pingcap/autoflow

autoflow is an open-source Knowledge Graph project built on top of TiDB Vector, LlamaIndex, and DSPy. The project, developed by Pingcap, offers an advanced conversational search experience through its comprehensive website crawler and perplexity-style search page. Users can edit the Knowledge Graph to ensure accurate and up-to-date information. The platform also includes an embeddable JavaScript snippet for seamless integration into websites.

The tech stack includes TiDB as the database for storing chat history, vector, JSON, and analytic data; LlamaIndex as the RAG (Recommender, Analyzer, and Graph) framework; DSPy for programming foundation models; Next.js as the framework; and shadcn/ui for design.

To deploy autoflow, users can follow the instructions provided in the documentation. The project is open-source under the Apache License, Version 2.0. For more information, visit the project’s live demo at TiDB.AI or the documentation page. Contributions from the community are welcome, and guidelines can be found in the CONTRIBUTING.md file.

The project’s performance statistics can be viewed on OSS Insight, and it is available on Docker Hub with versions for both the backend and frontend. The platform includes features such as an advanced built-in website crawler, conversational search, and an embeddable JavaScript snippet for easy integration into websites. Users can reach out to the team on Twitter for any inquiries.