Turbopuffer-backed RAG for your documents
Upload documents, parse with Docling, embed with your provider, and use Turbopuffer for semantic, keyword, and hybrid retrieval through one API.
Features
Everything you need for RAG
Any Document Format
PDF, DOCX, PPTX, HTML, Markdown, images with OCR, and more — powered by Docling for universal document parsing.
Any Embedding Model
OpenAI, Cohere, Voyage, and OpenAI-compatible models with per-collection configuration. Mix providers across collections.
Turbopuffer Search
Turbopuffer powers semantic, keyword, and hybrid search from the same chunk store, with Reciprocal Rank Fusion for mixed queries.
Namespace Isolation
Each collection maps to a Turbopuffer namespace, keeping vector writes, keyword indexes, exports, and deletes scoped.
Webhooks
HMAC-signed webhook payloads with automatic retries for collection and connector events.
TypeScript SDK
Zero-dependency TypeScript client for Node.js, browsers, Deno, and Bun. Full type safety and automatic retries.
Simple integration
Upload, embed, and search in minutes
bigRAG handles the entire RAG pipeline. Upload any document format, and it automatically parses, chunks, embeds, and indexes in Turbopuffer for vector and full-text search. Use the TypeScript SDK or REST API.
import { BigRAG } from "@bigrag/client";
const client = new BigRAG({
apiKey: "bigrag_sk_...",
baseUrl: "http://localhost:4000",
});
const { results } = await client.queries.query("knowledge_base", {
query: "What is the PTO policy?",
top_k: 5,
});15+
Document formats
PDF, DOCX, images, and more
1
Search backend
Turbopuffer
12+
Embedding models
OpenAI, Cohere, Voyage
5 min
To deploy
Docker Compose
Deploy bigRAG with Turbopuffer search
Run the API, admin UI, Postgres, and Redis yourself, then connect Turbopuffer for managed vector and full-text retrieval.
$ git clone https://github.com/bigint/rag.computer
$ cd rag.computer
$ docker compose up -d