# Data Documentation Checker

Data Documentation Checker organizes and checks user-provided data documentation so field meanings, types, and constraints can be understood consistently by people, systems, and AI agents.

## Tools

- `extract_data_dictionary`: extracts explicitly stated field definitions from supplied documentation, field tables, JSON Schema, or sample structures.
- `check_data_documentation`: checks whether supplied documentation clearly states key field information.
- `compare_schema_versions`: compares two supplied schema or documentation versions by exact field name.

## Technical Stack

- Cloudflare Workers
- TypeScript
- MCP over JSON-RPC at `POST /mcp`

## Install

```bash
npm install
```

## Local Run

```bash
npm run dev
```

## Typecheck

```bash
npm run typecheck
```

## Local MCP Regression

```bash
npm run test:mcp:local
```

The regression script starts a local Wrangler worker, waits for `/health`, runs route checks, MCP `initialize`, `tools/list`, `tools/call` coverage for all three tools, error contract cases, unknown-tool handling, annotation checks, and output schema alignment checks. It exits automatically and does not require manual JSON entry.

## Routes

- `GET /`: minimal HTML status page.
- `GET /health`: stable JSON health response.
- `POST /mcp`: only MCP endpoint.

## Boundaries

This app is read-only and stateless. It has no login, OAuth, database, external API calls, external system access, or external side effects.

## Out of Scope

The app does not inspect real databases, check live data quality, modify schemas, generate migrations, repair data models, or call LLM or external APIs.

## Support

Support Email: `sidcraigau@gmail.com`

## Current Gate State

This project has not been deployed. It has not entered Developer Mode testing and has not been submitted for review.