DB-GPT is an open-source agentic AI data assistant maintained by eosphoros-ai, positioned as an open-source tool for next-generation AI + data products. Its repository address is https://github.com/eosphoros-ai/DB-GPT。
In terms of purpose, DB-GPT is designed to help users and teams connect to databases, CSV / Excel files, data warehouses, and knowledge bases; ask questions in natural language, with AI autonomously writing SQL; run Python-based and code-driven analysis workflows; load and execute reusable skills to complete domain-specific tasks; generate charts, dashboards, HTML reports, and analysis summaries; and safely execute tasks in a sandbox environment. It is also a platform that can be used to build AI-native data agents, workflows, and applications, powered by agents, AWEL, RAG, and multi-model support.
Main capabilities include:
- Agentic data analysis: plan tasks, break work into steps, call tools, and complete analysis workflows end to end.
- Autonomous SQL + code execution: generate SQL and code to query data, clean datasets, compute metrics, and produce outputs.
- Multi-source data access: work across structured and unstructured sources, including databases, spreadsheets, documents, and knowledge bases.
- Skill-driven extensibility: package domain knowledge, analysis methods, and execution workflows into reusable skills.
- Sandboxed execution: run code and tools in an isolated environment, making analysis safer and more reliable.
In terms of usage, the official one-line installer is available for macOS and Linux. You can directly specify the profile and API key: for example, use Kimi 2.5 via the Moonshot API, or use MiniMax via an OpenAI-compatible API. If you already have a local DB-GPT checkout directory, you can reuse it instead of cloning it again to ~/.dbgpt/DB-GPT. After installation is complete, start the server using the generated profile configuration, then open http://localhost:5670 . The official option also supports reviewing the installation script first.
It can also be installed via PyPI: Python 3.10+ is required, as well as uv (recommended) or pip. The default installation includes the core framework (CLI, FastAPI, Agent), OpenAI-compatible LLM support, DashScope / Tongyi support, RAG document parsing, and ChromaDB vector storage. On first run, an interactive setup wizard starts, guiding you to select an LLM provider and enter an API key; after completion, the Web server starts automatically, and you can access http://localhost:5670 That is all. For advanced installation methods such as Docker, local GPU models (vLLM, llama.cpp), or manual installation from source, see the full documentation.
Applicable scenarios include: analyzing CSV / Excel files and generating visualization reports; connecting to databases and generating profiling reports; asking business questions in natural language and letting AI automatically write SQL; performing financial report analysis and generating code, charts, and narrative summaries; creating and reusing SQL analysis skills and domain workflows; combining code, SQL, retrieval, and tools in a single agent workflow; building next-generation AI + data assistants for teams or products. The official site also provides application development guides, AWEL Flow usage, development guides, AWEL, SMMF fine-tuning, and other advanced usage documentation, as well as information related to Text2SQL fine-tuning.
Regarding privacy and security, the official project protects data privacy and execution security through private model deployment, proxy desensitization, and sandbox execution mechanisms. The project uses the MIT license.
Regarding limitations, the materials do not provide specific performance data, charging or pricing information, a complete list of supported models, platform compatibility details, or deployment resource requirements; this content is not yet available.
