1// 1. Initialize document
2const doc = MDocument.fromText("Your document text here...");
3
4// 2. Create chunks
5const chunks = await doc.chunk({
6  strategy: "recursive",
7  size: 512,
8  overlap: 50,
9});
10
11// 3. Generate embeddings; we need to pass the text of each chunk
12const { embeddings } = await embedMany({
13    values: chunks.map((c) => c.text),
14    model: openai.embeddings("text-embedding-3-small"),
15  })
16
17// 4. Store in vector database
18const pgVector = new pgVector({
19    connectionString: process.env.POSTGRES_CONNECTION_STRING,
20})

Mastra RAG

Everything you need for RAG and context engineering under one framework

Mastra handles the complete RAG pipeline and enhances LLM outputs by incorporating relevant context from your data sources.

Retrieving documents from mastra-db

Build your full RAG pipeline

Chunk documents, generate embeddings, store them in a vector DB, and retrieve context when you need it - all with one tool.

.embed()

.query()

.rerank()

input

knowledge base

.embed()

embedding model

.query()

retrieval

vector
stores

.rerank()

llm

output

Give agents knowledge of your data

RAG in Mastra helps you enhance LLM outputs by incorporating relevant context from your own data sources, improving accuracy and grounding responses in real information

Fully Flexible

Control how documents are chunked, decide the embedding strategy, and store vectors in the database you prefer. Use filters for precise retrieval, tailoring the pipeline to your needs.

Chunking and Embedding

Split documents and transform into vectors

Vector Database

Store in your preferred favorite vector DB provider

Retrieval

Retrieve using advanced metadata filtering

Advanced RAG techniques

Mastra allows context engineering with agentic RAG, ReAG, Graph RAG, and other techniques.

Reasoning-Augmented Generation

Mastra lets you go beyond RAG with ReAG, enabling models to reason directly over your documents for more accurate, context-aware answers

Advanced Context Engineering

Shape context your way with memory, history, and RAG, ensuring your agents deliver accurate results grounded in real information.

Ready to get started with Mastra?

Docs Mastra 101 Book Blog