RAG Pipeline

Ingest documents, embed them, and retrieve relevant context during chat. Uses @agentskit/rag with any embedder and vector store.

Ingest documents, embed them, and retrieve relevant context during chat. Uses @agentskit/rag with any embedder and vector store.

Setup

import { createRAG } from '@agentskit/rag'
import { openaiEmbedder } from '@agentskit/adapters'

const rag = createRAG({
  embed: openaiEmbedder({ apiKey: process.env.OPENAI_API_KEY! }),
  store: yourVectorStore,  // SQLite, Redis, or in-memory
  chunkSize: 512,
  chunkOverlap: 50,
})

Ingest Documents

await rag.ingest([
  { id: 'readme', content: readFileSync('README.md', 'utf-8'), source: 'README.md' },
  { id: 'guide', content: readFileSync('docs/guide.md', 'utf-8'), source: 'guide.md' },
])

Search Directly

const results = await rag.search('how to configure tools', { topK: 3 })
results.forEach(doc => {
  console.log(`[${doc.source}] ${doc.content.slice(0, 100)}...`)
})

Use with Chat

createRAG returns a Retriever — pass it directly to useChat or the runtime:

import { useChat } from '@agentskit/react'

function RAGChat() {
  const chat = useChat({
    adapter: yourAdapter,
    retriever: rag,  // retrieved context auto-injected into system prompt
  })
  // ... render chat UI
}

Custom Chunking

const rag = createRAG({
  embed: yourEmbedder,
  store: yourStore,
  split: (text) => text.split('\n\n'),  // paragraph-based chunking
})

✎ Edit this page on GitHub·Found a problem? Open an issue →·How to contribute →