pingcap

autoflow

AI#RAG#GraphRAG#LlamaIndex#TiDB#Next.js

2,758

// summary

AutoFlow is an open-source knowledge base tool that utilizes graph RAG technology built on TiDB Vector, LlamaIndex, and DSPy. The platform provides a Perplexity-style conversational search experience powered by an advanced built-in website crawler. Users can also integrate a customizable search widget into their own websites using a simple JavaScript snippet.

// technical analysis

AutoFlow is an open-source knowledge base tool designed to implement GraphRAG by leveraging TiDB Vector, LlamaIndex, and DSPy. It solves the challenge of building intelligent, context-aware search interfaces by integrating advanced web crawling with structured knowledge graph retrieval. The project prioritizes a modular architecture that combines robust database storage with modern frontend frameworks, though it remains in early development stages as it transitions toward a more accessible Python package ecosystem.

// key highlights

Implements GraphRAG technology to provide more accurate and context-aware knowledge retrieval compared to standard RAG.

Features a built-in website crawler that scrapes sitemaps to keep knowledge bases updated with official documentation.

Provides a Perplexity-style conversational search interface to enhance user interaction and information discovery.

Offers an embeddable JavaScript snippet that allows developers to add a conversational search widget to any website easily.

Utilizes TiDB as a unified storage solution for chat history, vector embeddings, JSON data, and analytics.

Integrates DSPy to enable programmatic control over foundation models, moving beyond traditional prompt engineering.

// use cases

Perplexity-style conversational search with automated sitemap URL scraping

Embeddable JavaScript widget for instant product-related query responses on external sites

Knowledge base management using graph RAG and TiDB for storing chat history and vector data

// getting started

To begin using AutoFlow, you can deploy the application using Docker Compose, which requires a system with at least 4 CPU cores and 8GB of RAM. Detailed deployment instructions are available in the project's official documentation. You can also explore the live demo at tidb.ai to see the conversational search capabilities in action.