CensusChat Documentation
CensusChat Documentation
Natural language queries over US Census demographics, for healthcare strategy teams.
Pick the path that matches what you came for.
I want to run it
| Step | Document |
|---|---|
| 1. Get it running locally | Quick Start |
| 2. Configure API keys (Anthropic + Census) | API Key Setup |
| 3. Load the full dataset (states → block groups) | Quick Start: Expanded Data |
| 4. Load ACS data directly from the Census API | ACS Data Loading |
| Docker will not start | Docker Troubleshooting |
| Deploy to Railway | Railway Deployment |
| Deploy to production generally | Deployment Guide |
The Census API key is optional but raises the rate limit. Its own setup notes: Census API Key Setup.
I want to understand the architecture
| Question | Document |
|---|---|
| What is the system, end to end? | Architecture Overview |
| How do the services fit together? | System Architecture |
| What is the data model? | Data Architecture |
| How is SQL kept safe? | Security Architecture |
| What runs where? | Infrastructure Architecture |
| How is the backend laid out? | Backend Structure |
| How is the frontend laid out? | Frontend Architecture |
| What variables and geographies exist? | Marketing Analytics Expansion |
| What do the block-group columns mean? | Block Group Variables |
Deeper reference material: DuckDB and MCP references.
I want the API
| Topic | Document |
|---|---|
| Endpoint reference | API Overview |
| MCP tools and protocol | MCP API Documentation |
| Current MCP schema | MCP Schema Update |
| Run the MCP server over stdio | MCP stdio Setup |
| Build against the FDB MCP layer | FDB MCP Developer Guide |
| Request/response flow and validation | API Integration Guide |
| How MCP is implemented | MCP Implementation Summary |
I want to contribute
| Topic | Document |
|---|---|
| Contribution process and standards | CONTRIBUTING.md |
| Branching strategy | Git Branching Strategy |
| Commit conventions | Commit Sequence Examples |
| Running the test suite | Testing Guide |
| How the tests are built | Testing Infrastructure |
| Repo conventions for AI agents | CLAUDE.md |
| Reporting a vulnerability | SECURITY.md |
| Release history | CHANGELOG.md |
Data-pipeline internals, if you are changing the loaders: Data Loading System and Implementation Summary.
Not part of the published docs
These stay in the repository for history. They are excluded from the published site and are not maintained as current documentation.
internal/— session notes, epic planning, documentation-status snapshotsplans/— in-flight implementation plansarchive/— superseded documents, kept for the recordproject-management/— roadmap, personas, PRD, MVP status. Some of it is dated; read it as project history, not as a description of the current build.