Dockerfile

FROM python:3.11-slim

WORKDIR /app

# Build tools are a fallback in case a prebuilt llama-cpp-python wheel isn't
# available for this platform/Python and pip has to compile from source.
RUN apt-get update && apt-get install -y --no-install-recommends \
    build-essential cmake git \
    && rm -rf /var/lib/apt/lists/*

# llama.cpp (CPU) — prefer the prebuilt CPU wheel index, fall back to source build.
RUN pip install --no-cache-dir \
    llama-cpp-python \
    --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu

RUN pip install --no-cache-dir \
    fastapi \
    "uvicorn[standard]" \
    pydantic \
    jinja2 \
    huggingface_hub

COPY app.py index.html chat_template.jinja start.sh ./
RUN chmod +x start.sh

# The GGUF is pulled from HF at startup (see start.sh) into /app/model.
ENV MODEL_PATH=/app/model/puffyshirt-Q4_K_M.gguf
ENV PYTHONUNBUFFERED=1
ENV HF_HOME=/app/.cache/huggingface

# HuggingFace Spaces expects port 7860
EXPOSE 7860

CMD ["./start.sh"]