DB-GPT
Self-hosted AI data assistant that queries databases, analyzes files, and generates reports with local models or cloud APIs. Open source under MIT.
DB-GPT is a self-hosted AI data assistant for teams analyzing business data and developers building data applications. It turns plain-language requests into SQL queries and Python analysis, then produces charts, dashboards, or HTML reports. You can run it on macOS or Linux, with Docker deployment also supported.
Its agents can plan an analysis, break it into tasks, and execute code and tools in sandboxed environments. They can query relational databases and warehouses, clean CSV or Excel data, calculate metrics, and use documents or knowledge bases alongside structured data. That makes it useful for questions that need both database queries and supporting material, such as financial analysis or database profiling.
Model choice is flexible. DB-GPT supports local GPU models through vLLM and llama.cpp, including DeepSeek, Qwen, Llama, and Gemma families. It also connects to cloud providers through OpenAI-compatible APIs, Moonshot, and DashScope. The application runs on your own machine or server; choosing a cloud model sends model requests to that provider. Private model deployment is available for teams that want to keep inference under their control.
For repeatable work, reusable skills package domain knowledge and analysis methods into workflows. Developers can combine agents with document retrieval and AWEL workflow orchestration to build their own data assistants. DB-GPT is open source under the MIT license and supports Text2SQL fine-tuning for model families including Qwen and LLaMA.
