Dockerfile

# ✅ Base image
FROM python:3.10-slim

# ✅ Install system dependencies
RUN apt-get update && apt-get install -y gcc g++ make

# ✅ Create working directory with write permissions (Hugging Face specific)
WORKDIR /app
RUN mkdir -p /app/.cache /tmp/.triton /tmp/torchinductor_cache && chmod -R 777 /app/.cache /tmp/.triton /tmp/torchinductor_cache

# ✅ Environment variables for Hugging Face caching
ENV HF_HOME=/app/.cache \
    HF_DATASETS_CACHE=/app/.cache \
    HF_HUB_CACHE=/app/.cache \
    TRITON_CACHE_DIR=/tmp/.triton \
    TORCHINDUCTOR_CACHE_DIR=/tmp/torchinductor_cache

# ✅ Copy requirements and install
COPY requirements.txt ./
RUN pip install --no-cache-dir -r requirements.txt

# ✅ Copy all project files
COPY . .

# ✅ Expose FastAPI port
EXPOSE 7860

# ✅ Health check dummy server (important for Hugging Face Spaces)
# This will run as a background thread inside Python, not as a separate Docker CMD
# so we only run uvicorn in CMD below.

# ✅ Final startup command
CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]