| # DiagraMine — Streamlit demo + FastAPI image for the Hugging Face Space. | |
| # | |
| # Same as the repo's Dockerfile, plus what a Space needs: Hugging Face runs | |
| # the container as user 1000, so the app lives in that user's home (PaddleOCR | |
| # caches its models under ~/.paddleocr); torch comes from the CPU-only index | |
| # (the default wheels pull in CUDA, several GB); and the OCR models are fetched | |
| # at build time so the first /extract after a cold start doesn't download them. | |
| # nginx puts both apps on the one public port: the demo at /, the API at | |
| # /extract, /health and /docs (see nginx.conf and start.sh). | |
| FROM python:3.11-slim | |
| ENV PYTHONUNBUFFERED=1 \ | |
| PYTHONDONTWRITEBYTECODE=1 \ | |
| PIP_NO_CACHE_DIR=1 \ | |
| DEBIAN_FRONTEND=noninteractive | |
| # OS libs: OpenCV needs libGL + libglib; matplotlib needs libfreetype/libpng; | |
| # paddlepaddle's CPU build links the OpenMP runtime (libgomp). | |
| RUN apt-get update && apt-get install -y --no-install-recommends \ | |
| libgomp1 \ | |
| libgl1 \ | |
| libglib2.0-0 \ | |
| libsm6 \ | |
| libxext6 \ | |
| libxrender1 \ | |
| libfreetype6 \ | |
| libpng16-16 \ | |
| nginx \ | |
| && rm -rf /var/lib/apt/lists/* | |
| RUN useradd -m -u 1000 user | |
| USER user | |
| ENV HOME=/home/user \ | |
| PATH=/home/user/.local/bin:$PATH \ | |
| MPLCONFIGDIR=/home/user/.config/matplotlib | |
| WORKDIR /home/user/app | |
| # Install Python deps first for cacheability. | |
| COPY --chown=user requirements.txt ./ | |
| RUN pip install --user --upgrade pip \ | |
| && pip install --user --extra-index-url https://download.pytorch.org/whl/cpu -r requirements.txt | |
| # Source. | |
| COPY --chown=user src/ ./src/ | |
| COPY --chown=user diagram_analysis.py ./ | |
| COPY --chown=user data/icon_templates/ ./data/icon_templates/ | |
| COPY --chown=user app.py ui_theme.py search_interview_test.png ./ | |
| COPY --chown=user .streamlit/ ./.streamlit/ | |
| COPY --chown=user data/eval/ ./data/eval/ | |
| COPY --chown=user nginx.conf start.sh ./ | |
| # PaddleOCR's English det/rec/cls models (~15 MB), fetched now rather than on | |
| # the first request. torch first, as in src/text_detection/paddle_detector.py. | |
| RUN python -c "import torch; from paddleocr import PaddleOCR; PaddleOCR(use_angle_cls=True, lang='en', show_log=False)" | |
| EXPOSE 8000 | |
| CMD ["bash", "start.sh"] | |