Search stack replicates the Hamilton site pattern with pg_orrery-specific additions: - FastAPI REST API (chat SSE streaming, semantic search, health check) - FastMCP server at /mcp with doc search and live SQL query tools - pgvector + pgai vectorizer for 1024-dim document embeddings - Hybrid search (semantic cosine + text ILIKE with pg_trgm GIN) - Dual LLM backend: self-hosted qwen3 via GPU gateway or Anthropic Claude - Live read-only pg_orrery SQL execution with safety guardrails (SELECT-only validation, read-only transaction, 5s timeout, 100-row cap) - Convenience MCP tools: planet_position, sky_survey, satellite_pass - MDX content ingestion from docs/src/content/docs/ (50 pages) - Docker Compose: pg_orrery+pgvector DB, pgai, vectorizer-worker, API - Alembic async migrations, Makefile, .env.example
30 lines
1.0 KiB
Docker
30 lines
1.0 KiB
Docker
# syntax=docker/dockerfile:1
|
|
FROM ghcr.io/astral-sh/uv:python3.12-bookworm-slim AS base
|
|
WORKDIR /app
|
|
|
|
ENV UV_COMPILE_BYTECODE=1
|
|
ENV UV_LINK_MODE=copy
|
|
|
|
# Install dependencies first (cache layer)
|
|
COPY pyproject.toml ./
|
|
RUN --mount=type=cache,target=/root/.cache/uv \
|
|
uv pip install --system -r pyproject.toml
|
|
|
|
# Copy source
|
|
COPY . .
|
|
|
|
# --- Development (editable install for hot reload) ---
|
|
FROM base AS dev
|
|
RUN --mount=type=cache,target=/root/.cache/uv \
|
|
uv pip install --system --no-deps -e .
|
|
CMD ["uvicorn", "orrery_search.main:app", "--host", "0.0.0.0", "--port", "8000", "--reload", "--proxy-headers", "--forwarded-allow-ips", "*"]
|
|
|
|
# --- Production (non-editable, non-root) ---
|
|
FROM base AS prod
|
|
RUN --mount=type=cache,target=/root/.cache/uv \
|
|
uv pip install --system --no-deps .
|
|
RUN adduser --disabled-password --gecos "" orrery && \
|
|
mkdir -p /data && chown -R orrery:orrery /data
|
|
USER orrery
|
|
CMD ["uvicorn", "orrery_search.main:app", "--host", "0.0.0.0", "--port", "8000", "--workers", "2", "--proxy-headers", "--forwarded-allow-ips", "*"]
|