@abraca/schema

The canonical page-type schema spec for Abracadabra. One Zod declaration per doc-type generates TypeScript, a JSON-Schema bundle, Rust, Swift, and Markdown — so every language validates the same shapes. Zero runtime dependencies.

Documentation

Full, code-derived documentation lives in docs/ — the registry model, the universal meta object, all 16 page types, the query DSL, CRDT markers, the plugin manifest validator, and the codegen pipeline. It is the source of truth (more authoritative than any prose page-type guide).

At a glance

import { defineSchema, Kanban, Doc } from "@abraca/schema";
const schema = defineSchema({ types: [Kanban, Doc] });
schema.validateMeta("kanban", meta);   // {ok:true} for unknown types (Rule 4)

• 16 page types — doc, prose, kanban, gallery, table, outline, checklist, graph, timeline, calendar, map, dashboard, chart, sheets, slides, overview.
• Additive / opt-in — no aggregate schema; compose the subset you need; unknown types are never rejected, known types are closed (additionalProperties:false).
• query.ts is the wire contract the Rust server's query engine mirrors.

::

The generated JSON-Schema bundle (registry/v1.json) is the real cross-language contract — consumed by the Rust SchemaValidationExtension and MCP tool schemas. Never hand-edit src/generated/**; run pnpm schema:gen (CI verifies). yjs/zod are peer deps.

License

MIT.