@cunny-ai/core

Model registry, cached loading and engine for @cunny-ai/* task packages.

import { createEngine } from '@cunny-ai/core'

const engine = createEngine()
const model = await engine.loadModel('face-detect', {
  model: 'balanced', // tier alias, exact model id, or default
  onProgress: ({ loaded, total }) => console.log(`${loaded}/${total}`),
})

• Registry: task → models → variants, with license and size metadata
• Cache API storage, concurrent download dedupe, progress events
• Typed errors (ModelNotFoundError, ProviderMissingError)

Task packages use this internally. API reference. MIT.