Dockerfile

FROM python:3.11-slim

WORKDIR /app

# ── System dependencies (C++ compiler only — no vision / audio tools needed) ──
RUN apt-get update && apt-get install -y \
    build-essential \
    cmake \
    git \
    && rm -rf /var/lib/apt/lists/*

# ── Compile llama-cpp-python CPU-only (AVX2/512 disabled → runs on any HF CPU) ─
RUN pip install --no-cache-dir -U pip && \
    CMAKE_ARGS="-DGGML_BLAS=OFF -DGGML_CUDA=OFF -DGGML_METAL=OFF \
                -DGGML_AVX2=OFF -DGGML_AVX512=OFF" \
    FORCE_CMAKE=1 \
    pip install --no-cache-dir --timeout=1800 llama-cpp-python==0.2.90

# ── App dependencies ───────────────────────────────────────────────────────────
COPY requirements-hf.txt .
RUN pip install --no-cache-dir -r requirements-hf.txt

# ── Copy application source ────────────────────────────────────────────────────
COPY . .

# ── HF Spaces expects port 7860 ───────────────────────────────────────────────
EXPOSE 7860

ENV PORT=7860

CMD ["python", "app.py"]