eosphoros-ai/DB-GPT

What it solves

DB-GPT is an agentic AI data assistant designed to bridge the gap between natural language and complex data analysis. It solves the problem of needing specialized technical skills to query databases, analyze spreadsheets, or generate business reports by allowing users to interact with their data using plain English.

How it works

The system connects to various data sources—including relational databases, CSV/Excel files, and knowledge bases—and uses Large Language Models (LLMs) to autonomously plan tasks, write SQL queries, and execute Python code. To ensure security, it runs these operations in sandboxed environments. It can also utilize "skills" (packaged domain knowledge and workflows) to perform repeatable business analysis tasks and transform the results into visual charts, dashboards, and HTML reports.

Who it’s for

It is built for developers and enterprises looking to create AI-native data agents and workflows, as well as business users who need to perform complex data analysis and reporting without writing manual code or SQL.

Highlights

• Autonomous Data Interaction: Automatically generates and executes SQL and Python code to query and clean data.
• Multi-Source Integration: Works across structured databases and unstructured documents/knowledge bases.
• Sandboxed Execution: Runs code and tools in isolated environments for enhanced safety.
• Extensible Skill System: Allows packaging of domain-specific analysis methods into reusable skills.
• Broad Model Support: Compatible with a wide range of LLMs including DeepSeek, Qwen, Llama, and GLM.
• Text2SQL Finetuning: Provides support for fine-tuning models specifically for SQL generation.