@apisurf/canon

The canon command: author documents — ordered lists of blocks — in one local SQLite file.

npm i -g @apisurf/canon
canon help

Every command here is about one record — create it, read it, change it, move it, delete it — and every one of them is something an agent does. Turning the result into a website is the other half of the pair, @apisurf/canonui, which ships separately so that nothing an agent needs drags a static site generator along with it:

npm i -g @apisurf/canonui
canonui build retry-guide

Everything is documented in the CLI itself — canon help, then canon help <command> — and in the repository README.

What is in here

| File | What it is | | ------------ | ------------------------------------------------------------ | | bin.ts | Argument parsing and dispatch. One parser per command shape | | help.ts | Every help page. One screen each, never a wall of everything | | howto.ts | Worked examples: whole tasks, start to finish | | read.ts | get and ls — projected reads, bounded by default | | write.ts | new, set, rm, mv — one store call each | | cat.ts | A document, as one markdown document | | find.ts | Full-text search over every block | | refs.ts | Turning --document retry-guide into the id a column holds | | content.ts | Where a block's body comes from: inline, a file, or stdin | | format.ts | Tables, CSV and JSON for the terminal | | version.ts | The build's version, and the schema it expects |

The schema, the store and the entity definitions live in @canon/db. The markdown cat prints, and the dates every command spells, live in @canon/core — shared with the theme in @apisurf/canonui, which serves the same documents at /index.md. Both are workspace packages bundled into dist/bin.js at build time.