# pingcap/autoflow

pingcap/autoflow is a Graph RAG based and conversational knowledge base tool built with TiDB Serverless Vector Storage. Demo: https://tidb.ai

Repository: https://github.com/pingcap/autoflow
Canonical: https://ross.abutalabs.com/products/autoflow
Homepage: https://tidb.ai
Language: TypeScript
License: Apache-2.0
License Family: permissive
Topics: mysql, rag, serverless, vector-database, chatbot, graphrag, knowledge-graph, cot
Last push: 2026-04-27T13:55:14+00:00

## Health v2 (maintenance only)
Score: 63/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 79, release rhythm 40, longevity 67
- inputs: {"age_days": 940, "days_push": 128, "days_rel": 607, "gap_med": 25, "n_releases_24m": 2}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2971, forks 195 (observed 2026-08-28T04:07:32.775650+00:00)

## What it is
AutoFlow is an open-source Graph RAG knowledge base tool that builds knowledge graphs over crawled documentation and answers questions via conversational search, built on TiDB Serverless Vector Storage, LlamaIndex, and DSPy. It ships as a deployable web application (Docker Compose) with an embeddable JavaScript chat widget, demonstrated at tidb.ai.

## Use cases
- build a conversational knowledge base over my product docs
- add a Perplexity-style search page to my documentation site
- embed a chat widget that answers questions from my website content
- crawl a sitemap and index docs into a knowledge graph for RAG
- self-host a Graph RAG question-answering assistant
- answer user questions with citations from a knowledge graph

## When to choose
- you want a self-hosted Graph RAG assistant over documentation or websites
- you already use TiDB or want TiDB Serverless vector storage
- you need an embeddable conversational search widget for your site

## When to avoid
- you need a production-hardened, stable RAG platform (project is early stage)
- you want a simple vector-only RAG without knowledge graph overhead
- you cannot run Docker Compose with ~4 CPU cores and 8GB RAM

## Facets
- artifact type: application
- maturity: experimental
- function: rag, search-engine, chatbot, web-scraping, vector-database, llm-inference
- domain: large-language-models, chatbots
- platform: self-hosted, python
- tags: graphrag, knowledge-base, tidb, llamaindex, dspy, conversational-search, embeddable-widget, retrieval-augmented-generation, search, knowledge-graphs, docker, web-server, typescript

## Member repositories
- pingcap/autoflow (main) score 63

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:32.775650+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:31:43.476913+00:00, confidence not recorded.
  - readme: https://github.com/pingcap/autoflow (fetched 2026-08-28T04:07:32.775650+00:00, sha bf93d353d839)
  - homepage: https://tidb.ai (fetched 2026-08-29T09:47:11.467928+00:00, sha 741792aca215)
  - site_page: https://tidb.ai/api-docs (fetched 2026-08-29T09:47:11.477491+00:00, sha bfb8182743f0)
- Data as of 2026-08-30T08:39:29.467469+00:00.