vectorstores Provider

The vectorstores provider integrates vectorstores with the AI SDK, enabling AI workflows that retrieve and store data in vector databases.

Setup

The vectorstores provider is available in the @vectorstores/vercel module. You can install it with:

pnpm add @vectorstores/vercel @vectorstores/core

Document Indexing

Before you can use the vectorstores provider, you need to define an index for your documents. An index stores document embeddings in a vector database, enabling semantic search over your content.

An easy way to create an index is by using the VectorStoreIndex.fromDocuments function from the @vectorstores/core package. This function will automatically chunk your documents into smaller chunks, embed them, and store them in the vector database.

To be able to embed the documents, you need an embedding model. You can use the vercelEmbedding function (see Utilities) to use any AI SDK embedding model.

Once you have an index, you can use a retriever to find the most relevant documents based on similarity to a given query.

Utilities

The vectorstores provider offers two main utility functions for integrating AI SDK with vectorstores.

vercelEmbedding

The vercelEmbedding function adapts any AI SDK embedding model for use with vectorstores. This enables you to use embedding providers like OpenAI or Cohere within the vectorstores ecosystem.

Here is an example of how to create an index using the vercelEmbedding function:

import { openai } from '@ai-sdk/openai';
import { vercelEmbedding } from '@vectorstores/vercel';
import { VectorStoreIndex } from '@vectorstores/core';


const documents = [new Document({ text: 'This is a large document.' })];


const index = await VectorStoreIndex.fromDocuments(documents, {
  embedFunc: vercelEmbedding(openai.embedding('text-embedding-3-small')),
});

vercelTool

The vercelTool function wraps a vectorstores retriever as an AI SDK tool, enabling agents that can autonomously access data in vector databases.

import { generateText } from 'ai';
import { openai } from '@ai-sdk/openai';
import { vercelTool } from '@vectorstores/vercel';


const { text } = await generateText({
  model: openai('gpt-5-mini'),
  tools: {
    search: vercelTool({
      retriever: index.asRetriever(),
      description: 'Search the knowledge base for pricing information',
    }),
  },
  prompt: 'What is the price of a burrito?',
});

Configuration Options

OptionTypeRequiredDescription
retrieverBaseRetrieverYesThe vectorstores retriever instance to wrap
descriptionstringYesGuidance text helping the LLM understand when and how to use the tool
noResultsMessagestringNoCustom fallback message when no documents match (default: "No results found in documents.")

Use Cases

Agentic RAG

Ask questions over the knowledge stores in the vector database. The LLM autonomously decides whether to query the vector database using a tool, retrieves relevant content, and incorporates findings into responses.

View example →

Agent Memory Systems

Store and retrieve user-specific information across conversations by combining a tool storing memories in the vector database and a tool retrieving memories from the vector database.

View example →

References