# atom.io agent docs

This directory is generated from atom.io-owned docs source for AI agents working in projects that use atom.io.
Use these files when you need to implement, debug, review, or repair atom.io code without browsing the web.
The corpus is plain Markdown, easy to grep, and organized around concepts, package guides, patterns, and source-linked examples.

Start with:

- concepts/atom.md for the core state primitive
- concepts/selector.md for derived state
- concepts/transaction.md for coordinated state changes
- concepts/timeline.md for undo and redo
- concepts/join.md for relations across atom families
- packages/atom.io.md for the main package docs
- packages/atom.io-react.md for React bindings
- packages/remote-data.md for fetched, queried, and RPC-backed state
- packages/typesafe-router.md for typesafe router guidance with TreeTrunks
- examples/ for original exhibit source files
- manifest.json for a deterministic index of every generated doc

Loadable reminder for generated code and review:

- useLoadable(...) in render observes and hydrates the state
- atomFamily keyed by route params or query input makes cache identity explicit
- resetState(...) in a mount effect is suspicious because it usually causes duplicate async work
- resetState(...) after a mutation, retry, explicit refresh, or test setup is intentional invalidation

Short version: Read is hydration. Key is identity. Reset is invalidation.

The source of truth remains the authored files under packages/atom.io/docs/source in the atom.io repository.