Dockerfile

FROM python:3.12-slim

WORKDIR /app

# Install system dependencies for sentence-transformers + llama-cpp + SSL
RUN apt-get update && apt-get install -y --no-install-recommends \
    build-essential gcc g++ cmake ca-certificates openssl && \
    rm -rf /var/lib/apt/lists/*

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

# Copy app
COPY . .

# Pre-download bge-m3 + build FAISS index from KB at build time
# (single RUN layer so the model cache is reused)
RUN python -c "from sentence_transformers import SentenceTransformer; SentenceTransformer('BAAI/bge-m3', trust_remote_code=True)" && \
    python precompute_embeddings.py

# Download Llama 3.2 3B GGUF at build time
RUN mkdir -p /app/models && \
    python -c "from huggingface_hub import hf_hub_download; hf_hub_download('QuantFactory/Llama-3.2-3B-Instruct-GGUF', 'Llama-3.2-3B-Instruct.Q4_K_M.gguf', local_dir='/app/models')"

# Default path for local LLM (can be overridden by env)
ENV PORT=7860
ENV LLAMA_MODEL_PATH=/app/models/Llama-3.2-3B-Instruct.Q4_K_M.gguf
CMD ["python", "server.py"]