feat: switch deployment to Hugging Face Spaces (16GB free tier)

Render free tier OOMs at 512MB with sentence-transformers. HF Spaces
provides 16GB RAM free with Docker support.

- Dockerfile: user 1000 (HF requirement), port 7860, models pre-downloaded
- README: HF Spaces frontmatter, updated demo URL and curl examples
- Startup warmup re-enabled (16GB is plenty)
- Remove render.yaml

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

README.md

@@ -1,3 +1,12 @@
+---
+title: agent-bench
+emoji: "🔍"
+colorFrom: blue
+colorTo: purple
+sdk: docker
+app_port: 7860
+---
+
 # agent-bench
 
 ![CI](https://github.com/tyy0811/agent-bench/actions/workflows/ci.yaml/badge.svg)
@@ -27,21 +36,21 @@ Evaluated on 27 hand-crafted questions using **gpt-4o-mini** ($0.0004/query) ove
 
 ## Live Demo
 
-**https://agent-bench.onrender.com** (Frankfurt, free tier — first request after idle may take ~30-60s for cold start)
+**https://tyy0811-agent-bench.hf.space** (Hugging Face Spaces — first request after idle may take ~30s for cold start)
 
 ```bash
 # In-scope question (expect answer with sources)
-curl -X POST https://agent-bench.onrender.com/ask \
+curl -X POST https://tyy0811-agent-bench.hf.space/ask \
   -H "Content-Type: application/json" \
   -d '{"question": "How do I define a path parameter in FastAPI?"}'
 
 # Out-of-scope question (expect grounded refusal)
-curl -X POST https://agent-bench.onrender.com/ask \
+curl -X POST https://tyy0811-agent-bench.hf.space/ask \
   -H "Content-Type: application/json" \
   -d '{"question": "How do I cook pasta?"}'
 
 # Health check
-curl https://agent-bench.onrender.com/health
+curl https://tyy0811-agent-bench.hf.space/health
 ```
 
 ## Quick Start (Local)
@@ -156,7 +165,7 @@ See [DECISIONS.md](DECISIONS.md) for rationale on building from primitives, RRF
 | Retrieval precision | RRF only | RRF + cross-encoder | Reranking |
 | Provider resilience | None | Retry + backoff | Error handling |
 | Rate limiting | None | 10 RPM per IP | API hardening |
-| Cloud deployment | None | Render (Frankfurt) | Docker → production |
+| Cloud deployment | None | HF Spaces (Docker) | Docker → production |
 | CI/CD | None | GitHub Actions | Automated quality gates |
 
 See [DECISIONS.md](DECISIONS.md) for the reasoning behind each design choice.

agent_bench/serving/app.py

@@ -104,21 +104,16 @@ def create_app(config: AppConfig | None = None) -> FastAPI:
     app.add_middleware(RateLimitMiddleware, requests_per_minute=config.serving.rate_limit_rpm)
     app.include_router(router)
 
-    # Startup warmup: skip on constrained environments (Render free tier 512MB).
-    # Models load lazily on first request instead.
-    import os
-
-    if os.environ.get("AGENT_BENCH_ENV") != "production":
-
-        @app.on_event("startup")
-        async def warmup() -> None:
-            import structlog
-
-            log = structlog.get_logger()
-            log.info("warmup_start")
-            _ = embedder.embed("warmup")
-            if reranker is not None:
-                _ = reranker.model  # noqa: F841
-            log.info("warmup_complete")
+    # Startup warmup: eager-load models to reduce cold start latency
+    @app.on_event("startup")
+    async def warmup() -> None:
+        import structlog
+
+        log = structlog.get_logger()
+        log.info("warmup_start")
+        _ = embedder.embed("warmup")
+        if reranker is not None:
+            _ = reranker.model  # noqa: F841
+        log.info("warmup_complete")
 
     return app

docker/Dockerfile

@@ -1,12 +1,15 @@
 FROM python:3.11-slim
-WORKDIR /app
 
-# Copy all source before pip install (package needs agent_bench/ to build)
-COPY pyproject.toml .
-COPY agent_bench/ agent_bench/
-COPY configs/ configs/
-COPY data/ data/
-COPY scripts/ scripts/
+# HF Spaces requires user ID 1000
+RUN useradd -m -u 1000 user
+WORKDIR /home/user/app
+
+# Copy source and install
+COPY --chown=user pyproject.toml .
+COPY --chown=user agent_bench/ agent_bench/
+COPY --chown=user configs/ configs/
+COPY --chown=user data/ data/
+COPY --chown=user scripts/ scripts/
 
 RUN pip install --no-cache-dir .
 
@@ -17,5 +20,6 @@ RUN python -c "from sentence_transformers import CrossEncoder; CrossEncoder('cro
 # Run ingestion at build time so the store is ready
 RUN python scripts/ingest.py --doc-dir data/tech_docs/ --store-path .cache/store
 
-EXPOSE 8000
-CMD ["uvicorn", "agent_bench.serving.app:create_app", "--factory", "--host", "0.0.0.0", "--port", "8000"]
+USER user
+EXPOSE 7860
+CMD ["uvicorn", "agent_bench.serving.app:create_app", "--factory", "--host", "0.0.0.0", "--port", "7860"]

render.yaml

@@ -1,19 +0,0 @@
-services:
-  - type: web
-    name: agent-bench
-    runtime: python
-    region: frankfurt
-    plan: free
-    autoDeploy: true
-    buildCommand: pip install . && python scripts/ingest.py --doc-dir data/tech_docs/ --store-path .cache/store
-    startCommand: uvicorn agent_bench.serving.app:create_app --factory --host 0.0.0.0 --port $PORT
-    envVars:
-      - key: OPENAI_API_KEY
-        sync: false
-      - key: AGENT_BENCH_ENV
-        value: production
-      - key: PYTHON_VERSION
-        value: 3.11.10
-      - key: PYTHONUNBUFFERED
-        value: "1"
-    healthCheckPath: /health