@ifc-lite/extensions

Extension manifest, capability grammar, and slot registry for IFClite's user-customization system.

This package implements the non-UI half of the design described in docs/architecture/ai-customization/. It is host-agnostic — the same code is consumed by the browser viewer, the desktop app, the CLI, and the headless server.

What's here (v0.1.0 — Phase 0)

• Manifest — typed schema + hand-rolled validator producing structured { path, code, hint } errors.
• Capability grammar — parser + matcher + risk-badge computation
  • set-diff helpers. The OCAP capability vocabulary that gates every bridge call.
• Slot registry — in-memory pub/sub for contribution points. The host subscribes; the loader registers extensions; the registry composes contributions.
• when clause language — small expression grammar (boolean ops, comparisons, dotted identifiers, literals) the host evaluates to show/hide contributions.
• Bundle layout — walker that reads a directory or a .iflx archive into an in-memory Bundle value. Verifies referenced files exist.
• Manifest migrations — chain scaffold for forward-compatibility with future manifest versions.

Coming in later phases

• Phase 1 — IndexedDB storage, host loader, runtime activation, sandbox wiring, audit log, viewer-side slot binding.
• Phase 2 — Widget DSL renderer, AI authoring pipeline, repair loop.
• Phase 3 — Flavor data model, export/import, three-way merge.
• Phase 4 — Action log, pattern miner, prompt overlay, SDK-update repair.

Usage (Phase 0)

import {
  validateManifest,
  parseCapability,
  matchCapability,
  computeRisk,
  diffCapabilities,
  parseWhen,
  evaluateWhen,
  SlotRegistry,
  loadBundleFromDirectory,
} from '@ifc-lite/extensions';

// Validate a manifest
const result = validateManifest(manifestJson);
if (result.ok) {
  console.log(result.value.id);
} else {
  for (const err of result.errors) {
    console.error(`${err.path}: ${err.message}`);
  }
}

Design references

• docs/architecture/ai-customization/01-extension-model.md
• docs/architecture/ai-customization/02-security.md
• docs/architecture/ai-customization/03-ui-surface.md

Licensed under MPL-2.0.