DataAI (also referenced as DB-AI in repos and releases) is a Windows desktop database tool for developers, analysts, and DBAs. Connect locally to MySQL, PostgreSQL, Oracle, SQL Server, SQLite, MariaDB, and more; run queries and schema work using natural language or SQL; export results. It ships with an AI agent (LangGraph) and long-term memory so the more you use it on the same database, the better SQL generation matches your tables and habits.
Get it
- Gitee repository: https://gitee.com/CodeYG/db-ai-pro
- Releases (installers & notes): https://gitee.com/CodeYG/db-ai-pro/releases
Stars and Issues are welcome.
Why try it?
1. Natural language → runnable SQL, grounded in your schema
- Describe what you need in plain language; the agent picks tables, loads DDL, and can sample enum-like values before emitting SQL.
- Dialect-aware output for MySQL, PostgreSQL, Oracle, etc.
- If the editor already contains SQL, short follow-ups (e.g. "only active rows", "order by time desc") build on that context instead of guessing table names from scratch.
2. AI-assisted CREATE / ALTER aligned with your existing style
-
Multi-turn clarification for
CREATE TABLE/ALTER TABLE. - Uses existing tables as reference for naming, types, keys, and indexes—handy when extending a brownfield system.
3. Long-term memory
- Per connection + database: naming patterns, similar past queries, table-level summaries (compressed usage patterns), and more.
- Later prompts include "how tables are typically used" so you repeat business context less often.
- "AI memory" UI to inspect, clear, or regenerate table summaries manually.
4. A real client, not only a chat box
- SQL editor with highlighting, formatting, and schema-based completion; multiple statements and result tabs.
- Result grids: sort, filter, row edits with UPDATE generation, export to CSV / Excel / JSON (per current release).
-
Connection manager with encrypted storage; Navicat import (Windows registry /
.ncx) to reduce migration friction. - Table sync, Redis-related features, etc.—see project README and release notes.
5. Local-first, flexible models
- Runs locally; credentials and memory stay on your machine (AI calls still depend on your chosen API or local LLM).
- OpenAI-compatible endpoints—cloud APIs or Ollama and similar.
Who it's for
| Role | Typical use |
|---|---|
| Backend / full-stack | Daily querying, debugging, drafting complex JOINs/aggregates |
| Data analysts | Explore schemas without memorizing SQL dialects |
| DBAs / ops | Multi-DB management, structure review, scripting workflows |
| Tech leads | Balance productivity with data-boundary policies |
Stack (short)
Python 3.8+, PyQt6, SQLAlchemy, LangGraph query agent, OpenAI-compatible APIs. See docs/ in the repo for architecture notes.
Closing
If you want less boilerplate SQL and less mental load per table, but prefer not to upload your schema to a random web-only tool, DataAI combines a proper client with LLM assistance on your own machine.
Repo: https://gitee.com/CodeYG/db-ai-pro
Download: https://gitee.com/CodeYG/db-ai-pro/releases
When reposting, please keep the project links. Feedback from real workflows helps the project evolve.
Top comments (0)