Autoflow

Autoflow is an open source conversational knowledge base tool from PingCAP that answers questions over a body of documentation using graph based retrieval augmented generation. It builds a knowledge graph alongside vector embeddings stored in TiDB, then combines the two when retrieving context, so answers draw on how concepts relate rather than only on text similarity. The listed site, tidb.ai, is the live demo: an assistant that answers questions about the TiDB database. The tool fits teams that want a documentation assistant for their own product and are comfortable deploying it themselves. It is built on TiDB vector storage, LlamaIndex and DSPy. Features include a conversational search page in the style of Perplexity, a built in crawler that ingests official and documentation sites through sitemap URLs, and an embeddable JavaScript snippet that places the chat window on your own website, typically in the bottom right corner. Deployment uses Docker Compose, and the project recommends at least 4 CPU cores and 8GB of RAM. Autoflow is free under the Apache 2.0 license. There is no paid plan for the tool itself; running it requires a TiDB database, which can be TiDB Cloud (priced separately by PingCAP), plus the model providers you configure. The repository warns that Autoflow is still in early development, and PingCAP states the next step is to package it as an installable Python library for building RAG solutions.