Dockerfile

# ── Stage 1: Builder (downloads model) ──
FROM ghcr.io/ggml-org/llama.cpp:full AS builder

WORKDIR /app

RUN apt-get update && apt-get install -y --no-install-recommends \
    python3 python3-pip python3-venv \
    && rm -rf /var/lib/apt/lists/*

RUN python3 -m venv /opt/venv
ENV PATH="/opt/venv/bin:$PATH"

RUN pip install --no-cache-dir -U pip huggingface_hub

RUN python3 -c 'from huggingface_hub import hf_hub_download; \
    hf_hub_download(repo_id="Abiray/MiniCPM5-1B-GGUF", \
                    filename="minicpm5-1b-Q4_K_M.gguf", \
                    local_dir="/app")'

# ── Stage 2: Runtime ──
FROM ghcr.io/ggml-org/llama.cpp:full

# Create user FIRST (HF Spaces runs as UID 1000)
RUN useradd -m -u 1000 user

WORKDIR /app
ENV HOME=/home/user \
    PATH=/home/user/.local/bin:/opt/venv/bin:$PATH \
    PYTHONUNBUFFERED=1

# Minimal runtime deps
RUN apt-get update && apt-get install -y --no-install-recommends \
    python3 python3-pip python3-venv \
    ffmpeg \
    libgomp1 \
    && rm -rf /var/lib/apt/lists/*

# Copy venv from builder
COPY --from=builder /opt/venv /opt/venv

# Install runtime packages
RUN pip install --no-cache-dir flask edge-tts llama-cpp-python \
    && rm -rf /root/.cache/pip

# Copy model
COPY --from=builder /app/minicpm5-1b-Q4_K_M.gguf /app/minicpm5-1b-Q4_K_M.gguf

# ── CRITICAL: Copy app + assets with correct ownership ──
# The model/ and animation/ folders must be in your repo root
COPY --chown=user:user app.py /app/app.py
COPY --chown=user:user model /app/model
COPY --chown=user:user animation /app/animation

# Ensure user owns everything
RUN chown -R user:user /app

USER user

EXPOSE 7860

ENTRYPOINT []
CMD ["python3", "/app/app.py"]