Documentation

Everything here is rendered from the docs/ folder of the repository. To try Groundnote first, follow the quickstart in the README.

Run it

Configuration: The groundnote.json file, environment variables, Cloudflare Access and backup settings.
CLI: Every command of the groundnote CLI, doctor and the nightly backup run.
Backup: The ways to back up a Groundnote instance, and how to restore and verify it.
Providers: Which text goes to which provider, and the terms to read before you switch one on.

Architecture: The binding contract for the data model, the API and authentication.
Approval modes: Modes A, B and C, how proposals are classified, the mandatory exceptions and retract.
Knowledge graph: Entities and links: the tree that gives rules and sources their place in a scope.
Lexical search: How stage a searches, and what each step adds in the measurements.
Search stages b and c: Local embeddings and the JEV reranker: settings, limits and how they fit into the search.
Redaction: The proposed design for redaction, erasure and encrypted backups. Not built yet.

Guide for agents: How an AI agent uses the MCP tools and prompts, with complete example dialogues.

Threat model: What Groundnote defends against, and how.
Data at rest: How stored data and backups are protected. A design record: erasure is not built yet.

Roadmap: Proposals measured on real data, and what comes next.
Practice findings: What a practice test with real meetings, mails and chats showed.
Releasing: How a version gets from main to the image and the release page.
Deviations: Where Groundnote departs from the original German contract, and why.
Glossary: The German contract terms and their English names in the code.
Brand: The mark, the colours and the type, and how to use them.