Dockerfile.lambda

# syntax=docker/dockerfile:1.7
# ── Lambda CV Pipeline Container ──────────────────────────────
# Runs template matching (symbol search) and shape parsing (keynote
# detection) on batches of blueprint pages. Invoked by the web server
# via AWS Lambda fan-out for parallel page processing.
#
# Uses python:3.11-slim (Debian) + awslambdaric instead of the official
# Lambda base image (Amazon Linux) because:
#   1. Main app Dockerfile uses Debian slim — matching system libraries
#   2. tesseract-ocr installs at /usr/bin/tesseract via apt-get
#      (extract_keynotes.py line 22 hardcodes this path)
#   3. opencv-python-headless wheels are reliable on Debian
#
# Build:  docker build -f Dockerfile.lambda -t beaver-cv-lambda .
# ~400-500MB image. Cold start ~5-8s.

FROM python:3.11-slim

ENV DEBIAN_FRONTEND=noninteractive
ENV PYTHONUNBUFFERED=1

# System deps — same packages as main Dockerfile lines 45-57
RUN apt-get update && apt-get install -y --no-install-recommends \
    tesseract-ocr \
    tesseract-ocr-eng \
    libglib2.0-0 \
    libsm6 \
    libxext6 \
    libxrender1 \
    libgl1 \
    && rm -rf /var/lib/apt/lists/*

# Python deps + AWS Lambda Runtime Interface Client.
# --no-cache-dir removed: BuildKit cache mount keeps wheel cache outside the
# layer, so no image bloat and reinstalls reuse downloads.
RUN --mount=type=cache,target=/root/.cache/pip,sharing=locked \
    pip install \
    awslambdaric \
    numpy \
    opencv-python-headless \
    pytesseract \
    boto3

COPY scripts/template_match.py /opt/code/
COPY scripts/extract_keynotes.py /opt/code/
COPY scripts/lambda_handler.py /opt/code/

WORKDIR /opt/code

ENTRYPOINT ["python3", "-m", "awslambdaric"]
CMD ["lambda_handler.handler"]