# Aix-DB

A conversational data-analysis system that turns natural-language questions into validated SQL, runs them against a database, and renders the results as charts, using retrieval-augmented generation and a multi-agent pipeline.

Aix-DB is a data assistant built on LangChain and LangGraph that converts natural-language questions into database queries and visualizations end to end. A question is parsed for intent, then a hybrid retrieval step combining embeddings, BM25 and a Neo4j knowledge graph pulls the relevant table structures and business context before a Text2SQL engine generates and syntax-checks the query.

The generated SQL runs against the target data source, which can be any of more than eight supported relational, vector or graph database types, and the results are rendered automatically as ECharts or AntV charts. The system also supports table-level question answering, deeper multi-step data investigation, and an MCP-based multi-agent and skill mode for extending its capabilities.

The frontend is built with Vue 3 and TypeScript, the API gateway with the asynchronous Sanic framework and JWT authentication, and the whole stack deploys with Docker Compose alongside PostgreSQL and MinIO.

Website: https://github.com/apconw/Aix-DB