Open Census MCP Server

# --------------------------------------------- # Census‑MCP runtime image – version 2.0 # --------------------------------------------- # * Python 3.11‑slim as base # * R + tidycensus for data retrieval helpers # * Pre-cached sentence transformer model (all-mpnet-base-v2) # * Runtime‑only KB assets (ChromaDB + FAISS, concepts, geo scalars) # * No raw PDFs / build scripts → lean container # # Build: # docker build -t ghcr.io/brockwebb/open-census-mcp:2.0 . # --------------------------------------------- FROM python:3.11-slim AS base LABEL maintainer="brockwebb" \ org.opencontainers.image.source="https://github.com/brockwebb/open-census-mcp-server" # -------------------------------------------------- # 1️⃣ System‑level dependencies (R + dev headers) # -------------------------------------------------- RUN apt-get update && apt-get install -y --no-install-recommends \ r-base r-base-dev \ r-cran-dplyr r-cran-jsonlite \ libcurl4-openssl-dev libssl-dev libxml2-dev \ libgdal-dev libudunits2-dev libproj-dev libgeos-dev \ libfontconfig1-dev libharfbuzz-dev libfribidi-dev \ libfreetype6-dev libpng-dev libtiff5-dev libjpeg-dev \ build-essential \ cmake \ libabsl-dev \ git \ && rm -rf /var/lib/apt/lists/* # -------------------------------------------------- # 2️⃣ Python dependencies # -------------------------------------------------- WORKDIR /app COPY requirements.txt . RUN pip install --no-cache-dir -r requirements.txt # -------------------------------------------------- # 2.5️⃣ Pre-download sentence transformer model # -------------------------------------------------- RUN python -c "from sentence_transformers import SentenceTransformer; \ import os; \ os.makedirs('/app/model_cache', exist_ok=True); \ print('🔄 Downloading all-mpnet-base-v2 model (~400MB)...'); \ model = SentenceTransformer('sentence-transformers/all-mpnet-base-v2', \ cache_folder='/app/model_cache'); \ print(f'✅ Model cached: {model.get_sentence_embedding_dimension()} dimensions'); \ print(f' Max sequence length: {model.max_seq_length}'); \ print(' Container is now self-contained!');" # Set environment variable so the app knows where to find the cached model ENV SENTENCE_TRANSFORMERS_CACHE=/app/model_cache # -------------------------------------------------- # 3️⃣ R packages needed by tidycensus # -------------------------------------------------- # Install tidycensus (sf dependencies should now work) RUN R -e "options(repos = c(CRAN = 'https://cloud.r-project.org/')); \ install.packages('tidycensus', dependencies=TRUE); \ if (!require('tidycensus', quietly=TRUE)) { \ cat('TIDYCENSUS INSTALLATION FAILED\n'); \ q(status=1) \ } else { \ cat('tidycensus installed successfully\n') \ }" # Verify all packages work RUN R -e "library(jsonlite); library(dplyr); library(tidycensus); cat('All R packages loaded successfully\n')" # -------------------------------------------------- # 4️⃣ Runtime‑only KB assets + application code # -------------------------------------------------- # ChromaDB knowledge base (main vector database) COPY knowledge-base/vector-db/ /app/data/vector_db/ # Remove obsolete stats-index (kb_search.py was removed) # COPY knowledge-base/stats-index/ /app/stats-index/ # Concepts + geo scalars COPY knowledge-base/concepts/concept_backbone.ttl \ knowledge-base/concepts/variable_links.ttl \ knowledge-base/concepts/geo_similarity_scalars.json \ /app/data/concepts/ # Static geo limits for GeoAdvisor COPY data/table_geos.json /app/data/ # Core data files COPY knowledge-base/2023_ACS_Enriched_Universe.json /app/ COPY knowledge-base/scripts/COOS_Complete_Ontology.json /app/ # Application code COPY src/ /app/src/ # -------------------------------------------------- # 5️⃣ Environment configuration # -------------------------------------------------- ENV VECTOR_DB_TYPE=chromadb \ PYTHONPATH=/app:/app/src \ CENSUS_MCP_CONTAINER=true \ R_EXECUTABLE=/usr/bin/Rscript \ EMBEDDING_MODEL=sentence-transformers/all-mpnet-base-v2 # -------------------------------------------------- # 6️⃣ Drop root privileges # -------------------------------------------------- RUN useradd -m -u 1000 census && chown -R census:census /app USER census # -------------------------------------------------- # 7️⃣ Container health‑check & launch # -------------------------------------------------- HEALTHCHECK --interval=30s --timeout=10s --start-period=40s --retries=3 \ CMD python -c "import sys; sys.path.append('/app/src'); from census_mcp_server import health_check; exit(0 if health_check() else 1)" CMD ["python", "src/census_mcp_server.py"]