Home
Login

An AI-native data application development framework built on AWEL and multi-agent systems for data intelligence applications.

MITPython 16.8keosphoros-ai Last Updated: 2025-06-20

DB-GPT: AI-Native Data Application Development Framework

Project Overview

DB-GPT is an open-source AI-native data application development framework that integrates AWEL (Agentic Workflow Expression Language) and a multi-Agent system. This project aims to build infrastructure in the large language model (LLM) field by developing various technical capabilities, such as multi-model management (SMMF), Text2SQL effect optimization, RAG framework and optimization, multi-Agent framework collaboration, and AWEL (agent workflow orchestration), making data-driven LLM applications simpler and more convenient.

In the Data 3.0 era, based on models and databases, enterprises and developers can build their own customized applications with less code.

Core Architecture and Capabilities

Main Functional Modules

1. RAG (Retrieval-Augmented Generation)

RAG is currently the most practical and urgently needed area. DB-GPT has implemented a RAG-based framework, allowing users to build knowledge base applications using DB-GPT's RAG capabilities.

2. GBI (Generative Business Intelligence)

Generative BI is one of the core capabilities of the DB-GPT project, providing foundational data intelligence technology for building enterprise report analysis and business insights.

3. Fine-tuning Framework

Model fine-tuning is an indispensable capability for any enterprise to implement in vertical and segmented fields. DB-GPT provides a complete fine-tuning framework that seamlessly integrates with the DB-GPT project. In recent fine-tuning work, an accuracy rate of 82.5% was achieved based on the Spider dataset.

4. Data-Driven Multi-Agent Framework

DB-GPT provides a data-driven self-evolving multi-Agent framework designed to continuously make decisions and execute based on data.

5. Data Factory

The data factory is primarily responsible for cleaning and processing trusted knowledge and data in the era of large language models.

6. Data Source Integration

Integrate various data sources to seamlessly connect production business data to the core functions of DB-GPT.

Related Projects

DB-GPT-Hub

DB-GPT-Hub focuses on achieving high-performance Text-to-SQL workflows by applying supervised fine-tuning (SFT) on large language models (LLMs).

dbgpts

dbgpts is the official repository, containing data applications, AWEL operators, AWEL workflow templates, and agents built on DB-GPT.

DB-GPT-Plugins

DB-GPT plugins, which can directly run Auto-GPT plugins.

Supported Language Models

DB-GPT supports a wide range of large language models, including:

  • Open Source Models:

    • LLaMA / LLaMA-2 / LLaMA-3 / LLaMA-3.1
    • BLOOM / BLOOMZ
    • Falcon
    • Baichuan / Baichuan2
    • InternLM
    • Qwen series (Qwen2.5, Qwen3, etc.)
    • XVERSE
    • ChatGLM2 / GLM-4
    • DeepSeek series
    • Yi series
    • Gemma series
    • Phi-3
    • CodeQwen
    • Mixtral
    • SOLAR
  • API Models:

    • ERNIE Bot
    • Tongyi Qianwen
    • Zhipu AI
    • And other API services

Key Features

1. Private Domain Q&A and Data Processing

The DB-GPT project provides a series of features designed to improve knowledge base construction and enable efficient storage and retrieval of structured and unstructured data. These features include:

  • Built-in support for uploading multiple file formats
  • Ability to integrate custom data extraction plugins
  • Unified vector storage and retrieval capabilities

2. Multi-Data Source and GBI

The project facilitates seamless natural language interaction with diverse data sources, including Excel, databases, and data warehouses. It simplifies the process of querying and retrieving information from these sources, enabling users to conduct intuitive conversations and gain insights. Additionally, DB-GPT supports the generation of analytical reports.

3. Multi-Agent and Plugins

It provides support for custom plugins to perform various tasks and natively integrates the Auto-GPT plugin model. The Agent protocol follows the Agent Protocol standard.

4. Automated Text2SQL Fine-tuning

We have also developed an automated fine-tuning lightweight framework centered around large language models (LLMs), Text2SQL datasets, LoRA/QLoRA/Pturning, and other fine-tuning methods. This framework simplifies Text-to-SQL fine-tuning, making it as simple as an assembly line process.

5. SMMF (Service-Oriented Multi-Model Management Framework)

We provide extensive model support, including dozens of large language models (LLMs) from open-source and API agents, such as LLaMA/LLaMA2, Baichuan, ChatGLM, ERNIE Bot, Tongyi, Zhipu, etc.

Privacy and Security

We ensure data privacy and security by implementing various technologies, including private large models and agent desensitization.

Supported Data Sources

In the .env configuration file, modify the LANGUAGE parameter to switch to different languages. The default is English (Chinese: zh, English: en, other languages will be added later).

Technical Architecture

DB-GPT adopts a modular architecture design, mainly including:

  • AWEL Workflow Orchestration Layer: Provides the expression and orchestration capabilities of agent workflows
  • Multi-Model Management Layer: Unifies the management and scheduling of different large language models
  • Data Access Layer: Supports the access and processing of multiple data sources
  • Agent Collaboration Layer: Enables collaboration between multiple AI agents
  • Application Service Layer: Provides application services for end-users