Dockerfile

# PlaigLab — Hugging Face Spaces (Docker SDK) image.
# Always-on container: long-running investigations, background threads and the
# in-memory job table all work as written. Listens on 0.0.0.0:7860 (HF routes
# external traffic to that port).
FROM python:3.11-slim

# System libs some wheels expect at runtime (pymupdf, onnxruntime, numpy).
RUN apt-get update && apt-get install -y --no-install-recommends \
        libgl1 libglib2.0-0 \
    && rm -rf /var/lib/apt/lists/*

# HF Spaces run as uid 1000; give that user a writable home + app dir.
RUN useradd -m -u 1000 user
ENV HOME=/home/user \
    PATH=/home/user/.local/bin:$PATH \
    HOST=0.0.0.0 \
    PORT=7860 \
    PYTHONUNBUFFERED=1
WORKDIR /home/user/app

# Install Python deps first so this layer caches across code changes.
COPY --chown=user requirements.txt .
RUN pip install --no-cache-dir --upgrade pip \
    && pip install --no-cache-dir -r requirements.txt

# Copy the project (respects .dockerignore).
COPY --chown=user . .

# Fetch the Phase-A model stack (MiniLM + GPT-2/distilGPT-2 ONNX, ~250MB) at
# build time so the semantic + Binoculars AI ensemble is live on first request.
# Build still succeeds if a download hiccups — the app degrades to lexical mode.
RUN python scripts/get_models.py || echo "model fetch incomplete; running in lite mode"

USER user
EXPOSE 7860
CMD ["python", "webapp/server.py"]