Dockerfile

# Use a lightweight Python base
FROM python:3.10-slim

WORKDIR /app

# Install system dependencies
RUN apt-get update && apt-get install -y \
    libopenblas-dev \
    && rm -rf /var/lib/apt/lists/*

    
# Install llama-cpp-python from PREBUILT wheel (3 seconds vs 10+ minutes)
RUN pip install \
    https://huggingface.co/Luigi/llama-cpp-python-wheels-hf-spaces-free-cpu/resolve/main/llama_cpp_python-0.3.22-cp310-cp310-linux_x86_64.whl

# Copy requirements first for cache
COPY requirements.txt .

# Install remaining requirements
RUN pip install -r requirements.txt

# Copy project files
COPY . .

# Setup environment variables for compilation
# We need to find where pip installed llama-cpp-python to link against it
# In docker, it's usually /usr/local/lib/python3.12/site-packages
ENV SITE_PACKAGES=/usr/local/lib/python3.12/site-packages

# Compile the engine
WORKDIR /app/engine
RUN g++ -O2 -shared -fPIC -o libbatch.so batch_server.cpp \
    -I"${SITE_PACKAGES}/include" \
    -L"${SITE_PACKAGES}/llama_cpp/lib" \
    -lllama -Wl,-rpath,"${SITE_PACKAGES}/llama_cpp/lib"

# Setup Model (Download during build or mount volume? 
# Best practice: Download in build if small, or use script at runtime.
# Here we'll rely on the user mounting the model or running the setup script.
# But for "Tunnel Code Optimized", let's assume valid model is present or downloaded.
# We'll expose the setup script.)

WORKDIR /app
EXPOSE 8000

# Start command
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8000"]