Skip to content
FlowRAG

FlowRAG

Batch indexing, stateless queries, Lambda-friendly. No servers, no Python, no complexity.

Knowledge Graph First

Combines vector search with entity relationships for richer, more connected results. Trace data flows across your entire system.

Lambda-Friendly

Works as a library, not a server. Stateless queries with fast cold starts. Perfect for serverless architectures.

Git-Friendly Storage

File-based storage you can commit to your repo. JSON, SQLite, and LanceDB or sqlite-vec — no external databases required.

Pluggable Everything

Swap storage backends, embedders, extractors, rerankers, parsers, and evaluators. Local ONNX, Gemini, OpenAI, or AWS Bedrock — your choice.

Incremental Indexing

Only re-process changed documents. SHA-256 content hashing skips unchanged files automatically. Delete documents with automatic KG cleanup.

Observable & Evaluable

Built-in hooks for LLM calls, embeddings, and search. Plug in evaluators for precision, recall, and faithfulness metrics.

Install

One command for a complete local setup with presets:

Terminal window
npm install @flowrag/pipeline @flowrag/presets

Index and search

Define a schema, index your files, and query:

import { defineSchema } from '@flowrag/core';
import { createFlowRAG } from '@flowrag/pipeline';
import { createLocalStorage } from '@flowrag/presets';
const rag = createFlowRAG({
schema: defineSchema({
entityTypes: ['SERVICE', 'DATABASE', 'PROTOCOL'],
relationTypes: ['USES', 'PRODUCES', 'CONSUMES'],
}),
...createLocalStorage('./data'),
});
await rag.index('./content');
const results = await rag.search('how does authentication work');