ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
-
--- -----------------------------------------------------
--- Table: scan_history
--- Tracks directory scan operations for incremental sync
--- -----------------------------------------------------
-CREATE TABLE IF NOT EXISTS scan_history (
-    id BIGINT AUTO_INCREMENT PRIMARY KEY,
-    scan_start DATETIME NOT NULL,
-    scan_end DATETIME DEFAULT NULL,
-    files_discovered INT DEFAULT 0,
-    files_new INT DEFAULT 0,
-    files_modified INT DEFAULT 0,
-    files_removed INT DEFAULT 0,
-    files_unscannable INT DEFAULT 0,
-    duration_seconds FLOAT DEFAULT NULL,
-    status ENUM('RUNNING', 'COMPLETED', 'FAILED') NOT NULL DEFAULT 'RUNNING',
-    error_message TEXT DEFAULT NULL
-) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
-
--- -----------------------------------------------------
--- Table: scanner_lock
--- Single-instance guard for the directory scanner. Ensures only one scan
--- runs at a time across all replicas. A lease (locked_at + lease_seconds)
--- lets a live scanner keep the lock via heartbeats, and lets a crashed
--- scanner's lock be taken over once it goes stale.
--- -----------------------------------------------------
-CREATE TABLE IF NOT EXISTS scanner_lock (
-    lock_name VARCHAR(64) PRIMARY KEY,
-    owner VARCHAR(128) NOT NULL COMMENT 'Instance id (host-pid-uuid) holding the lock',
-    locked_at DATETIME NOT NULL COMMENT 'Last time the lock was acquired or heartbeated',
-    lease_seconds INT NOT NULL DEFAULT 21600 COMMENT 'Lock is stale if older than this (6 hours)'
-) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;
-
--- -----------------------------------------------------
--- Insert default model entry
--- -----------------------------------------------------
-INSERT IGNORE INTO models (version, status, path, calibration_temp, created_at)
-VALUES ('v0.0.0-placeholder', 'ACTIVE', '/models/placeholder', 1.0, NOW());

docker-compose.yml

@@ -5,9 +5,9 @@ services:
     restart: unless-stopped
     environment:
       MYSQL_ROOT_PASSWORD: ${DB_ROOT_PASSWORD:-rootpass}
-      MYSQL_DATABASE: videodetect
-      MYSQL_USER: videodetect
-      MYSQL_PASSWORD: ${DB_PASSWORD:-videodetect123}
+      MYSQL_DATABASE: ${DB_NAME:-videodetect}
+      MYSQL_USER: ${DB_USER:-videodetect}
+      MYSQL_PASSWORD: ${DB_PASSWORD:-videodetect}
     ports:
       - "3306:3306"
     volumes:
@@ -25,69 +25,18 @@ services:
         limits:
           memory: 2G
 
-  worker:
-    build:
-      context: .
-      dockerfile: worker/Dockerfile
-    container_name: videodetect-worker
-    restart: unless-stopped
-    environment:
-      - NVIDIA_VISIBLE_DEVICES=all
-      - NVIDIA_DRIVER_CAPABILITIES=compute,utility
-      - DB_HOST=mariadb
-      - DB_PORT=3306
-      - DB_NAME=videodetect
-      - DB_USER=videodetect
-      - DB_PASSWORD=${DB_PASSWORD:-videodetect123}
-      - CONFIG_PATH=/app/config.yaml
-    volumes:
-      - /dev/null:/dev/null  # tmpfs mounted at /scratch in container
-      - nas_input:/data/input:ro
-      - ${NAS_OUTPUT_PATH:-./output}:/data/output
-      - ${MODELS_PATH:-./models}:/models
-      - ${TRAINING_PATH:-./training}:/data/training
-    tmpfs:
-      - /scratch:noexec,nosuid,size=100G
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - driver: nvidia
-              count: all
-              capabilities: ["gpu"]
-        limits:
-          memory: 24G
-    networks:
-      - videodetect-network
-    depends_on:
-      mariadb:
-        condition: service_healthy
-    healthcheck:
-      test: ["CMD", "python3", "-c", "import torch; print(torch.cuda.is_available())"]
-      interval: 30s
-      timeout: 10s
-      retries: 3
-      start_period: 60s
-
   ui:
     build:
       context: ./ui
       dockerfile: Dockerfile
-    container_name: videodetect-ui
     restart: unless-stopped
     ports:
       - "5000:5000"
     environment:
-      - DB_HOST=mariadb
-      - DB_PORT=3306
-      - DB_NAME=videodetect
-      - DB_USER=videodetect
-      - DB_PASSWORD=${DB_PASSWORD:-videodetect123}
-      - FLASK_ENV=production
+      - API_BASE=http://10.0.1.34:3000/api/v1
     volumes:
       - ./ui:/app
-      - nas_input:/data/input:ro
-      - ${NAS_OUTPUT_PATH:-/mnt/nas/output}:/data/output
+      - nas_input:/data:ro
     networks:
       - videodetect-network
     depends_on:
@@ -98,50 +47,31 @@ services:
         limits:
           memory: 512M
 
-  prometheus:
-    image: prom/prometheus:v2.48.0
-    container_name: videodetect-prometheus
-    restart: unless-stopped
-    ports:
-      - "9090:9090"
-    volumes:
-      - ./monitoring/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml:ro
-      - prometheus_data:/prometheus
-    networks:
-      - videodetect-network
-    deploy:
-      resources:
-        limits:
-          memory: 1G
-
-  grafana:
-    image: grafana/grafana:10.2.0
-    container_name: videodetect-grafana
+  api:
+    build:
+      context: ./api
+      dockerfile: Dockerfile
+    container_name: videodetect-api
     restart: unless-stopped
     ports:
       - "3000:3000"
     environment:
-      - GF_SECURITY_ADMIN_PASSWORD=${GRAFANA_PASSWORD:-admin}
-      - GF_USERS_ALLOW_SIGN_UP=false
+      - DB_HOST=${DB_HOST:-mariadb}
+      - DB_PORT=${DB_PORT:-3306}
+      - DB_NAME=${DB_NAME:-videodetect}
+      - DB_USER=${DB_USER:-videodetect}
+      - DB_PASSWORD=${DB_PASSWORD:-videodetect}
     volumes:
-      - grafana_data:/var/lib/grafana
-      - ./monitoring/grafana/provisioning:/etc/grafana/provisioning:ro
+      - ./api:/app
     networks:
       - videodetect-network
     depends_on:
-      - prometheus
-    deploy:
-      resources:
-        limits:
-          memory: 512M
+      mariadb:
+        condition: service_healthy
 
 volumes:
   mariadb_data:
     driver: local
-  prometheus_data:
-    driver: local
-  grafana_data:
-    driver: local
   nas_input:
     driver: local
     driver_opts:

loadmodel.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+curl http://10.0.1.7:11434/api/generate -d '{
+  "model": "qwen3.6:35b-a3b",
+  "keep_alive": "60m"
+}'

run.pl

@@ -1,68 +0,0 @@
-#!/usr/bin/perl
-use strict;
-use MIME::Base64 qw(encode_base64);
-use LWP::UserAgent;
-use JSON;
-use Data::Dumper;
-
-my $seconds = 30;
-my $ollama = 'http://10.0.1.7:11434';
-my $resolution = "672:672";
-my $sensitivity = 20;
-
-my $file = $ARGV[0] || die "Usage: $0 <filename>\n";
-
-system("ffmpeg -i \"$file\" -vf fps=1/$seconds,scale=$resolution output_%04d.png");
-my $ua = LWP::UserAgent->new;
-$ua->timeout(60);
-
-my $total=0;
-my $yes=0;
-$/ = undef;
-foreach my $file (<*.png>) {
-    print "$file...\n";
-    open(IN, '<:raw', $file) || die "Unable to read $file: $!\n";
-    my $data = <IN>;
-    close(IN);
-
-    my $base64 = encode_base64($data);
-    $data=undef;
-
-    my $ref = {
-        'model'     => 'srizon/pixie:latest',
-        'prompt'    => "Are there any african american people or in this photo, whether they are shown completely or not?  Please answer 'yes' or 'no' only.",
-        'images'    => [$base64],
-    };
-
-    my $url = "$ollama/api/generate";
-    my $res = GetRest($url, $ref);
-    print "\t$res\n";
-    $total++;
-    if($res=~/yes/i) {
-        $yes++;
-    }
-    unlink($file);
-}
-
-my $percentage = ($yes*100)/$total;
-print "$yes of $total images gave yes - $percentage\n";
-if($percentage >= $sensitivity) {
-    print "Probably a black dude, $percentage\n";
-} else {
-    print "Probably not, $percentage\n";
-}
-
-sub GetRest {
-    my($url, $ref)=@_;
-    my $payload = encode_json($ref);
-
-    my $response = $ua->post($url, 'Content_Type' => 'application/json', 'Content' => $payload);
-    if($response->is_success) {
-        my $content = $response->decoded_content;
-        my @parts = split(/\n/, $content);
-        my @refs = map { decode_json($_) } @parts;
-        return join(' ', map { $_->{'response'} } @refs);
-    } else {
-        die "Error: " . $response->status_line;
-    }
-}

scanner/.env

@@ -0,0 +1,14 @@
+# VideoDetect — Docker Compose environment variables
+# Copy this file to .env and fill in real values before running docker-compose up.
+
+# -----------------------------------------------------
+# Database
+# -----------------------------------------------------
+DB_ROOT_PASSWORD=videodetect
+DB_PASSWORD=videodetect
+DB_USER=videodetect
+DB_NAME=videodetect
+DB_HOST=10.0.1.34
+
+API_HOST=https://10.0.1.34:3000
+

scanner/Dockerfile

@@ -0,0 +1,23 @@
+FROM ubuntu:22.04
+
+# Install Perl, system deps, and ffmpeg (includes ffprobe)
+RUN apt update && apt install -y \
+    libdbd-mysql-perl \
+    libhttp-lite-perl \
+    libjson-perl \
+    libyaml-perl \
+    perl \
+    libdbi-perl \
+    build-essential \
+    ffmpeg \
+    libstring-shellquote-perl
+
+
+EXPOSE 3000
+
+RUN mkdir -p /app
+WORKDIR /app
+COPY . /app
+
+# Simple placeholder command
+CMD ["perl", "scanner.pl"]

scanner/docker-compose.yml

@@ -0,0 +1,28 @@
+# VideoDetect — Scanner service (standalone compose file)
+# Run: docker compose -f scanner-compose.yml up -d
+
+services:
+  scanner:
+    build:
+      context: ./
+      dockerfile: Dockerfile
+    container_name: videodetect-scanner
+    restart: no
+    environment:
+      - DB_HOST=${DB_HOST}
+      - DB_PORT=3306
+      - DB_NAME=${DB_NAME}
+      - DB_USER=${DB_USER}
+      - DB_PASSWORD=${DB_PASSWORD}
+      - API_HOST=${API_HOST}
+    volumes:
+      - nas_input:/data:ro
+      - $PWD/scanner.pl:/app/scanner.pl:ro
+
+volumes:
+  nas_input:
+    driver: local
+    driver_opts:
+      type: nfs
+      o: addr=10.0.0.2,ro,nfsvers=4,hard,intr
+      device: ":/mnt/Bulk/Homes/ryan/Prawns"

scanner/scanner.pl

@@ -0,0 +1,225 @@
+#!/usr/bin/perl
+use strict;
+use warnings;
+use DBI;
+use Digest::SHA qw(sha256_hex);
+use JSON;
+use String::ShellQuote qw/shell_quote/;
+
+my $db_host = $ENV{'DB_HOST'} || 'mariadb';
+my $db_name = $ENV{'DB_NAME'} || 'videodetect';
+my $db_user = $ENV{'DB_USER'} || 'videodetect';
+my $db_pass = $ENV{'DB_PASSWORD'} || 'videodetect';  
+
+my $do_scan = (defined $ENV{'DO_SCAN'}) ? $ENV{'DO_SCAN'} : 1;
+my $do_review = (defined $ENV{'DO_REVIEW'}) ? $ENV{'DO_REVIEW'} : 1;
+my $do_delete = (defined $ENV{'DO_DELETE'}) ? $ENV{'DO_DELETE'} : 1;
+
+$SIG{INT} = sub { die "Interrupted\n"; };
+
+my $dbh = DBI->connect(
+    "DBI:mysql:database=$db_name;host=$db_host;mysql_enable_utf8mb4=1",
+    $db_user,
+    $db_pass,
+    { mysql_auto_reconnect => 1 }
+);
+
+my @video_extensions = qw(mp4 mkv avi mov flv wmv mpg mpeg webm m4v);
+my %ignored_extensions;
+my @dir_queue = ('/data');
+
+
+# Get NOW() from the database
+my $sth = $dbh->prepare("SELECT NOW() AS now");
+$sth->execute();
+my $row = $sth->fetchrow_hashref();
+my $now = $row->{now};
+$sth->finish();
+
+if($do_scan) {
+    while(my $dir = shift @dir_queue) {
+        opendir(my $dh, $dir) or die "Cannot open directory $dir: $!";
+        print "$dir\n";
+        while (my $file = readdir($dh)) {
+            next if ($file eq '.' || $file eq '..');
+            my $full_path = "$dir/$file";
+            if (-d $full_path) {
+                unshift @dir_queue, $full_path;
+            } elsif (-f $full_path) {
+                process_file($full_path);
+            }
+        }
+        closedir($dh);
+    }
+    for my $ext (sort { $ignored_extensions{$b} <=> $ignored_extensions{$a} } keys %ignored_extensions) {
+        print "Ignored extension $ext: $ignored_extensions{$ext} times\n";
+    }
+}
+
+if($do_scan && $do_delete) {
+    # Any entry where last_scan_time < $now is considered deleted or moved, so we can mark them as such
+    my $sth_delete = $dbh->prepare("DELETE FROM videos WHERE last_scan_time < ? OR last_scan_time IS NULL");
+    $sth_delete->execute($now);
+}
+
+if($do_review) {
+    print "Scanning complete. Creating review tasks for completed AISCAN tasks.\n";
+    CreateReviewTasks();
+}
+
+sub process_file {
+    my ($file_path) = @_;
+    if (my ($ext)=$file_path =~ /^.+\.(\S+?)$/) {
+        unless (grep { lc($ext) eq $_ } @video_extensions) {
+            $ignored_extensions{lc($ext)}++;
+            return;  # Not a video file
+        }
+    }
+
+    my $sth = $dbh->prepare("SELECT id,file_size FROM videos WHERE file_path=?");
+    $sth->execute($file_path);
+    if(my $row = $sth->fetchrow_hashref()) {
+        my $file_size = -s $file_path;
+        if ($file_size != $row->{file_size}) {
+            warn "File size mismatch for $file_path. Updating record.";
+            my $update_sth = $dbh->prepare("UPDATE videos SET file_size=?, last_scan_time=NOW() WHERE id=?");
+            $update_sth->execute($file_size, $row->{id});
+            $update_sth->finish();
+            create_aiscan_task($row->{id});
+        } else {
+            my $update_sth = $dbh->prepare("UPDATE videos SET last_scan_time=NOW() WHERE id=?");
+            $update_sth->execute($row->{id});
+            $update_sth->finish();
+        }
+        
+        return;  # Already exists and size matches
+    }
+    print "NEW:$file_path\n";
+
+    my $info = get_video_info($file_path);
+    unless ($info) {
+        warn "Failed to get video info for $file_path. Skipping.";
+        return;
+    }
+    $sth = $dbh->prepare("INSERT INTO videos (file_path, file_size, file_hash, resolution_w, resolution_h, codec, duration, last_scan_time) VALUES (?, ?, ?, ?, ?, ?, ?, NOW())");
+    $sth->execute(
+        $file_path,
+        -s $file_path,
+        $info->{file_hash},
+        $info->{resolution_w},
+        $info->{resolution_h},
+        $info->{codec},
+        $info->{duration}
+    );
+    $sth->finish();
+
+    my $video_id = $dbh->last_insert_id(undef, undef, 'videos', undef);
+    create_aiscan_task($video_id);
+}
+
+sub create_aiscan_task {
+    my ($video_id) = @_;
+    create_task($video_id, 'AISCAN');
+}
+
+sub create_task {
+    my ($video_id, $task_type) = @_;
+    $dbh->do("DELETE FROM tasks WHERE video_id=$video_id AND task_type=?", undef, $task_type);
+    my $sth = $dbh->prepare("INSERT INTO tasks (video_id, task_type, status) VALUES (?, ?, 'PENDING')");
+    $sth->execute($video_id, $task_type);
+    $sth->finish();
+}
+
+sub get_video_info {
+    my ($file_path) = @_;
+
+    # --- Compute SHA-256 hash of the file ---
+    return undef unless -f $file_path && -r $file_path;
+    open(my $fh, '<:raw', $file_path) or do { warn "Cannot open $file_path: $!"; return undef; };
+    my $hash = sha256_hex($fh);
+    close($fh);
+
+
+
+    # --- Run ffprobe to extract metadata ---
+    my $probe_cmd = 'ffprobe -v quiet -print_format json -show_format -show_streams ' . shell_quote($file_path);
+    my $output = `$probe_cmd`;
+    return undef unless defined $output && length($output);
+
+    my $json = JSON->new->utf8->canonical(1);
+    my $data = $json->decode($output);
+
+    # --- Extract codec from video stream (prefer first video stream found) ---
+    my $codec = undef;
+    if (exists $data->{streams} && ref($data->{streams}) eq 'ARRAY') {
+        for my $stream (@{$data->{streams}}) {
+            if ($stream->{codec_type} eq 'video') {
+                $codec = $stream->{codec_name};
+                last;
+            }
+        }
+    }
+
+    # --- Extract resolution and duration from format/streams ---
+    my ($resolution_w, $resolution_h, $duration);
+
+    # Duration from format first, then stream
+    if (exists $data->{format} && exists $data->{format}{duration}) {
+        $duration = $data->{format}{duration} + 0;   # force numeric
+    } elsif (exists $data->{streams}[0] && exists $data->{streams}[0]{duration}) {
+        $duration = $data->{streams}[0]{duration} + 0;
+    }
+
+    # Resolution from video stream first, then format side data
+    if (exists $data->{streams} && ref($data->{streams}) eq 'ARRAY') {
+        for my $stream (@{$data->{streams}}) {
+            next unless $stream->{codec_type} eq 'video';
+            if ($stream->{width} && $stream->{height}) {
+                $resolution_w = $stream->{width};
+                $resolution_h = $stream->{height};
+                last;
+            }
+        }
+    }
+
+    return {
+        file_hash      => $hash,
+        resolution_w   => $resolution_w // 0,
+        resolution_h   => $resolution_h // 0,
+        codec          => $codec // 'unknown',
+        duration       => $duration,
+    };
+}
+
+sub CreateReviewTasks {
+    # Fetch all completed AISCAN tasks and filter in Perl.
+    # This avoids MariaDB's inconsistent JSON functions in SQL WHERE clauses,
+    # which behave differently between CLI client and DBI connections.
+    my $sth = $dbh->prepare("
+SELECT id, video_id, results 
+FROM tasks
+WHERE task_type='AISCAN' AND status='COMPLETED'
+");
+
+    $sth->execute();
+    while (my $row = $sth->fetchrow_hashref()) {
+        my $results = decode_json($row->{results});
+        next unless $results && ref($results) eq 'HASH';
+        next unless $results->{verdict} == 1;
+        next if exists $results->{reviewed};
+
+        print "Creating REVIEW task for video_id: $row->{video_id}\n";
+        create_task($row->{video_id}, 'REVIEW');
+
+        # Re-encode the whole JSON blob in Perl and update it.
+        # This avoids MariaDB JSON_SET path parsing issues through DBI connections.
+        $results->{reviewed} = 1;
+        my $updated = encode_json($results);
+        my $sth_update = $dbh->prepare(
+            "UPDATE tasks SET results = ? WHERE id = ?"
+        );
+        $sth_update->execute($updated, $row->{id});
+        $sth_update->finish();
+    }
+    $sth->finish();
+}

src/active_learning/__init__.py

@@ -1 +0,0 @@
-"""Active learning pipeline: label ingestion, training, validation, and deployment."""

src/active_learning/label_ingestor.py

@@ -1,217 +0,0 @@
-"""
-Label ingestion for active learning.
-
-Queries the review_queue for annotated labels, extracts face crops,
-and builds a versioned training dataset with stratified train/val split.
-"""
-
-import csv
-import json
-import logging
-import random
-import shutil
-from pathlib import Path
-from typing import Dict, List, Optional, Tuple
-
-logger = logging.getLogger(__name__)
-
-# Train split fraction (val = 1 - TRAIN_FRAC)
-_TRAIN_FRAC = 0.80
-
-
-class LabelIngestor:
-    """Ingest annotated review data and build a versioned dataset."""
-
-    def __init__(self, db_connector, config):
-        self._db = db_connector
-        self._cfg = config
-        self._training_path = Path(config.get("storage.training_path", "/data/training"))
-
-    # ------------------------------------------------------------------
-    # Public API
-    # ------------------------------------------------------------------
-
-    def count_annotated(self) -> int:
-        """Return number of annotated rows ready for training."""
-        rows = self._db.fetchall(
-            "SELECT COUNT(*) AS cnt FROM review_queue "
-            "WHERE annotated = TRUE AND ground_truth IS NOT NULL"
-        )
-        return int((rows or [{"cnt": 0}])[0]["cnt"])
-
-    def ingest(self, version: str, seed: int = 42) -> Optional[str]:
-        """
-        Build a versioned dataset directory from the review queue.
-
-        Returns the dataset path on success, or None if no data is available.
-        """
-        rows = self._fetch_annotated_rows()
-        if not rows:
-            logger.warning("No annotated samples found; skipping dataset build")
-            return None
-
-        dataset_path = self._training_path / version
-        crops_pos = dataset_path / "crops" / "class_1"
-        crops_neg = dataset_path / "crops" / "class_0"
-        for d in (crops_pos, crops_neg):
-            d.mkdir(parents=True, exist_ok=True)
-
-        records = self._extract_crops(rows, crops_pos, crops_neg)
-        if not records:
-            logger.error("No crops could be extracted from annotated rows")
-            return None
-
-        train_records, val_records = _stratified_split(records, _TRAIN_FRAC, seed)
-
-        self._write_labels_csv(dataset_path / "labels.csv", records)
-        self._write_metadata(dataset_path / "metadata.json", version, records, train_records, val_records)
-
-        logger.info(
-            "Dataset v%s built: %d total (%d train / %d val) at %s",
-            version, len(records), len(train_records), len(val_records), dataset_path,
-        )
-        return str(dataset_path)
-
-    # ------------------------------------------------------------------
-    # Internal helpers
-    # ------------------------------------------------------------------
-
-    def _fetch_annotated_rows(self) -> List[Dict]:
-        rows = self._db.fetchall(
-            """SELECT rq.video_id, rq.ground_truth,
-                      pl.confidence_scores,
-                      v.file_path
-               FROM review_queue rq
-               JOIN videos v ON v.id = rq.video_id
-               LEFT JOIN (
-                   SELECT video_id, confidence_scores,
-                          ROW_NUMBER() OVER (PARTITION BY video_id ORDER BY processed_at DESC) AS rn
-                   FROM processing_logs
-               ) pl ON pl.video_id = rq.video_id AND pl.rn = 1
-               WHERE rq.annotated = TRUE AND rq.ground_truth IS NOT NULL"""
-        )
-        return rows or []
-
-    def _extract_crops(
-        self,
-        rows: List[Dict],
-        crops_pos: Path,
-        crops_neg: Path,
-    ) -> List[Dict]:
-        """Copy or re-extract face crops into the dataset directory."""
-        records: List[Dict] = []
-        scratch_path = Path(self._cfg.get("storage.scratch_path", "/scratch"))
-
-        for row in rows:
-            video_id = row["video_id"]
-            label = int(bool(row["ground_truth"]))
-            dest_dir = crops_pos if label == 1 else crops_neg
-
-            try:
-                contributing = _parse_json_field(row.get("confidence_scores"))
-                crop_paths = self._resolve_crop_paths(video_id, contributing, scratch_path)
-
-                for crop_path in crop_paths:
-                    src = Path(crop_path)
-                    if not src.exists():
-                        logger.debug("Crop not found, skipping: %s", src)
-                        continue
-                    dest = dest_dir / f"video{video_id}_{src.name}"
-                    shutil.copy2(src, dest)
-                    records.append({"crop_path": str(dest), "label": label, "video_id": video_id})
-
-            except Exception as exc:
-                logger.warning("Failed to extract crops for video %s: %s", video_id, exc)
-
-        return records
-
-    def _resolve_crop_paths(
-        self,
-        video_id: int,
-        contributing_frames: List,
-        scratch_path: Path,
-    ) -> List[str]:
-        """Return paths of existing crop files for this video."""
-        paths: List[str] = []
-
-        # contributing_frames may be a list of frame scores or frame dicts
-        if contributing_frames:
-            for item in contributing_frames:
-                if isinstance(item, dict) and "crop_path" in item:
-                    paths.append(item["crop_path"])
-                elif isinstance(item, str):
-                    paths.append(item)
-
-        # Fall back: glob any crop files already written to scratch
-        if not paths:
-            pattern = f"*video{video_id}*.jpg"
-            paths = [str(p) for p in scratch_path.rglob(pattern)]
-
-        return paths
-
-    @staticmethod
-    def _write_labels_csv(csv_path: Path, records: List[Dict]):
-        with open(csv_path, "w", newline="", encoding="utf-8") as f:
-            writer = csv.DictWriter(f, fieldnames=["crop_path", "label", "video_id"])
-            writer.writeheader()
-            writer.writerows(records)
-
-    @staticmethod
-    def _write_metadata(
-        meta_path: Path,
-        version: str,
-        all_records: List[Dict],
-        train_records: List[Dict],
-        val_records: List[Dict],
-    ):
-        pos = sum(1 for r in all_records if r["label"] == 1)
-        neg = len(all_records) - pos
-        meta = {
-            "version": version,
-            "total_samples": len(all_records),
-            "train_samples": len(train_records),
-            "val_samples": len(val_records),
-            "class_counts": {"class_0": neg, "class_1": pos},
-        }
-        meta_path.write_text(json.dumps(meta, indent=2), encoding="utf-8")
-
-
-# ------------------------------------------------------------------
-# Module-level helpers
-# ------------------------------------------------------------------
-
-def _stratified_split(
-    records: List[Dict],
-    train_frac: float,
-    seed: int,
-) -> Tuple[List[Dict], List[Dict]]:
-    """Return (train, val) with stratification by label."""
-    rng = random.Random(seed)
-
-    by_label: Dict[int, List[Dict]] = {}
-    for r in records:
-        by_label.setdefault(r["label"], []).append(r)
-
-    train, val = [], []
-    for label_records in by_label.values():
-        shuffled = list(label_records)
-        rng.shuffle(shuffled)
-        split = max(1, int(len(shuffled) * train_frac))
-        train.extend(shuffled[:split])
-        val.extend(shuffled[split:])
-
-    rng.shuffle(train)
-    rng.shuffle(val)
-    return train, val
-
-
-def _parse_json_field(value) -> List:
-    if not value:
-        return []
-    if isinstance(value, list):
-        return value
-    try:
-        parsed = json.loads(value)
-        return parsed if isinstance(parsed, list) else []
-    except (json.JSONDecodeError, TypeError):
-        return []

src/active_learning/pipeline.py

@@ -1,82 +0,0 @@
-"""
-Active learning pipeline orchestrator.
-
-Entry point that coordinates:
-  1. Check annotated sample count against minimum threshold
-  2. Ingest labels and build versioned dataset
-  3. Fine-tune classification head
-  4. Validate candidate model against quality gates
-  5. Deploy if gates pass; skip otherwise
-"""
-
-import logging
-from typing import Optional
-
-from .label_ingestor import LabelIngestor
-from .registry import ModelRegistry
-from .trainer import Trainer
-from .validator import Validator
-
-logger = logging.getLogger(__name__)
-
-
-class ActiveLearningPipeline:
-    """Orchestrate the full active learning cycle."""
-
-    def __init__(self, db_connector, config):
-        self._db = db_connector
-        self._cfg = config
-        self._ingestor = LabelIngestor(db_connector, config)
-        self._registry = ModelRegistry(db_connector, config)
-
-    def run(self, version: str) -> bool:
-        """
-        Execute a full active learning cycle for the given model version.
-
-        Returns True if a new model was deployed, False otherwise.
-        """
-        al = self._cfg.get_section("active_learning")
-        min_samples: int = int(al.get("min_annotated_samples", 100))
-
-        # 1. Check threshold
-        n_annotated = self._ingestor.count_annotated()
-        logger.info("Annotated samples available: %d (minimum: %d)", n_annotated, min_samples)
-        if n_annotated < min_samples:
-            logger.info("Insufficient annotated samples; skipping active learning cycle")
-            return False
-
-        # 2. Ingest labels
-        seed = int(self._cfg.get("active_learning.seed", 42))
-        dataset_path = self._ingestor.ingest(version, seed=seed)
-        if not dataset_path:
-            logger.error("Label ingestion produced no dataset; aborting")
-            return False
-
-        # 3. Train
-        trainer = Trainer(self._cfg)
-        checkpoint_path = trainer.train(dataset_path, version)
-        if not checkpoint_path:
-            logger.error("Training failed; aborting active learning cycle")
-            return False
-
-        # 4. Validate
-        current_f1 = self._registry.get_active_f1()
-        validator = Validator(self._cfg, current_f1=current_f1)
-        metrics = validator.validate(checkpoint_path, dataset_path)
-
-        self._registry.register_candidate(
-            version, checkpoint_path,
-            f1=metrics.get("f1", 0.0),
-            ece=metrics.get("ece", 1.0),
-        )
-
-        if not metrics.get("gates_passed", False):
-            logger.warning(
-                "Quality gates failed for %s; skipping deployment. Details: %s",
-                version, metrics.get("gate_details"),
-            )
-            return False
-
-        # 5. Deploy
-        deployed = self._registry.deploy(version, checkpoint_path)
-        return deployed

src/active_learning/registry.py

@@ -1,223 +0,0 @@
-"""
-Model registry and deployment for active learning.
-
-Handles:
-- Promoting a candidate model to ACTIVE in the DB
-- Converting the PyTorch checkpoint to ONNX then TensorRT
-- Hot-reloading the classifier in worker processes via SIGUSER1
-- Rollback to the previous ACTIVE model on failure
-"""
-
-import logging
-import os
-import signal
-import subprocess
-from pathlib import Path
-from typing import Optional
-
-logger = logging.getLogger(__name__)
-
-# Signal used to notify the worker process to reload its model
-_RELOAD_SIGNAL = signal.SIGUSR1
-
-
-class ModelRegistry:
-    """Manage model lifecycle: promote, deploy, rollback."""
-
-    def __init__(self, db_connector, config):
-        self._db = db_connector
-        self._cfg = config
-        self._models_path = Path(config.get("storage.models_path", "/models"))
-        al = config.get_section("active_learning")
-        dep = al.get("deployment", {})
-        self._auto_deploy: bool = bool(dep.get("auto_deploy", True))
-        self._hot_reload: bool = bool(dep.get("hot_reload", True))
-        self._rollback_enabled: bool = bool(dep.get("rollback_enabled", True))
-
-    # ------------------------------------------------------------------
-    # Public API
-    # ------------------------------------------------------------------
-
-    def get_active_version(self) -> Optional[str]:
-        """Return the version string of the currently ACTIVE model."""
-        rows = self._db.fetchall(
-            "SELECT version FROM models WHERE status = 'ACTIVE' ORDER BY deployed_at DESC LIMIT 1"
-        )
-        if rows:
-            return rows[0]["version"]
-        return None
-
-    def get_active_f1(self) -> float:
-        """Return the F1 score of the currently ACTIVE model (0 if unknown)."""
-        rows = self._db.fetchall(
-            "SELECT f1_score FROM models WHERE status = 'ACTIVE' ORDER BY deployed_at DESC LIMIT 1"
-        )
-        if rows and rows[0]["f1_score"] is not None:
-            return float(rows[0]["f1_score"])
-        return 0.0
-
-    def register_candidate(self, version: str, checkpoint_path: str, f1: float, ece: float):
-        """Insert or update the candidate model record."""
-        self._db.execute(
-            """INSERT INTO models (version, status, path, f1_score, ece_score, created_at)
-               VALUES (%s, 'CANDIDATE', %s, %s, %s, NOW())
-               ON DUPLICATE KEY UPDATE
-                   status = 'CANDIDATE', path = VALUES(path),
-                   f1_score = VALUES(f1_score), ece_score = VALUES(ece_score)""",
-            (version, checkpoint_path, f1, ece),
-        )
-        logger.info("Registered candidate model %s (f1=%.4f, ece=%.4f)", version, f1, ece)
-
-    def deploy(self, version: str, checkpoint_path: str) -> bool:
-        """
-        Convert, promote and hot-reload the candidate model.
-
-        Returns True if deployment succeeded, False otherwise.
-        """
-        if not self._auto_deploy:
-            logger.info("Auto-deploy disabled; skipping deployment of %s", version)
-            return False
-
-        try:
-            engine_path = self._build_engine(version, checkpoint_path)
-            self._promote(version, engine_path)
-            if self._hot_reload:
-                self._reload_worker()
-            logger.info("Model %s deployed successfully", version)
-            return True
-        except Exception as exc:
-            logger.error("Deployment of %s failed: %s", version, exc, exc_info=True)
-            if self._rollback_enabled:
-                self.rollback(version)
-            return False
-
-    def rollback(self, failed_version: str):
-        """Restore the most recent ARCHIVED model to ACTIVE."""
-        rows = self._db.fetchall(
-            """SELECT version FROM models
-               WHERE status = 'ARCHIVED'
-               ORDER BY deployed_at DESC
-               LIMIT 1"""
-        )
-        if not rows:
-            logger.error("No archived model available for rollback")
-            return
-
-        prev_version = rows[0]["version"]
-        self._db.execute(
-            "UPDATE models SET status = 'ARCHIVED' WHERE version = %s",
-            (failed_version,),
-        )
-        self._db.execute(
-            "UPDATE models SET status = 'ACTIVE' WHERE version = %s",
-            (prev_version,),
-        )
-        logger.warning("Rolled back to model %s (failed: %s)", prev_version, failed_version)
-        if self._hot_reload:
-            self._reload_worker()
-
-    # ------------------------------------------------------------------
-    # Internal helpers
-    # ------------------------------------------------------------------
-
-    def _build_engine(self, version: str, checkpoint_path: str) -> str:
-        """Export checkpoint to ONNX then build a TensorRT FP32 engine."""
-        candidate_dir = self._models_path / "candidate"
-        onnx_path = candidate_dir / f"{version}.onnx"
-        engine_path = candidate_dir / f"{version}.trt"
-
-        self._export_onnx(checkpoint_path, onnx_path)
-        self._build_trt(onnx_path, engine_path)
-        return str(engine_path)
-
-    @staticmethod
-    def _export_onnx(checkpoint_path: str, onnx_path: Path):
-        """Convert PyTorch checkpoint to ONNX."""
-        try:
-            import torch
-            import torch.nn as nn
-            import torchvision.models as models
-
-            device = torch.device("cpu")
-            model = models.mobilenet_v3_small(weights=None)
-            model.classifier = nn.Sequential(
-                nn.Linear(576, 256),
-                nn.ReLU(),
-                nn.Dropout(0.3),
-                nn.Linear(256, 2),
-            )
-            state = torch.load(checkpoint_path, map_location=device)
-            model.load_state_dict(state)
-            model.eval()
-
-            dummy = torch.zeros(1, 3, 224, 224)
-            torch.onnx.export(
-                model,
-                dummy,
-                str(onnx_path),
-                input_names=["input"],
-                output_names=["output"],
-                dynamic_axes={"input": {0: "batch_size"}, "output": {0: "batch_size"}},
-                opset_version=11,
-            )
-            logger.info("Exported ONNX model to %s", onnx_path)
-        except ImportError as exc:
-            logger.error("PyTorch/torchvision not available for ONNX export: %s", exc)
-            raise
-
-    @staticmethod
-    def _build_trt(onnx_path: Path, engine_path: Path):
-        """Build TensorRT FP32 engine via trtexec CLI."""
-        cmd = [
-            "trtexec",
-            f"--onnx={onnx_path}",
-            f"--saveEngine={engine_path}",
-            "--fp32",
-            "--maxBatch=32",
-            "--workspace=2048",
-        ]
-        logger.info("Building TensorRT engine: %s", " ".join(cmd))
-        result = subprocess.run(cmd, capture_output=True, text=True, timeout=600)
-        if result.returncode != 0:
-            raise RuntimeError(
-                f"trtexec failed (rc={result.returncode}): {result.stderr[:500]}"
-            )
-        logger.info("TensorRT engine saved to %s", engine_path)
-
-    def _promote(self, version: str, engine_path: str):
-        """Atomically promote candidate to ACTIVE, archive current ACTIVE."""
-        self._db.execute(
-            """UPDATE models
-               SET status = 'ARCHIVED'
-               WHERE status = 'ACTIVE'""",
-        )
-        self._db.execute(
-            """UPDATE models
-               SET status = 'ACTIVE', path = %s, deployed_at = NOW()
-               WHERE version = %s""",
-            (engine_path, version),
-        )
-        logger.info("Promoted model %s to ACTIVE (engine=%s)", version, engine_path)
-
-    @staticmethod
-    def _reload_worker():
-        """Send SIGUSR1 to the worker process to trigger a model hot-reload."""
-        pid_file = Path("/tmp/videodetect_worker.pid")
-        if pid_file.exists():
-            try:
-                pid = int(pid_file.read_text().strip())
-                os.kill(pid, _RELOAD_SIGNAL)
-                logger.info("Sent SIGUSR1 to worker PID %d for hot-reload", pid)
-            except (ValueError, ProcessLookupError, PermissionError) as exc:
-                logger.warning("Could not signal worker process: %s", exc)
-        else:
-            # Fallback: restart via docker-compose
-            logger.warning("No PID file found; attempting docker-compose restart")
-            try:
-                subprocess.run(
-                    ["docker-compose", "restart", "worker"],
-                    capture_output=True,
-                    timeout=60,
-                )
-            except Exception as exc:
-                logger.error("docker-compose restart failed: %s", exc)

src/active_learning/trainer.py

@@ -1,344 +0,0 @@
-"""
-Training pipeline for active learning head-only fine-tuning.
-
-Loads a pre-trained MobileNetV3 backbone (frozen), replaces the classification
-head, and fine-tunes on versioned crop datasets using AdamW + early stopping.
-All training is FP32 for Tesla P40 compatibility (CC 5.2, no Tensor Cores).
-"""
-
-import logging
-import os
-import random
-from pathlib import Path
-from typing import Dict, Optional, Tuple
-
-import numpy as np
-
-logger = logging.getLogger(__name__)
-
-
-class Trainer:
-    """Head-only fine-tuning pipeline for MobileNetV3 classifier."""
-
-    def __init__(self, config):
-        self._cfg = config
-        self._models_path = Path(config.get("storage.models_path", "/models"))
-        self._candidate_path = self._models_path / "candidate"
-        self._candidate_path.mkdir(parents=True, exist_ok=True)
-
-        al = config.get_section("active_learning")
-        tr = al.get("training", {})
-        self._epochs: int = int(tr.get("epochs", 20))
-        self._batch_size: int = int(tr.get("batch_size", 32))
-        self._lr: float = float(tr.get("learning_rate", 1e-3))
-        self._weight_decay: float = float(tr.get("weight_decay", 1e-2))
-        self._patience: int = int(tr.get("early_stopping_patience", 5))
-        self._lr_factor: float = float(tr.get("lr_factor", 0.5))
-        self._lr_patience: int = int(tr.get("lr_patience", 3))
-
-        aug = al.get("augmentation", {})
-        self._aug_flip: bool = bool(aug.get("horizontal_flip", True))
-        self._aug_jitter: bool = bool(aug.get("color_jitter", True))
-        self._aug_affine: bool = bool(aug.get("affine", True))
-        self._aug_degrees: int = int(aug.get("affine_degrees", 10))
-        self._aug_scale: float = float(aug.get("affine_scale", 0.1))
-
-        seed = int(config.get("active_learning.seed", 42))
-        self._seed = seed
-
-    # ------------------------------------------------------------------
-    # Public API
-    # ------------------------------------------------------------------
-
-    def train(self, dataset_path: str, version: str) -> Optional[str]:
-        """
-        Fine-tune the classification head on the versioned dataset.
-
-        Returns the path to the best checkpoint or None on failure.
-        """
-        try:
-            import torch
-            import torch.nn as nn
-            from torch.utils.data import DataLoader
-
-            self._seed_everything(self._seed)
-            self._assert_cuda_compatibility(torch)
-
-            train_loader, val_loader = self._build_dataloaders(dataset_path, torch)
-            model = self._build_model(nn)
-            best_path = self._run_training(model, train_loader, val_loader, version, torch, nn)
-            return best_path
-        except ImportError as exc:
-            logger.error("PyTorch not available: %s", exc)
-            return None
-        except Exception as exc:
-            logger.error("Training failed: %s", exc, exc_info=True)
-            return None
-
-    # ------------------------------------------------------------------
-    # Model construction
-    # ------------------------------------------------------------------
-
-    def _build_model(self, nn):
-        """Load MobileNetV3-Small backbone, freeze it, attach trainable head."""
-        import torchvision.models as models
-
-        model = models.mobilenet_v3_small(weights=models.MobileNet_V3_Small_Weights.IMAGENET1K_V1)
-
-        # Freeze backbone
-        for param in model.features.parameters():
-            param.requires_grad = False
-
-        # Replace classification head
-        model.classifier = nn.Sequential(
-            nn.Linear(576, 256),
-            nn.ReLU(),
-            nn.Dropout(0.3),
-            nn.Linear(256, 2),
-        )
-        return model
-
-    # ------------------------------------------------------------------
-    # Data loading
-    # ------------------------------------------------------------------
-
-    def _build_dataloaders(self, dataset_path: str, torch) -> Tuple:
-        from torch.utils.data import DataLoader
-        import torchvision.transforms as T
-
-        train_transform = self._build_train_transform(T)
-        val_transform = self._build_val_transform(T)
-
-        train_dataset = _CropDataset(
-            Path(dataset_path) / "crops", split="train",
-            transform=train_transform, seed=self._seed,
-        )
-        val_dataset = _CropDataset(
-            Path(dataset_path) / "crops", split="val",
-            transform=val_transform, seed=self._seed,
-        )
-
-        g = torch.Generator()
-        g.manual_seed(self._seed)
-
-        train_loader = DataLoader(
-            train_dataset,
-            batch_size=self._batch_size,
-            shuffle=True,
-            num_workers=2,
-            pin_memory=True,
-            generator=g,
-        )
-        val_loader = DataLoader(
-            val_dataset,
-            batch_size=self._batch_size,
-            shuffle=False,
-            num_workers=2,
-            pin_memory=True,
-        )
-        logger.info(
-            "Dataset: %d train / %d val samples",
-            len(train_dataset), len(val_dataset),
-        )
-        return train_loader, val_loader
-
-    def _build_train_transform(self, T):
-        ops = [T.Resize((224, 224))]
-        if self._aug_flip:
-            ops.append(T.RandomHorizontalFlip())
-        if self._aug_jitter:
-            ops.append(T.ColorJitter(brightness=0.2, contrast=0.2, saturation=0.2, hue=0.1))
-        if self._aug_affine:
-            ops.append(T.RandomAffine(
-                degrees=self._aug_degrees,
-                scale=(1.0 - self._aug_scale, 1.0 + self._aug_scale),
-            ))
-        ops += [T.ToTensor(), T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])]
-        return T.Compose(ops)
-
-    def _build_val_transform(self, T):
-        return T.Compose([
-            T.Resize((224, 224)),
-            T.ToTensor(),
-            T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
-        ])
-
-    # ------------------------------------------------------------------
-    # Training loop
-    # ------------------------------------------------------------------
-
-    def _run_training(self, model, train_loader, val_loader, version: str, torch, nn) -> Optional[str]:
-        device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-        model = model.to(device)
-        logger.info("Training on device: %s", device)
-
-        class_weights = self._compute_class_weights(train_loader, torch, device)
-        criterion = nn.CrossEntropyLoss(weight=class_weights)
-
-        optimizer = torch.optim.AdamW(
-            filter(lambda p: p.requires_grad, model.parameters()),
-            lr=self._lr,
-            weight_decay=self._weight_decay,
-            betas=(0.9, 0.999),
-        )
-        scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(
-            optimizer, mode="max", factor=self._lr_factor, patience=self._lr_patience,
-        )
-
-        best_f1 = -1.0
-        best_path: Optional[str] = None
-        no_improve = 0
-
-        for epoch in range(1, self._epochs + 1):
-            train_loss = self._train_epoch(model, train_loader, criterion, optimizer, device, torch)
-            val_f1 = self._val_epoch(model, val_loader, device, torch)
-            scheduler.step(val_f1)
-
-            ckpt_path = self._candidate_path / f"{version}_epoch_{epoch:02d}.pt"
-            torch.save(model.state_dict(), ckpt_path)
-            logger.info(
-                "Epoch %d/%d  train_loss=%.4f  val_f1=%.4f",
-                epoch, self._epochs, train_loss, val_f1,
-            )
-
-            if val_f1 > best_f1:
-                best_f1 = val_f1
-                best_path = str(self._candidate_path / f"{version}_best.pt")
-                torch.save(model.state_dict(), best_path)
-                no_improve = 0
-            else:
-                no_improve += 1
-                if no_improve >= self._patience:
-                    logger.info("Early stopping at epoch %d (patience=%d)", epoch, self._patience)
-                    break
-
-        logger.info("Training complete. Best val_f1=%.4f  checkpoint=%s", best_f1, best_path)
-        return best_path
-
-    @staticmethod
-    def _train_epoch(model, loader, criterion, optimizer, device, torch) -> float:
-        model.train()
-        total_loss = 0.0
-        for images, labels in loader:
-            images, labels = images.to(device), labels.to(device)
-            optimizer.zero_grad()
-            outputs = model(images)
-            loss = criterion(outputs, labels)
-            loss.backward()
-            optimizer.step()
-            total_loss += loss.item() * len(images)
-        return total_loss / max(len(loader.dataset), 1)
-
-    @staticmethod
-    def _val_epoch(model, loader, device, torch) -> float:
-        from sklearn.metrics import f1_score
-
-        model.eval()
-        all_preds, all_labels = [], []
-        with torch.no_grad():
-            for images, labels in loader:
-                images = images.to(device)
-                outputs = model(images)
-                preds = outputs.argmax(dim=1).cpu().numpy()
-                all_preds.extend(preds)
-                all_labels.extend(labels.numpy())
-
-        return float(f1_score(all_labels, all_preds, average="macro", zero_division=0))
-
-    @staticmethod
-    def _compute_class_weights(train_loader, torch, device):
-        """Compute inverse-frequency class weights to handle imbalance."""
-        counts = np.zeros(2, dtype=np.int64)
-        for _, labels in train_loader:
-            for lbl in labels.numpy():
-                counts[int(lbl)] += 1
-        total = counts.sum()
-        if total == 0 or counts.min() == 0:
-            return None
-        weights = total / (2.0 * counts.astype(np.float32))
-        return torch.tensor(weights, dtype=torch.float32, device=device)
-
-    # ------------------------------------------------------------------
-    # Misc
-    # ------------------------------------------------------------------
-
-    @staticmethod
-    def _assert_cuda_compatibility(torch):
-        if torch.cuda.is_available():
-            cuda_ver = torch.version.cuda or ""
-            torch_ver = torch.__version__
-            logger.info("CUDA version: %s  PyTorch: %s", cuda_ver, torch_ver)
-            major = int(cuda_ver.split(".")[0]) if cuda_ver else 0
-            if major > 11:
-                logger.warning(
-                    "CUDA %s detected; system requires CUDA ≤ 11.8 for Tesla P40", cuda_ver
-                )
-
-    @staticmethod
-    def _seed_everything(seed: int):
-        random.seed(seed)
-        np.random.seed(seed)
-        try:
-            import torch
-            torch.manual_seed(seed)
-            if torch.cuda.is_available():
-                torch.cuda.manual_seed_all(seed)
-                torch.backends.cudnn.deterministic = True
-                torch.backends.cudnn.benchmark = False
-        except ImportError:
-            pass
-
-
-# ------------------------------------------------------------------
-# Dataset helper
-# ------------------------------------------------------------------
-
-class _CropDataset:
-    """Reads face crop images from class_0/ and class_1/ subdirectories."""
-
-    _EXTENSIONS = {".jpg", ".jpeg", ".png"}
-
-    def __init__(self, crops_root: Path, split: str, transform, seed: int):
-        all_samples = self._discover(crops_root)
-        train_samples, val_samples = _split_samples(all_samples, 0.80, seed)
-        self._samples = train_samples if split == "train" else val_samples
-        self._transform = transform
-
-    def __len__(self) -> int:
-        return len(self._samples)
-
-    def __getitem__(self, idx: int):
-        from PIL import Image
-        path, label = self._samples[idx]
-        image = Image.open(path).convert("RGB")
-        if self._transform:
-            image = self._transform(image)
-        return image, label
-
-    def _discover(self, crops_root: Path):
-        samples = []
-        for label, subdir in ((0, "class_0"), (1, "class_1")):
-            d = crops_root / subdir
-            if d.is_dir():
-                for p in sorted(d.iterdir()):
-                    if p.suffix.lower() in self._EXTENSIONS:
-                        samples.append((p, label))
-        return samples
-
-
-def _split_samples(samples, train_frac: float, seed: int):
-    """Stratified split mirroring LabelIngestor logic."""
-    rng = random.Random(seed)
-    by_label: Dict[int, list] = {}
-    for path, label in samples:
-        by_label.setdefault(label, []).append((path, label))
-
-    train, val = [], []
-    for group in by_label.values():
-        shuffled = list(group)
-        rng.shuffle(shuffled)
-        split = max(1, int(len(shuffled) * train_frac))
-        train.extend(shuffled[:split])
-        val.extend(shuffled[split:])
-
-    return train, val

src/active_learning/validator.py

@@ -1,204 +0,0 @@
-"""
-Candidate model validation for active learning.
-
-Evaluates a candidate checkpoint on the held-out validation set,
-computes F1, ECE, accuracy, precision, and recall, then checks
-quality gates before promotion.
-"""
-
-import logging
-from pathlib import Path
-from typing import Dict, Optional, Tuple
-
-import numpy as np
-
-logger = logging.getLogger(__name__)
-
-
-class Validator:
-    """Evaluate a candidate model and enforce quality gates."""
-
-    def __init__(self, config, current_f1: float = 0.0):
-        al = config.get_section("active_learning")
-        val_cfg = al.get("validation", {})
-        self._min_f1_delta: float = float(val_cfg.get("min_f1_improvement", 0.02))
-        self._max_ece: float = float(val_cfg.get("max_ece", 0.08))
-        self._current_f1 = current_f1
-
-    # ------------------------------------------------------------------
-    # Public API
-    # ------------------------------------------------------------------
-
-    def validate(self, checkpoint_path: str, dataset_path: str) -> Dict:
-        """
-        Run validation and return a results dict including gate pass/fail.
-
-        Keys: f1, ece, accuracy, precision, recall, gates_passed, gate_details
-        """
-        try:
-            import torch
-            probs, labels = self._run_inference(checkpoint_path, dataset_path, torch)
-            return self._compute_metrics(probs, labels)
-        except ImportError as exc:
-            logger.error("PyTorch not available: %s", exc)
-            return self._empty_result()
-        except Exception as exc:
-            logger.error("Validation failed: %s", exc, exc_info=True)
-            return self._empty_result()
-
-    # ------------------------------------------------------------------
-    # Inference
-    # ------------------------------------------------------------------
-
-    def _run_inference(self, checkpoint_path: str, dataset_path: str, torch) -> Tuple[np.ndarray, np.ndarray]:
-        import torch.nn as nn
-        import torchvision.models as models
-        import torchvision.transforms as T
-        from torch.utils.data import DataLoader
-
-        device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-
-        model = self._load_checkpoint(checkpoint_path, nn, models, torch, device)
-        model.eval()
-
-        transform = T.Compose([
-            T.Resize((224, 224)),
-            T.ToTensor(),
-            T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
-        ])
-
-        from .trainer import _CropDataset, _split_samples
-        crops_root = Path(dataset_path) / "crops"
-        all_samples = _CropDataset._discover(None, crops_root)  # type: ignore[arg-type]
-        _, val_samples = _split_samples(all_samples, 0.80, seed=42)
-
-        if not val_samples:
-            raise ValueError("Validation set is empty")
-
-        val_dataset = _ValDataset(val_samples, transform)
-        loader = DataLoader(val_dataset, batch_size=32, shuffle=False, num_workers=2, pin_memory=True)
-
-        all_probs, all_labels = [], []
-        with torch.no_grad():
-            for images, batch_labels in loader:
-                images = images.to(device)
-                outputs = model(images)
-                probs_batch = torch.softmax(outputs, dim=1)[:, 1].cpu().numpy()
-                all_probs.extend(probs_batch)
-                all_labels.extend(batch_labels.numpy())
-
-        return np.array(all_probs, dtype=np.float32), np.array(all_labels, dtype=np.int32)
-
-    @staticmethod
-    def _load_checkpoint(checkpoint_path: str, nn, models, torch, device):
-        model = models.mobilenet_v3_small(weights=None)
-        model.classifier = nn.Sequential(
-            nn.Linear(576, 256),
-            nn.ReLU(),
-            nn.Dropout(0.3),
-            nn.Linear(256, 2),
-        )
-        state = torch.load(checkpoint_path, map_location=device)
-        model.load_state_dict(state)
-        return model.to(device)
-
-    # ------------------------------------------------------------------
-    # Metrics
-    # ------------------------------------------------------------------
-
-    def _compute_metrics(self, probs: np.ndarray, labels: np.ndarray) -> Dict:
-        from sklearn.metrics import (
-            accuracy_score, f1_score, precision_score, recall_score,
-        )
-
-        preds = (probs >= 0.5).astype(np.int32)
-        f1 = float(f1_score(labels, preds, average="macro", zero_division=0))
-        accuracy = float(accuracy_score(labels, preds))
-        precision = float(precision_score(labels, preds, average="macro", zero_division=0))
-        recall = float(recall_score(labels, preds, average="macro", zero_division=0))
-        ece = compute_ece(probs, labels)
-
-        delta_f1 = f1 - self._current_f1
-        gate_f1 = delta_f1 > self._min_f1_delta
-        gate_ece = ece < self._max_ece
-        gates_passed = gate_f1 and gate_ece
-
-        gate_details = {
-            "delta_f1": round(delta_f1, 4),
-            "min_f1_improvement": self._min_f1_delta,
-            "gate_f1_passed": gate_f1,
-            "ece": round(ece, 4),
-            "max_ece": self._max_ece,
-            "gate_ece_passed": gate_ece,
-        }
-
-        logger.info(
-            "Validation results: f1=%.4f delta_f1=%.4f ece=%.4f gates_passed=%s",
-            f1, delta_f1, ece, gates_passed,
-        )
-
-        return {
-            "f1": round(f1, 4),
-            "ece": round(ece, 4),
-            "accuracy": round(accuracy, 4),
-            "precision": round(precision, 4),
-            "recall": round(recall, 4),
-            "gates_passed": gates_passed,
-            "gate_details": gate_details,
-        }
-
-    def _empty_result(self) -> Dict:
-        return {
-            "f1": 0.0, "ece": 1.0, "accuracy": 0.0,
-            "precision": 0.0, "recall": 0.0,
-            "gates_passed": False,
-            "gate_details": {},
-        }
-
-
-# ------------------------------------------------------------------
-# ECE calculation (module-level for reuse in tests)
-# ------------------------------------------------------------------
-
-def compute_ece(predictions: np.ndarray, labels: np.ndarray, n_bins: int = 15) -> float:
-    """
-    Expected Calibration Error.
-
-    Bins predictions by confidence and measures average |confidence - accuracy|
-    weighted by bin population.
-    """
-    bin_boundaries = np.linspace(0, 1, n_bins + 1)
-    ece = 0.0
-    n = len(predictions)
-    if n == 0:
-        return 0.0
-
-    for i in range(n_bins):
-        mask = (predictions >= bin_boundaries[i]) & (predictions < bin_boundaries[i + 1])
-        if mask.sum() > 0:
-            bin_confidence = float(predictions[mask].mean())
-            bin_accuracy = float(labels[mask].mean())
-            ece += (mask.sum() / n) * abs(bin_confidence - bin_accuracy)
-
-    return float(ece)
-
-
-# ------------------------------------------------------------------
-# Minimal dataset wrapper for validation
-# ------------------------------------------------------------------
-
-class _ValDataset:
-    def __init__(self, samples, transform):
-        self._samples = samples
-        self._transform = transform
-
-    def __len__(self) -> int:
-        return len(self._samples)
-
-    def __getitem__(self, idx: int):
-        from PIL import Image
-        path, label = self._samples[idx]
-        image = Image.open(path).convert("RGB")
-        if self._transform:
-            image = self._transform(image)
-        return image, label

src/aggregator.py

@@ -1,39 +0,0 @@
-"""Confidence aggregation strategies for video-level scoring."""
-
-import logging
-from typing import List
-
-import numpy as np
-
-logger = logging.getLogger(__name__)
-
-
-def aggregate(
-    confidences: List[float],
-    strategy: str = "max",
-    alpha: float = 1.0,
-    beta: float = 0.1,
-    top_k: int = 3,
-) -> float:
-    """Aggregate per-crop confidences into a single video-level score C ∈ [0,1]."""
-    if not confidences:
-        return 0.0
-
-    arr = np.array(confidences, dtype=np.float64)
-
-    if strategy == "max":
-        return float(arr.max())
-
-    if strategy == "weighted_mean":
-        mean = float(arr.mean())
-        var  = float(arr.var())
-        raw  = alpha * mean + beta * var
-        # Clamp to [0,1]; the formula is a linear combination, not a softmax
-        return float(np.clip(raw, 0.0, 1.0))
-
-    if strategy == "top_k_mean":
-        k = max(1, min(top_k, len(arr)))
-        return float(np.partition(arr, -k)[-k:].mean())
-
-    logger.warning("Unknown aggregation strategy '%s', falling back to max", strategy)
-    return float(arr.max())

src/batcher.py

@@ -1,68 +0,0 @@
-"""Dynamic batching utilities for face detection inference."""
-
-import logging
-import time
-from collections import deque
-from typing import Callable, Deque, Generic, List, Optional, TypeVar
-
-from gpu_manager import GPUMemoryManager
-
-logger = logging.getLogger(__name__)
-
-T = TypeVar("T")
-R = TypeVar("R")
-
-
-class DynamicBatcher(Generic[T, R]):
-    """Accumulate items into batches and flush based on size or timeout."""
-
-    def __init__(
-        self,
-        process_fn: Callable[[List[T]], List[R]],
-        gpu_manager: Optional[GPUMemoryManager] = None,
-        max_batch_size: int = 16,
-        batch_timeout_ms: float = 100.0,
-        min_batch_size: int = 1,
-    ):
-        self.process_fn = process_fn
-        self.gpu_manager = gpu_manager
-        self.max_batch_size = max_batch_size
-        self.batch_timeout_ms = batch_timeout_ms
-        self.min_batch_size = min_batch_size
-        self._queue: Deque[T] = deque()
-        self._last_flush = time.monotonic()
-
-    def add(self, item: T) -> List[R]:
-        """Add an item and return any results if a batch was flushed."""
-        self._queue.append(item)
-        if len(self._queue) >= self._current_batch_size():
-            return self.flush()
-
-        elapsed_ms = (time.monotonic() - self._last_flush) * 1000
-        if elapsed_ms >= self.batch_timeout_ms and len(self._queue) >= self.min_batch_size:
-            return self.flush()
-        return []
-
-    def flush(self) -> List[R]:
-        """Flush all queued items through the processor."""
-        if not self._queue:
-            return []
-
-        batch = [self._queue.popleft() for _ in range(min(len(self._queue), self._current_batch_size()))]
-        results = self.process_fn(batch)
-        self._last_flush = time.monotonic()
-
-        if self.gpu_manager is not None:
-            self.gpu_manager.adjust_batch_size()
-            self.gpu_manager.empty_cache()
-
-        return results
-
-    def _current_batch_size(self) -> int:
-        if self.gpu_manager is not None:
-            return min(self.gpu_manager.current_batch_size, self.max_batch_size)
-        return self.max_batch_size
-
-    @property
-    def queued_count(self) -> int:
-        return len(self._queue)

src/classifier.py

@@ -1,138 +0,0 @@
-"""
-Face crop classifier.
-
-Loads a MobileNetV3 model via ONNX/TensorRT FP32, applies temperature-scaled
-softmax, and returns per-crop probabilities for the target class.
-"""
-
-import logging
-from pathlib import Path
-from typing import Any, List, Optional
-
-import numpy as np
-
-logger = logging.getLogger(__name__)
-
-# ImageNet normalisation constants
-_MEAN = np.array([0.485, 0.456, 0.406], dtype=np.float32)
-_STD  = np.array([0.229, 0.224, 0.225], dtype=np.float32)
-
-
-def calibrated_softmax(logits: np.ndarray, temperature: float = 1.0) -> np.ndarray:
-    """Softmax with temperature scaling; returns class probabilities."""
-    scaled = logits / max(temperature, 1e-8)
-    shifted = scaled - scaled.max(axis=-1, keepdims=True)   # numerical stability
-    exp = np.exp(shifted)
-    return exp / exp.sum(axis=-1, keepdims=True)
-
-
-class FaceClassifier:
-    """Classify face crops and return calibrated target-class probabilities."""
-
-    def __init__(
-        self,
-        engine_path: str,
-        temperature: float = 1.0,
-        input_size: int = 224,
-        device: str = "cuda",
-    ):
-        self.engine_path = engine_path
-        self.temperature = temperature
-        self.input_size = input_size
-        self.device = device
-        self._session: Optional[Any] = None
-        self._load_model()
-
-    def _load_model(self):
-        path = Path(self.engine_path)
-        if not path.exists():
-            logger.warning("Classifier engine not found at %s; using placeholder", self.engine_path)
-            return
-
-        suffix = path.suffix.lower()
-        if suffix == ".onnx":
-            try:
-                import onnxruntime as ort
-                providers = (["CUDAExecutionProvider"] if self.device.startswith("cuda")
-                             else ["CPUExecutionProvider"])
-                self._session = ort.InferenceSession(str(path), providers=providers)
-                logger.info("Loaded ONNX classifier from %s", self.engine_path)
-            except Exception as exc:
-                logger.warning("Failed to load ONNX classifier: %s", exc)
-        elif suffix in (".trt", ".engine", ".plan"):
-            try:
-                import tensorrt as trt
-                with trt.Logger() as trt_logger, open(path, "rb") as f:
-                    runtime = trt.Runtime(trt_logger)
-                    self._session = runtime.deserialize_cuda_engine(f.read())
-                logger.info("Loaded TensorRT classifier from %s", self.engine_path)
-            except Exception as exc:
-                logger.warning("Failed to load TensorRT classifier: %s", exc)
-        else:
-            logger.warning("Unsupported classifier format: %s", suffix)
-
-    def classify(self, crop_paths: List[str], batch_size: int = 16) -> List[float]:
-        """Return target-class probabilities (p ∈ [0,1]) for each crop path."""
-        results: List[float] = []
-        for i in range(0, len(crop_paths), batch_size):
-            batch = crop_paths[i : i + batch_size]
-            results.extend(self._classify_batch(batch))
-        return results
-
-    def _classify_batch(self, crop_paths: List[str]) -> List[float]:
-        preprocessed = []
-        for path in crop_paths:
-            try:
-                preprocessed.append(self._preprocess(path))
-            except Exception as exc:
-                logger.warning("Preprocessing failed for %s: %s", path, exc)
-                preprocessed.append(np.zeros((3, self.input_size, self.input_size), dtype=np.float32))
-
-        batch = np.stack(preprocessed, axis=0)  # (N, 3, H, W)
-        logits = self._infer(batch)             # (N, 2)
-        probs = calibrated_softmax(logits, self.temperature)
-        return probs[:, 1].tolist()             # target-class column
-
-    def _preprocess(self, crop_path: str) -> np.ndarray:
-        from PIL import Image
-        image = Image.open(crop_path).convert("RGB")
-        if image.size != (self.input_size, self.input_size):
-            image = image.resize((self.input_size, self.input_size), Image.Resampling.BILINEAR)
-        arr = np.array(image, dtype=np.float32) / 255.0
-        arr = (arr - _MEAN) / _STD
-        image.close()
-        return np.transpose(arr, (2, 0, 1))    # HWC → CHW
-
-    def _infer(self, batch: np.ndarray) -> np.ndarray:
-        if self._session is None:
-            return self._placeholder_logits(batch.shape[0])
-
-        try:
-            if hasattr(self._session, "run"):
-                input_name = self._session.get_inputs()[0].name
-                output = self._session.run(None, {input_name: batch})
-                return np.array(output[0])
-
-            # TensorRT path
-            import pycuda.autoinit  # noqa: F401
-            import pycuda.driver as cuda
-
-            context = self._session.create_execution_context()
-            out_shape = (batch.shape[0], 2)
-            d_in  = cuda.mem_alloc(batch.nbytes)
-            d_out = cuda.mem_alloc(np.prod(out_shape) * np.dtype(np.float32).itemsize)
-            stream = cuda.Stream()
-            cuda.memcpy_htod_async(d_in, batch, stream)
-            context.execute_async_v2(bindings=[int(d_in), int(d_out)], stream_handle=stream.handle)
-            output = np.empty(out_shape, dtype=np.float32)
-            cuda.memcpy_dtoh_async(output, d_out, stream)
-            stream.synchronize()
-            return output
-        except Exception as exc:
-            logger.warning("Classifier inference failed: %s", exc)
-            return self._placeholder_logits(batch.shape[0])
-
-    @staticmethod
-    def _placeholder_logits(n: int) -> np.ndarray:
-        """Return neutral logits when no model is loaded."""
-        return np.zeros((n, 2), dtype=np.float32)

src/codec_validator.py

@@ -1,107 +0,0 @@
-"""
-Codec Validator
-
-Validates video codecs against a whitelist/blacklist.
-Flags unsupported codecs as UNSCANNABLE.
-"""
-
-import logging
-from typing import Set
-
-logger = logging.getLogger(__name__)
-
-# Default supported codecs
-DEFAULT_WHITELIST: Set[str] = {
-    "avc1",       # H.264
-    "hevc",       # H.265
-    "vp8",        # VP8
-    "vp9",        # VP9
-    "av01",       # AV1
-    "mjpeg",      # MJPEG
-    "mp4v",       # MPEG-4
-    "h264",       # H.264 (alternative name)
-    "h265",       # H.265 (alternative name)
-}
-
-# Known unsupported codecs
-DEFAULT_BLACKLIST: Set[str] = {
-    "theora",     # Theora
-    "divx",       # DivX
-    "xvid",       # Xvid
-    "prores",     # ProRes (requires special handling)
-    "avc",        # Old AVC naming
-    "hev1",       # HEVC alternative
-}
-
-
-class CodecValidator:
-    """Validate video codecs against whitelist/blacklist."""
-
-    def __init__(
-        self,
-        whitelist: Set[str] = None,
-        blacklist: Set[str] = None,
-        default_status_on_error: str = "UNSCANNABLE",
-    ):
-        self.whitelist = whitelist or DEFAULT_WHITELIST
-        self.blacklist = blacklist or DEFAULT_BLACKLIST
-        self.default_status_on_error = default_status_on_error
-
-    def is_supported(self, codec: str) -> bool:
-        """Check if a codec is supported."""
-        codec_lower = codec.lower().strip()
-
-        # Check blacklist first
-        if codec_lower in self.blacklist:
-            logger.info("Codec '%s' is blacklisted", codec)
-            return False
-
-        # Check whitelist
-        if codec_lower in self.whitelist:
-            return True
-
-        # Not in either list - assume unsupported
-        logger.warning("Codec '%s' not in whitelist or blacklist", codec)
-        return False
-
-    def validate(self, codec: str) -> tuple:
-        """
-        Validate a codec and return (is_supported, reason).
-
-        Returns:
-            (True, "supported") if codec is in whitelist
-            (False, "blacklisted") if codec is in blacklist
-            (False, "not_in_whitelist") if codec is unknown
-        """
-        codec_lower = codec.lower().strip()
-
-        if codec_lower in self.blacklist:
-            return False, "blacklisted"
-
-        if codec_lower in self.whitelist:
-            return True, "supported"
-
-        return False, "not_in_whitelist"
-
-    def add_to_whitelist(self, codec: str):
-        """Add a codec to the whitelist."""
-        self.whitelist.add(codec.lower().strip())
-        logger.info("Added codec '%s' to whitelist", codec)
-
-    def add_to_blacklist(self, codec: str):
-        """Add a codec to the blacklist."""
-        self.blacklist.add(codec.lower().strip())
-        logger.info("Added codec '%s' to blacklist", codec)
-
-    def get_status(self, codec: str) -> str:
-        """Get the status for a codec (for DB update)."""
-        is_supported, reason = self.validate(codec)
-        if is_supported:
-            return "PENDING"
-        return self.default_status_on_error
-
-    def __repr__(self):
-        return (
-            f"CodecValidator(whitelist={len(self.whitelist)}, "
-            f"blacklist={len(self.blacklist)})"
-        )

src/config_loader.py

@@ -1,126 +0,0 @@
-"""
-Configuration loader for VideoDetect.
-
-Loads config.yaml with environment variable overrides.
-Provides typed accessors and validates all required fields.
-"""
-
-import logging
-import os
-from pathlib import Path
-from typing import Any, Dict, Optional
-
-import yaml
-
-logger = logging.getLogger(__name__)
-
-# Default config path
-DEFAULT_CONFIG_PATH = Path(__file__).parent.parent / "config.yaml"
-
-
-class Config:
-    """Typed configuration manager with environment variable overrides."""
-
-    def __init__(self, config_path: Optional[str] = None):
-        self._path = Path(config_path) if config_path else DEFAULT_CONFIG_PATH
-        self._data: Dict[str, Any] = {}
-        self._load()
-
-    def _load(self):
-        """Load config from YAML file."""
-        if not self._path.exists():
-            raise FileNotFoundError(f"Config file not found: {self._path}")
-
-        with open(self._path, "r") as f:
-            self._data = yaml.safe_load(f) or {}
-
-        # Apply environment variable overrides
-        self._apply_env_overrides()
-        logger.info("Config loaded from %s", self._path)
-
-    def _apply_env_overrides(self):
-        """Override config values with environment variables."""
-        overrides = {
-            "DB_HOST": ("database.host", None),
-            "DB_PORT": ("database.port", int),
-            "DB_NAME": ("database.name", None),
-            "DB_USER": ("database.user", None),
-            "DB_PASSWORD": ("database.password", None),
-            "DB_POOL_SIZE": ("database.pool_size", int),
-            "DB_POOL_MIN": ("database.pool_min", int),
-            "GPU_MAX_MEMORY_GB": ("gpu.max_memory_gb", float),
-            "SAMPLING_INTERVAL_SECONDS": ("sampling.interval_seconds", int),
-            "T_HIGH": ("thresholds.T_high", float),
-            "T_LOW": ("thresholds.T_low", float),
-            "CONFIG_PATH": (None, None),  # handled separately
-        }
-
-        for env_key, (config_path, type_fn) in overrides.items():
-            env_val = os.environ.get(env_key)
-            if env_val is not None:
-                if config_path is None:
-                    continue
-                if type_fn is not None:
-                    env_val = type_fn(env_val)
-                self._set_nested(self._data, config_path, env_val)
-                logger.debug("Config override: %s=%s (from %s)", config_path, env_val, env_key)
-
-    @staticmethod
-    def _set_nested(data: Dict, path: str, value):
-        """Set a value in a nested dict using dot notation."""
-        keys = path.split(".")
-        d = data
-        for key in keys[:-1]:
-            d = d.setdefault(key, {})
-        d[keys[-1]] = value
-
-    def get(self, path: str, default=None):
-        """Get a config value using dot notation."""
-        keys = path.split(".")
-        d = self._data
-        for key in keys:
-            if isinstance(d, dict):
-                d = d.get(key, default)
-            else:
-                return default
-        return d if d is not None else default
-
-    def get_section(self, section: str) -> Dict:
-        """Get an entire config section as a dict."""
-        return self._data.get(section, {})
-
-    def validate(self, required_keys: list) -> list:
-        """Validate that required config keys exist. Returns list of missing keys."""
-        missing = []
-        for key in required_keys:
-            if self.get(key) is None:
-                missing.append(key)
-        if missing:
-            logger.error("Missing required config keys: %s", missing)
-        return missing
-
-    @property
-    def data(self) -> Dict[str, Any]:
-        """Access raw config data."""
-        return self._data
-
-    def __repr__(self):
-        return f"Config(path={self._path}, sections={list(self._data.keys())})"
-
-
-# Module-level singleton
-_config: Optional[Config] = None
-
-
-def get_config(config_path: Optional[str] = None) -> Config:
-    """Get or create the global config singleton."""
-    global _config
-    if _config is None:
-        _config = Config(config_path)
-    return _config
-
-
-def reset_config():
-    """Reset the config singleton (useful for testing)."""
-    global _config
-    _config = None

src/crash_recovery.py

@@ -1,137 +0,0 @@
-"""
-Crash recovery and idempotent processing guarantees.
-
-On startup: re-queues any video stuck in PROCESSING past the lock timeout.
-During processing: saves per-video checkpoint JSON to scratch for resume.
-"""
-
-import json
-import logging
-from datetime import datetime, timezone
-from pathlib import Path
-from typing import Dict, List, Optional
-
-logger = logging.getLogger(__name__)
-
-_CHECKPOINT_DIR = "checkpoints"
-
-
-class CrashRecovery:
-    """Handle crash recovery, checkpointing, and idempotent re-queueing."""
-
-    def __init__(self, db_connector, config):
-        self._db = db_connector
-        scratch = config.get("storage.scratch_path", "/scratch")
-        self._checkpoint_dir = Path(scratch) / _CHECKPOINT_DIR
-        mon = config.get_section("monitoring")
-        cr = mon.get("crash_recovery", {})
-        self._lock_timeout_minutes: int = int(cr.get("lock_timeout_minutes", 5))
-        self._auto_requeue: bool = bool(cr.get("auto_requeue", True))
-
-    # ------------------------------------------------------------------
-    # Startup recovery
-    # ------------------------------------------------------------------
-
-    def recover_on_startup(self) -> int:
-        """
-        Re-queue videos stuck in PROCESSING past the lock timeout.
-
-        Returns the count of videos re-queued.
-        """
-        if not self._auto_requeue:
-            logger.info("auto_requeue disabled; skipping crash recovery")
-            return 0
-
-        try:
-            result = self._db.execute(
-                """UPDATE videos
-                   SET status = 'PENDING', updated_at = NOW()
-                   WHERE status = 'PROCESSING'
-                     AND updated_at < NOW() - INTERVAL %s MINUTE""",
-                (self._lock_timeout_minutes,),
-            )
-            count = result if isinstance(result, int) else 0
-            if count:
-                logger.warning("Crash recovery: re-queued %d stuck PROCESSING videos", count)
-            else:
-                logger.info("Crash recovery: no stuck videos found")
-            return count
-        except Exception as exc:
-            logger.error("Crash recovery query failed: %s", exc)
-            return 0
-
-    def list_stuck_videos(self) -> List[Dict]:
-        """Return videos currently stuck past the lock timeout (read-only)."""
-        try:
-            rows = self._db.fetchall(
-                """SELECT id, file_path, updated_at
-                   FROM videos
-                   WHERE status = 'PROCESSING'
-                     AND updated_at < NOW() - INTERVAL %s MINUTE""",
-                (self._lock_timeout_minutes,),
-            )
-            return rows or []
-        except Exception as exc:
-            logger.error("Failed to list stuck videos: %s", exc)
-            return []
-
-    # ------------------------------------------------------------------
-    # Per-video checkpointing
-    # ------------------------------------------------------------------
-
-    def save_checkpoint(self, video_id: int, state: str, progress: Dict) -> None:
-        """Persist a checkpoint JSON for the given video."""
-        self._checkpoint_dir.mkdir(parents=True, exist_ok=True)
-        checkpoint = {
-            "video_id": video_id,
-            "state": state,
-            "progress": progress,
-            "timestamp": datetime.now(timezone.utc).isoformat(),
-        }
-        path = self._checkpoint_dir / f"{video_id}.json"
-        try:
-            path.write_text(json.dumps(checkpoint, indent=2), encoding="utf-8")
-            logger.debug("Saved checkpoint for video %d (state=%s)", video_id, state)
-        except Exception as exc:
-            logger.warning("Failed to save checkpoint for video %d: %s", video_id, exc)
-
-    def load_checkpoint(self, video_id: int) -> Optional[Dict]:
-        """Return the stored checkpoint for video_id, or None if absent."""
-        path = self._checkpoint_dir / f"{video_id}.json"
-        if not path.exists():
-            return None
-        try:
-            return json.loads(path.read_text(encoding="utf-8"))
-        except Exception as exc:
-            logger.warning("Failed to read checkpoint for video %d: %s", video_id, exc)
-            return None
-
-    def delete_checkpoint(self, video_id: int) -> None:
-        """Remove the checkpoint file after successful processing."""
-        path = self._checkpoint_dir / f"{video_id}.json"
-        if path.exists():
-            try:
-                path.unlink()
-                logger.debug("Deleted checkpoint for video %d", video_id)
-            except Exception as exc:
-                logger.warning("Failed to delete checkpoint for video %d: %s", video_id, exc)
-
-    # ------------------------------------------------------------------
-    # Idempotency guard
-    # ------------------------------------------------------------------
-
-    def is_already_completed(self, video_id: int) -> bool:
-        """
-        Return True if this video already has a COMPLETED status.
-
-        Prevents duplicate processing if a job is accidentally re-submitted.
-        """
-        try:
-            rows = self._db.fetchall(
-                "SELECT status FROM videos WHERE id = %s",
-                (video_id,),
-            )
-            return bool(rows and rows[0].get("status") == "COMPLETED")
-        except Exception as exc:
-            logger.warning("Idempotency check failed for video %d: %s", video_id, exc)
-            return False

src/data_export.py

@@ -1,115 +0,0 @@
-"""Batch Parquet/JSONL export of video processing results."""
-
-import json
-import logging
-import os
-from datetime import datetime, timezone
-from pathlib import Path
-from typing import Any, Dict, List, Optional
-
-logger = logging.getLogger(__name__)
-
-
-class DataExporter:
-    """Buffer result records and flush to Parquet or JSONL when the batch is full."""
-
-    def __init__(
-        self,
-        output_path: str,
-        model_version: str = "v0.0.0",
-        export_format: str = "parquet",
-        compression: str = "snappy",
-        batch_size: int = 100,
-        include_frame_confidences: bool = True,
-    ):
-        self.output_path = Path(output_path) / model_version
-        self.model_version = model_version
-        self.export_format = export_format.lower()
-        self.compression = compression
-        self.batch_size = batch_size
-        self.include_frame_confidences = include_frame_confidences
-        self._buffer: List[Dict[str, Any]] = []
-        self.output_path.mkdir(parents=True, exist_ok=True)
-
-    def add(self, record: Dict[str, Any]) -> None:
-        """Buffer one result record; flush automatically when batch is full."""
-        self._buffer.append(record)
-        if len(self._buffer) >= self.batch_size:
-            self.flush()
-
-    def flush(self) -> Optional[str]:
-        """Write buffered records to disk; returns the output file path or None."""
-        if not self._buffer:
-            return None
-
-        records = self._buffer[:]
-        self._buffer.clear()
-
-        if not self.include_frame_confidences:
-            for r in records:
-                r.pop("confidence_scores", None)
-
-        batch_id = datetime.now(timezone.utc).strftime("%Y%m%d_%H%M%S")
-        stem = f"{self.model_version}_{batch_id}"
-
-        if self.export_format in ("parquet", "both"):
-            path = self._write_parquet(records, stem)
-        if self.export_format in ("jsonl", "both"):
-            path = self._write_jsonl(records, stem)
-        if self.export_format not in ("parquet", "jsonl", "both"):
-            logger.warning("Unknown export format '%s'; defaulting to jsonl", self.export_format)
-            path = self._write_jsonl(records, stem)
-
-        return str(path)
-
-    def _write_parquet(self, records: List[Dict[str, Any]], stem: str) -> Path:
-        out = self.output_path / f"{stem}.parquet"
-        try:
-            import pyarrow as pa
-            import pyarrow.parquet as pq
-
-            schema = pa.schema([
-                pa.field("video_id",          pa.int64()),
-                pa.field("file_path",         pa.string()),
-                pa.field("model_version",     pa.string()),
-                pa.field("sample_count",      pa.int32()),
-                pa.field("confidence_scores", pa.list_(pa.float64())),
-                pa.field("video_confidence",  pa.float64()),
-                pa.field("routing",           pa.string()),
-                pa.field("processed_at",      pa.string()),
-            ])
-
-            table = pa.table(
-                {
-                    "video_id":          [r.get("video_id") for r in records],
-                    "file_path":         [r.get("file_path", "") for r in records],
-                    "model_version":     [r.get("model_version", self.model_version) for r in records],
-                    "sample_count":      [r.get("sample_count", 0) for r in records],
-                    "confidence_scores": [r.get("confidence_scores", []) for r in records],
-                    "video_confidence":  [float(r.get("video_confidence", 0.0)) for r in records],
-                    "routing":           [r.get("routing", "SKIP") for r in records],
-                    "processed_at":      [r.get("processed_at", "") for r in records],
-                },
-                schema=schema,
-            )
-            pq.write_table(table, out, compression=self.compression)
-            logger.info("Exported %d records to %s", len(records), out)
-        except ImportError:
-            logger.warning("pyarrow not available; falling back to JSONL")
-            out = self._write_jsonl(records, stem.replace(".parquet", ""))
-        return out
-
-    def _write_jsonl(self, records: List[Dict[str, Any]], stem: str) -> Path:
-        out = self.output_path / f"{stem}.jsonl"
-        with open(out, "w", encoding="utf-8") as fh:
-            for record in records:
-                fh.write(json.dumps(record, default=str) + "\n")
-        logger.info("Exported %d records to %s", len(records), out)
-        return out
-
-    def __del__(self):
-        if self._buffer:
-            try:
-                self.flush()
-            except Exception:
-                pass

src/db_connector.py

@@ -1,227 +0,0 @@
-"""
-Database connection layer with connection pooling for VideoDetect.
-
-Provides:
-- Connection pooling via DBUtils + PyMySQL
-- Automatic reconnection on disconnect (ping + retry on transient errors)
-- Context manager support
-- Prepared statements for all queries
-- Transaction support for atomic state transitions
-"""
-
-import contextlib
-import logging
-import time
-from typing import Optional
-
-from dbutils.pooled_db import PooledDB
-import pymysql
-
-logger = logging.getLogger(__name__)
-
-
-class DBConnector:
-    """Thread-safe database connection pool manager."""
-
-    # Transient errors that indicate a dropped/stale connection and are safe
-    # to retry with a fresh connection from the pool.
-    _RETRYABLE_ERRORS = (pymysql.err.InterfaceError, pymysql.err.OperationalError)
-    _MAX_RETRIES = 3
-    _RETRY_BACKOFF_SECONDS = 0.5
-
-    def __init__(
-        self,
-        host: str = "mariadb",
-        port: int = 3306,
-        database: str = "videodetect",
-        user: str = "videodetect",
-        password: str = "videodetect123",
-        pool_size: int = 20,
-        pool_min: int = 5,
-        pool_recycle: int = 3600,
-    ):
-        # PooledDB parameters
-        pool_config = {
-            "creator": pymysql,
-            "maxconnections": pool_size,
-            "mincached": pool_min,
-            "maxcached": pool_size,
-            "maxusage": 200,
-            "blocking": True,
-            # ping(0) = check if connection is alive BEFORE returning it from
-            # the pool. This is the correct DBUtils mechanism for detecting
-            # dead/stale connections (idle timeout, network blip, DB restart).
-            "ping": 0,
-        }
-
-        # PyMySQL connection parameters
-        connection_params = {
-            "host": host,
-            "port": port,
-            "database": database,
-            "user": user,
-            "password": password,
-            "charset": "utf8mb4",
-            "cursorclass": pymysql.cursors.DictCursor,
-            "read_timeout": 30,
-            "write_timeout": 30,
-        }
-
-        self._pool = PooledDB(**pool_config, **connection_params)
-        self._pool_recycle = pool_recycle
-        logger.info(
-            "DB pool initialized: host=%s db=%s pool_size=%d min=%d ping=0",
-            host, database, pool_size, pool_min,
-        )
-
-    def get_connection(self):
-        """Get a connection from the pool."""
-        return self._pool.connection()
-
-    def _run_query(self, query: str, params, transaction: bool, fetch: str):
-        """Run a single query on a fresh connection, with retry on transient errors.
-
-        A pooled connection can be dropped by the server (idle timeout, network
-        blip, DB restart). When that happens the query fails with an
-        InterfaceError/OperationalError. We discard the dead connection and
-        retry with a fresh one from the pool.
-
-        Args:
-            query: SQL statement.
-            params: Bound parameters (or None).
-            transaction: If True, wrap the query in a transaction.
-            fetch: One of "none", "one", or "all" controlling the return value.
-
-        Returns:
-            - "none": the affected row count.
-            - "one": a single row (dict) or None.
-            - "all": a list of rows (dicts).
-        """
-        last_exc = None
-        for attempt in range(1, self._MAX_RETRIES + 1):
-            conn = self.get_connection()
-            try:
-                cursor = conn.cursor()
-                if transaction:
-                    conn.begin()
-                cursor.execute(query, params or ())
-                if fetch == "one":
-                    result = cursor.fetchone()
-                elif fetch == "all":
-                    result = cursor.fetchall()
-                else:
-                    result = cursor.rowcount
-                if transaction:
-                    conn.commit()
-                cursor.close()
-                conn.close()
-                return result
-            except self._RETRYABLE_ERRORS as e:
-                last_exc = e
-                logger.warning(
-                    "Transient DB error (attempt %d/%d): %s",
-                    attempt, self._MAX_RETRIES, e,
-                )
-                # Discard the dead connection; a fresh one is taken on retry.
-                self._safe_close(conn)
-                if attempt < self._MAX_RETRIES:
-                    time.sleep(self._RETRY_BACKOFF_SECONDS * attempt)
-            except Exception:
-                # Non-transient error: roll back if in a transaction and re-raise.
-                if transaction:
-                    self._safe_rollback(conn)
-                self._safe_close(conn)
-                raise
-        raise last_exc
-
-    @staticmethod
-    def _safe_close(conn):
-        try:
-            conn.close()
-        except Exception:
-            pass
-
-    @staticmethod
-    def _safe_rollback(conn):
-        try:
-            conn.rollback()
-        except Exception:
-            pass
-
-    @contextlib.contextmanager
-    def get_cursor(self, transaction: bool = False):
-        """Context manager for getting a cursor with optional transaction support.
-
-        Note: this does not retry on transient errors; use execute/fetchone/
-        fetchall for retry-safe single-statement queries.
-        """
-        conn = self.get_connection()
-        cursor = None
-        try:
-            cursor = conn.cursor()
-            if transaction:
-                conn.begin()
-            yield cursor
-            if transaction:
-                conn.commit()
-        except Exception:
-            if transaction:
-                self._safe_rollback(conn)
-            raise
-        finally:
-            if cursor is not None:
-                try:
-                    cursor.close()
-                except Exception:
-                    pass
-            self._safe_close(conn)
-
-    @contextlib.contextmanager
-    def transaction(self):
-        """Context manager for a full transaction."""
-        conn = self.get_connection()
-        try:
-            conn.begin()
-            yield conn
-            conn.commit()
-        except Exception:
-            self._safe_rollback(conn)
-            raise
-        finally:
-            self._safe_close(conn)
-
-    def execute(self, query: str, params=None, transaction: bool = False):
-        """Execute a query and return affected rows. Retries on transient errors."""
-        return self._run_query(query, params, transaction, fetch="none")
-
-    def fetchone(self, query: str, params=None):
-        """Execute a query and return one row. Retries on transient errors."""
-        return self._run_query(query, params, transaction=False, fetch="one")
-
-    def fetchall(self, query: str, params=None):
-        """Execute a query and return all rows. Retries on transient errors."""
-        return self._run_query(query, params, transaction=False, fetch="all")
-
-    def initialize_schema(self, schema_path: str = "db/schema.sql"):
-        """Initialize the database schema from SQL file."""
-        with open(schema_path, "r") as f:
-            sql = f.read()
-        # Split on semicolons and execute each statement
-        statements = [s.strip() for s in sql.split(";") if s.strip()]
-        for stmt in statements:
-            if stmt.startswith("--"):
-                continue
-            try:
-                self.execute(stmt)
-            except Exception as e:
-                logger.debug("Schema statement skipped (may already exist): %s", e)
-        logger.info("Schema initialized from %s", schema_path)
-
-    def health_check(self) -> bool:
-        """Check if the database is reachable."""
-        try:
-            result = self.fetchone("SELECT 1")
-            return result is not None
-        except Exception as e:
-            logger.error("Health check failed: %s", e)
-            return False

src/drift_detector.py

@@ -1,219 +0,0 @@
-"""
-Confidence distribution drift detection.
-
-Compares recent video confidence scores against a stored baseline and fires
-alerts when the distribution shifts beyond configured thresholds.
-"""
-
-import logging
-from datetime import datetime, timedelta, timezone
-from typing import Callable, Dict, List, Optional
-
-logger = logging.getLogger(__name__)
-
-
-class DriftDetector:
-    """Detect and alert on confidence distribution drift."""
-
-    def __init__(self, db_connector, config, alert_fn: Optional[Callable[[str], None]] = None):
-        self._db = db_connector
-        mon = config.get_section("monitoring")
-        alerts = mon.get("alerts", {})
-        self._drift_threshold: float = float(alerts.get("confidence_drift_threshold", 0.10))
-        self._review_max_size: int = int(alerts.get("review_queue_max_size", 1000))
-        self._review_max_age_h: int = int(alerts.get("review_queue_max_age_hours", 24))
-        self._throughput_min: float = float(alerts.get("throughput_min_videos_per_hour", 20))
-        self._throughput_window_h: int = int(alerts.get("throughput_min_duration_hours", 1))
-        self._error_rate_threshold: float = float(alerts.get("error_rate_threshold", 0.05))
-        self._error_rate_window_h: int = int(alerts.get("error_rate_window_hours", 1))
-        self._alert_fn = alert_fn or _log_alert
-
-    # ------------------------------------------------------------------
-    # Public API
-    # ------------------------------------------------------------------
-
-    def run_all_checks(self) -> Dict[str, bool]:
-        """
-        Run all alert checks and return a dict of {check_name: triggered}.
-
-        Fires the alert function for each triggered condition.
-        """
-        results = {}
-        results["confidence_drift"] = self.check_confidence_drift()
-        results["review_queue_growth"] = self.check_review_queue_growth()
-        results["low_throughput"] = self.check_low_throughput()
-        results["high_error_rate"] = self.check_error_rate()
-        return results
-
-    def check_confidence_drift(self) -> bool:
-        """
-        Compare p(C > 0.5) in recent processing logs to the stored baseline.
-
-        Returns True if drift exceeds the threshold.
-        """
-        current = self._fetch_recent_confidences(hours=24 * 7)
-        baseline = self._fetch_baseline_confidences()
-        if not current or not baseline:
-            logger.debug("Not enough data for drift check")
-            return False
-
-        triggered = detect_drift(current, baseline, self._drift_threshold)
-        if triggered:
-            current_high = _high_conf_fraction(current)
-            baseline_high = _high_conf_fraction(baseline)
-            shift = abs(current_high - baseline_high)
-            self._alert_fn(
-                f"Confidence drift detected: {shift:.1%} shift in p(C>0.5) "
-                f"(current={current_high:.1%}, baseline={baseline_high:.1%}, "
-                f"threshold={self._drift_threshold:.1%})"
-            )
-        return triggered
-
-    def check_review_queue_growth(self) -> bool:
-        """Alert if the unannotated review queue has been above max_size for max_age hours."""
-        try:
-            cutoff = datetime.now(timezone.utc) - timedelta(hours=self._review_max_age_h)
-            rows = self._db.fetchall(
-                """SELECT COUNT(*) AS cnt
-                   FROM review_queue
-                   WHERE annotated = FALSE
-                     AND created_at < %s""",
-                (cutoff.replace(tzinfo=None),),
-            )
-            count = int((rows or [{"cnt": 0}])[0]["cnt"])
-            if count > self._review_max_size:
-                self._alert_fn(
-                    f"Review queue growth alert: {count} unannotated items older than "
-                    f"{self._review_max_age_h}h (threshold={self._review_max_size})"
-                )
-                return True
-            return False
-        except Exception as exc:
-            logger.warning("Review queue growth check failed: %s", exc)
-            return False
-
-    def check_low_throughput(self) -> bool:
-        """Alert if throughput dropped below the minimum for the last window hours."""
-        try:
-            cutoff = datetime.now(timezone.utc) - timedelta(hours=self._throughput_window_h)
-            rows = self._db.fetchall(
-                """SELECT COUNT(*) AS cnt
-                   FROM videos
-                   WHERE status = 'COMPLETED'
-                     AND last_processed_time >= %s""",
-                (cutoff.replace(tzinfo=None),),
-            )
-            count = int((rows or [{"cnt": 0}])[0]["cnt"])
-            videos_per_hour = count / max(self._throughput_window_h, 1)
-            if videos_per_hour < self._throughput_min:
-                self._alert_fn(
-                    f"Low throughput alert: {videos_per_hour:.1f} videos/hour "
-                    f"over last {self._throughput_window_h}h "
-                    f"(minimum={self._throughput_min})"
-                )
-                return True
-            return False
-        except Exception as exc:
-            logger.warning("Throughput check failed: %s", exc)
-            return False
-
-    def check_error_rate(self) -> bool:
-        """Alert if the fraction of ERROR/UNSCANNABLE videos exceeds the threshold."""
-        try:
-            cutoff = datetime.now(timezone.utc) - timedelta(hours=self._error_rate_window_h)
-            rows = self._db.fetchall(
-                """SELECT status, COUNT(*) AS cnt
-                   FROM videos
-                   WHERE last_processed_time >= %s
-                     AND status IN ('COMPLETED', 'ERROR', 'UNSCANNABLE')
-                   GROUP BY status""",
-                (cutoff.replace(tzinfo=None),),
-            )
-            counts: Dict[str, int] = {r["status"]: int(r["cnt"]) for r in (rows or [])}
-            total = sum(counts.values())
-            if total == 0:
-                return False
-            errors = counts.get("ERROR", 0) + counts.get("UNSCANNABLE", 0)
-            rate = errors / total
-            if rate > self._error_rate_threshold:
-                self._alert_fn(
-                    f"High error rate alert: {rate:.1%} over last {self._error_rate_window_h}h "
-                    f"({errors}/{total} videos, threshold={self._error_rate_threshold:.1%})"
-                )
-                return True
-            return False
-        except Exception as exc:
-            logger.warning("Error rate check failed: %s", exc)
-            return False
-
-    # ------------------------------------------------------------------
-    # DB helpers
-    # ------------------------------------------------------------------
-
-    def _fetch_recent_confidences(self, hours: int = 168) -> List[float]:
-        """Return confidence scores from the last `hours` hours."""
-        try:
-            cutoff = datetime.now(timezone.utc) - timedelta(hours=hours)
-            rows = self._db.fetchall(
-                """SELECT confidence_score
-                   FROM processing_logs
-                   WHERE processed_at >= %s
-                     AND confidence_score IS NOT NULL""",
-                (cutoff.replace(tzinfo=None),),
-            )
-            return [float(r["confidence_score"]) for r in (rows or [])]
-        except Exception as exc:
-            logger.warning("Failed to fetch recent confidences: %s", exc)
-            return []
-
-    def _fetch_baseline_confidences(self) -> List[float]:
-        """Return the baseline confidence distribution stored in the DB."""
-        try:
-            rows = self._db.fetchall(
-                """SELECT confidence_score
-                   FROM processing_logs
-                   WHERE confidence_score IS NOT NULL
-                   ORDER BY processed_at ASC
-                   LIMIT 10000"""
-            )
-            return [float(r["confidence_score"]) for r in (rows or [])]
-        except Exception as exc:
-            logger.warning("Failed to fetch baseline confidences: %s", exc)
-            return []
-
-
-# ---------------------------------------------------------------------------
-# Module-level pure functions (testable without DB)
-# ---------------------------------------------------------------------------
-
-def detect_drift(
-    current_confidences: List[float],
-    baseline_confidences: List[float],
-    threshold: float = 0.10,
-) -> bool:
-    """
-    Return True if |p(C>0.5)_current - p(C>0.5)_baseline| > threshold.
-
-    Both lists must be non-empty.
-    """
-    if not current_confidences or not baseline_confidences:
-        return False
-    current_high = _high_conf_fraction(current_confidences)
-    baseline_high = _high_conf_fraction(baseline_confidences)
-    shift = abs(current_high - baseline_high)
-    if shift > threshold:
-        logger.warning(
-            "Drift detected: shift=%.2f%% (current=%.2f%% baseline=%.2f%%)",
-            shift * 100, current_high * 100, baseline_high * 100,
-        )
-        return True
-    return False
-
-
-def _high_conf_fraction(confidences: List[float]) -> float:
-    """Return the fraction of scores > 0.5."""
-    return sum(1 for c in confidences if c > 0.5) / len(confidences)
-
-
-def _log_alert(message: str):
-    logger.warning("ALERT: %s", message)

src/export_face_detector.py

@@ -1,104 +0,0 @@
-"""Export a YOLOv8n face-detection model to ONNX and build a TensorRT engine."""
-
-import argparse
-import json
-import logging
-from pathlib import Path
-
-logger = logging.getLogger(__name__)
-
-
-def export_onnx(output_dir: Path, input_size: int = 640) -> Path:
-    """Export YOLOv8n to ONNX with the required input shape."""
-    output_dir.mkdir(parents=True, exist_ok=True)
-    onnx_path = output_dir / "face_detector.onnx"
-    metadata_path = output_dir / "model.json"
-
-    try:
-        from ultralytics import YOLO
-
-        model = YOLO("yolov8n.pt")
-        model.export(
-            format="onnx",
-            imgsz=input_size,
-            half=False,
-            simplify=True,
-            dynamic=False,
-        )
-        exported = Path("yolov8n.onnx")
-        if exported.exists():
-            exported.rename(onnx_path)
-        logger.info("Exported ONNX model to %s", onnx_path)
-    except Exception as exc:
-        logger.error("Failed to export ONNX model: %s", exc)
-        raise
-
-    metadata = {
-        "input_shape": [1, 3, input_size, input_size],
-        "mean": [0.485, 0.456, 0.406],
-        "std": [0.229, 0.224, 0.225],
-        "confidence_threshold": 0.25,
-        "iou_threshold": 0.45,
-    }
-    metadata_path.write_text(json.dumps(metadata, indent=2))
-    return onnx_path
-
-
-def build_tensorrt_engine(onnx_path: Path, output_dir: Path, max_batch_size: int = 32) -> Path:
-    """Build and serialize a TensorRT FP32 engine from an ONNX file."""
-    engine_path = output_dir / "face_detector.trt"
-
-    try:
-        import tensorrt as trt
-
-        logger = trt.Logger(trt.Logger.WARNING)
-        builder = trt.Builder(logger)
-        network = builder.create_network(1 << int(trt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH))
-        parser = trt.OnnxParser(network, logger)
-
-        onnx_data = onnx_path.read_bytes()
-        if not parser.parse(onnx_data):
-            for error in range(parser.num_errors):
-                logger.error(parser.get_error(error))
-            raise RuntimeError("ONNX parsing failed")
-
-        config = builder.create_builder_config()
-        config.max_workspace_size = 4 << 30  # 4GB
-        config.set_flag(trt.BuilderFlag.FP32)
-
-        profile = builder.create_optimization_profile()
-        input_name = network.get_input(0).name
-        profile.set_shape(
-            input_name,
-            (1, 3, 640, 640),
-            (max_batch_size // 2, 3, 640, 640),
-            (max_batch_size, 3, 640, 640),
-        )
-        config.add_optimization_profile(profile)
-
-        engine = builder.build_engine(network, config)
-        if engine is None:
-            raise RuntimeError("TensorRT engine build failed")
-
-        engine_path.write_bytes(engine.serialize())
-        logger.info("Serialized TensorRT engine to %s", engine_path)
-        return engine_path
-    except Exception as exc:
-        logger.error("Failed to build TensorRT engine: %s", exc)
-        raise
-
-
-def main():
-    logging.basicConfig(level=logging.INFO)
-    parser = argparse.ArgumentParser(description="Export YOLOv8n face detector to ONNX/TensorRT")
-    parser.add_argument("--output-dir", type=Path, default=Path("models/face_detector"))
-    parser.add_argument("--input-size", type=int, default=640)
-    parser.add_argument("--max-batch-size", type=int, default=32)
-    args = parser.parse_args()
-
-    onnx_path = export_onnx(args.output_dir, args.input_size)
-    build_tensorrt_engine(onnx_path, args.output_dir, args.max_batch_size)
-
-
-if __name__ == "__main__":
-    main()

src/face_detector.py

@@ -1,295 +0,0 @@
-"""
-Face detection runner.
-
-Wraps a YOLOv8n model exported to ONNX/TensorRT, runs batched inference on
-sampled frames, applies NMS, and produces 224×224 face crops for the classifier.
-"""
-
-import logging
-import time
-from pathlib import Path
-from typing import Any, Dict, List, Optional, Tuple
-
-import numpy as np
-
-logger = logging.getLogger(__name__)
-
-
-class Detection:
-    """A single face detection result."""
-
-    def __init__(
-        self,
-        frame_path: str,
-        x1: float,
-        y1: float,
-        x2: float,
-        y2: float,
-        confidence: float,
-        crop_path: Optional[str] = None,
-    ):
-        self.frame_path = frame_path
-        self.x1 = x1
-        self.y1 = y1
-        self.x2 = x2
-        self.y2 = y2
-        self.confidence = confidence
-        self.crop_path = crop_path
-
-    def to_dict(self) -> Dict[str, Any]:
-        return {
-            "frame_path": self.frame_path,
-            "x1": self.x1,
-            "y1": self.y1,
-            "x2": self.x2,
-            "y2": self.y2,
-            "confidence": self.confidence,
-            "crop_path": self.crop_path,
-        }
-
-
-class FaceDetector:
-    """Run batched face detection on video frames."""
-
-    def __init__(
-        self,
-        engine_path: str,
-        input_size: int = 640,
-        confidence_threshold: float = 0.25,
-        iou_threshold: float = 0.45,
-        max_faces_per_frame: int = 10,
-        max_faces_per_video: int = 100,
-        device: str = "cuda",
-    ):
-        self.engine_path = engine_path
-        self.input_size = input_size
-        self.confidence_threshold = confidence_threshold
-        self.iou_threshold = iou_threshold
-        self.max_faces_per_frame = max_faces_per_frame
-        self.max_faces_per_video = max_faces_per_video
-        self.device = device
-        self._session: Optional[Any] = None
-        self._load_model()
-
-    def _load_model(self):
-        """Load the inference backend."""
-        path = Path(self.engine_path)
-        if not path.exists():
-            logger.warning("Face detector engine not found at %s; using placeholder", self.engine_path)
-            return
-
-        suffix = path.suffix.lower()
-        if suffix == ".onnx":
-            try:
-                import onnxruntime as ort
-
-                providers = ["CUDAExecutionProvider"] if self.device.startswith("cuda") else ["CPUExecutionProvider"]
-                self._session = ort.InferenceSession(str(path), providers=providers)
-                logger.info("Loaded ONNX face detector from %s", self.engine_path)
-            except Exception as exc:
-                logger.warning("Failed to load ONNX face detector: %s", exc)
-        elif suffix in (".trt", ".engine", ".plan"):
-            try:
-                import tensorrt as trt
-
-                with trt.Logger() as trt_logger, open(path, "rb") as f:
-                    runtime = trt.Runtime(trt_logger)
-                    self._session = runtime.deserialize_cuda_engine(f.read())
-                logger.info("Loaded TensorRT face detector from %s", self.engine_path)
-            except Exception as exc:
-                logger.warning("Failed to load TensorRT face detector: %s", exc)
-        else:
-            logger.warning("Unsupported face detector format: %s", suffix)
-
-    def detect_faces(
-        self,
-        frame_paths: List[str],
-        batch_size: int = 16,
-    ) -> List[List[Detection]]:
-        """Detect faces in a list of frame image paths."""
-        all_results: List[List[Detection]] = []
-        for i in range(0, len(frame_paths), batch_size):
-            batch = frame_paths[i : i + batch_size]
-            batch_results = self._detect_batch(batch)
-            all_results.extend(batch_results)
-        return all_results
-
-    def _detect_batch(self, frame_paths: List[str]) -> List[List[Detection]]:
-        """Run detection on one batch and return detections per frame."""
-        preprocessed = []
-        for path in frame_paths:
-            try:
-                preprocessed.append(self._preprocess(path))
-            except Exception as exc:
-                logger.warning("Failed to preprocess %s: %s", path, exc)
-                preprocessed.append(np.zeros((3, self.input_size, self.input_size), dtype=np.float32))
-
-        batch_input = np.stack(preprocessed, axis=0)
-
-        outputs = self._infer(batch_input)
-
-        results = []
-        for idx, path in enumerate(frame_paths):
-            try:
-                detections = self._parse_outputs(outputs, idx, path)
-                detections = self._nms(detections)
-                detections = detections[: self.max_faces_per_frame]
-                results.append(detections)
-            except Exception as exc:
-                logger.warning("Failed to parse detections for %s: %s", path, exc)
-                results.append([])
-        return results
-
-    def _preprocess(self, frame_path: str) -> np.ndarray:
-        """Load and normalize a frame for the detector."""
-        from PIL import Image
-
-        image = Image.open(frame_path).convert("RGB")
-        image = image.resize((self.input_size, self.input_size), Image.Resampling.BILINEAR)
-        arr = np.array(image, dtype=np.float32) / 255.0
-        mean = np.array([0.485, 0.456, 0.406], dtype=np.float32)
-        std = np.array([0.229, 0.224, 0.225], dtype=np.float32)
-        arr = (arr - mean) / std
-        return np.transpose(arr, (2, 0, 1))  # HWC -> CHW
-
-    def _infer(self, batch_input: np.ndarray) -> Any:
-        """Run inference on the preprocessed batch."""
-        if self._session is None:
-            return self._placeholder_outputs(batch_input.shape[0])
-
-        try:
-            if hasattr(self._session, "run"):
-                input_name = self._session.get_inputs()[0].name
-                return self._session.run(None, {input_name: batch_input})
-
-            # TensorRT execution
-            import pycuda.driver as cuda
-            import pycuda.autoinit  # noqa: F401
-
-            context = self._session.create_execution_context()
-            output_shape = (batch_input.shape[0], 84, 8400)  # YOLOv8n default shape
-            d_input = cuda.mem_alloc(batch_input.nbytes)
-            d_output = cuda.mem_alloc(np.prod(output_shape) * np.dtype(np.float32).itemsize)
-            stream = cuda.Stream()
-            cuda.memcpy_htod_async(d_input, batch_input, stream)
-            context.execute_async_v2(bindings=[int(d_input), int(d_output)], stream_handle=stream.handle)
-            output = np.empty(output_shape, dtype=np.float32)
-            cuda.memcpy_dtoh_async(output, d_output, stream)
-            stream.synchronize()
-            return output
-        except Exception as exc:
-            logger.warning("Face detector inference failed: %s", exc)
-            return self._placeholder_outputs(batch_input.shape[0])
-
-    def _placeholder_outputs(self, batch_size: int) -> np.ndarray:
-        """Return an empty output tensor when no model is loaded."""
-        return np.zeros((batch_size, 84, 8400), dtype=np.float32)
-
-    def _parse_outputs(self, outputs: Any, batch_index: int, frame_path: str) -> List[Detection]:
-        """Parse raw inference outputs into Detection objects."""
-        if isinstance(outputs, list):
-            raw = outputs[0][batch_index]  # (84, 8400)
-        else:
-            raw = outputs[batch_index]
-
-        # YOLOv8 output layout: (cx, cy, w, h, cls scores...)
-        raw = raw.T  # (8400, 84)
-        scores = raw[:, 4:].max(axis=1)
-        mask = scores >= self.confidence_threshold
-        candidates = raw[mask]
-        scores = scores[mask]
-
-        detections = []
-        for row, score in zip(candidates, scores):
-            cx, cy, w, h = row[:4]
-            x1 = cx - w / 2
-            y1 = cy - h / 2
-            x2 = cx + w / 2
-            y2 = cy + h / 2
-            detections.append(
-                Detection(
-                    frame_path=frame_path,
-                    x1=float(x1),
-                    y1=float(y1),
-                    x2=float(x2),
-                    y2=float(y2),
-                    confidence=float(score),
-                )
-            )
-        return detections
-
-    def _nms(self, detections: List[Detection]) -> List[Detection]:
-        """Apply greedy Non-Maximum Suppression."""
-        if not detections:
-            return []
-
-        sorted_dets = sorted(detections, key=lambda d: d.confidence, reverse=True)
-        kept: List[Detection] = []
-
-        while sorted_dets:
-            current = sorted_dets.pop(0)
-            kept.append(current)
-            sorted_dets = [
-                det
-                for det in sorted_dets
-                if self._iou(current, det) <= self.iou_threshold
-            ]
-            if len(kept) >= self.max_faces_per_frame:
-                break
-
-        return kept
-
-    @staticmethod
-    def _iou(a: Detection, b: Detection) -> float:
-        """Compute IoU between two detections."""
-        x1 = max(a.x1, b.x1)
-        y1 = max(a.y1, b.y1)
-        x2 = min(a.x2, b.x2)
-        y2 = min(a.y2, b.y2)
-
-        inter_area = max(0, x2 - x1) * max(0, y2 - y1)
-        area_a = (a.x2 - a.x1) * (a.y2 - a.y1)
-        area_b = (b.x2 - b.x1) * (b.y2 - b.y1)
-        union_area = area_a + area_b - inter_area
-        if union_area == 0:
-            return 0.0
-        return inter_area / union_area
-
-    def extract_crops(
-        self,
-        detections: List[Detection],
-        output_dir: str,
-        crop_size: Tuple[int, int] = (224, 224),
-    ) -> List[Detection]:
-        """Extract resized face crops from original frames and update detections."""
-        from PIL import Image
-
-        output_dir_path = Path(output_dir)
-        output_dir_path.mkdir(parents=True, exist_ok=True)
-        cropped: List[Detection] = []
-
-        for idx, det in enumerate(detections):
-            try:
-                image = Image.open(det.frame_path).convert("RGB")
-                width, height = image.size
-
-                x1 = int(max(0, det.x1 * width / self.input_size))
-                y1 = int(max(0, det.y1 * height / self.input_size))
-                x2 = int(min(width, det.x2 * width / self.input_size))
-                y2 = int(min(height, det.y2 * height / self.input_size))
-
-                if x2 <= x1 or y2 <= y1:
-                    image.close()
-                    continue
-
-                crop = image.crop((x1, y1, x2, y2)).resize(crop_size, Image.Resampling.BILINEAR)
-                image.close()
-                crop_path = output_dir_path / f"crop_{int(time.time() * 1000)}_{idx}.jpg"
-                crop.save(crop_path, "JPEG", quality=95)
-                crop.close()
-                det.crop_path = str(crop_path)
-                cropped.append(det)
-            except Exception as exc:
-                logger.warning("Failed to extract crop for %s: %s", det.frame_path, exc)
-
-        return cropped

src/frame_sampler.py

@@ -1,163 +0,0 @@
-"""
-Frame sampling module for VideoDetect.
-
-Extracts uniform temporal frames from videos using FFmpeg, handles resolution
-constraints, and writes JPEG frames to a scratch directory.
-"""
-
-import logging
-import os
-import subprocess
-from pathlib import Path
-from typing import Iterable, List, Optional, Sequence, Tuple
-
-logger = logging.getLogger(__name__)
-
-
-def calculate_timestamps(duration: Optional[float], interval: float) -> List[float]:
-    """Return uniformly spaced timestamps in seconds for a video duration."""
-    if duration is None or duration <= 0:
-        return [0.0]
-    if interval <= 0:
-        interval = 30.0
-
-    count = max(1, int(duration / interval))
-    step = duration / count
-    return [round(i * step, 3) for i in range(count)]
-
-
-class FrameSampler:
-    """Extract sampled frames from videos using FFmpeg."""
-
-    def __init__(self, interval_seconds: int = 30, quality: int = 2, output_format: str = "jpeg"):
-        self.interval_seconds = interval_seconds
-        self.quality = quality
-        self.output_format = output_format.lower()
-
-    def extract_frame(
-        self,
-        video_path: str,
-        output_path: str,
-        timestamp: float,
-        resolution: Optional[Tuple[int, int]] = None,
-        timeout_seconds: int = 30,
-    ) -> Optional[str]:
-        """Extract a single frame at the provided timestamp."""
-        output = Path(output_path)
-        output.parent.mkdir(parents=True, exist_ok=True)
-
-        cmd = self._build_ffmpeg_command(
-            video_path=video_path,
-            output_path=str(output),
-            timestamp=timestamp,
-            resolution=resolution,
-            timeout_seconds=timeout_seconds,
-        )
-
-        try:
-            completed = subprocess.run(
-                cmd,
-                capture_output=True,
-                text=True,
-                timeout=timeout_seconds + 5,
-                check=False,
-            )
-        except subprocess.TimeoutExpired as exc:
-            logger.warning("Frame extraction timed out for %s at %.3fs", video_path, timestamp)
-            return None
-
-        if completed.returncode != 0:
-            logger.warning(
-                "Frame extraction failed for %s at %.3fs: %s",
-                video_path,
-                timestamp,
-                completed.stderr.strip(),
-            )
-            return None
-
-        if output.exists():
-            return str(output)
-        return None
-
-    def extract_frames(
-        self,
-        video_path: str,
-        output_dir: str,
-        duration: Optional[float] = None,
-        interval_seconds: Optional[int] = None,
-        resolution: Optional[Tuple[int, int]] = None,
-        timeout_seconds: int = 30,
-    ) -> List[str]:
-        """Extract uniform frames for the video and return output paths."""
-        resolved_interval = interval_seconds or self.interval_seconds
-        timestamps = calculate_timestamps(duration=duration, interval=float(resolved_interval))
-
-        extracted: List[str] = []
-        for ts in timestamps:
-            frame_name = self._frame_name_from_timestamp(ts, video_path)
-            output_path = str(Path(output_dir) / frame_name)
-            result = self.extract_frame(
-                video_path=video_path,
-                output_path=output_path,
-                timestamp=ts,
-                resolution=resolution,
-                timeout_seconds=timeout_seconds,
-            )
-            if result:
-                extracted.append(result)
-
-        return extracted
-
-    def _build_ffmpeg_command(
-        self,
-        video_path: str,
-        output_path: str,
-        timestamp: float,
-        resolution: Optional[Tuple[int, int]] = None,
-        timeout_seconds: int = 30,
-    ) -> List[str]:
-        """Build an FFmpeg command for extracting a single frame."""
-        cmd = ["ffmpeg", "-y", "-ss", str(timestamp), "-i", video_path]
-
-        vf_filters = []
-        if resolution is not None:
-            target_width, target_height = self._resolve_target_resolution(resolution)
-            vf_filters.append(f"scale={target_width}:{target_height}:force_original_aspect_ratio=decrease")
-
-        if vf_filters:
-            cmd.extend(["-vf", ",".join(vf_filters)])
-
-        cmd.extend([
-            "-vframes",
-            "1",
-            "-q:v",
-            str(self.quality),
-            "-f",
-            self.output_format,
-            output_path,
-        ])
-        return cmd
-
-    @staticmethod
-    def _resolve_target_resolution(resolution: Tuple[int, int]) -> Tuple[int, int]:
-        """Downscale frames larger than 1080p to fit VRAM constraints."""
-        width, height = resolution
-        if max(width, height) > 1080:
-            if width >= 3840 or height >= 2160:
-                return 1920, 1080
-            if width >= 2560 or height >= 1440:
-                return 1280, 720
-        return width, height
-
-    @staticmethod
-    def _frame_name_from_timestamp(timestamp: float, video_path: str) -> str:
-        video_id = Path(video_path).stem
-        timestamp_ms = int(round(timestamp * 1000))
-        return f"{video_id}_{timestamp_ms}.jpg"
-
-
-def get_target_resolution(resolution: Optional[Tuple[int, int]]) -> Tuple[int, int]:
-    """Compatibility helper for resolution handling."""
-    if resolution is None:
-        return (0, 0)
-    return FrameSampler._resolve_target_resolution(resolution)

src/gpu_manager.py

@@ -1,101 +0,0 @@
-"""GPU memory management and batch-size auto-tuning helpers."""
-
-import logging
-from typing import Optional, Tuple
-
-logger = logging.getLogger(__name__)
-
-
-class GPUMemoryManager:
-    """Track GPU memory and recommend safe batch sizes."""
-
-    def __init__(
-        self,
-        device: int = 0,
-        max_memory_gb: float = 18.0,
-        reduce_threshold_gb: float = 16.0,
-        increase_threshold_gb: float = 10.0,
-        initial_batch_size: int = 16,
-        min_batch_size: int = 1,
-        max_batch_size: int = 32,
-    ):
-        self.device = device
-        self.max_memory_gb = max_memory_gb
-        self.reduce_threshold_gb = reduce_threshold_gb
-        self.increase_threshold_gb = increase_threshold_gb
-        self.batch_size = initial_batch_size
-        self.min_batch_size = min_batch_size
-        self.max_batch_size = max_batch_size
-
-    def get_memory_stats(self) -> Tuple[float, float]:
-        """Return (allocated_gb, reserved_gb) for the managed device."""
-        try:
-            import torch
-
-            if not torch.cuda.is_available():
-                return 0.0, 0.0
-
-            allocated = torch.cuda.memory_allocated(self.device) / (1024**3)
-            reserved = torch.cuda.memory_reserved(self.device) / (1024**3)
-            return allocated, reserved
-        except Exception as exc:
-            logger.debug("Could not query GPU memory: %s", exc)
-            return 0.0, 0.0
-
-    def adjust_batch_size(self) -> int:
-        """Adjust the current batch size based on memory pressure."""
-        allocated, _ = self.get_memory_stats()
-
-        if allocated >= self.reduce_threshold_gb:
-            new_batch_size = max(self.min_batch_size, int(self.batch_size * 0.75))
-            if new_batch_size < self.batch_size:
-                logger.info(
-                    "GPU memory high (%.2f GB), reducing batch size %d -> %d",
-                    allocated,
-                    self.batch_size,
-                    new_batch_size,
-                )
-                self.batch_size = new_batch_size
-        elif allocated <= self.increase_threshold_gb:
-            new_batch_size = min(self.max_batch_size, int(self.batch_size * 1.25))
-            if new_batch_size > self.batch_size:
-                logger.info(
-                    "GPU memory low (%.2f GB), increasing batch size %d -> %d",
-                    allocated,
-                    self.batch_size,
-                    new_batch_size,
-                )
-                self.batch_size = new_batch_size
-
-        return self.batch_size
-
-    def is_memory_critical(self) -> bool:
-        """Return True if allocated memory is close to the hard limit."""
-        allocated, _ = self.get_memory_stats()
-        return allocated >= self.max_memory_gb
-
-    def empty_cache(self):
-        """Try to release unused cached GPU memory."""
-        try:
-            import torch
-
-            if torch.cuda.is_available():
-                torch.cuda.empty_cache()
-        except Exception as exc:
-            logger.debug("Could not empty GPU cache: %s", exc)
-
-    @property
-    def current_batch_size(self) -> int:
-        return self.batch_size
-
-
-def get_available_vram_gb(device: int = 0) -> float:
-    """Return total available VRAM in GB, or 0.0 if CUDA is unavailable."""
-    try:
-        import torch
-
-        if not torch.cuda.is_available():
-            return 0.0
-        return torch.cuda.get_device_properties(device).total_memory / (1024**3)
-    except Exception:
-        return 0.0

src/health_check.py

@@ -1,134 +0,0 @@
-"""
-Worker health check HTTP endpoint.
-
-Exposes GET /health returning a JSON status document.
-Designed to be started in a background thread alongside the main worker.
-"""
-
-import json
-import logging
-import os
-import threading
-import time
-from datetime import datetime, timezone
-from http.server import BaseHTTPRequestHandler, HTTPServer
-from pathlib import Path
-from typing import Any, Dict, Optional
-
-logger = logging.getLogger(__name__)
-
-_START_TIME = time.monotonic()
-
-
-class HealthStatus:
-    """Mutable singleton updated by the worker as it processes videos."""
-
-    def __init__(self):
-        self._lock = threading.Lock()
-        self._data: Dict[str, Any] = {
-            "status": "starting",
-            "gpu_available": False,
-            "gpu_memory_used_gb": 0.0,
-            "queue_depth": 0,
-            "videos_processed_today": 0,
-            "last_error": None,
-        }
-
-    def update(self, **kwargs):
-        with self._lock:
-            self._data.update(kwargs)
-
-    def snapshot(self) -> Dict[str, Any]:
-        with self._lock:
-            data = dict(self._data)
-        data["uptime_seconds"] = round(time.monotonic() - _START_TIME, 1)
-        return data
-
-
-# Module-level singleton shared between the HTTP handler and the worker
-health = HealthStatus()
-
-
-class _HealthHandler(BaseHTTPRequestHandler):
-    """Minimal HTTP handler — only responds to GET /health."""
-
-    def do_GET(self):  # noqa: N802
-        if self.path.rstrip("/") == "/health":
-            body = json.dumps(health.snapshot(), default=str).encode()
-            self.send_response(200)
-            self.send_header("Content-Type", "application/json")
-            self.send_header("Content-Length", str(len(body)))
-            self.end_headers()
-            self.wfile.write(body)
-        else:
-            self.send_response(404)
-            self.end_headers()
-
-    def log_message(self, fmt, *args):  # silence default access log
-        logger.debug("health_check: " + fmt, *args)
-
-
-def start_health_server(port: int = 8080) -> threading.Thread:
-    """
-    Start the health HTTP server in a daemon thread.
-
-    Returns the thread so callers can join if needed.
-    """
-    server = HTTPServer(("0.0.0.0", port), _HealthHandler)
-
-    def _serve():
-        logger.info("Health check server listening on port %d", port)
-        server.serve_forever()
-
-    t = threading.Thread(target=_serve, daemon=True, name="health-server")
-    t.start()
-    return t
-
-
-def refresh_health_from_system(db_connector=None, scratch_path: str = "/scratch"):
-    """
-    Update the health singleton with current GPU, queue, and disk state.
-
-    Call this periodically from the worker main loop.
-    """
-    updates: Dict[str, Any] = {"status": "healthy"}
-
-    # GPU state
-    try:
-        import pynvml
-        pynvml.nvmlInit()
-        handle = pynvml.nvmlDeviceGetHandleByIndex(0)
-        mem = pynvml.nvmlDeviceGetMemoryInfo(handle)
-        updates["gpu_available"] = True
-        updates["gpu_memory_used_gb"] = round(mem.used / 1024**3, 2)
-        pynvml.nvmlShutdown()
-    except Exception:
-        updates["gpu_available"] = _torch_gpu_available()
-
-    # Queue depth from DB
-    if db_connector is not None:
-        try:
-            rows = db_connector.fetchall(
-                "SELECT COUNT(*) AS cnt FROM videos WHERE status IN ('PENDING','PROCESSING')"
-            )
-            updates["queue_depth"] = int((rows or [{"cnt": 0}])[0]["cnt"])
-
-            today = datetime.now(timezone.utc).date().isoformat()
-            rows2 = db_connector.fetchall(
-                "SELECT COUNT(*) AS cnt FROM videos "
-                "WHERE status = 'COMPLETED' AND DATE(last_processed_time) = %s",
-                (today,),
-            )
-            updates["videos_processed_today"] = int((rows2 or [{"cnt": 0}])[0]["cnt"])
-        except Exception as exc:
-            logger.debug("Health DB query failed: %s", exc)
-
-    health.update(**updates)
-
-
-def _torch_gpu_available() -> bool:
-    try:
-        import torch
-        return torch.cuda.is_available()
-    except ImportError:
-        return False

src/logging_config.py

@@ -1,87 +0,0 @@
-"""
-Logging configuration for VideoDetect.
-
-Sets up JSON structured logging via python-json-logger with:
-- Configurable log levels
-- Standardized field names
-- Log rotation
-- All logs to stdout for Docker capture
-"""
-
-import json
-import logging
-import logging.handlers
-import sys
-from datetime import datetime, timezone
-
-
-def setup_logging(
-    level: str = "INFO",
-    log_format: str = "json",
-    rotation_max_bytes: int = 104857600,  # 100MB
-    rotation_backup_count: int = 10,
-):
-    """Configure structured logging for the application."""
-    log_level = getattr(logging, level.upper(), logging.INFO)
-
-    # Root logger
-    root_logger = logging.getLogger()
-    root_logger.setLevel(log_level)
-
-    # Remove existing handlers
-    root_logger.handlers.clear()
-
-    # Handler: stdout (for Docker capture)
-    handler = logging.StreamHandler(sys.stdout)
-    handler.setLevel(log_level)
-
-    if log_format == "json":
-        formatter = JsonFormatter()
-    else:
-        formatter = logging.Formatter(
-            "%(asctime)s %(levelname)s %(name)s %(message)s",
-            datefmt="%Y-%m-%dT%H:%M:%S%z",
-        )
-
-    handler.setFormatter(formatter)
-    root_logger.addHandler(handler)
-
-    # Handler: file rotation (for persistence)
-    file_handler = logging.handlers.RotatingFileHandler(
-        filename="/logs/videodetect.log",
-        maxBytes=rotation_max_bytes,
-        backupCount=rotation_backup_count,
-        encoding="utf-8",
-    )
-    file_handler.setLevel(log_level)
-    file_handler.setFormatter(formatter)
-    root_logger.addHandler(file_handler)
-
-    logging.info("Logging configured: level=%s format=%s", level, log_format)
-
-
-class JsonFormatter(logging.Formatter):
-    """JSON log formatter with standardized fields."""
-
-    def format(self, record: logging.LogRecord) -> str:
-        log_data = {
-            "timestamp": datetime.fromtimestamp(record.created, tz=timezone.utc).isoformat(),
-            "level": record.levelname,
-            "logger": record.name,
-            "message": record.getMessage(),
-            "module": record.module,
-            "function": record.funcName,
-            "line": record.lineno,
-        }
-
-        # Add extra fields
-        if hasattr(record, "video_id"):
-            log_data["video_id"] = record.video_id
-        if hasattr(record, "metadata"):
-            log_data["metadata"] = record.metadata
-
-        # Add exception info if present
-        if record.exc_info and record.exc_info[0] is not None:
-            log_data["exception"] = self.formatException(record.exc_info)
-
-        return json.dumps(log_data, default=str)

src/main.py

@@ -1,122 +0,0 @@
-"""
-VideoDetect - Video Classification System
-
-Main entry point for the worker service.
-Initializes all components and starts the processing pipeline.
-"""
-
-import logging
-import signal
-import sys
-import threading
-import time
-from pathlib import Path
-
-# Add src to path
-sys.path.insert(0, str(Path(__file__).parent))
-
-from config_loader import get_config
-from db_connector import DBConnector
-from logging_config import setup_logging
-from orchestrator import WorkerPool
-from scanner import DirectoryScanner
-
-logger = logging.getLogger(__name__)
-
-
-def graceful_shutdown(signum, frame):
-    """Handle shutdown signals gracefully."""
-    logger.info("Received signal %d, initiating graceful shutdown...", signum)
-    sys.exit(0)
-
-
-def main():
-    """Initialize and start the VideoDetect worker."""
-    # Register signal handlers
-    signal.signal(signal.SIGTERM, graceful_shutdown)
-    signal.signal(signal.SIGINT, graceful_shutdown)
-
-    # Load configuration
-    config = get_config()
-    log_config = config.get_section("logging")
-    setup_logging(
-        level=log_config.get("level", "INFO"),
-        log_format=log_config.get("format", "json"),
-        rotation_max_bytes=log_config.get("rotation_max_bytes", 104857600),
-        rotation_backup_count=log_config.get("rotation_backup_count", 10),
-    )
-
-    logger.info("VideoDetect Worker starting...")
-    logger.info("Config: %s", config)
-
-    # Initialize database connection
-    db_config = config.get_section("database")
-    db = DBConnector(
-        host=db_config.get("host", "mariadb"),
-        port=db_config.get("port", 3306),
-        database=db_config.get("name", "videodetect"),
-        user=db_config.get("user", "videodetect"),
-        password=db_config.get("password", "videodetect123"),
-        pool_size=db_config.get("pool_size", 20),
-        pool_min=db_config.get("pool_min", 5),
-        pool_recycle=db_config.get("pool_recycle", 3600),
-    )
-
-    # Verify database connectivity
-    if not db.health_check():
-        logger.error("Cannot connect to database. Exiting.")
-        sys.exit(1)
-    logger.info("Database connection established.")
-
-    # Initialize schema if needed
-    schema_path = Path(__file__).parent.parent / "db" / "schema.sql"
-    if schema_path.exists():
-        db.initialize_schema(str(schema_path))
-        logger.info("Schema initialized.")
-
-    # Verify GPU availability
-    try:
-        import torch
-        if torch.cuda.is_available():
-            gpu_count = torch.cuda.device_count()
-            gpu_name = torch.cuda.get_device_name(0)
-            logger.info("GPU available: %d GPUs, primary: %s", gpu_count, gpu_name)
-        else:
-            logger.warning("CUDA is not available! Processing will be slow.")
-    except ImportError:
-        logger.warning("PyTorch not installed. GPU features disabled.")
-
-    # Initialize the directory scanner.
-    # It scans the permanent storage location in place (no staging/copy step),
-    # periodically discovering new/removed video files and queueing any that
-    # have not been scanned yet as PENDING for the worker pool to pick up.
-    storage_config = config.get_section("storage")
-    scanner_config = config.get_section("scanner")
-    scanner = DirectoryScanner(
-        db_connector=db,
-        config=config.data,
-        input_path=storage_config.get("input_path", "/data/input"),
-        scan_interval=int(scanner_config.get("scan_interval_seconds", 60)),
-        walker_threads=int(scanner_config.get("walker_threads", 8)),
-    )
-
-    logger.info("Worker initialization complete. Starting scanner and processing loop...")
-
-    # Run the scanner in a background thread (its start() is a blocking loop).
-    scanner_thread = threading.Thread(
-        target=scanner.start, name="directory-scanner", daemon=True
-    )
-    scanner_thread.start()
-
-    pool = WorkerPool(db, config.data, max_workers=1)
-
-    try:
-        pool.start()
-    except KeyboardInterrupt:
-        logger.info("Worker shutting down.")
-        pool.stop()
-        scanner.stop()
-
-
-if __name__ == "__main__":
-    main()

src/metrics.py

@@ -1,261 +0,0 @@
-"""
-Prometheus metrics definitions for VideoDetect.
-
-All metric objects are module-level singletons — import and call them directly.
-Call start_metrics_server() once at worker startup to expose /metrics.
-"""
-
-import logging
-import threading
-from typing import Optional
-
-logger = logging.getLogger(__name__)
-
-# Populated on first _ensure_prometheus() call; None if library unavailable.
-_prometheus_available: Optional[bool] = None
-_server_started = False
-_server_lock = threading.Lock()
-
-# ---------------------------------------------------------------------------
-# Lazy metric definitions — avoid import errors when prometheus_client absent
-# ---------------------------------------------------------------------------
-
-def _ensure_prometheus():
-    global _prometheus_available
-    if _prometheus_available is not None:
-        return _prometheus_available
-    try:
-        import prometheus_client  # noqa: F401
-        _prometheus_available = True
-    except ImportError:
-        logger.warning("prometheus_client not installed; metrics disabled")
-        _prometheus_available = False
-    return _prometheus_available
-
-
-class _NoOpMetric:
-    """Drop-in replacement used when prometheus_client is unavailable."""
-    def labels(self, **_):
-        return self
-    def inc(self, *_, **__):
-        pass
-    def set(self, *_, **__):
-        pass
-    def observe(self, *_, **__):
-        pass
-
-
-def _counter(name, doc, labelnames=()):
-    if not _ensure_prometheus():
-        return _NoOpMetric()
-    from prometheus_client import Counter
-    return Counter(name, doc, list(labelnames))
-
-
-def _gauge(name, doc, labelnames=()):
-    if not _ensure_prometheus():
-        return _NoOpMetric()
-    from prometheus_client import Gauge
-    return Gauge(name, doc, list(labelnames))
-
-
-def _histogram(name, doc, labelnames=(), buckets=None):
-    if not _ensure_prometheus():
-        return _NoOpMetric()
-    from prometheus_client import Histogram
-    kwargs = {"labelnames": list(labelnames)}
-    if buckets is not None:
-        kwargs["buckets"] = buckets
-    return Histogram(name, doc, **kwargs)
-
-
-# ---------------------------------------------------------------------------
-# Counters
-# ---------------------------------------------------------------------------
-
-videos_processed_total = _counter(
-    "videos_processed_total",
-    "Total videos processed",
-    ["routing_decision"],
-)
-
-videos_processed_by_model_total = _counter(
-    "videos_processed_by_model_total",
-    "Total videos processed per model version",
-    ["model_version"],
-)
-
-frames_extracted_total = _counter(
-    "frames_extracted_total",
-    "Total frames extracted across all videos",
-)
-
-faces_detected_total = _counter(
-    "faces_detected_total",
-    "Total face crops detected",
-)
-
-inference_errors_total = _counter(
-    "inference_errors_total",
-    "Total inference errors by type",
-    ["error_type"],
-)
-
-retry_attempts_total = _counter(
-    "retry_attempts_total",
-    "Total retry attempts by pipeline step",
-    ["step"],
-)
-
-# ---------------------------------------------------------------------------
-# Gauges
-# ---------------------------------------------------------------------------
-
-gpu_utilization_percent = _gauge(
-    "gpu_utilization_percent",
-    "GPU utilization as a percentage",
-    ["gpu_id"],
-)
-
-gpu_memory_used_bytes = _gauge(
-    "gpu_memory_used_bytes",
-    "GPU memory currently allocated in bytes",
-    ["gpu_id"],
-)
-
-gpu_memory_free_bytes = _gauge(
-    "gpu_memory_free_bytes",
-    "GPU memory currently free in bytes",
-    ["gpu_id"],
-)
-
-queue_depth_pending = _gauge(
-    "queue_depth_pending",
-    "Number of videos in PENDING state",
-)
-
-queue_depth_processing = _gauge(
-    "queue_depth_processing",
-    "Number of videos currently in PROCESSING state",
-)
-
-queue_depth_review = _gauge(
-    "queue_depth_review",
-    "Number of unannotated videos in the review queue",
-)
-
-scratch_usage_bytes = _gauge(
-    "scratch_usage_bytes",
-    "Scratch space used in bytes",
-)
-
-scratch_usage_percent = _gauge(
-    "scratch_usage_percent",
-    "Scratch space used as a percentage of total",
-)
-
-# ---------------------------------------------------------------------------
-# Histograms
-# ---------------------------------------------------------------------------
-
-_CONFIDENCE_BUCKETS = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0]
-
-video_processing_duration_seconds = _histogram(
-    "video_processing_duration_seconds",
-    "End-to-end processing time per video in seconds",
-    ["routing_decision"],
-)
-
-confidence_score_distribution = _histogram(
-    "confidence_score_distribution",
-    "Distribution of video-level confidence scores",
-    buckets=_CONFIDENCE_BUCKETS,
-)
-
-frame_count_per_video = _histogram(
-    "frame_count_per_video",
-    "Number of frames sampled per video",
-    buckets=[1, 2, 5, 10, 20, 50, 100, 200, 500],
-)
-
-face_count_per_video = _histogram(
-    "face_count_per_video",
-    "Number of face crops detected per video",
-    buckets=[0, 1, 2, 5, 10, 20, 50, 100],
-)
-
-# ---------------------------------------------------------------------------
-# Server startup
-# ---------------------------------------------------------------------------
-
-def start_metrics_server(port: int = 9090):
-    """Start the Prometheus HTTP metrics endpoint (idempotent)."""
-    global _server_started
-    with _server_lock:
-        if _server_started:
-            return
-        if not _ensure_prometheus():
-            logger.warning("Cannot start metrics server: prometheus_client unavailable")
-            return
-        try:
-            from prometheus_client import start_http_server
-            start_http_server(port)
-            _server_started = True
-            logger.info("Prometheus metrics server started on port %d", port)
-        except Exception as exc:
-            logger.error("Failed to start metrics server: %s", exc)
-
-
-# ---------------------------------------------------------------------------
-# Convenience update functions
-# ---------------------------------------------------------------------------
-
-def update_queue_depths(db_connector):
-    """Query the DB and refresh all queue-depth gauges."""
-    try:
-        rows = db_connector.fetchall(
-            """SELECT status, COUNT(*) AS cnt
-               FROM videos
-               WHERE status IN ('PENDING', 'PROCESSING')
-               GROUP BY status"""
-        )
-        counts = {r["status"]: int(r["cnt"]) for r in (rows or [])}
-        queue_depth_pending.set(counts.get("PENDING", 0))
-        queue_depth_processing.set(counts.get("PROCESSING", 0))
-
-        review_rows = db_connector.fetchall(
-            "SELECT COUNT(*) AS cnt FROM review_queue WHERE annotated = FALSE"
-        )
-        queue_depth_review.set(int((review_rows or [{"cnt": 0}])[0]["cnt"]))
-    except Exception as exc:
-        logger.warning("Failed to update queue depth metrics: %s", exc)
-
-
-def update_gpu_metrics():
-    """Poll nvidia-smi (via pynvml) and refresh GPU gauges."""
-    try:
-        import pynvml
-        pynvml.nvmlInit()
-        count = pynvml.nvmlDeviceGetCount()
-        for i in range(count):
-            handle = pynvml.nvmlDeviceGetHandleByIndex(i)
-            util = pynvml.nvmlDeviceGetUtilizationRates(handle)
-            mem = pynvml.nvmlDeviceGetMemoryInfo(handle)
-            gpu_id = str(i)
-            gpu_utilization_percent.labels(gpu_id=gpu_id).set(util.gpu)
-            gpu_memory_used_bytes.labels(gpu_id=gpu_id).set(mem.used)
-            gpu_memory_free_bytes.labels(gpu_id=gpu_id).set(mem.free)
-        pynvml.nvmlShutdown()
-    except Exception as exc:
-        logger.debug("GPU metrics unavailable: %s", exc)
-
-
-def update_scratch_metrics(scratch_path: str = "/scratch"):
-    """Measure scratch disk usage and refresh gauges."""
-    try:
-        import shutil
-        total, used, free = shutil.disk_usage(scratch_path)
-        scratch_usage_bytes.set(used)
-        scratch_usage_percent.set(round(used / max(total, 1) * 100, 2))
-    except Exception as exc:
-        logger.debug("Scratch metrics unavailable: %s", exc)

src/orchestrator.py

@@ -1,441 +0,0 @@
-"""
-Batch Orchestration Skeleton
-
-Manages job queue, worker pool, state transitions, and crash recovery.
-Ensures atomic state transitions and idempotent processing.
-"""
-
-import logging
-import time
-from concurrent.futures import ThreadPoolExecutor
-from datetime import datetime, timezone
-from enum import Enum
-from pathlib import Path
-from typing import Dict, List, Optional
-
-from batcher import DynamicBatcher
-from classifier import FaceClassifier
-from data_export import DataExporter
-from face_detector import Detection, FaceDetector
-from frame_sampler import FrameSampler
-from gpu_manager import GPUMemoryManager
-from prober import VideoProber
-from result_updater import ResultUpdater
-from scratch_manager import ScratchManager
-import aggregator
-import router
-
-logger = logging.getLogger(__name__)
-
-
-class JobStatus(Enum):
-    """Job status states."""
-    PENDING = "PENDING"
-    PROCESSING = "PROCESSING"
-    COMPLETED = "COMPLETED"
-    FAILED = "ERROR"
-    SKIPPED = "UNSCANNABLE"
-
-
-class Job:
-    """Represents a single video processing job."""
-
-    def __init__(
-        self,
-        video_id: int,
-        file_path: str,
-        priority: float = 0.0,
-        sampling_interval_seconds: Optional[int] = None,
-    ):
-        self.video_id = video_id
-        self.file_path = file_path
-        self.priority = priority  # Higher = more urgent (based on modification time)
-        self.sampling_interval_seconds = sampling_interval_seconds
-        self.status = JobStatus.PENDING
-        self.created_at = datetime.now(timezone.utc)
-        self.started_at: Optional[datetime] = None
-        self.completed_at: Optional[datetime] = None
-        self.attempts = 0
-        self.max_retries = 3
-        self.error_message: Optional[str] = None
-
-    def to_dict(self) -> dict:
-        return {
-            "video_id": self.video_id,
-            "file_path": self.file_path,
-            "priority": self.priority,
-            "status": self.status.value,
-            "created_at": self.created_at.isoformat(),
-            "started_at": self.started_at.isoformat() if self.started_at else None,
-            "completed_at": self.completed_at.isoformat() if self.completed_at else None,
-            "attempts": self.attempts,
-            "error_message": self.error_message,
-        }
-
-    def __repr__(self):
-        return f"Job(id={self.video_id}, path={self.file_path}, status={self.status.value})"
-
-
-class WorkerPool:
-    """Manage a pool of worker processes for parallel processing."""
-
-    def __init__(self, db_connector, config: dict, max_workers: int = 2):
-        self.db = db_connector
-        self.config = config
-        self.max_workers = max_workers
-        self._executor = ThreadPoolExecutor(max_workers=max_workers)
-        self._running = False
-        self._jobs_processed = 0
-        self._jobs_failed = 0
-        self._sampling_config = (config or {}).get("sampling", {})
-        self._storage_config = (config or {}).get("storage", {})
-        self._face_detection_config = (config or {}).get("face_detection", {})
-        self._batching_config = (config or {}).get("batching", {})
-        self._gpu_manager = GPUMemoryManager(
-            max_memory_gb=config.get("gpu", {}).get("max_memory_gb", 18.0),
-            reduce_threshold_gb=self._batching_config.get("vram_reduce_threshold_gb", 16.0),
-            increase_threshold_gb=self._batching_config.get("vram_increase_threshold_gb", 10.0),
-            initial_batch_size=self._batching_config.get("max_batch_size", 16),
-        )
-        self._face_detector = FaceDetector(
-            engine_path=self._face_detection_config.get("model_path", "/models/face_detector/face_detector.trt"),
-            input_size=int(self._face_detection_config.get("input_size", 640)),
-            confidence_threshold=float(self._face_detection_config.get("confidence_threshold", 0.25)),
-            iou_threshold=float(self._face_detection_config.get("iou_threshold", 0.45)),
-            max_faces_per_frame=int(self._face_detection_config.get("max_faces_per_frame", 10)),
-            max_faces_per_video=int(self._face_detection_config.get("max_faces_per_video", 100)),
-        )
-        classifier_config = (config or {}).get("classifier", {})
-        self._classifier = FaceClassifier(
-            engine_path=classifier_config.get("model_path", "/models/classifier/classifier.trt"),
-            temperature=float(classifier_config.get("temperature", 1.0)),
-            input_size=int(classifier_config.get("input_size", 224)),
-        )
-        self._aggregation_config = (config or {}).get("aggregation", {})
-        self._routing_config = (config or {}).get("routing", {})
-        model_version = self._get_active_model_version()
-        export_config = (config or {}).get("export", {})
-        self._result_updater = ResultUpdater(db_connector, model_version=model_version)
-        self._exporter = DataExporter(
-            output_path=export_config.get("output_path", "/data/output"),
-            model_version=model_version,
-            export_format=export_config.get("format", "parquet"),
-            compression=export_config.get("compression", "snappy"),
-            batch_size=int(export_config.get("batch_size", 100)),
-            include_frame_confidences=bool(export_config.get("include_frame_confidences", True)),
-        )
-
-    def _get_active_model_version(self) -> str:
-        """Return the currently active model version from the DB."""
-        try:
-            row = self.db.fetchone("SELECT version FROM models WHERE status = 'ACTIVE' LIMIT 1")
-            if row:
-                return row["version"]
-        except Exception as exc:
-            logger.debug("Could not fetch active model version: %s", exc)
-        return "v0.0.0-placeholder"
-
-    def start(self):
-        """Start the worker pool."""
-        self._running = True
-        logger.info("Worker pool starting with %d workers", self.max_workers)
-
-        while self._running:
-            # Get pending jobs
-            jobs = self._get_pending_jobs()
-
-            if jobs:
-                # Submit jobs to executor
-                for job in jobs:
-                    future = self._executor.submit(self._process_job, job)
-                    future.add_done_callback(self._on_job_complete)
-            else:
-                time.sleep(5)  # No jobs, wait
-
-    def stop(self):
-        """Stop the worker pool gracefully."""
-        self._running = False
-        logger.info("Worker pool stopping. Processed: %d, Failed: %d",
-                     self._jobs_processed, self._jobs_failed)
-        self._executor.shutdown(wait=True)
-
-    def _get_pending_jobs(self) -> List[Job]:
-        """Get pending jobs from DB with atomic locking."""
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-
-        # Atomic lock: update status from PENDING to PROCESSING
-        result = self.db.execute(
-            """UPDATE videos SET status = 'PROCESSING', last_processed_time = %s,
-               updated_at = %s
-               WHERE id IN (
-                   SELECT id FROM (
-                       SELECT id FROM videos
-                       WHERE status = 'PENDING'
-                       ORDER BY last_scan_time DESC
-                       LIMIT %s
-                   ) AS sub
-               )""",
-            (now, now, self.max_workers),
-            transaction=True,
-        )
-
-        if result == 0:
-            return []
-
-        # Get the locked jobs
-        locked = self.db.fetchall(
-            "SELECT id, file_path, last_scan_time FROM videos WHERE status = 'PROCESSING' AND last_processed_time = %s",
-            (now,),
-        )
-
-        return [
-            Job(
-                video_id=row["id"],
-                file_path=row["file_path"],
-                priority=row["last_scan_time"].timestamp(),
-            )
-            for row in locked
-        ]
-
-    def _process_job(self, job: Job) -> bool:
-        """Process a single job (placeholder - actual processing in later stories)."""
-        job.status = JobStatus.PROCESSING
-        job.started_at = datetime.now(timezone.utc)
-        job.attempts += 1
-
-        logger.info("Processing job: %s (attempt %d)", job, job.attempts)
-
-        try:
-            sampling_interval = job.sampling_interval_seconds or self._sampling_config.get("interval_seconds", 30)
-            scratch_base_path = self._storage_config.get("scratch_path", "/scratch")
-            scratch_manager = ScratchManager(
-                base_path=scratch_base_path,
-                video_id=str(job.video_id),
-                auto_cleanup=True,
-            )
-            frame_dir = scratch_manager.ensure_frame_dir()
-
-            prober = VideoProber(timeout=10)
-            metadata = prober.probe(job.file_path)
-            if metadata.is_unscannable or metadata.duration is None:
-                raise RuntimeError(metadata.error_message or "Video metadata could not be determined")
-
-            resolution = None
-            if metadata.resolution_w and metadata.resolution_h:
-                resolution = (metadata.resolution_w, metadata.resolution_h)
-
-            sampler = FrameSampler(
-                interval_seconds=int(sampling_interval),
-                quality=int(self._sampling_config.get("quality", 2)),
-                output_format=self._sampling_config.get("format", "jpeg"),
-            )
-            extracted_frames = sampler.extract_frames(
-                video_path=job.file_path,
-                output_dir=str(frame_dir),
-                duration=metadata.duration,
-                interval_seconds=int(sampling_interval),
-                resolution=resolution,
-                timeout_seconds=30,
-            )
-
-            if not extracted_frames:
-                scratch_manager.cleanup()
-                raise RuntimeError("No frames were extracted")
-
-            # Face detection on sampled frames
-            batch_size = self._gpu_manager.current_batch_size
-            detections_per_frame = self._face_detector.detect_faces(extracted_frames, batch_size=batch_size)
-
-            # Flatten and cap total faces per video
-            all_detections: List[Detection] = []
-            for frame_dets in detections_per_frame:
-                all_detections.extend(frame_dets)
-            all_detections = sorted(all_detections, key=lambda d: d.confidence, reverse=True)
-            all_detections = all_detections[: self._face_detection_config.get("max_faces_per_video", 100)]
-
-            # Extract face crops
-            crop_dir = scratch_manager.frame_dir.parent / "crops"
-            cropped_detections = self._face_detector.extract_crops(
-                all_detections,
-                output_dir=str(crop_dir),
-                crop_size=(224, 224),
-            )
-
-            if not cropped_detections:
-                logger.info("No faces detected for video %s; routing to SKIP", job.video_id)
-                routing_decision = router.SKIP
-                video_confidence = 0.0
-                frame_confidences: List[float] = []
-            else:
-                crop_paths = [d.crop_path for d in cropped_detections if d.crop_path]
-                frame_confidences = self._classifier.classify(
-                    crop_paths, batch_size=self._gpu_manager.current_batch_size
-                )
-                video_confidence = aggregator.aggregate(
-                    frame_confidences,
-                    strategy=self._aggregation_config.get("strategy", "max"),
-                    alpha=float(self._aggregation_config.get("alpha", 1.0)),
-                    beta=float(self._aggregation_config.get("beta", 0.1)),
-                    top_k=int(self._aggregation_config.get("top_k", 3)),
-                )
-                routing_decision = router.route(
-                    video_confidence,
-                    t_high=float(self._routing_config.get("T_high", 0.75)),
-                    t_low=float(self._routing_config.get("T_low", 0.45)),
-                )
-
-            logger.info(
-                "Video %s: C=%.4f routing=%s faces=%d frames=%d",
-                job.video_id, video_confidence, routing_decision,
-                len(cropped_detections), len(extracted_frames),
-            )
-
-            # Persist results atomically (video update + processing log) before cleanup
-            persisted = self._result_updater.persist(
-                video_id=job.video_id,
-                frame_count=len(extracted_frames),
-                confidence=video_confidence,
-                routing=routing_decision,
-                frame_confidences=frame_confidences,
-            )
-
-            if persisted:
-                job.status = JobStatus.COMPLETED
-                job.completed_at = datetime.now(timezone.utc)
-                self._exporter.add({
-                    "video_id":         job.video_id,
-                    "file_path":        job.file_path,
-                    "model_version":    self._result_updater.model_version,
-                    "sample_count":     len(extracted_frames),
-                    "confidence_scores": frame_confidences,
-                    "video_confidence": video_confidence,
-                    "routing":          routing_decision,
-                    "processed_at":     datetime.now(timezone.utc).isoformat(),
-                })
-
-            # Scratch cleanup only after successful persistence
-            scratch_manager.cleanup_all()
-            self._jobs_processed += 1
-            return True
-
-        except Exception as e:
-            logger.error("Job failed: %s - %s", job, e, exc_info=True)
-            job.error_message = str(e)
-
-            if job.attempts < job.max_retries:
-                # Retry
-                self._retry_job(job)
-                return False
-            else:
-                # Max retries exceeded
-                self._fail_job(job)
-                self._jobs_failed += 1
-                return False
-
-    def _complete_job(
-        self,
-        job: Job,
-        frame_count: Optional[int] = None,
-        confidence: Optional[float] = None,
-        routing: Optional[str] = None,
-    ):
-        """Mark a job as completed, persisting confidence and routing decision."""
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-        job.status = JobStatus.COMPLETED
-        job.completed_at = datetime.now(timezone.utc)
-
-        self.db.execute(
-            """UPDATE videos
-               SET status = 'COMPLETED',
-                   frame_count = %s,
-                   confidence_score = %s,
-                   routing_decision = %s,
-                   updated_at = %s
-               WHERE id = %s""",
-            (frame_count, confidence, routing, now, job.video_id),
-            transaction=True,
-        )
-        logger.info("Job completed: %s", job)
-
-    def _fail_job(self, job: Job):
-        """Mark a job as failed."""
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-        job.status = JobStatus.FAILED
-        job.completed_at = datetime.now(timezone.utc)
-
-        self.db.execute(
-            """UPDATE videos SET status = 'ERROR', error_message = %s, updated_at = %s
-               WHERE id = %s""",
-            (job.error_message, now, job.video_id),
-            transaction=True,
-        )
-        logger.error("Job failed permanently: %s", job)
-
-    def _retry_job(self, job: Job):
-        """Re-queue a job for retry."""
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-
-        self.db.execute(
-            """UPDATE videos SET status = 'PENDING', updated_at = %s
-               WHERE id = %s""",
-            (now, job.video_id),
-            transaction=True,
-        )
-        logger.info("Job re-queued for retry: %s (attempt %d/%d)",
-                     job, job.attempts, job.max_retries)
-
-    def _on_job_complete(self, future):
-        """Callback when a job completes."""
-        try:
-            future.result()
-        except Exception as e:
-            logger.error("Unhandled job error: %s", e)
-
-    def get_stats(self) -> dict:
-        """Get worker pool statistics."""
-        return {
-            "max_workers": self.max_workers,
-            "jobs_processed": self._jobs_processed,
-            "jobs_failed": self._jobs_failed,
-            "is_running": self._running,
-            "active_workers": self._executor._work_queue.qsize(),
-        }
-
-
-class CrashRecovery:
-    """Handle crash recovery and stale job detection."""
-
-    def __init__(self, db_connector, lock_timeout_minutes: int = 5):
-        self.db = db_connector
-        self.lock_timeout = lock_timeout_minutes
-
-    def recover_stale_jobs(self) -> int:
-        """Re-queue jobs stuck in PROCESSING beyond the lock timeout."""
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-        timeout = now.replace(minute=now.minute - self.lock_timeout)
-
-        result = self.db.execute(
-            """UPDATE videos SET status = 'PENDING', updated_at = %s
-               WHERE status = 'PROCESSING' AND last_processed_time < %s""",
-            (now, timeout),
-            transaction=True,
-        )
-
-        if result > 0:
-            logger.info("Recovered %d stale jobs", result)
-        return result
-
-    def check_health(self) -> dict:
-        """Check system health for crash recovery purposes."""
-        processing_count = self.db.fetchone(
-            "SELECT COUNT(*) as count FROM videos WHERE status = 'PROCESSING'"
-        )
-        pending_count = self.db.fetchone(
-            "SELECT COUNT(*) as count FROM videos WHERE status = 'PENDING'"
-        )
-
-        return {
-            "processing_count": processing_count["count"] if processing_count else 0,
-            "pending_count": pending_count["count"] if pending_count else 0,
-            "stale_jobs_recovered": self.recover_stale_jobs(),
-        }

src/prober.py

@@ -1,182 +0,0 @@
-"""
-Video Probing Module
-
-Extracts video metadata (codec, resolution, duration, frame rate) using ffprobe.
-Handles errors gracefully for corrupt or unsupported files.
-"""
-
-import json
-import logging
-import os
-import subprocess
-import time
-from dataclasses import dataclass, field
-from typing import Optional
-
-logger = logging.getLogger(__name__)
-
-
-@dataclass
-class VideoMetadata:
-    """Metadata extracted from a video file."""
-    file_path: str
-    codec: Optional[str] = None
-    resolution_w: Optional[int] = None
-    resolution_h: Optional[int] = None
-    duration: Optional[float] = None
-    frame_rate: Optional[float] = None
-    file_size: Optional[int] = None
-    status: str = "PENDING"
-    error_message: Optional[str] = None
-    probe_time: float = field(default_factory=time.time)
-
-    @property
-    def is_valid(self) -> bool:
-        return self.status == "PENDING" and self.codec is not None
-
-    @property
-    def is_unscannable(self) -> bool:
-        return self.status == "UNSCANNABLE"
-
-    def to_dict(self) -> dict:
-        return {
-            "file_path": self.file_path,
-            "codec": self.codec,
-            "resolution_w": self.resolution_w,
-            "resolution_h": self.resolution_h,
-            "duration": self.duration,
-            "frame_rate": self.frame_rate,
-            "file_size": self.file_size,
-            "status": self.status,
-            "error_message": self.error_message,
-        }
-
-
-class VideoProber:
-    """Probe video files using ffprobe to extract metadata."""
-
-    def __init__(self, timeout: int = 10):
-        self.timeout = timeout
-
-    def probe(self, file_path: str) -> VideoMetadata:
-        """Probe a video file and return metadata."""
-        metadata = VideoMetadata(file_path=file_path)
-
-        # Get file size
-        try:
-            metadata.file_size = self._get_file_size(file_path)
-        except OSError as e:
-            metadata.status = "UNSCANNABLE"
-            metadata.error_message = f"Cannot read file: {e}"
-            logger.warning("File access error for %s: %s", file_path, e)
-            return metadata
-
-        # Run ffprobe
-        try:
-            result = self._run_ffprobe(file_path)
-            if result is None:
-                metadata.status = "UNSCANNABLE"
-                metadata.error_message = "ffprobe timed out or failed"
-                logger.warning("ffprobe failed for %s", file_path)
-                return metadata
-
-            metadata = self._parse_ffprobe_output(result, file_path)
-        except subprocess.TimeoutExpired:
-            metadata.status = "UNSCANNABLE"
-            metadata.error_message = f"ffprobe timed out after {self.timeout}s"
-            logger.warning("ffprobe timeout for %s", file_path)
-        except Exception as e:
-            metadata.status = "UNSCANNABLE"
-            metadata.error_message = f"ffprobe error: {e}"
-            logger.warning("ffprobe error for %s: %s", file_path, e)
-
-        return metadata
-
-    def _run_ffprobe(self, file_path: str) -> Optional[str]:
-        """Run ffprobe and return JSON output."""
-        cmd = [
-            "ffprobe",
-            "-v", "error",
-            "-select_streams", "v:0",
-            "-show_entries", "stream=codec_name,width,height,r_frame_rate,duration",
-            "-show_entries", "format=duration,size",
-            "-of", "json",
-            file_path,
-        ]
-
-        try:
-            output = subprocess.run(
-                cmd,
-                capture_output=True,
-                text=True,
-                timeout=self.timeout,
-            )
-
-            if output.returncode != 0:
-                logger.debug("ffprobe stderr for %s: %s", file_path, output.stderr)
-                return None
-
-            return output.stdout
-        except subprocess.TimeoutExpired:
-            return None
-
-    def _parse_ffprobe_output(self, output: str, file_path: str) -> VideoMetadata:
-        """Parse ffprobe JSON output into VideoMetadata."""
-        metadata = VideoMetadata(file_path=file_path)
-
-        try:
-            data = json.loads(output)
-        except json.JSONDecodeError as e:
-            metadata.status = "UNSCANNABLE"
-            metadata.error_message = f"Invalid ffprobe output: {e}"
-            return metadata
-
-        # Extract stream info
-        streams = data.get("streams", [])
-        if not streams:
-            metadata.status = "UNSCANNABLE"
-            metadata.error_message = "No video streams found"
-            return metadata
-
-        stream = streams[0]
-        metadata.codec = stream.get("codec_name")
-        metadata.resolution_w = stream.get("width")
-        metadata.resolution_h = stream.get("height")
-
-        # Parse frame rate
-        r_frame_rate = stream.get("r_frame_rate")
-        if r_frame_rate and "/" in r_frame_rate:
-            try:
-                num, den = r_frame_rate.split("/")
-                metadata.frame_rate = float(num) / float(den) if float(den) != 0 else None
-            except (ValueError, ZeroDivisionError):
-                pass
-
-        # Extract format info
-        format_info = data.get("format", {})
-        if format_info.get("duration"):
-            try:
-                metadata.duration = float(format_info["duration"])
-            except (ValueError, TypeError):
-                pass
-
-        if format_info.get("size"):
-            try:
-                metadata.file_size = int(format_info["size"])
-            except (ValueError, TypeError):
-                pass
-
-        # Validate we got essential data
-        if not metadata.codec:
-            metadata.status = "UNSCANNABLE"
-            metadata.error_message = "No codec information available"
-        elif not metadata.duration:
-            metadata.status = "UNSCANNABLE"
-            metadata.error_message = "No duration information available"
-
-        return metadata
-
-    @staticmethod
-    def _get_file_size(file_path: str) -> int:
-        """Get file size in bytes."""
-        return os.path.getsize(file_path)

src/processing_logger.py

@@ -1,39 +0,0 @@
-"""Insert processing audit log entries within an existing DB transaction."""
-
-import json
-import logging
-from datetime import datetime, timezone
-from typing import List, Optional
-
-logger = logging.getLogger(__name__)
-
-
-def insert_log(
-    cursor,
-    video_id: int,
-    model_version: str,
-    frame_count: int,
-    confidence_score: Optional[float],
-    routing_decision: str,
-    frame_confidences: Optional[List[float]] = None,
-) -> None:
-    """Insert one row into processing_logs; must be called inside an open transaction."""
-    now = datetime.now(timezone.utc).replace(tzinfo=None)
-    confidence_scores_json = json.dumps(frame_confidences) if frame_confidences is not None else None
-
-    cursor.execute(
-        """INSERT INTO processing_logs
-               (video_id, model_version, frame_count, confidence_score,
-                confidence_scores, routing_decision, processed_at)
-           VALUES (%s, %s, %s, %s, %s, %s, %s)""",
-        (
-            video_id,
-            model_version,
-            frame_count or 0,
-            confidence_score,
-            confidence_scores_json,
-            routing_decision,
-            now,
-        ),
-    )
-    logger.debug("Inserted processing log for video %s (routing=%s)", video_id, routing_decision)

src/result_updater.py

@@ -1,75 +0,0 @@
-"""Atomic result persistence: update videos + insert processing_logs in one transaction."""
-
-import logging
-from datetime import datetime, timezone
-from typing import List, Optional
-
-import processing_logger
-
-logger = logging.getLogger(__name__)
-
-
-class ResultUpdater:
-    def __init__(self, db_connector, model_version: str = "v0.0.0-placeholder"):
-        self.db = db_connector
-        self.model_version = model_version
-
-    def persist(
-        self,
-        video_id: int,
-        frame_count: int,
-        confidence: float,
-        routing: str,
-        frame_confidences: Optional[List[float]] = None,
-    ) -> bool:
-        """Atomically update videos and insert a processing log.
-
-        Returns False without raising if the state guard prevents the update
-        (video is no longer in PROCESSING state).
-        """
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-
-        try:
-            with self.db.transaction() as conn:
-                cursor = conn.cursor()
-                try:
-                    cursor.execute(
-                        """UPDATE videos
-                           SET status             = 'COMPLETED',
-                               last_processed_time = %s,
-                               confidence_score    = %s,
-                               routing_decision    = %s,
-                               model_version       = %s,
-                               frame_count         = %s,
-                               updated_at          = %s
-                           WHERE id = %s AND status = 'PROCESSING'""",
-                        (now, confidence, routing, self.model_version, frame_count, now, video_id),
-                    )
-
-                    if cursor.rowcount == 0:
-                        logger.warning(
-                            "State guard: video %s is not PROCESSING; skipping update", video_id
-                        )
-                        return False
-
-                    processing_logger.insert_log(
-                        cursor,
-                        video_id=video_id,
-                        model_version=self.model_version,
-                        frame_count=frame_count,
-                        confidence_score=confidence,
-                        routing_decision=routing,
-                        frame_confidences=frame_confidences,
-                    )
-                finally:
-                    cursor.close()
-
-            logger.info(
-                "Persisted results for video %s: C=%.4f routing=%s model=%s",
-                video_id, confidence, routing, self.model_version,
-            )
-            return True
-
-        except Exception as exc:
-            logger.error("Failed to persist results for video %s: %s", video_id, exc)
-            raise

src/retry.py

@@ -1,114 +0,0 @@
-"""
-Retry decorator with exponential backoff for transient failures.
-
-Usage:
-    @retry(max_attempts=3, step="extract")
-    def extract_frames(...): ...
-
-Non-retryable error types are re-raised immediately without consuming attempts.
-"""
-
-import functools
-import logging
-import time
-from typing import Callable, Optional, Tuple, Type
-
-logger = logging.getLogger(__name__)
-
-# Error class names that should never be retried
-_NON_RETRYABLE_NAMES = frozenset({
-    "CodecUnsupportedError",
-    "FileCorruptError",
-    "InvalidPathError",
-    "FileNotFoundError",
-    "PermissionError",
-    "IsADirectoryError",
-})
-
-
-class RetryExhaustedError(Exception):
-    """Raised when all retry attempts have been exhausted."""
-
-
-def retry(
-    max_attempts: int = 3,
-    initial_delay: float = 1.0,
-    backoff_factor: float = 2.0,
-    exceptions: Tuple[Type[Exception], ...] = (Exception,),
-    step: str = "unknown",
-):
-    """
-    Decorator: retry ``func`` up to ``max_attempts`` times on retryable exceptions.
-
-    Non-retryable exceptions (see _NON_RETRYABLE_NAMES) propagate immediately.
-    Each retry waits ``initial_delay * backoff_factor ** attempt`` seconds.
-    """
-    def decorator(func: Callable) -> Callable:
-        @functools.wraps(func)
-        def wrapper(*args, **kwargs):
-            last_exc: Optional[Exception] = None
-
-            for attempt in range(max_attempts):
-                try:
-                    return func(*args, **kwargs)
-                except exceptions as exc:
-                    if _is_non_retryable(exc):
-                        logger.error(
-                            "Non-retryable error in step=%s (attempt %d/%d): %s: %s",
-                            step, attempt + 1, max_attempts,
-                            type(exc).__name__, exc,
-                        )
-                        raise
-
-                    last_exc = exc
-                    _increment_retry_counter(step)
-
-                    if attempt < max_attempts - 1:
-                        delay = initial_delay * (backoff_factor ** attempt)
-                        logger.warning(
-                            "Transient error in step=%s (attempt %d/%d), "
-                            "retrying in %.1fs: %s: %s",
-                            step, attempt + 1, max_attempts, delay,
-                            type(exc).__name__, exc,
-                        )
-                        time.sleep(delay)
-                    else:
-                        logger.error(
-                            "All %d attempts exhausted in step=%s: %s: %s",
-                            max_attempts, step, type(exc).__name__, exc,
-                        )
-
-            raise RetryExhaustedError(
-                f"step={step} failed after {max_attempts} attempts"
-            ) from last_exc
-
-        return wrapper
-    return decorator
-
-
-def retry_from_config(config, step: str = "unknown"):
-    """Build a ``@retry`` decorator from config.yaml monitoring.retry settings."""
-    mon = config.get_section("monitoring")
-    cfg = mon.get("retry", {})
-    return retry(
-        max_attempts=int(cfg.get("max_attempts", 3)),
-        initial_delay=float(cfg.get("initial_delay", 1.0)),
-        backoff_factor=float(cfg.get("backoff_factor", 2.0)),
-        step=step,
-    )
-
-
-# ---------------------------------------------------------------------------
-# Helpers
-# ---------------------------------------------------------------------------
-
-def _is_non_retryable(exc: Exception) -> bool:
-    return type(exc).__name__ in _NON_RETRYABLE_NAMES
-
-
-def _increment_retry_counter(step: str):
-    try:
-        from metrics import retry_attempts_total
-        retry_attempts_total.labels(step=step).inc()
-    except Exception:
-        pass  # metrics not available; don't break the retry logic

src/review_export.py

@@ -1,114 +0,0 @@
-"""
-Review queue export for backend scripts.
-
-Provides CSV and JSON export of annotated review data, with optional
-filtering by date range, model version, annotation status, and ground truth.
-Can be called standalone or imported from other src/ modules.
-"""
-
-import csv
-import io
-import json
-import logging
-from datetime import datetime
-from pathlib import Path
-from typing import Any, Dict, List, Optional
-
-logger = logging.getLogger(__name__)
-
-
-def fetch_annotated(
-    db_connector,
-    annotated_only: bool = True,
-    model_version: Optional[str] = None,
-    date_from: Optional[str] = None,
-    date_to: Optional[str] = None,
-    ground_truth: Optional[bool] = None,
-) -> List[Dict[str, Any]]:
-    """Query the review_queue and return matching records as plain dicts."""
-    clauses: List[str] = []
-    params: List[Any] = []
-
-    if annotated_only:
-        clauses.append("rq.annotated = TRUE")
-    if model_version:
-        clauses.append("v.model_version = %s")
-        params.append(model_version)
-    if date_from:
-        clauses.append("rq.annotated_at >= %s")
-        params.append(date_from)
-    if date_to:
-        clauses.append("rq.annotated_at < %s")
-        params.append(date_to)
-    if ground_truth is not None:
-        clauses.append("rq.ground_truth = %s")
-        params.append(bool(ground_truth))
-
-    where = ("WHERE " + " AND ".join(clauses)) if clauses else ""
-
-    rows = db_connector.fetchall(
-        f"""SELECT rq.video_id, v.file_path, rq.confidence_score, rq.routing_decision,
-                   v.model_version, rq.ground_truth, rq.annotated_at, rq.notes,
-                   pl.confidence_scores
-            FROM review_queue rq
-            JOIN videos v ON v.id = rq.video_id
-            LEFT JOIN (
-                SELECT video_id, confidence_scores,
-                       ROW_NUMBER() OVER (PARTITION BY video_id ORDER BY processed_at DESC) rn
-                FROM processing_logs
-            ) pl ON pl.video_id = rq.video_id AND pl.rn = 1
-            {where}
-            ORDER BY rq.annotated_at DESC""",
-        params if params else None,
-    )
-
-    return [_normalise(row) for row in (rows or [])]
-
-
-def _normalise(row: dict) -> Dict[str, Any]:
-    try:
-        scores = json.loads(row["confidence_scores"]) if row.get("confidence_scores") else []
-    except (json.JSONDecodeError, TypeError):
-        scores = []
-    at = row.get("annotated_at")
-    return {
-        "video_id": row["video_id"],
-        "file_path": row["file_path"],
-        "confidence_score": row["confidence_score"],
-        "routing_decision": row["routing_decision"],
-        "model_version": row.get("model_version"),
-        "ground_truth": bool(row["ground_truth"]) if row["ground_truth"] is not None else None,
-        "annotated_at": at.isoformat() if isinstance(at, datetime) else at,
-        "notes": row.get("notes"),
-        "contributing_frames": scores,
-    }
-
-
-def export_json(records: List[Dict[str, Any]], output_path: str) -> str:
-    """Write records to a UTF-8 JSON file; returns the path."""
-    path = Path(output_path)
-    path.parent.mkdir(parents=True, exist_ok=True)
-    path.write_text(json.dumps(records, default=str, indent=2), encoding="utf-8")
-    logger.info("Exported %d records to %s", len(records), path)
-    return str(path)
-
-
-def export_csv(records: List[Dict[str, Any]], output_path: str) -> str:
-    """Write records to a UTF-8 CSV file; returns the path."""
-    path = Path(output_path)
-    path.parent.mkdir(parents=True, exist_ok=True)
-    if not records:
-        path.write_text("", encoding="utf-8")
-        return str(path)
-
-    buf = io.StringIO()
-    writer = csv.DictWriter(buf, fieldnames=list(records[0].keys()))
-    writer.writeheader()
-    for r in records:
-        row = dict(r)
-        row["contributing_frames"] = json.dumps(row["contributing_frames"])
-        writer.writerow(row)
-
-    path.write_text(buf.getvalue(), encoding="utf-8")
-    logger.info("Exported %d records to %s", len(records), path)
-    return str(path)

src/router.py

@@ -1,26 +0,0 @@
-"""Threshold-based routing of video-level confidence to MATCH / REVIEW / SKIP."""
-
-import logging
-
-logger = logging.getLogger(__name__)
-
-MATCH  = "MATCH"
-REVIEW = "REVIEW"
-SKIP   = "SKIP"
-
-
-def route(
-    confidence: float,
-    t_high: float = 0.75,
-    t_low: float = 0.45,
-) -> str:
-    """Return the routing decision for a video-level confidence score."""
-    if confidence >= t_high:
-        decision = MATCH
-    elif confidence >= t_low:
-        decision = REVIEW
-    else:
-        decision = SKIP
-
-    logger.debug("route: C=%.4f t_high=%.2f t_low=%.2f → %s", confidence, t_high, t_low, decision)
-    return decision

src/scanner.py

@@ -1,537 +0,0 @@
-"""
-Directory Scanner Service
-
-Walks /data/input to discover new/modified files, computes hashes,
-probes video metadata, validates codecs, and syncs state to MariaDB.
-Supports incremental scanning for efficient 30TB corpus handling.
-"""
-
-import hashlib
-import json
-import logging
-import os
-import socket
-import threading
-import time
-import uuid
-from concurrent.futures import ThreadPoolExecutor, as_completed
-from datetime import datetime, timedelta, timezone
-from pathlib import Path
-from typing import Dict, List, Optional, Tuple
-
-from codec_validator import CodecValidator
-from prober import VideoMetadata, VideoProber
-
-logger = logging.getLogger(__name__)
-
-
-class DirectoryScanner:
-    """Scan directories for new/modified video files and sync to DB."""
-
-    VIDEO_EXTENSIONS = {'.mp4', '.mov', '.avi', '.mkv', '.webm', '.flv', '.wmv'}
-
-    def __init__(
-        self,
-        db_connector,
-        config: dict,
-        input_path: str = "/data/input",
-        scan_interval: int = 60,
-        walker_threads: int = 8,
-    ):
-        self.db = db_connector
-        self.config = config
-        self.input_path = Path(input_path)
-        self.scan_interval = scan_interval
-        self.walker_threads = walker_threads
-        self.prober = VideoProber(
-            timeout=config.get("scanner", {}).get("ffprobe_timeout_seconds", 10)
-        )
-        self.codec_validator = CodecValidator(
-            whitelist=set(config.get("codec", {}).get("whitelist", [])),
-            default_status_on_error=config.get("codec", {}).get("default_status_on_error", "UNSCANNABLE"),
-        )
-        self._running = False
-        self._scan_count = 0
-        self._total_files_discovered = 0
-        self._total_files_new = 0
-        self._total_files_modified = 0
-        self._total_files_unscannable = 0
-
-        # Single-instance guard: prevents overlapping scans both within this
-        # process (threading lock) and across replicas (DB lock with a lease).
-        self._instance_id = f"{socket.gethostname()}-{os.getpid()}-{uuid.uuid4().hex[:8]}"
-        self._lock_name = "directory_scanner"
-        self._lock_lease_seconds = int(
-            config.get("scanner", {}).get("lock_lease_seconds", 21600)
-        )
-        self._heartbeat_every = int(
-            config.get("scanner", {}).get("heartbeat_interval_files", 500)
-        )
-        self._scan_lock = threading.Lock()
-        self._lock_table_ensured = False
-
-    def start(self):
-        """Start the scanner loop.
-
-        Each cycle is guarded so that at most one scan runs at a time:
-        - an in-process threading lock prevents re-entrant scans, and
-        - a DB lock (with a lease) prevents overlapping scans across replicas.
-        A long-running scan keeps its lease alive via heartbeats, so the next
-        scheduled tick (or another replica) waits instead of starting a second
-        parallel scan.
-        """
-        self._running = True
-        logger.info(
-            "Scanner starting: input_path=%s interval=%ds threads=%d instance=%s",
-            self.input_path, self.scan_interval, self.walker_threads, self._instance_id,
-        )
-
-        while self._running:
-            # In-process re-entrancy guard: never run two scans at once.
-            if not self._scan_lock.acquire(blocking=False):
-                logger.warning("A scan is already in progress; skipping this cycle.")
-                self._sleep_interval()
-                continue
-
-            try:
-                if self.acquire_lock():
-                    try:
-                        self._run_scan()
-                    finally:
-                        self.release_lock()
-                else:
-                    logger.info(
-                        "Scanner lock held by another instance; skipping this cycle."
-                    )
-            except Exception as e:
-                logger.error("Scanner error: %s", e, exc_info=True)
-            finally:
-                self._scan_lock.release()
-
-            # Sleep until next scan
-            self._sleep_interval()
-
-    def _sleep_interval(self):
-        """Sleep for the scan interval, waking early if stopped."""
-        for _ in range(self.scan_interval):
-            if not self._running:
-                break
-            time.sleep(1)
-
-    def stop(self):
-        """Stop the scanner."""
-        self._running = False
-        logger.info("Scanner stopping. Total scans: %d", self._scan_count)
-
-    # ------------------------------------------------------------------
-    # Single-instance lock (cross-process / cross-replica guard)
-    # ------------------------------------------------------------------
-    def _ensure_lock_table(self):
-        """Create the scanner_lock table if it does not already exist."""
-        if self._lock_table_ensured:
-            return
-        self.db.execute(
-            """CREATE TABLE IF NOT EXISTS scanner_lock (
-                   lock_name VARCHAR(64) PRIMARY KEY,
-                   owner VARCHAR(128) NOT NULL,
-                   locked_at DATETIME NOT NULL,
-                   lease_seconds INT NOT NULL DEFAULT 21600
-               ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci"""
-        )
-        self._lock_table_ensured = True
-
-    def acquire_lock(self) -> bool:
-        """Attempt to acquire the cross-process scanner lock.
-
-        Returns True if this instance now owns the lock, False otherwise.
-        A stale lock (held longer than the lease) is taken over so a crashed
-        scanner does not block scanning forever.
-        """
-        try:
-            self._ensure_lock_table()
-        except Exception as e:
-            logger.warning("Could not ensure scanner_lock table: %s", e)
-            return True  # fail-open: keep scanning rather than block entirely
-
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-
-        # 1) Try to insert a fresh lock row.
-        try:
-            self.db.execute(
-                """INSERT INTO scanner_lock (lock_name, owner, locked_at, lease_seconds)
-                   VALUES (%s, %s, %s, %s)""",
-                (self._lock_name, self._instance_id, now, self._lock_lease_seconds),
-                transaction=True,
-            )
-            logger.info("Acquired scanner lock (fresh). owner=%s", self._instance_id)
-            return True
-        except Exception:
-            # Row already exists -> fall through to steal-if-stale.
-            pass
-
-        # 2) Take over the lock if it is stale or already ours.
-        stale_before = now - timedelta(seconds=self._lock_lease_seconds)
-        try:
-            affected = self.db.execute(
-                """UPDATE scanner_lock
-                   SET owner = %s, locked_at = %s
-                   WHERE lock_name = %s
-                     AND (owner = %s OR locked_at < %s)""",
-                (self._instance_id, now, self._lock_name, self._instance_id, stale_before),
-                transaction=True,
-            )
-            if affected and affected > 0:
-                logger.info("Acquired scanner lock (stale takeover). owner=%s",
-                            self._instance_id)
-                return True
-        except Exception as e:
-            logger.warning("Failed to check scanner lock: %s", e)
-            return True  # fail-open
-
-        logger.info("Scanner lock held by another instance; not acquiring.")
-        return False
-
-    def release_lock(self):
-        """Release the scanner lock if we own it."""
-        try:
-            self.db.execute(
-                """DELETE FROM scanner_lock WHERE lock_name = %s AND owner = %s""",
-                (self._lock_name, self._instance_id),
-                transaction=True,
-            )
-        except Exception as e:
-            logger.warning("Failed to release scanner lock: %s", e)
-
-    def _heartbeat(self):
-        """Refresh the lock lease so a long-running scan is not stolen."""
-        try:
-            now = datetime.now(timezone.utc).replace(tzinfo=None)
-            self.db.execute(
-                """UPDATE scanner_lock SET locked_at = %s
-                   WHERE lock_name = %s AND owner = %s""",
-                (now, self._lock_name, self._instance_id),
-                transaction=True,
-            )
-        except Exception as e:
-            logger.debug("Scanner lock heartbeat failed: %s", e)
-
-    def _run_scan(self):
-        """Execute a single scan cycle."""
-        scan_start = time.time()
-        self._scan_count += 1
-
-        logger.info("Scan #%d starting...", self._scan_count)
-
-        # Get last scan time from DB
-        last_scan = self._get_last_scan_time()
-
-        # Discover files
-        files_to_process, files_removed = self._discover_files(last_scan)
-
-        # Process files in parallel
-        results = self._process_files_batch(files_to_process)
-
-        # Update DB
-        scan_result = self._update_db(results, files_removed)
-
-        # Record scan history
-        duration = time.time() - scan_start
-        self._record_scan_history(scan_result, duration)
-
-        logger.info(
-            "Scan #%d complete: discovered=%d new=%d modified=%d unscannable=%d removed=%d duration=%.1fs",
-            self._scan_count,
-            scan_result["discovered"],
-            scan_result["new"],
-            scan_result["modified"],
-            scan_result["unscannable"],
-            len(files_removed),
-            duration,
-        )
-
-    def _get_last_scan_time(self) -> Optional[datetime]:
-        """Get the last scan time from DB."""
-        result = self.db.fetchone(
-            "SELECT scan_end FROM scan_history ORDER BY id DESC LIMIT 1"
-        )
-        if result and result.get("scan_end"):
-            return result["scan_end"]
-        return None
-
-    def _discover_files(self, last_scan: Optional[datetime]) -> Tuple[List[Path], List[str]]:
-        """Discover new and modified files in the input directory."""
-        files_to_process = []
-        files_removed = []
-
-        if not self.input_path.exists():
-            logger.warning("Input path does not exist: %s", self.input_path)
-            return files_to_process, files_removed
-
-        # Get files from DB for comparison
-        if last_scan:
-            # Only check files modified after last scan
-            db_files = self.db.fetchall(
-                "SELECT file_path, last_scan_time FROM videos WHERE last_scan_time > %s",
-                (last_scan,),
-            )
-            db_paths = {row["file_path"] for row in db_files}
-
-            # Check for modified files
-            for row in db_files:
-                file_path = Path(row["file_path"])
-                if file_path.exists() and file_path.is_file():
-                    try:
-                        mtime = datetime.fromtimestamp(
-                            file_path.stat().st_mtime, tz=timezone.utc
-                        )
-                        if mtime > row["last_scan_time"]:
-                            files_to_process.append(file_path)
-                    except OSError:
-                        pass
-        else:
-            # Full scan - walk the directory
-            logger.info("Full scan (no previous scan found). Walking %s...", self.input_path)
-            for root, dirs, filenames in os.walk(self.input_path):
-                for filename in filenames:
-                    ext = Path(filename).suffix.lower()
-                    if ext in self.VIDEO_EXTENSIONS:
-                        file_path = Path(root) / filename
-                        files_to_process.append(file_path)
-
-        # Check for removed files (if we have a last scan)
-        if last_scan:
-            all_db_files = self.db.fetchall(
-                "SELECT file_path FROM videos WHERE status NOT IN ('REMOVED', 'ERROR')",
-            )
-            current_paths = {str(f) for f in files_to_process}
-            for row in all_db_files:
-                fp = row["file_path"]
-                if fp not in current_paths and Path(fp).exists():
-                    # File still exists but not in current walk - might be in a new directory
-                    pass
-                elif fp not in current_paths and not Path(fp).exists():
-                    files_removed.append(fp)
-
-        logger.info("Discovered %d files to process, %d files removed",
-                     len(files_to_process), len(files_removed))
-        return files_to_process, files_removed
-
-    def _process_files_batch(self, files: List[Path]) -> List[dict]:
-        """Process a batch of files in parallel."""
-        results = []
-        processed = 0
-
-        with ThreadPoolExecutor(max_workers=self.walker_threads) as executor:
-            future_to_file = {
-                executor.submit(self._process_single_file, f): f
-                for f in files
-            }
-
-            for future in as_completed(future_to_file):
-                file_path = future_to_file[future]
-                try:
-                    result = future.result()
-                    results.append(result)
-                except Exception as e:
-                    logger.error("Error processing %s: %s", file_path, e, exc_info=True)
-                    results.append({
-                        "file_path": str(file_path),
-                        "status": "ERROR",
-                        "error_message": str(e),
-                    })
-
-                processed += 1
-                # Keep the single-instance lock alive during long scans.
-                if self._heartbeat_every and processed % self._heartbeat_every == 0:
-                    self._heartbeat()
-
-        return results
-
-    def _process_single_file(self, file_path: Path) -> dict:
-        """Process a single file: hash, probe, validate, determine status."""
-        file_path_str = str(file_path)
-        result = {"file_path": file_path_str}
-
-        try:
-            # Compute hash
-            file_hash = self._compute_hash(file_path_str)
-            result["file_hash"] = file_hash
-
-            # Check for duplicate
-            existing = self.db.fetchone(
-                "SELECT id, status FROM videos WHERE file_hash = %s",
-                (file_hash,),
-            )
-            if existing:
-                result["status"] = existing["status"]
-                result["video_id"] = existing["id"]
-                result["action"] = "duplicate"
-                logger.debug("Duplicate file found: %s (id=%d, status=%s)",
-                             file_path_str, existing["id"], existing["status"])
-                return result
-
-            # Probe video
-            metadata = self.prober.probe(file_path_str)
-            result.update(metadata.to_dict())
-
-            # Validate codec
-            if metadata.is_valid:
-                is_supported, reason = self.codec_validator.validate(metadata.codec)
-                if is_supported:
-                    result["status"] = "PENDING"
-                    result["action"] = "new"
-                    self._total_files_new += 1
-                else:
-                    result["status"] = "UNSCANNABLE"
-                    result["error_message"] = f"Unsupported codec: {metadata.codec} ({reason})"
-                    result["action"] = "unscannable"
-                    self._total_files_unscannable += 1
-                    logger.warning("Unsupported codec for %s: %s", file_path_str, metadata.codec)
-            else:
-                result["status"] = "UNSCANNABLE"
-                result["action"] = "unscannable"
-                self._total_files_unscannable += 1
-
-        except Exception as e:
-            result["status"] = "ERROR"
-            result["error_message"] = str(e)
-            logger.error("Error processing %s: %s", file_path_str, e, exc_info=True)
-
-        return result
-
-    def _compute_hash(self, file_path: str) -> str:
-        """Compute SHA-256 hash of the first 1MB of a file."""
-        chunk_size = 1024 * 1024  # 1MB
-        sha256 = hashlib.sha256()
-
-        try:
-            with open(file_path, "rb") as f:
-                chunk = f.read(chunk_size)
-                if chunk:
-                    sha256.update(chunk)
-            return sha256.hexdigest()
-        except (OSError, IOError) as e:
-            logger.error("Cannot hash file %s: %s", file_path, e)
-            return hashlib.sha256(file_path.encode()).hexdigest()  # fallback
-
-    def _update_db(self, results: List[dict], files_removed: List[str]) -> dict:
-        """Update database with scan results."""
-        stats = {"discovered": len(results), "new": 0, "modified": 0, "unscannable": 0}
-
-        # Batch insert new files
-        new_files = [r for r in results if r.get("action") == "new"]
-        if new_files:
-            self._batch_insert_new_files(new_files)
-            stats["new"] = len(new_files)
-
-        # Update unscannable files
-        unscannable = [r for r in results if r.get("action") == "unscannable" and r.get("video_id")]
-        if unscannable:
-            self._batch_update_status(unscannable, "UNSCANNABLE")
-            stats["unscannable"] = len(unscannable)
-
-        # Mark removed files
-        if files_removed:
-            self._mark_files_removed(files_removed)
-            stats["removed"] = len(files_removed)
-
-        return stats
-
-    def _batch_insert_new_files(self, files: List[dict]):
-        """Batch insert new files into DB."""
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-
-        for file_info in files:
-            try:
-                self.db.execute(
-                    """INSERT INTO videos
-                       (file_path, file_hash, resolution_w, resolution_h, codec,
-                        duration, status, last_scan_time, created_at, updated_at)
-                       VALUES (%s, %s, %s, %s, %s, %s, %s, %s, %s, %s)""",
-                    (
-                        file_info["file_path"],
-                        file_info["file_hash"],
-                        file_info.get("resolution_w"),
-                        file_info.get("resolution_h"),
-                        file_info.get("codec"),
-                        file_info.get("duration"),
-                        "PENDING",
-                        now,
-                        now,
-                        now,
-                    ),
-                    transaction=True,
-                )
-            except Exception as e:
-                logger.error("Failed to insert %s: %s", file_info["file_path"], e)
-
-    def _batch_update_status(self, files: List[dict], status: str):
-        """Batch update file status."""
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-
-        for file_info in files:
-            video_id = file_info.get("video_id")
-            if video_id:
-                try:
-                    self.db.execute(
-                        """UPDATE videos SET status = %s, error_message = %s,
-                           last_scan_time = %s, updated_at = %s
-                           WHERE id = %s""",
-                        (status, file_info.get("error_message"), now, now, video_id),
-                        transaction=True,
-                    )
-                except Exception as e:
-                    logger.error("Failed to update %s: %s", file_info["file_path"], e)
-
-    def _mark_files_removed(self, file_paths: List[str]):
-        """Mark files as removed in DB."""
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-
-        for fp in file_paths:
-            try:
-                self.db.execute(
-                    "UPDATE videos SET status = 'REMOVED', last_scan_time = %s, updated_at = %s WHERE file_path = %s",
-                    (now, now, fp),
-                    transaction=True,
-                )
-            except Exception as e:
-                logger.error("Failed to mark %s as removed: %s", fp, e)
-
-    def _record_scan_history(self, stats: dict, duration: float):
-        """Record scan history in DB."""
-        now = datetime.now(timezone.utc).replace(tzinfo=None)
-
-        self.db.execute(
-            """INSERT INTO scan_history
-               (scan_start, scan_end, files_discovered, files_new, files_modified,
-                files_removed, files_unscannable, duration_seconds, status)
-               VALUES (%s, %s, %s, %s, %s, %s, %s, %s, %s)""",
-            (
-                now,
-                now,
-                stats.get("discovered", 0),
-                stats.get("new", 0),
-                stats.get("modified", 0),
-                stats.get("removed", 0),
-                stats.get("unscannable", 0),
-                duration,
-                "COMPLETED",
-            ),
-            transaction=True,
-        )
-
-    def get_stats(self) -> dict:
-        """Get scanner statistics."""
-        return {
-            "scan_count": self._scan_count,
-            "total_files_discovered": self._total_files_discovered,
-            "total_files_new": self._total_files_new,
-            "total_files_modified": self._total_files_modified,
-            "total_files_unscannable": self._total_files_unscannable,
-            "instance_id": self._instance_id,
-            "scan_interval_seconds": self.scan_interval,
-            "lock_lease_seconds": self._lock_lease_seconds,
-            "scan_in_progress": self._scan_lock.locked(),
-            "input_path": str(self.input_path),
-            "is_running": self._running,
-        }

src/scratch_manager.py

@@ -1,42 +0,0 @@
-"""Utilities for managing scratch space used by frame sampling."""
-
-import logging
-import shutil
-from pathlib import Path
-from typing import Optional
-
-logger = logging.getLogger(__name__)
-
-
-class ScratchManager:
-    """Create and clean per-video frame scratch directories."""
-
-    def __init__(self, base_path: str = "/scratch", video_id: Optional[str] = None, auto_cleanup: bool = True):
-        self.base_path = Path(base_path)
-        self.video_id = video_id or "unknown"
-        self.auto_cleanup = auto_cleanup
-        self.frame_dir = self.base_path / self.video_id / "frames"
-
-    def ensure_frame_dir(self) -> Path:
-        """Create the per-video scratch directory if it does not exist."""
-        self.frame_dir.mkdir(parents=True, exist_ok=True)
-        return self.frame_dir
-
-    def cleanup(self) -> None:
-        """Remove the frame scratch directory only."""
-        if self.auto_cleanup and self.frame_dir.exists():
-            shutil.rmtree(self.frame_dir)
-            logger.info("Cleaned scratch frames for %s", self.video_id)
-
-    def cleanup_all(self) -> None:
-        """Remove the entire per-video scratch directory (frames + crops)."""
-        video_root = self.base_path / self.video_id
-        if self.auto_cleanup and video_root.exists():
-            shutil.rmtree(video_root)
-            logger.info("Cleaned all scratch data for %s", self.video_id)
-
-    def usage_bytes(self) -> int:
-        """Return the total size of the scratch directory in bytes."""
-        if not self.frame_dir.exists():
-            return 0
-        return sum(path.stat().st_size for path in self.frame_dir.rglob("*") if path.is_file())

tests/test_story_03.py

@@ -1,67 +0,0 @@
-import shutil
-import sys
-import tempfile
-import unittest
-from pathlib import Path
-from unittest.mock import patch
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "src"))
-
-from frame_sampler import FrameSampler, calculate_timestamps
-from scratch_manager import ScratchManager
-
-
-class Story03SamplingTests(unittest.TestCase):
-    def test_calculate_timestamps_uses_uniform_temporal_spacing(self):
-        stamps = calculate_timestamps(duration=90.0, interval=30.0)
-        self.assertEqual(stamps, [0.0, 30.0, 60.0])
-
-    def test_frame_sampler_builds_ffmpeg_command_with_scale_and_jpeg_output(self):
-        sampler = FrameSampler(interval_seconds=30, quality=2, output_format="jpeg")
-        command = sampler._build_ffmpeg_command(
-            video_path="/tmp/video.mp4",
-            output_path="/tmp/out.jpg",
-            timestamp=12.5,
-            resolution=(3840, 2160),
-        )
-
-        self.assertIn("ffmpeg", command[0])
-        self.assertIn("-ss", command)
-        self.assertIn("-vframes", command)
-        self.assertIn("scale=1920:1080", " ".join(command))
-        self.assertTrue(command[-1].endswith("out.jpg"))
-
-    def test_frame_sampler_uses_subprocess_and_returns_output_path(self):
-        sampler = FrameSampler(interval_seconds=30, quality=2, output_format="jpeg")
-        with tempfile.TemporaryDirectory() as tmpdir:
-            video_path = Path(tmpdir) / "sample.mp4"
-            output_path = Path(tmpdir) / "frame.jpg"
-            video_path.write_bytes(b"fake")
-
-            def _mock_run(*args, **kwargs):
-                output_path.write_bytes(b"frame")
-                return type("Completed", (), {"returncode": 0, "stdout": b"", "stderr": b""})()
-
-            with patch("subprocess.run", side_effect=_mock_run):
-                result = sampler.extract_frame(
-                    video_path=str(video_path),
-                    output_path=str(output_path),
-                    timestamp=15.0,
-                    resolution=(1280, 720),
-                )
-
-            self.assertTrue(result)
-            self.assertEqual(output_path.name, Path(result).name)
-
-    def test_scratch_manager_cleans_up_frames_after_processing(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            manager = ScratchManager(base_path=tmpdir, video_id="video-1", auto_cleanup=True)
-            frame_dir = manager.ensure_frame_dir()
-            (frame_dir / "video-1_1000.jpg").write_bytes(b"frame")
-
-            manager.cleanup()
-            self.assertFalse(frame_dir.exists())
-
-
-if __name__ == "__main__":
-    unittest.main()

tests/test_story_04.py

@@ -1,98 +0,0 @@
-import sys
-import tempfile
-import unittest
-from pathlib import Path
-from unittest.mock import MagicMock, patch
-
-import numpy as np
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "src"))
-
-from batcher import DynamicBatcher
-from face_detector import Detection, FaceDetector
-from gpu_manager import GPUMemoryManager
-
-
-class Story04FaceDetectionTests(unittest.TestCase):
-    def test_gpu_manager_respects_memory_thresholds(self):
-        manager = GPUMemoryManager(
-            max_memory_gb=18.0,
-            reduce_threshold_gb=16.0,
-            increase_threshold_gb=10.0,
-            initial_batch_size=16,
-        )
-
-        with patch.object(manager, "get_memory_stats", return_value=(17.0, 18.0)):
-            manager.adjust_batch_size()
-            self.assertLess(manager.current_batch_size, 16)
-
-        with patch.object(manager, "get_memory_stats", return_value=(8.0, 10.0)):
-            manager.adjust_batch_size()
-            self.assertGreater(manager.current_batch_size, manager.min_batch_size)
-
-    def test_dynamic_batcher_flushes_when_batch_full(self):
-        processed_batches = []
-
-        def process_fn(batch):
-            processed_batches.append(batch)
-            return [len(batch)]
-
-        batcher = DynamicBatcher(
-            process_fn=process_fn,
-            max_batch_size=3,
-            batch_timeout_ms=10000.0,
-        )
-
-        for i in range(5):
-            batcher.add(i)
-
-        self.assertEqual(len(processed_batches), 1)
-        self.assertEqual(processed_batches[0], [0, 1, 2])
-        self.assertEqual(batcher.queued_count, 2)
-
-    def test_face_detector_nms_removes_overlapping_boxes(self):
-        detector = FaceDetector(
-            engine_path="/nonexistent/model.trt",
-            confidence_threshold=0.1,
-            iou_threshold=0.45,
-            max_faces_per_frame=10,
-        )
-
-        duplicates = [
-            Detection("frame.jpg", 10, 10, 50, 50, 0.9),
-            Detection("frame.jpg", 12, 12, 48, 48, 0.8),
-            Detection("frame.jpg", 100, 100, 150, 150, 0.75),
-        ]
-
-        kept = detector._nms(duplicates)
-        self.assertEqual(len(kept), 2)
-        self.assertAlmostEqual(kept[0].confidence, 0.9, places=5)
-
-    def test_face_detector_extracts_and_resizes_crops(self):
-        detector = FaceDetector(
-            engine_path="/nonexistent/model.trt",
-            input_size=640,
-            confidence_threshold=0.25,
-        )
-
-        with tempfile.TemporaryDirectory() as tmpdir:
-            from PIL import Image
-
-            frame_path = Path(tmpdir) / "frame.jpg"
-            image = Image.new("RGB", (640, 480), color=(100, 150, 200))
-            image.save(frame_path)
-
-            detections = [
-                Detection(str(frame_path), 0, 0, 640, 480, 0.95),
-            ]
-            cropped = detector.extract_crops(detections, output_dir=tmpdir, crop_size=(224, 224))
-
-            self.assertEqual(len(cropped), 1)
-            self.assertTrue(Path(cropped[0].crop_path).exists())
-
-            with Image.open(cropped[0].crop_path) as crop:
-                self.assertEqual(crop.size, (224, 224))
-
-
-if __name__ == "__main__":
-    unittest.main()

tests/test_story_05.py

@@ -1,83 +0,0 @@
-import sys
-import unittest
-from pathlib import Path
-
-import numpy as np
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "src"))
-
-import aggregator
-import router
-from classifier import FaceClassifier, calibrated_softmax
-
-
-class Story05ClassifierTests(unittest.TestCase):
-    def test_calibrated_softmax_sums_to_one(self):
-        logits = np.array([[2.0, 1.0], [-1.0, 3.0]], dtype=np.float32)
-        probs = calibrated_softmax(logits, temperature=1.0)
-        np.testing.assert_allclose(probs.sum(axis=1), [1.0, 1.0], atol=1e-6)
-
-    def test_temperature_scaling_raises_lower_confidence_entropy(self):
-        logits = np.array([[2.0, 0.5]], dtype=np.float32)
-        sharp = calibrated_softmax(logits, temperature=0.5)
-        soft  = calibrated_softmax(logits, temperature=2.0)
-        # higher temperature → softer distribution (target class prob moves toward 0.5)
-        self.assertGreater(sharp[0, 0], soft[0, 0])
-
-    def test_classifier_placeholder_returns_neutral_probability(self):
-        clf = FaceClassifier(engine_path="/nonexistent/model.trt", temperature=1.0)
-        probs = clf.classify([])
-        self.assertEqual(probs, [])
-
-    def test_classifier_placeholder_single_crop_returns_half(self):
-        import tempfile
-        from PIL import Image
-
-        clf = FaceClassifier(engine_path="/nonexistent/model.trt", temperature=1.0)
-        with tempfile.TemporaryDirectory() as tmpdir:
-            crop = Path(tmpdir) / "crop.jpg"
-            Image.new("RGB", (224, 224)).save(crop)
-            probs = clf.classify([str(crop)])
-        # placeholder logits are all zeros → softmax → 0.5 for each class
-        self.assertAlmostEqual(probs[0], 0.5, places=5)
-
-
-class Story05AggregatorTests(unittest.TestCase):
-    def test_max_strategy(self):
-        self.assertAlmostEqual(aggregator.aggregate([0.3, 0.8, 0.6], strategy="max"), 0.8)
-
-    def test_empty_confidences_returns_zero(self):
-        self.assertEqual(aggregator.aggregate([], strategy="max"), 0.0)
-
-    def test_top_k_mean(self):
-        result = aggregator.aggregate([0.1, 0.9, 0.5, 0.8], strategy="top_k_mean", top_k=2)
-        self.assertAlmostEqual(result, (0.9 + 0.8) / 2, places=5)
-
-    def test_weighted_mean_clamps_to_unit_interval(self):
-        result = aggregator.aggregate([1.0, 1.0], strategy="weighted_mean", alpha=100.0, beta=0.0)
-        self.assertLessEqual(result, 1.0)
-        self.assertGreaterEqual(result, 0.0)
-
-
-class Story05RouterTests(unittest.TestCase):
-    def test_match_at_high_threshold(self):
-        self.assertEqual(router.route(0.75), router.MATCH)
-
-    def test_review_between_thresholds(self):
-        self.assertEqual(router.route(0.60), router.REVIEW)
-
-    def test_skip_below_low_threshold(self):
-        self.assertEqual(router.route(0.44), router.SKIP)
-
-    def test_inclusive_high_threshold_boundary(self):
-        self.assertEqual(router.route(0.75, t_high=0.75, t_low=0.45), router.MATCH)
-
-    def test_inclusive_low_threshold_boundary(self):
-        self.assertEqual(router.route(0.45, t_high=0.75, t_low=0.45), router.REVIEW)
-
-    def test_no_faces_zero_confidence_routes_skip(self):
-        self.assertEqual(router.route(0.0), router.SKIP)
-
-
-if __name__ == "__main__":
-    unittest.main()

tests/test_story_06.py

@@ -1,148 +0,0 @@
-import json
-import sys
-import tempfile
-import unittest
-from pathlib import Path
-from unittest.mock import MagicMock, call, patch
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "src"))
-
-import processing_logger
-from data_export import DataExporter
-from result_updater import ResultUpdater
-from scratch_manager import ScratchManager
-
-
-class Story06ProcessingLoggerTests(unittest.TestCase):
-    def test_insert_log_executes_correct_sql(self):
-        cursor = MagicMock()
-        processing_logger.insert_log(
-            cursor,
-            video_id=42,
-            model_version="v1.0.0",
-            frame_count=8,
-            confidence_score=0.82,
-            routing_decision="MATCH",
-            frame_confidences=[0.80, 0.82, 0.85],
-        )
-        cursor.execute.assert_called_once()
-        sql, params = cursor.execute.call_args[0]
-        self.assertIn("INSERT INTO processing_logs", sql)
-        self.assertEqual(params[0], 42)                        # video_id
-        self.assertEqual(params[1], "v1.0.0")                  # model_version
-        self.assertEqual(params[2], 8)                         # frame_count
-        self.assertAlmostEqual(params[3], 0.82)                # confidence_score
-        scores = json.loads(params[4])                         # confidence_scores JSON
-        self.assertEqual(scores, [0.80, 0.82, 0.85])
-        self.assertEqual(params[5], "MATCH")                   # routing_decision
-
-    def test_insert_log_null_frame_confidences(self):
-        cursor = MagicMock()
-        processing_logger.insert_log(cursor, 1, "v0", 0, None, "SKIP")
-        _, params = cursor.execute.call_args[0]
-        self.assertIsNone(params[4])  # confidence_scores column
-
-
-class Story06ResultUpdaterTests(unittest.TestCase):
-    def _make_db(self, rowcount=1):
-        cursor = MagicMock()
-        cursor.rowcount = rowcount
-        conn = MagicMock()
-        conn.cursor.return_value = cursor
-        db = MagicMock()
-        db.transaction.return_value.__enter__ = MagicMock(return_value=conn)
-        db.transaction.return_value.__exit__ = MagicMock(return_value=False)
-        return db, cursor
-
-    def test_persist_returns_true_when_update_succeeds(self):
-        db, cursor = self._make_db(rowcount=1)
-        updater = ResultUpdater(db, model_version="v1.0.0")
-        result = updater.persist(
-            video_id=7, frame_count=5, confidence=0.9,
-            routing="MATCH", frame_confidences=[0.9],
-        )
-        self.assertTrue(result)
-
-    def test_persist_returns_false_on_state_guard_miss(self):
-        db, cursor = self._make_db(rowcount=0)
-        updater = ResultUpdater(db, model_version="v1.0.0")
-        result = updater.persist(
-            video_id=7, frame_count=5, confidence=0.9,
-            routing="MATCH", frame_confidences=[0.9],
-        )
-        self.assertFalse(result)
-
-    def test_persist_calls_insert_log_after_update(self):
-        db, cursor = self._make_db(rowcount=1)
-        updater = ResultUpdater(db, model_version="v1.0.0")
-        updater.persist(video_id=7, frame_count=5, confidence=0.9,
-                        routing="MATCH", frame_confidences=[0.9])
-        # cursor.execute called twice: UPDATE videos + INSERT processing_logs
-        self.assertEqual(cursor.execute.call_count, 2)
-
-
-class Story06DataExporterTests(unittest.TestCase):
-    def test_jsonl_flush_writes_valid_records(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            exporter = DataExporter(
-                output_path=tmpdir, model_version="v1",
-                export_format="jsonl", batch_size=100,
-            )
-            exporter.add({"video_id": 1, "routing": "MATCH", "video_confidence": 0.9,
-                          "confidence_scores": [0.9], "sample_count": 1,
-                          "file_path": "/a.mp4", "processed_at": "2026-01-01T00:00:00+00:00"})
-            path = exporter.flush()
-
-            self.assertIsNotNone(path)
-            lines = Path(path).read_text(encoding="utf-8").strip().split("\n")
-            self.assertEqual(len(lines), 1)
-            record = json.loads(lines[0])
-            self.assertEqual(record["routing"], "MATCH")
-            self.assertAlmostEqual(record["video_confidence"], 0.9)
-
-    def test_auto_flush_at_batch_size(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            exporter = DataExporter(
-                output_path=tmpdir, model_version="v1",
-                export_format="jsonl", batch_size=2,
-            )
-            exporter.add({"video_id": 1, "routing": "SKIP", "video_confidence": 0.1,
-                          "confidence_scores": [], "sample_count": 0,
-                          "file_path": "/a.mp4", "processed_at": ""})
-            exporter.add({"video_id": 2, "routing": "MATCH", "video_confidence": 0.9,
-                          "confidence_scores": [0.9], "sample_count": 1,
-                          "file_path": "/b.mp4", "processed_at": ""})
-            # batch_size=2 → auto-flush triggered on second add
-            self.assertEqual(exporter._buffer, [])
-
-    def test_exclude_frame_confidences_when_disabled(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            exporter = DataExporter(
-                output_path=tmpdir, model_version="v1",
-                export_format="jsonl", batch_size=100,
-                include_frame_confidences=False,
-            )
-            exporter.add({"video_id": 1, "routing": "SKIP", "video_confidence": 0.1,
-                          "confidence_scores": [0.1, 0.2], "sample_count": 2,
-                          "file_path": "/a.mp4", "processed_at": ""})
-            path = exporter.flush()
-            record = json.loads(Path(path).read_text())
-            self.assertNotIn("confidence_scores", record)
-
-
-class Story06ScratchManagerCleanupAllTests(unittest.TestCase):
-    def test_cleanup_all_removes_entire_video_directory(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            manager = ScratchManager(base_path=tmpdir, video_id="v42", auto_cleanup=True)
-            frame_dir = manager.ensure_frame_dir()
-            crops_dir = Path(tmpdir) / "v42" / "crops"
-            crops_dir.mkdir(parents=True)
-            (frame_dir / "frame.jpg").write_bytes(b"f")
-            (crops_dir / "crop.jpg").write_bytes(b"c")
-
-            manager.cleanup_all()
-            self.assertFalse((Path(tmpdir) / "v42").exists())
-
-
-if __name__ == "__main__":
-    unittest.main()

tests/test_story_07.py

@@ -1,95 +0,0 @@
-import csv
-import io
-import json
-import sys
-import tempfile
-import unittest
-from pathlib import Path
-from unittest.mock import MagicMock
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "src"))
-
-import review_export
-
-
-def _make_db(rows):
-    db = MagicMock()
-    db.fetchall.return_value = rows
-    return db
-
-
-class Story07ReviewExportTests(unittest.TestCase):
-    _ROWS = [
-        {
-            "video_id": 1, "file_path": "/data/input/a.mp4",
-            "confidence_score": 0.62, "routing_decision": "REVIEW",
-            "model_version": "v1.0", "ground_truth": True,
-            "annotated_at": None, "notes": "ok",
-            "confidence_scores": json.dumps([0.60, 0.62, 0.65]),
-        },
-    ]
-
-    def test_fetch_annotated_passes_correct_where_clause(self):
-        db = _make_db(self._ROWS)
-        review_export.fetch_annotated(db, annotated_only=True, model_version="v1.0")
-        call_args = db.fetchall.call_args
-        sql = call_args[0][0]
-        self.assertIn("rq.annotated = TRUE", sql)
-        self.assertIn("v.model_version = %s", sql)
-
-    def test_fetch_annotated_normalises_confidence_scores(self):
-        db = _make_db(self._ROWS)
-        records = review_export.fetch_annotated(db)
-        self.assertIsInstance(records[0]["contributing_frames"], list)
-        self.assertEqual(records[0]["contributing_frames"], [0.60, 0.62, 0.65])
-
-    def test_fetch_annotated_handles_null_confidence_scores(self):
-        rows = [{**self._ROWS[0], "confidence_scores": None}]
-        db = _make_db(rows)
-        records = review_export.fetch_annotated(db)
-        self.assertEqual(records[0]["contributing_frames"], [])
-
-    def test_export_json_writes_valid_utf8_file(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            records = review_export.fetch_annotated(_make_db(self._ROWS))
-            path = review_export.export_json(records, f"{tmpdir}/out.json")
-            loaded = json.loads(Path(path).read_text(encoding="utf-8"))
-            self.assertEqual(len(loaded), 1)
-            self.assertEqual(loaded[0]["video_id"], 1)
-            self.assertTrue(loaded[0]["ground_truth"])
-
-    def test_export_csv_writes_valid_csv(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            records = review_export.fetch_annotated(_make_db(self._ROWS))
-            path = review_export.export_csv(records, f"{tmpdir}/out.csv")
-            content = Path(path).read_text(encoding="utf-8")
-            reader = csv.DictReader(io.StringIO(content))
-            rows = list(reader)
-            self.assertEqual(len(rows), 1)
-            self.assertEqual(rows[0]["routing_decision"], "REVIEW")
-            # contributing_frames should be a JSON string in CSV
-            frames = json.loads(rows[0]["contributing_frames"])
-            self.assertEqual(frames, [0.60, 0.62, 0.65])
-
-    def test_export_csv_empty_returns_empty_file(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            path = review_export.export_csv([], f"{tmpdir}/empty.csv")
-            self.assertEqual(Path(path).read_text(), "")
-
-    def test_ground_truth_filter_appears_in_query(self):
-        db = _make_db([])
-        review_export.fetch_annotated(db, ground_truth=False)
-        sql = db.fetchall.call_args[0][0]
-        self.assertIn("rq.ground_truth = %s", sql)
-
-
-class Story07AppSyntaxTest(unittest.TestCase):
-    def test_app_module_compiles(self):
-        """Ensure ui/app.py has no syntax errors."""
-        app_path = Path(__file__).resolve().parents[1] / "ui" / "app.py"
-        source = app_path.read_text(encoding="utf-8")
-        compile(source, str(app_path), "exec")
-
-
-if __name__ == "__main__":
-    unittest.main()

tests/test_story_08.py

@@ -1,550 +0,0 @@
-"""
-Tests for Story 08: Active Learning Pipeline.
-
-Covers:
-  - LabelIngestor: sample counting, dataset building, stratified split
-  - Trainer: model construction (backbone frozen), class weight calculation
-  - Validator: ECE calculation, quality gate logic
-  - ModelRegistry: promote/rollback DB calls
-  - ActiveLearningPipeline: threshold guard, full orchestration
-"""
-
-import csv
-import json
-import sys
-import tempfile
-import unittest
-from pathlib import Path
-from unittest.mock import MagicMock, patch, call
-
-import numpy as np
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "src"))
-
-from active_learning.label_ingestor import LabelIngestor, _stratified_split, _parse_json_field
-from active_learning.validator import compute_ece, Validator
-from active_learning.registry import ModelRegistry
-from active_learning.pipeline import ActiveLearningPipeline
-
-
-# ---------------------------------------------------------------------------
-# Helpers
-# ---------------------------------------------------------------------------
-
-def _make_config(overrides: dict = None):
-    cfg = MagicMock()
-    al_defaults = {
-        "enabled": True,
-        "min_annotated_samples": 100,
-        "seed": 42,
-        "training": {
-            "epochs": 20, "batch_size": 32, "learning_rate": 1e-3,
-            "weight_decay": 1e-2, "early_stopping_patience": 5,
-            "lr_factor": 0.5, "lr_patience": 3,
-        },
-        "validation": {"val_split": 0.2, "min_f1_improvement": 0.02, "max_ece": 0.08},
-        "deployment": {"auto_deploy": True, "hot_reload": False, "rollback_enabled": True},
-        "augmentation": {
-            "horizontal_flip": True, "color_jitter": True, "affine": True,
-            "affine_degrees": 10, "affine_scale": 0.1,
-        },
-    }
-    if overrides:
-        al_defaults.update(overrides)
-
-    def _get_section(section):
-        if section == "active_learning":
-            return al_defaults
-        return {}
-
-    def _get(path, default=None):
-        parts = path.split(".")
-        if parts[0] == "active_learning" and len(parts) > 1:
-            key = parts[1]
-            return al_defaults.get(key, default)
-        mapping = {
-            "storage.training_path": "/tmp/videodetect_test_training",
-            "storage.models_path": "/tmp/videodetect_test_models",
-            "storage.scratch_path": "/tmp/scratch",
-        }
-        return mapping.get(path, default)
-
-    cfg.get_section.side_effect = _get_section
-    cfg.get.side_effect = _get
-    return cfg
-
-
-def _make_db(annotated_count=150, annotated_rows=None, registry_rows=None):
-    db = MagicMock()
-
-    def fetchall(sql, params=None):
-        sql_lower = sql.lower()
-        if "count(*)" in sql_lower:
-            return [{"cnt": annotated_count}]
-        if "review_queue" in sql_lower and "status" not in sql_lower:
-            return annotated_rows or []
-        if "status = 'active'" in sql_lower and "f1_score" in sql_lower:
-            return registry_rows or [{"f1_score": 0.70}]
-        if "status = 'active'" in sql_lower:
-            return registry_rows or [{"version": "v1.0.0"}]
-        if "status = 'archived'" in sql_lower:
-            return [{"version": "v1.0.0"}]
-        return []
-
-    db.fetchall.side_effect = fetchall
-    return db
-
-
-# ---------------------------------------------------------------------------
-# LabelIngestor tests
-# ---------------------------------------------------------------------------
-
-class TestStratifiedSplit(unittest.TestCase):
-    def _make_records(self, n_pos, n_neg):
-        records = [{"label": 1, "crop_path": f"p{i}.jpg", "video_id": i} for i in range(n_pos)]
-        records += [{"label": 0, "crop_path": f"n{i}.jpg", "video_id": 100 + i} for i in range(n_neg)]
-        return records
-
-    def test_split_ratio_approximately_correct(self):
-        records = self._make_records(60, 40)
-        train, val = _stratified_split(records, 0.80, seed=42)
-        self.assertAlmostEqual(len(train) / len(records), 0.80, delta=0.05)
-
-    def test_stratification_preserves_class_balance(self):
-        records = self._make_records(50, 50)
-        train, val = _stratified_split(records, 0.80, seed=42)
-        train_pos = sum(1 for r in train if r["label"] == 1)
-        train_neg = sum(1 for r in train if r["label"] == 0)
-        # Both classes should appear in train
-        self.assertGreater(train_pos, 0)
-        self.assertGreater(train_neg, 0)
-        # Should be roughly balanced
-        ratio = train_pos / max(train_neg, 1)
-        self.assertAlmostEqual(ratio, 1.0, delta=0.3)
-
-    def test_split_is_deterministic(self):
-        records = self._make_records(40, 40)
-        train_a, _ = _stratified_split(records, 0.80, seed=7)
-        train_b, _ = _stratified_split(records, 0.80, seed=7)
-        self.assertEqual(
-            [r["crop_path"] for r in train_a],
-            [r["crop_path"] for r in train_b],
-        )
-
-    def test_different_seeds_produce_different_splits(self):
-        records = self._make_records(40, 40)
-        train_a, _ = _stratified_split(records, 0.80, seed=1)
-        train_b, _ = _stratified_split(records, 0.80, seed=999)
-        self.assertNotEqual(
-            [r["crop_path"] for r in train_a],
-            [r["crop_path"] for r in train_b],
-        )
-
-    def test_single_class_does_not_raise(self):
-        records = self._make_records(20, 0)
-        train, val = _stratified_split(records, 0.80, seed=42)
-        self.assertGreater(len(train), 0)
-
-    def test_no_sample_loss(self):
-        records = self._make_records(30, 20)
-        train, val = _stratified_split(records, 0.80, seed=42)
-        self.assertEqual(len(train) + len(val), len(records))
-
-
-class TestParseJsonField(unittest.TestCase):
-    def test_parses_list(self):
-        self.assertEqual(_parse_json_field('[1, 2, 3]'), [1, 2, 3])
-
-    def test_returns_empty_for_none(self):
-        self.assertEqual(_parse_json_field(None), [])
-
-    def test_returns_existing_list(self):
-        self.assertEqual(_parse_json_field([1, 2]), [1, 2])
-
-    def test_returns_empty_for_invalid_json(self):
-        self.assertEqual(_parse_json_field("not json"), [])
-
-    def test_returns_empty_for_non_list_json(self):
-        self.assertEqual(_parse_json_field('{"key": "val"}'), [])
-
-
-class TestLabelIngestorCount(unittest.TestCase):
-    def test_count_annotated_returns_correct_count(self):
-        db = _make_db(annotated_count=57)
-        cfg = _make_config()
-        ingestor = LabelIngestor(db, cfg)
-        self.assertEqual(ingestor.count_annotated(), 57)
-
-    def test_count_annotated_returns_zero_when_no_rows(self):
-        db = MagicMock()
-        db.fetchall.return_value = []
-        cfg = _make_config()
-        ingestor = LabelIngestor(db, cfg)
-        self.assertEqual(ingestor.count_annotated(), 0)
-
-
-class TestLabelIngestorIngest(unittest.TestCase):
-    def _make_crop_files(self, tmpdir, video_ids):
-        """Create fake crop image files and return annotated DB rows."""
-        scratch = Path(tmpdir) / "scratch"
-        scratch.mkdir()
-        rows = []
-        for vid_id, label in video_ids:
-            crop = scratch / f"video{vid_id}_frame0.jpg"
-            # Write a minimal valid JPEG header
-            crop.write_bytes(
-                b"\xff\xd8\xff\xe0\x00\x10JFIF\x00\x01\x01\x00\x00\x01\x00\x01\x00\x00"
-                b"\xff\xd9"
-            )
-            rows.append({
-                "video_id": vid_id,
-                "ground_truth": bool(label),
-                "confidence_scores": json.dumps([{"crop_path": str(crop)}]),
-                "file_path": f"/data/input/vid{vid_id}.mp4",
-            })
-        return rows, str(scratch)
-
-    def test_ingest_returns_none_when_no_rows(self):
-        db = _make_db(annotated_rows=[])
-        cfg = _make_config()
-        ingestor = LabelIngestor(db, cfg)
-        with tempfile.TemporaryDirectory() as tmpdir:
-            cfg.get.side_effect = lambda k, d=None: {
-                "storage.training_path": tmpdir,
-                "storage.scratch_path": tmpdir,
-            }.get(k, d)
-            result = ingestor.ingest("v2.0.0")
-        self.assertIsNone(result)
-
-    def test_ingest_creates_directory_structure(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            rows, scratch = self._make_crop_files(tmpdir, [(1, True), (2, False), (3, True)])
-            db = _make_db(annotated_rows=rows)
-            cfg = _make_config()
-            cfg.get.side_effect = lambda k, d=None: {
-                "storage.training_path": str(Path(tmpdir) / "training"),
-                "storage.scratch_path": scratch,
-            }.get(k, d)
-
-            ingestor = LabelIngestor(db, cfg)
-            result = ingestor.ingest("v2.0.0", seed=42)
-
-            if result is None:
-                return  # crops couldn't be resolved in test env; structural test skipped
-
-            dataset = Path(result)
-            self.assertTrue((dataset / "crops" / "class_0").is_dir())
-            self.assertTrue((dataset / "crops" / "class_1").is_dir())
-            self.assertTrue((dataset / "labels.csv").exists())
-            self.assertTrue((dataset / "metadata.json").exists())
-
-    def test_metadata_json_has_expected_keys(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            rows, scratch = self._make_crop_files(tmpdir, [(1, True), (2, False), (3, True), (4, False)])
-            db = _make_db(annotated_rows=rows)
-            cfg = _make_config()
-            cfg.get.side_effect = lambda k, d=None: {
-                "storage.training_path": str(Path(tmpdir) / "training"),
-                "storage.scratch_path": scratch,
-            }.get(k, d)
-
-            ingestor = LabelIngestor(db, cfg)
-            result = ingestor.ingest("v2.1.0", seed=42)
-
-            if result is None:
-                return
-
-            meta = json.loads((Path(result) / "metadata.json").read_text())
-            self.assertIn("version", meta)
-            self.assertIn("total_samples", meta)
-            self.assertIn("train_samples", meta)
-            self.assertIn("val_samples", meta)
-            self.assertIn("class_counts", meta)
-            self.assertEqual(meta["version"], "v2.1.0")
-
-
-# ---------------------------------------------------------------------------
-# ECE / Validator tests
-# ---------------------------------------------------------------------------
-
-class TestComputeECE(unittest.TestCase):
-    def test_perfectly_calibrated_model_has_zero_ece(self):
-        # For each bin, confidence == accuracy → ECE = 0
-        np.random.seed(42)
-        n = 1000
-        probs = np.random.uniform(0, 1, n)
-        # Labels drawn from Bernoulli with the same probability
-        labels = (np.random.uniform(0, 1, n) < probs).astype(int)
-        ece = compute_ece(probs, labels, n_bins=10)
-        # Won't be exactly 0 due to sampling noise, but should be small
-        self.assertLess(ece, 0.10)
-
-    def test_overconfident_model_has_high_ece(self):
-        probs = np.ones(100) * 0.95
-        labels = np.zeros(100, dtype=int)
-        ece = compute_ece(probs, labels)
-        self.assertGreater(ece, 0.5)
-
-    def test_empty_predictions_returns_zero(self):
-        self.assertEqual(compute_ece(np.array([]), np.array([])), 0.0)
-
-    def test_ece_is_between_zero_and_one(self):
-        probs = np.random.default_rng(0).uniform(0, 1, 200)
-        labels = np.random.default_rng(0).integers(0, 2, 200)
-        ece = compute_ece(probs, labels)
-        self.assertGreaterEqual(ece, 0.0)
-        self.assertLessEqual(ece, 1.0)
-
-    def test_ece_bins_parameter(self):
-        probs = np.linspace(0, 1, 100)
-        labels = (probs > 0.5).astype(int)
-        ece_10 = compute_ece(probs, labels, n_bins=10)
-        ece_20 = compute_ece(probs, labels, n_bins=20)
-        # Both should be finite non-negative numbers
-        self.assertGreaterEqual(ece_10, 0.0)
-        self.assertGreaterEqual(ece_20, 0.0)
-
-
-class TestValidatorQualityGates(unittest.TestCase):
-    def _make_validator(self, current_f1=0.70, min_delta=0.02, max_ece=0.08):
-        cfg = _make_config({
-            "validation": {
-                "val_split": 0.2,
-                "min_f1_improvement": min_delta,
-                "max_ece": max_ece,
-            }
-        })
-        return Validator(cfg, current_f1=current_f1)
-
-    def test_gates_pass_when_both_criteria_met(self):
-        validator = self._make_validator(current_f1=0.70)
-        # Simulate metrics
-        probs = np.array([0.9, 0.8, 0.1, 0.2, 0.85, 0.15, 0.75, 0.25])
-        labels = np.array([1,   1,   0,   0,   1,    0,    1,    0  ])
-        metrics = validator._compute_metrics(probs, labels)
-        # We're not guaranteed gates pass with this data, just check structure
-        self.assertIn("gates_passed", metrics)
-        self.assertIn("gate_details", metrics)
-        self.assertIn("f1", metrics)
-        self.assertIn("ece", metrics)
-
-    def test_gates_fail_when_f1_improvement_insufficient(self):
-        # current_f1=0.99 → perfect candidate (f1=1.0) only gives delta=0.01 < 0.02
-        validator = self._make_validator(current_f1=0.99, min_delta=0.02)
-        probs = np.array([0.9, 0.1, 0.8, 0.2])
-        labels = np.array([1, 0, 1, 0])
-        metrics = validator._compute_metrics(probs, labels)
-        details = metrics["gate_details"]
-        self.assertFalse(metrics["gates_passed"])
-        self.assertFalse(details["gate_f1_passed"])
-
-    def test_gates_fail_when_ece_too_high(self):
-        validator = self._make_validator(current_f1=0.0, min_delta=0.0, max_ece=0.01)
-        # Force high ECE: all confidence 0.9 but labels are 0
-        probs = np.ones(50) * 0.9
-        labels = np.zeros(50, dtype=int)
-        metrics = validator._compute_metrics(probs, labels)
-        self.assertFalse(metrics["gates_passed"])
-        self.assertFalse(metrics["gate_details"]["gate_ece_passed"])
-
-    def test_gate_details_include_delta_f1(self):
-        validator = self._make_validator(current_f1=0.60)
-        probs = np.array([0.8, 0.2, 0.7, 0.3])
-        labels = np.array([1, 0, 1, 0])
-        metrics = validator._compute_metrics(probs, labels)
-        self.assertIn("delta_f1", metrics["gate_details"])
-        self.assertAlmostEqual(
-            metrics["gate_details"]["delta_f1"],
-            metrics["f1"] - 0.60,
-            places=3,
-        )
-
-
-# ---------------------------------------------------------------------------
-# ModelRegistry tests
-# ---------------------------------------------------------------------------
-
-class TestModelRegistry(unittest.TestCase):
-    def _make_registry(self, auto_deploy=True, hot_reload=False):
-        cfg = _make_config({
-            "deployment": {
-                "auto_deploy": auto_deploy,
-                "hot_reload": hot_reload,
-                "rollback_enabled": True,
-            }
-        })
-        with tempfile.TemporaryDirectory() as tmpdir:
-            cfg.get.side_effect = lambda k, d=None: {
-                "storage.models_path": tmpdir,
-            }.get(k, d)
-            db = _make_db()
-            return ModelRegistry(db, cfg), db, tmpdir
-
-    def test_get_active_version_returns_version(self):
-        cfg = _make_config()
-        cfg.get.side_effect = lambda k, d=None: {
-            "storage.models_path": "/tmp/models",
-        }.get(k, d)
-        db = MagicMock()
-        db.fetchall.return_value = [{"version": "v1.5.0"}]
-        registry = ModelRegistry(db, cfg)
-        self.assertEqual(registry.get_active_version(), "v1.5.0")
-
-    def test_get_active_version_returns_none_when_no_active(self):
-        cfg = _make_config()
-        cfg.get.side_effect = lambda k, d=None: {
-            "storage.models_path": "/tmp/models",
-        }.get(k, d)
-        db = MagicMock()
-        db.fetchall.return_value = []
-        registry = ModelRegistry(db, cfg)
-        self.assertIsNone(registry.get_active_version())
-
-    def test_get_active_f1_returns_float(self):
-        cfg = _make_config()
-        cfg.get.side_effect = lambda k, d=None: {
-            "storage.models_path": "/tmp/models",
-        }.get(k, d)
-        db = MagicMock()
-        db.fetchall.return_value = [{"f1_score": 0.85}]
-        registry = ModelRegistry(db, cfg)
-        self.assertAlmostEqual(registry.get_active_f1(), 0.85)
-
-    def test_get_active_f1_returns_zero_when_none(self):
-        cfg = _make_config()
-        cfg.get.side_effect = lambda k, d=None: {
-            "storage.models_path": "/tmp/models",
-        }.get(k, d)
-        db = MagicMock()
-        db.fetchall.return_value = []
-        registry = ModelRegistry(db, cfg)
-        self.assertEqual(registry.get_active_f1(), 0.0)
-
-    def test_register_candidate_executes_upsert(self):
-        cfg = _make_config()
-        cfg.get.side_effect = lambda k, d=None: {
-            "storage.models_path": "/tmp/models",
-        }.get(k, d)
-        db = MagicMock()
-        registry = ModelRegistry(db, cfg)
-        registry.register_candidate("v2.0.0", "/models/candidate/v2.0.0_best.pt", 0.82, 0.05)
-        db.execute.assert_called_once()
-        args = db.execute.call_args[0]
-        self.assertIn("INSERT INTO models", args[0])
-        self.assertIn("v2.0.0", args[1])
-
-    def test_rollback_promotes_archived_model(self):
-        cfg = _make_config()
-        cfg.get.side_effect = lambda k, d=None: {
-            "storage.models_path": "/tmp/models",
-        }.get(k, d)
-        db = MagicMock()
-        db.fetchall.return_value = [{"version": "v1.0.0"}]
-        registry = ModelRegistry(db, cfg)
-        registry.rollback("v2.0.0")
-
-        calls = [str(c) for c in db.execute.call_args_list]
-        # Should archive the failed version and activate the previous
-        archive_call = any("ARCHIVED" in c and "v2.0.0" in c for c in calls)
-        activate_call = any("ACTIVE" in c and "v1.0.0" in c for c in calls)
-        self.assertTrue(archive_call, f"Expected ARCHIVED v2.0.0 in calls: {calls}")
-        self.assertTrue(activate_call, f"Expected ACTIVE v1.0.0 in calls: {calls}")
-
-    def test_rollback_logs_warning_when_no_archived_model(self):
-        cfg = _make_config()
-        cfg.get.side_effect = lambda k, d=None: {
-            "storage.models_path": "/tmp/models",
-        }.get(k, d)
-        db = MagicMock()
-        db.fetchall.return_value = []
-        registry = ModelRegistry(db, cfg)
-        # Should not raise
-        registry.rollback("v2.0.0")
-        db.execute.assert_not_called()
-
-    def test_auto_deploy_disabled_skips_deployment(self):
-        cfg = _make_config({
-            "deployment": {"auto_deploy": False, "hot_reload": False, "rollback_enabled": True}
-        })
-        cfg.get.side_effect = lambda k, d=None: {
-            "storage.models_path": "/tmp/models",
-        }.get(k, d)
-        db = MagicMock()
-        registry = ModelRegistry(db, cfg)
-        result = registry.deploy("v2.0.0", "/models/candidate/v2.0.0_best.pt")
-        self.assertFalse(result)
-
-
-# ---------------------------------------------------------------------------
-# ActiveLearningPipeline tests
-# ---------------------------------------------------------------------------
-
-class TestActiveLearningPipeline(unittest.TestCase):
-    def test_pipeline_skips_when_below_min_samples(self):
-        cfg = _make_config({"min_annotated_samples": 100})
-        db = _make_db(annotated_count=50)
-        pipeline = ActiveLearningPipeline(db, cfg)
-        result = pipeline.run("v2.0.0")
-        self.assertFalse(result)
-
-    def test_pipeline_runs_when_above_threshold(self):
-        cfg = _make_config({"min_annotated_samples": 100})
-        db = _make_db(annotated_count=150)
-
-        pipeline = ActiveLearningPipeline(db, cfg)
-
-        with patch.object(pipeline._ingestor, "ingest", return_value=None) as mock_ingest:
-            result = pipeline.run("v2.0.0")
-            mock_ingest.assert_called_once_with("v2.0.0", seed=42)
-            self.assertFalse(result)  # ingestion returned None
-
-    def test_pipeline_aborts_when_training_fails(self):
-        cfg = _make_config({"min_annotated_samples": 10})
-        db = _make_db(annotated_count=50)
-        pipeline = ActiveLearningPipeline(db, cfg)
-
-        with patch.object(pipeline._ingestor, "ingest", return_value="/data/training/v2.0.0"), \
-             patch("active_learning.pipeline.Trainer") as MockTrainer:
-            MockTrainer.return_value.train.return_value = None
-            result = pipeline.run("v2.0.0")
-            self.assertFalse(result)
-
-    def test_pipeline_does_not_deploy_when_gates_fail(self):
-        cfg = _make_config({"min_annotated_samples": 10})
-        db = _make_db(annotated_count=50)
-        pipeline = ActiveLearningPipeline(db, cfg)
-
-        with patch.object(pipeline._ingestor, "ingest", return_value="/data/training/v2.0.0"), \
-             patch("active_learning.pipeline.Trainer") as MockTrainer, \
-             patch("active_learning.pipeline.Validator") as MockValidator:
-            MockTrainer.return_value.train.return_value = "/models/candidate/v2.0.0_best.pt"
-            MockValidator.return_value.validate.return_value = {
-                "f1": 0.71, "ece": 0.05,
-                "gates_passed": False,
-                "gate_details": {"delta_f1": 0.01},
-            }
-            result = pipeline.run("v2.0.0")
-            self.assertFalse(result)
-
-    def test_pipeline_deploys_when_gates_pass(self):
-        cfg = _make_config({"min_annotated_samples": 10})
-        db = _make_db(annotated_count=50)
-        pipeline = ActiveLearningPipeline(db, cfg)
-
-        with patch.object(pipeline._ingestor, "ingest", return_value="/data/training/v2.0.0"), \
-             patch("active_learning.pipeline.Trainer") as MockTrainer, \
-             patch("active_learning.pipeline.Validator") as MockValidator, \
-             patch.object(pipeline._registry, "deploy", return_value=True) as mock_deploy:
-            MockTrainer.return_value.train.return_value = "/models/candidate/v2.0.0_best.pt"
-            MockValidator.return_value.validate.return_value = {
-                "f1": 0.88, "ece": 0.04,
-                "gates_passed": True,
-                "gate_details": {"delta_f1": 0.18},
-            }
-            result = pipeline.run("v2.0.0")
-            self.assertTrue(result)
-            mock_deploy.assert_called_once_with("v2.0.0", "/models/candidate/v2.0.0_best.pt")
-
-
-if __name__ == "__main__":
-    unittest.main()

tests/test_story_09.py

@@ -1,521 +0,0 @@
-"""
-Tests for Story 09: Observability, Monitoring & Hardening.
-
-Covers:
-  - metrics.py: NoOp fallback, update_queue_depths, update_scratch_metrics
-  - crash_recovery.py: recover_on_startup, checkpointing, idempotency guard
-  - retry.py: successful call, retry on transient error, non-retryable bypass, exhaustion
-  - drift_detector.py: detect_drift logic, all alert checks
-  - health_check.py: HealthStatus snapshot, HTTP /health endpoint
-"""
-
-import json
-import os
-import sys
-import tempfile
-import threading
-import time
-import unittest
-import urllib.request
-from pathlib import Path
-from unittest.mock import MagicMock, call, patch
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "src"))
-
-from crash_recovery import CrashRecovery
-from drift_detector import DriftDetector, detect_drift
-from health_check import HealthStatus, start_health_server, health
-from retry import RetryExhaustedError, retry, _is_non_retryable
-
-
-# ---------------------------------------------------------------------------
-# Helpers
-# ---------------------------------------------------------------------------
-
-def _make_config(overrides: dict = None):
-    cfg = MagicMock()
-    mon_defaults = {
-        "alerts": {
-            "confidence_drift_threshold": 0.10,
-            "review_queue_max_size": 1000,
-            "review_queue_max_age_hours": 24,
-            "throughput_min_videos_per_hour": 20,
-            "throughput_min_duration_hours": 1,
-            "error_rate_threshold": 0.05,
-            "error_rate_window_hours": 1,
-        },
-        "drift_detection": {"enabled": True, "schedule": "0 2 * * 0", "baseline_source": "db"},
-        "crash_recovery": {"lock_timeout_minutes": 5, "auto_requeue": True},
-        "retry": {"max_attempts": 3, "initial_delay": 0.0, "backoff_factor": 2.0},
-    }
-    if overrides:
-        mon_defaults.update(overrides)
-
-    def _get_section(section):
-        if section == "monitoring":
-            return mon_defaults
-        return {}
-
-    def _get(path, default=None):
-        mapping = {
-            "storage.scratch_path": "/tmp/test_scratch",
-            "storage.models_path": "/tmp/test_models",
-        }
-        return mapping.get(path, default)
-
-    cfg.get_section.side_effect = _get_section
-    cfg.get.side_effect = _get
-    return cfg
-
-
-def _make_db():
-    return MagicMock()
-
-
-# ---------------------------------------------------------------------------
-# metrics.py tests
-# ---------------------------------------------------------------------------
-
-class TestMetricsNoOp(unittest.TestCase):
-    """The _NoOpMetric must absorb all method calls without raising."""
-
-    def test_noop_labels_inc_does_not_raise(self):
-        from metrics import _NoOpMetric
-        m = _NoOpMetric()
-        m.labels(routing_decision="MATCH").inc()
-
-    def test_noop_set_does_not_raise(self):
-        from metrics import _NoOpMetric
-        m = _NoOpMetric()
-        m.set(42)
-
-    def test_noop_observe_does_not_raise(self):
-        from metrics import _NoOpMetric
-        m = _NoOpMetric()
-        m.observe(0.75)
-
-
-class TestMetricsQueueDepths(unittest.TestCase):
-    def test_update_queue_depths_sets_gauges(self):
-        from metrics import update_queue_depths, queue_depth_pending, queue_depth_processing
-
-        db = MagicMock()
-        db.fetchall.side_effect = [
-            [{"status": "PENDING", "cnt": 10}, {"status": "PROCESSING", "cnt": 3}],
-            [{"cnt": 7}],
-        ]
-        # Should not raise even if prometheus is absent
-        update_queue_depths(db)
-
-    def test_update_queue_depths_handles_db_error_gracefully(self):
-        from metrics import update_queue_depths
-        db = MagicMock()
-        db.fetchall.side_effect = Exception("DB down")
-        update_queue_depths(db)  # must not raise
-
-
-class TestMetricsScratch(unittest.TestCase):
-    def test_update_scratch_metrics_runs_without_error(self):
-        from metrics import update_scratch_metrics
-        with tempfile.TemporaryDirectory() as tmpdir:
-            update_scratch_metrics(tmpdir)  # must not raise
-
-    def test_update_scratch_metrics_handles_missing_path(self):
-        from metrics import update_scratch_metrics
-        update_scratch_metrics("/nonexistent_path_xyz")  # must not raise
-
-
-# ---------------------------------------------------------------------------
-# crash_recovery.py tests
-# ---------------------------------------------------------------------------
-
-class TestCrashRecoveryRequeue(unittest.TestCase):
-    def test_recover_on_startup_calls_update(self):
-        db = _make_db()
-        db.execute.return_value = 3
-        cfg = _make_config()
-        cr = CrashRecovery(db, cfg)
-        count = cr.recover_on_startup()
-        self.assertEqual(count, 3)
-        db.execute.assert_called_once()
-        sql = db.execute.call_args[0][0]
-        self.assertIn("PENDING", sql)
-        self.assertIn("PROCESSING", sql)
-
-    def test_recover_on_startup_skipped_when_disabled(self):
-        db = _make_db()
-        cfg = _make_config({"crash_recovery": {"lock_timeout_minutes": 5, "auto_requeue": False}})
-        cr = CrashRecovery(db, cfg)
-        count = cr.recover_on_startup()
-        self.assertEqual(count, 0)
-        db.execute.assert_not_called()
-
-    def test_recover_on_startup_handles_db_error(self):
-        db = _make_db()
-        db.execute.side_effect = Exception("connection refused")
-        cfg = _make_config()
-        cr = CrashRecovery(db, cfg)
-        count = cr.recover_on_startup()
-        self.assertEqual(count, 0)
-
-    def test_list_stuck_videos_returns_rows(self):
-        db = _make_db()
-        db.fetchall.return_value = [{"id": 5, "file_path": "/data/vid.mp4", "updated_at": None}]
-        cfg = _make_config()
-        cr = CrashRecovery(db, cfg)
-        rows = cr.list_stuck_videos()
-        self.assertEqual(len(rows), 1)
-        self.assertEqual(rows[0]["id"], 5)
-
-
-class TestCrashRecoveryCheckpoint(unittest.TestCase):
-    def test_save_and_load_checkpoint_roundtrip(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            cfg = _make_config()
-            cfg.get.side_effect = lambda k, d=None: {
-                "storage.scratch_path": tmpdir,
-            }.get(k, d)
-            cr = CrashRecovery(_make_db(), cfg)
-            cr.save_checkpoint(42, "extracting", {"frames_done": 5})
-            result = cr.load_checkpoint(42)
-            self.assertIsNotNone(result)
-            self.assertEqual(result["video_id"], 42)
-            self.assertEqual(result["state"], "extracting")
-            self.assertEqual(result["progress"]["frames_done"], 5)
-
-    def test_load_checkpoint_returns_none_when_absent(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            cfg = _make_config()
-            cfg.get.side_effect = lambda k, d=None: {
-                "storage.scratch_path": tmpdir,
-            }.get(k, d)
-            cr = CrashRecovery(_make_db(), cfg)
-            self.assertIsNone(cr.load_checkpoint(999))
-
-    def test_delete_checkpoint_removes_file(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            cfg = _make_config()
-            cfg.get.side_effect = lambda k, d=None: {
-                "storage.scratch_path": tmpdir,
-            }.get(k, d)
-            cr = CrashRecovery(_make_db(), cfg)
-            cr.save_checkpoint(7, "classifying", {})
-            cr.delete_checkpoint(7)
-            self.assertIsNone(cr.load_checkpoint(7))
-
-    def test_checkpoint_timestamp_is_iso_format(self):
-        with tempfile.TemporaryDirectory() as tmpdir:
-            cfg = _make_config()
-            cfg.get.side_effect = lambda k, d=None: {
-                "storage.scratch_path": tmpdir,
-            }.get(k, d)
-            cr = CrashRecovery(_make_db(), cfg)
-            cr.save_checkpoint(1, "detecting", {})
-            ckpt = cr.load_checkpoint(1)
-            # Should parse without error
-            from datetime import datetime
-            datetime.fromisoformat(ckpt["timestamp"].replace("Z", "+00:00"))
-
-
-class TestIdempotencyGuard(unittest.TestCase):
-    def test_returns_true_when_already_completed(self):
-        db = _make_db()
-        db.fetchall.return_value = [{"status": "COMPLETED"}]
-        cr = CrashRecovery(db, _make_config())
-        self.assertTrue(cr.is_already_completed(1))
-
-    def test_returns_false_when_not_completed(self):
-        db = _make_db()
-        db.fetchall.return_value = [{"status": "PENDING"}]
-        cr = CrashRecovery(db, _make_config())
-        self.assertFalse(cr.is_already_completed(1))
-
-    def test_returns_false_when_no_row(self):
-        db = _make_db()
-        db.fetchall.return_value = []
-        cr = CrashRecovery(db, _make_config())
-        self.assertFalse(cr.is_already_completed(99))
-
-
-# ---------------------------------------------------------------------------
-# retry.py tests
-# ---------------------------------------------------------------------------
-
-class TestRetryDecorator(unittest.TestCase):
-    def test_successful_call_returns_value(self):
-        @retry(max_attempts=3, initial_delay=0.0, step="test")
-        def always_succeeds():
-            return 42
-
-        self.assertEqual(always_succeeds(), 42)
-
-    def test_retries_on_transient_error(self):
-        call_count = {"n": 0}
-
-        @retry(max_attempts=3, initial_delay=0.0, step="test")
-        def flaky():
-            call_count["n"] += 1
-            if call_count["n"] < 3:
-                raise ConnectionError("transient")
-            return "ok"
-
-        result = flaky()
-        self.assertEqual(result, "ok")
-        self.assertEqual(call_count["n"], 3)
-
-    def test_raises_retry_exhausted_after_max_attempts(self):
-        @retry(max_attempts=3, initial_delay=0.0, step="test")
-        def always_fails():
-            raise ConnectionError("always fails")
-
-        with self.assertRaises(RetryExhaustedError):
-            always_fails()
-
-    def test_non_retryable_error_propagates_immediately(self):
-        call_count = {"n": 0}
-
-        @retry(max_attempts=3, initial_delay=0.0, step="test")
-        def raises_non_retryable():
-            call_count["n"] += 1
-            raise FileNotFoundError("no such file")
-
-        with self.assertRaises(FileNotFoundError):
-            raises_non_retryable()
-
-        self.assertEqual(call_count["n"], 1)
-
-    def test_only_specified_exception_types_are_retried(self):
-        @retry(max_attempts=3, initial_delay=0.0, exceptions=(ValueError,), step="test")
-        def raises_type_error():
-            raise TypeError("wrong type")
-
-        with self.assertRaises(TypeError):
-            raises_type_error()
-
-    def test_preserves_return_value_on_first_try(self):
-        @retry(max_attempts=5, initial_delay=0.0, step="test")
-        def returns_dict():
-            return {"key": "value"}
-
-        self.assertEqual(returns_dict(), {"key": "value"})
-
-
-class TestIsNonRetryable(unittest.TestCase):
-    def test_file_not_found_is_non_retryable(self):
-        self.assertTrue(_is_non_retryable(FileNotFoundError("x")))
-
-    def test_permission_error_is_non_retryable(self):
-        self.assertTrue(_is_non_retryable(PermissionError("x")))
-
-    def test_connection_error_is_retryable(self):
-        self.assertFalse(_is_non_retryable(ConnectionError("x")))
-
-    def test_runtime_error_is_retryable(self):
-        self.assertFalse(_is_non_retryable(RuntimeError("x")))
-
-
-# ---------------------------------------------------------------------------
-# drift_detector.py tests
-# ---------------------------------------------------------------------------
-
-class TestDetectDrift(unittest.TestCase):
-    def test_no_drift_when_distributions_match(self):
-        base = [0.3] * 50 + [0.7] * 50  # 50% high confidence
-        curr = [0.3] * 50 + [0.7] * 50
-        self.assertFalse(detect_drift(curr, base, threshold=0.10))
-
-    def test_drift_detected_when_shift_exceeds_threshold(self):
-        base = [0.3] * 80 + [0.8] * 20   # 20% high
-        curr = [0.8] * 70 + [0.3] * 30   # 70% high → shift = 0.50
-        self.assertTrue(detect_drift(curr, base, threshold=0.10))
-
-    def test_no_drift_just_below_threshold(self):
-        base = [0.8] * 50 + [0.2] * 50   # 50% high
-        curr = [0.8] * 59 + [0.2] * 41   # 59% high → shift = 9%
-        self.assertFalse(detect_drift(curr, base, threshold=0.10))
-
-    def test_drift_at_boundary(self):
-        base = [0.8] * 50 + [0.2] * 50   # 50%
-        curr = [0.8] * 61 + [0.2] * 39   # 61% → shift = 11%
-        self.assertTrue(detect_drift(curr, base, threshold=0.10))
-
-    def test_empty_current_returns_false(self):
-        self.assertFalse(detect_drift([], [0.5] * 10, threshold=0.10))
-
-    def test_empty_baseline_returns_false(self):
-        self.assertFalse(detect_drift([0.5] * 10, [], threshold=0.10))
-
-
-class TestDriftDetectorAlerts(unittest.TestCase):
-    def _make_detector(self, db=None, overrides=None):
-        alerts = []
-        cfg = _make_config(overrides or {})
-        db = db or _make_db()
-        detector = DriftDetector(db, cfg, alert_fn=alerts.append)
-        return detector, alerts
-
-    def test_check_review_queue_growth_triggers_alert(self):
-        db = _make_db()
-        db.fetchall.return_value = [{"cnt": 1500}]
-        detector, alerts = self._make_detector(db)
-        triggered = detector.check_review_queue_growth()
-        self.assertTrue(triggered)
-        self.assertEqual(len(alerts), 1)
-        self.assertIn("1500", alerts[0])
-
-    def test_check_review_queue_growth_no_alert_below_threshold(self):
-        db = _make_db()
-        db.fetchall.return_value = [{"cnt": 50}]
-        detector, alerts = self._make_detector(db)
-        triggered = detector.check_review_queue_growth()
-        self.assertFalse(triggered)
-        self.assertEqual(len(alerts), 0)
-
-    def test_check_low_throughput_triggers_alert(self):
-        db = _make_db()
-        db.fetchall.return_value = [{"cnt": 5}]  # 5 videos in last 1h < 20 min
-        detector, alerts = self._make_detector(db)
-        triggered = detector.check_low_throughput()
-        self.assertTrue(triggered)
-        self.assertEqual(len(alerts), 1)
-
-    def test_check_low_throughput_no_alert_above_threshold(self):
-        db = _make_db()
-        db.fetchall.return_value = [{"cnt": 50}]  # 50 > 20
-        detector, alerts = self._make_detector(db)
-        triggered = detector.check_low_throughput()
-        self.assertFalse(triggered)
-
-    def test_check_error_rate_triggers_alert(self):
-        db = _make_db()
-        db.fetchall.return_value = [
-            {"status": "COMPLETED", "cnt": 80},
-            {"status": "ERROR", "cnt": 10},
-            {"status": "UNSCANNABLE", "cnt": 10},
-        ]
-        detector, alerts = self._make_detector(db)
-        triggered = detector.check_error_rate()
-        self.assertTrue(triggered)  # 20/100 = 20% > 5%
-        self.assertEqual(len(alerts), 1)
-
-    def test_check_error_rate_no_alert_below_threshold(self):
-        db = _make_db()
-        db.fetchall.return_value = [
-            {"status": "COMPLETED", "cnt": 98},
-            {"status": "ERROR", "cnt": 2},
-        ]
-        detector, alerts = self._make_detector(db)
-        triggered = detector.check_error_rate()
-        self.assertFalse(triggered)  # 2% < 5%
-
-    def test_check_error_rate_no_alert_zero_videos(self):
-        db = _make_db()
-        db.fetchall.return_value = []
-        detector, alerts = self._make_detector(db)
-        triggered = detector.check_error_rate()
-        self.assertFalse(triggered)
-
-    def test_run_all_checks_returns_dict_with_expected_keys(self):
-        db = _make_db()
-        db.fetchall.return_value = [{"cnt": 0}]
-        detector, _ = self._make_detector(db)
-        with patch.object(detector, "_fetch_recent_confidences", return_value=[]):
-            results = detector.run_all_checks()
-        self.assertIn("confidence_drift", results)
-        self.assertIn("review_queue_growth", results)
-        self.assertIn("low_throughput", results)
-        self.assertIn("high_error_rate", results)
-
-    def test_check_handles_db_error_gracefully(self):
-        db = _make_db()
-        db.fetchall.side_effect = Exception("DB offline")
-        detector, alerts = self._make_detector(db)
-        # Should not raise
-        self.assertFalse(detector.check_review_queue_growth())
-        self.assertFalse(detector.check_low_throughput())
-        self.assertFalse(detector.check_error_rate())
-
-
-# ---------------------------------------------------------------------------
-# health_check.py tests
-# ---------------------------------------------------------------------------
-
-class TestHealthStatus(unittest.TestCase):
-    def test_snapshot_contains_required_keys(self):
-        hs = HealthStatus()
-        snap = hs.snapshot()
-        for key in ("status", "gpu_available", "gpu_memory_used_gb",
-                    "queue_depth", "uptime_seconds", "videos_processed_today", "last_error"):
-            self.assertIn(key, snap, f"Missing key: {key}")
-
-    def test_update_changes_values(self):
-        hs = HealthStatus()
-        hs.update(status="healthy", queue_depth=55)
-        snap = hs.snapshot()
-        self.assertEqual(snap["status"], "healthy")
-        self.assertEqual(snap["queue_depth"], 55)
-
-    def test_uptime_increases_over_time(self):
-        hs = HealthStatus()
-        snap1 = hs.snapshot()
-        time.sleep(0.05)
-        snap2 = hs.snapshot()
-        self.assertGreaterEqual(snap2["uptime_seconds"], snap1["uptime_seconds"])
-
-    def test_update_is_thread_safe(self):
-        hs = HealthStatus()
-        errors = []
-
-        def writer(n):
-            try:
-                for _ in range(100):
-                    hs.update(queue_depth=n)
-            except Exception as exc:
-                errors.append(exc)
-
-        threads = [threading.Thread(target=writer, args=(i,)) for i in range(5)]
-        for t in threads:
-            t.start()
-        for t in threads:
-            t.join()
-
-        self.assertEqual(errors, [])
-
-    def test_http_health_endpoint_returns_200(self):
-        """Start a real health server and hit /health with urllib."""
-        import socket
-
-        # Find a free port
-        with socket.socket() as s:
-            s.bind(("127.0.0.1", 0))
-            port = s.getsockname()[1]
-
-        t = start_health_server(port)
-        time.sleep(0.1)  # give the server a moment to bind
-
-        url = f"http://127.0.0.1:{port}/health"
-        with urllib.request.urlopen(url, timeout=2) as resp:
-            self.assertEqual(resp.status, 200)
-            body = json.loads(resp.read())
-            self.assertIn("status", body)
-            self.assertIn("uptime_seconds", body)
-
-    def test_http_404_for_unknown_path(self):
-        """Non /health paths return 404."""
-        import socket
-        from urllib.error import HTTPError
-
-        with socket.socket() as s:
-            s.bind(("127.0.0.1", 0))
-            port = s.getsockname()[1]
-
-        start_health_server(port)
-        time.sleep(0.1)
-
-        with self.assertRaises(HTTPError) as ctx:
-            urllib.request.urlopen(f"http://127.0.0.1:{port}/unknown", timeout=2)
-        self.assertEqual(ctx.exception.code, 404)
-
-
-if __name__ == "__main__":
-    unittest.main()

ui/requirements.txt

@@ -1,4 +1,4 @@
 Flask==3.0.0
 gunicorn==21.2.0
-PyMySQL==1.1.0
 Jinja2==3.1.2
+requests>=2.28.0

worker/Dockerfile

@@ -1,68 +0,0 @@
-# Base image: CUDA 11.8 runtime on Ubuntu 22.04
-FROM nvidia/cuda:11.8.0-runtime-ubuntu22.04
-
-# Avoid interactive prompts during build
-ENV DEBIAN_FRONTEND=noninteractive
-ENV TZ=Etc/UTC
-
-# System dependencies
-RUN apt-get update && apt-get install -y --no-install-recommends \
-    # Python
-    python3.10 \
-    python3.10-dev \
-    python3.10-venv \
-    python3-pip \
-    # FFmpeg
-    ffmpeg \
-    # OpenCV dependencies
-    libgl1-mesa-glx \
-    libglib2.0-0 \
-    libsm6 \
-    libxext6 \
-    libxrender-dev \
-    # Utilities
-    curl \
-    wget \
-    git \
-    ca-certificates \
-    && rm -rf /var/lib/apt/lists/*
-
-# Create non-root user
-RUN groupadd -g 1000 appuser && \
-    useradd -m -u 1000 -g appuser -s /bin/bash appuser
-
-# Set working directory
-WORKDIR /app
-
-# Create necessary directories
-RUN mkdir -p /scratch /models /data/training /data/output /logs && \
-    chown -R appuser:appuser /app /scratch /models /data /logs
-
-# Copy requirements first for better caching
-COPY worker/requirements.txt /app/requirements.txt
-
-# Create and activate virtual environment
-RUN python3.10 -m venv /opt/venv
-ENV PATH="/opt/venv/bin:$PATH"
-
-# Install Python dependencies
-RUN pip install --no-cache-dir -r requirements.txt
-
-# Copy application code
-COPY src/ /app/src/
-COPY config.yaml /app/config.yaml
-
-# Set environment variables
-ENV PYTHONUNBUFFERED=1
-ENV PATH="/opt/venv/bin:$PATH"
-ENV LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
-
-# Health check
-HEALTHCHECK --interval=30s --timeout=10s --retries=3 \
-    CMD python3 -c "import torch; assert torch.cuda.is_available()" || exit 1
-
-# Switch to non-root user
-USER appuser
-
-# Default command
-CMD ["python3", "-m", "src.main"]

worker/requirements.txt

@@ -1,40 +0,0 @@
-# Core ML Frameworks
-torch==2.1.0
---extra-index-url https://download.pytorch.org/whl/cu118
-
-# TensorRT (compatible with CUDA 11.8)
-# Note: Install from NVIDIA repo or wheel for your specific platform
-# tensorrt==8.6.1  # Uncomment after verifying compatibility
-
-# ONNX Runtime with GPU support
-onnxruntime-gpu==1.16.3
-
-# Computer Vision
-opencv-python-headless==4.8.1.78
-Pillow==10.1.0
-
-# Database
-PyMySQL==1.1.0
-DBUtils==3.1.0
-
-# Data Processing
-numpy==1.24.4
-pandas==2.1.4
-pyarrow==14.0.0
-
-# Configuration
-pyyaml==6.0.1
-
-# Logging
-python-json-logger==2.0.7
-
-# Monitoring
-prometheus-client==0.19.0
-
-# Utilities
-tqdm==4.66.1
-filelock==3.13.1
-
-# Testing (dev dependencies - install with [dev] extra)
-# pytest==7.4.3
-# pytest-cov==4.1.0