FROM python:3.11-slim

WORKDIR /app

RUN apt-get update && apt-get install -y git && rm -rf /var/lib/apt/lists/*

COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
RUN pip install --no-cache-dir huggingface_hub

# Set cache directory
ENV HF_HOME=/data/hf_cache

# Create offload directory with correct permissions
RUN mkdir -p /data/model_offload && chmod -R 777 /data/model_offload


# Create cache directory with writable permissions for runtime
RUN mkdir -p /data/hf_cache/hub/models--microsoft--Phi-3-mini-4k-instruct/blobs && chmod -R 777 /data/hf_cache

# Pre-download base model to cache
RUN python -c "from huggingface_hub import snapshot_download; \
               snapshot_download(repo_id='microsoft/Phi-3-mini-4k-instruct', local_dir='/data/hf_cache/models--microsoft--Phi-3-mini-4k-instruct')"

COPY . .

EXPOSE 7860

CMD ["uvicorn", "main:app", "--host=0.0.0.0", "--port=7860"]