TL;DR: Connect to MySQL, PostgreSQL, Oracle, SQL Server, SQLite, MariaDB, describe what you want in plain language, get runnable SQL. CREATE/ALTER with multi-turn chat, LangGraph agent, long-term memory per connection + database — so it gets smarter on your schema without uploading it to a random website.
If you do backend or data work, you have done this: copy table definitions into a browser tab, hope the model guesses the right dialect, then paste SQL back into DBeaver/Navicat. It works, but it is fragile and feels wrong for anything sensitive.
DataAI (also called DB-AI in releases) is a PyQt6 desktop app that keeps credentials and memory local and still gives you tool-using agents (schema fetch, enum sampling, dialect-aware generation). Below is why it is worth a download, not just a skim.
Three reasons to click through to the releases page
- NL → SQL that knows your catalog — The agent picks tables, loads DDL, can sample values, outputs syntax for your engine.
- Schema changes that match existing style — Multi-turn CREATE/ALTER with reference to tables you already have (naming, types, keys).
- Memory that is not a one-off chat — Naming patterns, similar queries, compressed table-level summaries; optional AI memory UI to inspect or regenerate.
Plus: real SQL editor (highlighting, completion), multi-tab results, sort/filter/export, Navicat import (Windows registry / .ncx). OpenAI-compatible APIs or local LLMs (e.g. Ollama).
Who it is for
| Role | Use case |
|---|---|
| Backend / full-stack | Daily queries, debugging, draft heavy JOINs |
| Analysts | Explore data without memorizing dialects |
| DBAs | Multi-DB ops, structure review |
Get it (this post is the ad — the binary is the proof)
- Repo: https://gitee.com/CodeYG/db-ai-pro
- Releases / installers: https://gitee.com/CodeYG/db-ai-pro/releases
Star the repo if it saves you time; open an Issue if something breaks. Share with anyone still doing the DDL-paste dance.
DataAI is an open-source project; links above are canonical.
Top comments (0)