@agentskit/rag

@agentskit/rag

Add your own knowledge to any agent in 3 lines — coming soon.

Why

• Your data, your agent — chunk documents, embed them, and have agents retrieve exactly the right context at query time; no fine-tuning required
• Works with what you already have — plugs directly into @agentskit/memory vector backends and any embedder from @agentskit/adapters, so nothing new to learn
• 3-line integration — rag.ingest(docs), rag.retrieve(query), and a drop-in useRAGChat() hook mean you go from idea to working knowledge base in minutes

Coming soon. This package is scaffolded and will be implemented in a future release. See #13.

Planned API

import { createRAG } from '@agentskit/rag'
import { fileVectorMemory } from '@agentskit/memory'
import { anthropic } from '@agentskit/adapters'

const rag = createRAG({
  vectorMemory: fileVectorMemory({ path: './vectors' }),
  embedder: anthropic({ apiKey: process.env.ANTHROPIC_API_KEY }),
})

await rag.ingest([{ id: 'doc-1', content: 'Your knowledge base content here...' }])
const context = await rag.retrieve('What does the docs say about X?')

Docs

Full documentation