ChatBI that remembers corrections and respects workspace boundaries

What it does

SQLBot is a conversational analytics tool—ChatBI—that connects to databases and translates plain-language questions into SQL, charts, and summaries. It uses large language models for understanding and RAG to anchor generated queries in your actual schema and business terms. The DataEase team ships it as a containerized web application with an admin console ready to go.

The interesting bit

Most Text-to-SQL demos stop at the query; SQLBot treats accuracy as a maintenance task. Operators can curate terminology libraries, edit SQL examples, and refine prompts so the model learns from corrections rather than repeating hallucinations. Add workspace-level isolation and fine-grained data permissions, and you get a system built for teams that want AI access without the chaos of a shared chat window.

Key highlights

• Plugs into a broad roster of LLMs—DeepSeek, Gemini, Kimi, OpenAI, and major Chinese cloud providers—through an OpenAI-compatible API layer.
• Embeds into existing stacks via web components, pop-ups, or MCP calls, with named support for n8n, Dify, MaxKB, and DataEase.
• Workspace-level resource isolation and configurable data permissions handle the unglamorous but critical work of multi-tenant safety.
• An iterative feedback mechanism: custom prompts, terminology configs, and maintained SQL examples let the system tighten its output based on real user interactions.

Caveats

• The license is essentially GPLv3 but forbids removing or altering the SQLBot logo and copyright, and requires derivatives to remain open source; commercial use needs explicit permission.
• The README is primarily in Chinese, with English documentation available but not as prominent.
• Specific supported database engines or data source types are not listed in the README.

Verdict

A solid candidate if you need a self-hosted, embeddable natural-language layer for analytics with tenant isolation. Look elsewhere if you want a fully unbranded fork or if you need clear documentation on supported database engines before installing.