Dockerfile

# syntax=docker/dockerfile:1
# ──────────────────────────────────────────────────────────────────────────────
# Vantage — single-container demo image.
#   • Stage 1 (web):  build the React/Vite frontend → static dist/
#   • Stage 2 (app):  FastAPI + agent + baked ML models, serves API *and* the dist
# Listens on 7860 (Hugging Face Spaces' expected port; override with $PORT).
# Secrets (Qdrant/Neo4j/OpenAI/Langfuse) are injected as env vars at run time —
# never baked in. See .env.example for the variable names.
# ──────────────────────────────────────────────────────────────────────────────

# ── Stage 1: build the frontend ───────────────────────────────────────────────
FROM node:20-slim AS web
WORKDIR /web
COPY apps/web/package.json apps/web/package-lock.json ./
RUN npm ci
COPY apps/web/ ./
RUN npm run build          # → /web/dist


# ── Stage 2: python API + models + static frontend ───────────────────────────
FROM python:3.11-slim AS app

ENV PYTHONUNBUFFERED=1 \
    PYTHONDONTWRITEBYTECODE=1 \
    PIP_NO_CACHE_DIR=1 \
    HF_HOME=/models \
    PORT=7860

WORKDIR /app

# curl is used by the container HEALTHCHECK below.
RUN apt-get update && apt-get install -y --no-install-recommends curl \
    && rm -rf /var/lib/apt/lists/*

# CPU-only torch first — avoids pulling the multi-GB CUDA build that
# sentence-transformers would otherwise drag in.
RUN pip install --index-url https://download.pytorch.org/whl/cpu torch==2.12.0

# Pinned runtime deps (matches the working py311 env), then the local package
# with --no-deps so its version ranges can't bump anything we just pinned.
COPY requirements.docker.txt ./
RUN pip install -r requirements.docker.txt
COPY packages/ ./packages/
RUN pip install --no-deps -e ./packages/vantage_core

# Bake the ML models into the image so runtime stays fully offline
# (the adapters open them with local_files_only=True).
RUN python - <<'PY'
from sentence_transformers import SentenceTransformer, CrossEncoder
from fastembed import SparseTextEmbedding
SentenceTransformer("all-MiniLM-L6-v2")
CrossEncoder("cross-encoder/ms-marco-MiniLM-L-6-v2")
list(SparseTextEmbedding("Qdrant/bm25").embed(["warmup"]))
print("models baked")
PY

# Application code + built frontend.
COPY apps/ ./apps/
COPY --from=web /web/dist ./apps/web/dist

EXPOSE 7860

HEALTHCHECK --interval=30s --timeout=5s --start-period=120s --retries=3 \
    CMD curl -fsS "http://localhost:${PORT}/api/health" || exit 1

# shell form so ${PORT} expands (HF/Railway/Render set their own PORT).
CMD uvicorn apps.api.main:app --host 0.0.0.0 --port ${PORT}