chore(infra): add deployment and benchmark tooling

Add Docker Compose with AMD GPU device wiring, AMD Developer Cloud deployment notes, and an API benchmark script for CPU versus MI300X timing comparisons.

docker-compose.yml

@@ -0,0 +1,40 @@
+services:
+  redis:
+    image: redis:7-alpine
+    ports:
+      - "6379:6379"
+
+  backend:
+    build:
+      context: ./backend
+      args:
+        ROCM_PYTORCH_IMAGE: ${ROCM_PYTORCH_IMAGE:-rocm/pytorch:latest}
+    env_file:
+      - .env.example
+    environment:
+      STORAGE_DIR: /app/data
+      REDIS_URL: redis://redis:6379/0
+      FRONTEND_ORIGIN: http://localhost:5173
+    volumes:
+      - ./backend/data:/app/data
+    ports:
+      - "8000:8000"
+    depends_on:
+      - redis
+    devices:
+      - /dev/kfd
+      - /dev/dri
+    group_add:
+      - video
+    ipc: host
+    shm_size: 16gb
+
+  frontend:
+    build:
+      context: ./frontend
+    environment:
+      VITE_API_BASE_URL: http://localhost:8000
+    ports:
+      - "5173:5173"
+    depends_on:
+      - backend

infra/amd-cloud.md

@@ -0,0 +1,52 @@
+# AMD Developer Cloud Deployment
+
+## Instance
+
+Use an AMD Instinct MI300X instance with ROCm 6.x. The backend expects ROCm-enabled PyTorch; on ROCm, PyTorch exposes AMD GPUs through the CUDA-compatible `torch.cuda` API and reports the HIP version in `torch.version.hip`.
+
+## Environment
+
+Set:
+
+```bash
+DEMO_MODE=false
+HF_TOKEN=...
+WHISPER_MODEL_ID=openai/whisper-large-v3
+QWEN_TEXT_MODEL_ID=Qwen/Qwen2.5-7B-Instruct
+QWEN_VL_MODEL_ID=Qwen/Qwen2-VL-7B-Instruct
+FFMPEG_VIDEO_CODEC=h264_amf
+```
+
+Build with the ROCm inference extras:
+
+```bash
+docker compose build --build-arg INSTALL_EXTRAS=.[ai,rocm-inference] backend
+docker compose up
+```
+
+The Docker compose file mounts `/dev/kfd` and `/dev/dri`, adds the `video` group, and uses host IPC for large model inference.
+
+## Inference Notes
+
+- Whisper Large V3 runs through Hugging Face `transformers` with ROCm PyTorch.
+- Qwen2.5 highlight detection is wired for `vLLM` with ROCm backend.
+- Qwen2-VL has a service boundary in `backend/app/services/multimodal.py`; add frame sampling there when demo time allows.
+- Keep `preferred_torch_dtype=bfloat16` on MI300X.
+
+## Benchmark
+
+Run the same source twice:
+
+1. CPU baseline: set `DEMO_MODE=false`, force CPU by hiding GPUs, and run `scripts/benchmark.py`.
+2. AMD GPU run: expose MI300X devices and run the same command.
+
+Capture:
+
+- `input`
+- `transcription`
+- `highlight_detection`
+- `multimodal_analysis`
+- `clip_generation`
+- `total`
+
+For the presentation, show the API timing JSON and the finished clips side by side.

scripts/benchmark.py

@@ -0,0 +1,66 @@
+import argparse
+import json
+import time
+import urllib.error
+import urllib.request
+
+
+def request_json(url: str, method: str = "GET", payload: dict | None = None) -> dict:
+    body = None
+    headers = {}
+    if payload is not None:
+        body = json.dumps(payload).encode("utf-8")
+        headers["Content-Type"] = "application/json"
+    request = urllib.request.Request(url, data=body, headers=headers, method=method)
+    try:
+        with urllib.request.urlopen(request, timeout=30) as response:
+            return json.loads(response.read().decode("utf-8"))
+    except urllib.error.HTTPError as exc:
+        detail = exc.read().decode("utf-8")
+        raise RuntimeError(f"{exc.code}: {detail}") from exc
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Run an AI Clip Studio API benchmark.")
+    parser.add_argument("--api", default="http://localhost:8000")
+    parser.add_argument("--youtube-url", required=True)
+    parser.add_argument("--language", default="Thai")
+    parser.add_argument("--style", default="informative")
+    parser.add_argument("--niche", default="education")
+    parser.add_argument("--clip-length", type=int, default=60)
+    args = parser.parse_args()
+
+    payload = {
+        "youtube_url": args.youtube_url,
+        "profile": {
+            "niche": args.niche,
+            "clip_style": args.style,
+            "clip_length_seconds": args.clip_length,
+            "primary_language": args.language,
+            "target_platform": "tiktok",
+        },
+    }
+    started = time.perf_counter()
+    job = request_json(f"{args.api}/api/jobs/youtube", "POST", payload)
+    while job["status"] in {"queued", "running"}:
+        time.sleep(2)
+        job = request_json(f"{args.api}/api/jobs/{job['id']}")
+
+    elapsed = round(time.perf_counter() - started, 3)
+    print(
+        json.dumps(
+            {
+                "job_id": job["id"],
+                "status": job["status"],
+                "elapsed_wall_seconds": elapsed,
+                "clips": len(job.get("clips", [])),
+                "timings": job.get("timings", {}),
+                "error": job.get("error"),
+            },
+            indent=2,
+        )
+    )
+
+
+if __name__ == "__main__":
+    main()