DB-GPT

DB-GPT is a self-hosted AI data assistant for teams analyzing business data and developers building data applications. It connects plain-language questions to database queries and code-based analysis.

• SQL and Python analysis: Agents plan tasks, query databases, clean datasets, and calculate metrics through generated SQL and Python code.
• Mixed data sources: Work with relational databases, warehouses, CSV and Excel files, documents, and knowledge bases in the same analysis workflow.
• Visual reports: Generate charts, dashboards, HTML reports, and summaries for tasks such as database profiling or financial analysis.
• Reusable workflows: Package domain knowledge and analysis methods into skills, or build custom data assistants with agents, document retrieval, and AWEL workflow orchestration.
• Sandboxed execution: Run generated code and tools in isolated environments.

DB-GPT is open source under the MIT license. It runs on macOS and Linux and supports privately deployed models, including DeepSeek, Qwen, Llama, and Gemma families. Cloud connections include Moonshot and OpenAI-compatible APIs; requests using those models go to the chosen provider. Text2SQL fine-tuning supports model families including Qwen and LLaMA.