autoflow

Why we included this project

Building a knowledge base that answers questions in plain language usually means wiring a crawler, a retrieval layer, and a chat front end together yourself. AutoFlow bundles all three: a website crawler pulls in sitemaps and documentation pages, a knowledge-graph-backed retrieval layer built on TiDB Serverless Vector Storage with LlamaIndex and DSPy handles the lookup, and a Perplexity-style conversational page sits on top. Because TiDB is MySQL compatible, teams already running that stack can add RAG without standing up a separate vector database. A small embeddable JavaScript snippet lets you put the assistant on an existing site instead of pushing users into a new app. The project is early, and the maintainers plan to release it as a pip-installable package, so it rewards engineers who like reading and adapting code over teams that need a polished product today.