Dockerfile

# Container image for local development, Kubernetes, and any container host.
#
# This is NOT what runs on Vercel — Vercel builds from api/index.py and
# requirements.txt directly. The two differ deliberately: this image installs
# requirements-local.txt as well, so the real cross-encoder reranker and the
# local embedding model are available. That is what makes benchmark configs D
# and E runnable at all.

FROM python:3.12-slim AS base

ENV PYTHONUNBUFFERED=1 \
    PYTHONDONTWRITEBYTECODE=1 \
    PIP_NO_CACHE_DIR=1 \
    PIP_DISABLE_PIP_VERSION_CHECK=1

WORKDIR /app

# curl is for the healthcheck; build-essential is needed by a few wheels that
# still compile on slim.
RUN apt-get update \
 && apt-get install -y --no-install-recommends curl build-essential \
 && rm -rf /var/lib/apt/lists/*

# ---------------------------------------------------------------------------
# Dependencies in their own layer so a source change does not reinstall torch.
# ---------------------------------------------------------------------------
COPY requirements.txt requirements-local.txt ./

ARG INCLUDE_LOCAL_MODELS=true
RUN pip install --upgrade pip \
 && if [ "$INCLUDE_LOCAL_MODELS" = "true" ]; then \
        pip install -r requirements-local.txt; \
    else \
        pip install -r requirements.txt; \
    fi

COPY . .

# Run as a non-root user. HF_HOME must be writable or the local embedder cannot
# download weights on first use.
RUN useradd --create-home --uid 10001 documind \
 && mkdir -p /home/documind/.cache/huggingface \
 && chown -R documind:documind /app /home/documind
USER documind

ENV HF_HOME=/home/documind/.cache/huggingface \
    ENVIRONMENT=docker \
    PORT=8000

EXPOSE 8000

HEALTHCHECK --interval=30s --timeout=5s --start-period=40s --retries=3 \
    CMD curl -fsS "http://localhost:${PORT}/live" || exit 1

# Single worker by default: the process holds a database pool and, when the
# local models are installed, model weights in memory. Scale with replicas, not
# with workers.
CMD ["sh", "-c", "uvicorn app.main:app --host 0.0.0.0 --port ${PORT} --workers 1 --timeout-keep-alive 75"]