Simple RAG

Set up semantic search with document ingestion and vector-based retrieval.

Install the SDK

Install the TypeGraph SDK, a database adapter, an embedding provider, and an LLM provider when graph extraction is enabled.

terminal

npm install @typegraph-ai/sdk @typegraph-ai/adapter-pgvector @ai-sdk/gateway @neondatabase/serverless

Initialize TypeGraph

Use typegraphDeploy() to provision Postgres, then create a tenant-scoped client with typegraphInit(). The same vectorStore + embedding + searchEmbedding + llm configuration supports document search, graph extraction, graph APIs, and memory APIs.

Initialize TypeGraph example

import { gateway } from '@ai-sdk/gateway'
import { neon } from '@neondatabase/serverless'
import { PgVectorAdapter } from '@typegraph-ai/adapter-pgvector'
import { TenantId, typegraphDeploy, typegraphInit } from '@typegraph-ai/sdk'

const sql = neon(process.env.DATABASE_URL!)
const vectorStore = new PgVectorAdapter({ sql })

const embedding = {
  model: gateway.embeddingModel('openai/text-embedding-3-small'),
  dimensions: 1536,
}

await typegraphDeploy({
  tenantId: TenantId('tenant_acme'),
  vectorStore,
  embedding,
  searchEmbedding: embedding,
  llm: { model: gateway('openai/gpt-5.4-mini') },
})

const typegraph = await typegraphInit({
  tenantId: TenantId('tenant_acme'),
  vectorStore,
  embedding,
  searchEmbedding: embedding,
  llm: { model: gateway('openai/gpt-5.4-mini') },
})

Ingest Documents

Documents are durable long-form content. Use document.ingest() for one document or an array of documents. The default bucket is created automatically; pass bucketId when you want a named collection.

Document ingest example

const result = await typegraph.document.ingest(
  [{
    id: 'docs:getting-started',
    name: 'Getting Started',
    description: 'Introductory setup guide for TypeGraph.',
    content: 'TypeGraph is a unified RAG pipeline SDK...',
    metadata: { product: 'typegraph' },
  }],
  {
    graphExtraction: false,
  },
)

console.log('Ingested:', result)

Search Documents

typegraph.search() separates what to search (resources) from how to score (weights). Use promptBuilder when you want TypeGraph to assemble the top results into an LLM-ready prompt string on the response.

Search example

const response = await typegraph.search('How do I get started?', {
  resources: ['documents'],
  weights: { semantic: 1, bm25: 0.7, graph: false, recency: 0.3 },
  limit: 5,
  promptBuilder: { format: 'xml' },
})

for (const chunk of response.results.chunks) {
  console.log(chunk.scores.output.fused, chunk.content)
}

console.log(response.prompt)

Example: Feed the Prompt to an LLM

Example only. TypeGraph does not require an LLM for vector-only RAG. If you want grounded answer generation, pass response.prompt to any model in your application.

rag-example

import { generateText } from 'ai'
import { openai } from '@ai-sdk/openai'

const { text } = await generateText({
  model: openai('gpt-5.4-mini'),
  prompt: `Answer the question based on the following context:

${response.prompt}

Question: How do I get started?
`,
})

console.log(text)

Back to All Guides

import { gateway } from '@ai-sdk/gateway' import { neon } from '@neondatabase/serverless' import { PgVectorAdapter } from '@typegraph-ai/adapter-pgvector' import { TenantId, typegraphDeploy, typegraphInit } from '@typegraph-ai/sdk' const sql = neon(process.env.DATABASE_URL!) const vectorStore = new PgVectorAdapter({ sql }) const embedding = { model: gateway.embeddingModel('openai/text-embedding-3-small'), dimensions: 1536, } await typegraphDeploy({ tenantId: TenantId('tenant_acme'), vectorStore, embedding, searchEmbedding: embedding, llm: { model: gateway('openai/gpt-5.4-mini') }, }) const typegraph = await typegraphInit({ tenantId: TenantId('tenant_acme'), vectorStore, embedding, searchEmbedding: embedding, llm: { model: gateway('openai/gpt-5.4-mini') }, })

const result = await typegraph.document.ingest( [{ id: 'docs:getting-started', name: 'Getting Started', description: 'Introductory setup guide for TypeGraph.', content: 'TypeGraph is a unified RAG pipeline SDK...', metadata: { product: 'typegraph' }, }], { graphExtraction: false, }, ) console.log('Ingested:', result)

const response = await typegraph.search('How do I get started?', { resources: ['documents'], weights: { semantic: 1, bm25: 0.7, graph: false, recency: 0.3 }, limit: 5, promptBuilder: { format: 'xml' }, }) for (const chunk of response.results.chunks) { console.log(chunk.scores.output.fused, chunk.content) } console.log(response.prompt)

import { generateText } from 'ai' import { openai } from '@ai-sdk/openai' const { text } = await generateText({ model: openai('gpt-5.4-mini'), prompt: `Answer the question based on the following context: ${response.prompt} Question: How do I get started? `, }) console.log(text)