Use the Next.js RAG template

The Next.js RAG template wires a streaming chat endpoint to a Postgres-backed vector store. It is the fastest path to a working RAG product on DeesseJS.

1. Scaffold

deessejs init nextjs-rag

2. Configure the embedding model

Set EMBEDDING_MODEL and EMBEDDING_DIMENSIONS in the env. Defaults target OpenAI's text-embedding-3-small at 1536 dimensions.

3. Index your corpus

pnpm rag:index ./content

The script chunks, embeds, and upserts into the documents table.

4. Stream a query

const answer = await rag.stream("What does ADR-013 say about topic tags?")
for await (const chunk of answer) {
  process.stdout.write(chunk.text)
}

What's next

• Build your first agent stack — combine RAG with tool use for an agent that can both retrieve and act.
• Manage Postgres migrations — version the documents table as the schema evolves.