Dockerfile

# =============================================================
#  Multi-stage Docker build for Contextual Similarity Engine
#  Single container: React frontend + FastAPI backend
#  Deploys to: HuggingFace Spaces (Docker SDK), local, Railway
# =============================================================

# Stage 1: Build frontend
FROM node:22-slim AS frontend-build
WORKDIR /app/frontend
COPY frontend/package.json frontend/package-lock.json ./
RUN npm ci
COPY frontend/ ./
RUN npm run build && node inline-assets.cjs

# Stage 2: Python runtime
FROM python:3.12-slim AS runtime

# Create non-root user (required by HF Spaces)
RUN useradd -m -u 1000 appuser
WORKDIR /app

# System deps for faiss-cpu and torch
RUN apt-get update && apt-get install -y --no-install-recommends \
    build-essential \
    && rm -rf /var/lib/apt/lists/*

# Install uv for fast dependency resolution
COPY --from=ghcr.io/astral-sh/uv:latest /uv /usr/local/bin/uv

# Copy dependency files first (cache layer)
COPY --chown=appuser pyproject.toml uv.lock ./

# Install Python dependencies
RUN uv sync --frozen --no-dev

# Copy backend source
COPY --chown=appuser *.py ./

# Copy pre-built frontend
COPY --chown=appuser --from=frontend-build /app/frontend/dist ./frontend/dist

# Data directories (HF cache, gensim/GloVe cache, engine state, trained models)
RUN mkdir -p /data/huggingface /data/gensim-data /data/engine_state /data/w2v_state /data/trained_model \
    && chown -R appuser:appuser /app /data

ENV HF_HOME=/data/huggingface
ENV TRANSFORMERS_CACHE=/data/huggingface
ENV ENGINE_STATE_DIR=/data/engine_state
ENV W2V_STATE_DIR=/data/w2v_state
# GloVe background model cache for anomaly detection — keep it under /data so it
# sits on persistent storage and isn't re-downloaded (~128MB) on every cold start.
ENV GENSIM_DATA_DIR=/data/gensim-data
# Override the background model (e.g. glove-wiki-gigaword-50) without code changes.
ENV BACKGROUND_MODEL=glove-wiki-gigaword-100

# Bake the background model into the image so anomaly detection works offline,
# with no runtime download or cold-start wait (adds ~128MB to the image).
# return_path=True just ensures the download/cache without loading it into RAM.
RUN uv run python -c "import os, gensim.downloader as d; d.load(os.environ['BACKGROUND_MODEL'], return_path=True)" \
    && chown -R appuser:appuser /data/gensim-data

# Switch to non-root user
USER appuser

# Expose port (HF Spaces expects 7860, override via PORT env)
EXPOSE 7860

# Run the server — HOST and PORT configurable via env
ENV HOST=0.0.0.0
ENV PORT=7860
CMD ["uv", "run", "python", "server.py"]