DOCSEARCHbyAlgolia
Submit Docs

DB-GPT

repository·main·Indexed 12 days ago

https://github.com/eosphoros-ai/db-gpt

An open-source agentic AI data assistant that automates data analysis by connecting to various data sources, writing SQL/Python code, and executing tasks in sandboxed environments. It features a middleware system for custom agent logic, a skill management system using SKILL.md, and an accelerator module (dbgpt-acc-auto) for cross-platform model deployment. Version 0.8.1 supports Text2SQL finetuning and integrates with LLM profiles such as OpenAI, Kimi, and MiniMax.

Tokens
265.3K
Snippets
693
Records
1.1K
Agent score
92%

What's inside DB-GPT

  1. What is DB-GPT?

    main

    DB-GPT is an open-source AI-native data application development framework featuring AWEL (Agentic Workflow Expression Language) and Agents. It is designed to build infrastructure for Large Language Model (LLM) applications involving data, enabling developers to create customized applications with minimal code.

    Core technical capabilities include:

    • RAG (Retrieval Augmented Generation): A framework for building knowledge-based applications.
    • GBI (Generative Business Intelligence): Provides fundamental data intelligence for enterprise report analysis and business insights.
    • Fine-tuning Framework: A complete framework for fine-tuning models on vertical and key domains (e.g., achieving 82.5% accuracy on the Spider dataset).
    • Data-driven Multi-Agents Framework: A self-evolving framework that continuously executes tasks based on data results.
    • Data Factory: Tools for cleaning and processing reliable knowledge and data.
    • Multi-source Data Access: Integration with various data sources including databases, spreadsheets, documents, and knowledge bases.

  2. Overview of DB-GPT Application Service Layer APIs

    main

    The Application Service Layer API is exposed on port 5670 after starting the DB-GPT webserver. These APIs focus on high-level application logic and are categorized into several functional areas.

    Note: Stable APIs use versioned paths like /api/v1/ or /api/v2/. Because the project is evolving rapidly, newer versions may not be fully backward compatible.

    For full interactive documentation and detailed parameter descriptions, navigate to http://127.0.0.1:5670/docs after the server is running.

  3. Overview of app Configuration classes

    main

    The app configuration type in DB-GPT is composed of several specialized configuration classes, each tailored to a specific interaction mode or functional capability. When configuring the application, you will use one of the following classes depending on your use case:

    • ChatDashboardConfig: Used for configuring the Chat Dashboard.
    • ChatExcelConfig: Used for configuring Chat Excel capabilities.
    • ChatKnowledgeConfig: Used for configuring Chat Knowledge (RAG/Knowledge Base) interactions.
    • ChatNormalConfig: Used for standard/normal chat configurations.
    • ChatWithDBExecuteConfig: Used for configurations involving direct database execution via chat.
    • ChatWithDBQAConfig: Used for configurations involving database Question & Answering (QA) via chat.

  5. Overview of DB-GPT

    main

    DB-GPT is an open-source agentic AI data assistant framework designed for AI-native data application development. It enables developers to build applications using LLM-based agents and specialized workflows for data tasks.

    Key capabilities include:

    • Agentic Data Analysis: Using AI agents to perform complex data reasoning.
    • Autonomous SQL + Code Execution: Automatically generating and executing SQL queries or Python code to interact with data.
    • Multi-source Data Access: Connecting to and querying various data sources.
    • Skills-driven Extensibility: Extending the system's capabilities through modular skills.
    • Sandboxed Execution: Running generated code in a secure, isolated environment.

  6. Overview of the dbgpt CLI

    main

    The dbgpt command-line interface provides tools for model deployment, service management, knowledge base operations, and debugging.

    Core commands include:

    • install: Install dependencies and plugins.
    • knowledge: Manage knowledge bases (viewing, deleting, loading documents).
    • model: Manage model serving and instances.
    • start/stop: Manage specific server instances.
    • trace: Analyze and visualize trace spans.
    ~ dbgpt --help
Already connect 'dbgpt'
Usage: dbgpt [OPTIONS] COMMAND [ARGS]...

Options:
  --log-level TEXT  Log level
  --version         Show the version and exit.
  --help            Show this message and exit.

Commands:
  install    Install dependencies, plugins, etc.
  knowledge  Knowledge command line tool
  model      Clients that manage model serving
  start      Start specific server.
  stop       Stop specific server.
  trace      Analyze and visualize trace spans.

  7. What is DB-GPT?

    main

    DB-GPT is an open-source AI-native data application development framework designed for interacting with databases using private LLM technology. It utilizes AWEL (Agent Workflow Expression Language) and specialized agents to simplify the development of large-scale model applications using data.

    Key capabilities include:

    • RAG (Retrieval Augmented Generation): Framework for building knowledge-based applications.
    • GBI (Generative Business Intelligence): Core technology for report analysis and business insight generation.
    • Fine-tuning Framework: Seamless integration for vertical/niche domain model optimization (e.g., achieving high Text2SQL accuracy on the Spider dataset).
    • Data-driven Multi-agent Framework: Self-evolving agents that make decisions and execute tasks based on data.
    • Data Factory: Tools for cleaning and processing reliable knowledge and data.
    • Multi-source Data Access: Connects to structured and unstructured data including databases, spreadsheets, documents, and knowledge bases.

  8. Overview of DB-GPT-Web

    main

    DB-GPT-Web is an open-source Chat UI built with Tailwind CSS and Next.js. It serves as a frontend for DB-GPT and provides a LLM to Vision solution.

    Key features include:

    • Enhanced Markdown Rendering: Beautifies standard markdown elements like table, thead, th, td, code, h1, h2, ul, li, a, and img.
    • AI-Specific Custom Labels: Supports specialized UI elements for AI workflows, such as plugin running, knowledge name, and Chart view.
    • Tech Stack: Uses ant-design for UI components, next.js for server-side rendering, and @antv/g2 for chart rendering.

  9. Overview of DB-GPT Configuration Types

    main

    DB-GPT uses a structured configuration system organized into several functional types. Developers can configure various components of the agentic data assistant, including data sources, LLMs, embedding models, and vector stores.

    Key configuration categories include:

    • app: Application-specific configurations (e.g., ChatNormalConfig, ChatWithDBQAConfig).
    • datasource: Connection parameters for various databases (e.g., MySQLParameters, PostgreSQLParameters, DuckDbConnectorParameters).
    • llm: Large Language Model deployment parameters (e.g., OpenAICompatibleDeployModelParameters, OllamaDeployModelParameters, VLLMDeployModelParameters).
    • embedding: Models used for generating embeddings (e.g., OllamaEmbeddingDeployModelParameters, HFEmbeddingDeployModelParameters).
    • vector_store: Configurations for vector databases (e.g., ChromaVectorConfig, PGVectorConfig, QdrantVectorConfig).
    • graph_store: Configurations for graph databases (e.g., Neo4jStoreConfig, TuGraphStoreConfig).
    • reranker: Parameters for reranking embedding results.
    • memory: Configuration for agent memory management (e.g., TokenBufferGPTsAppMemoryConfig).
    • serve: Settings related to the serving layer.
    • utils: Utility configurations like S3StorageConfig or OSSStorageConfig.

  11. What is TuGraph?

    main

    TuGraph is a graph database technology suite co-developed by Ant Group and Tsinghua University. It provides a complete graph technology system including graph storage, graph computing, graph learning, and graph development platforms. It is designed for real-time processing of massive multi-source relational data.

    TuGraph Product Family:

    • Open Source Products: TuGraph DB, TuGraph Analytics, OSGraph, and ChatTuGraph.
    • Internal Products: GeaBase, GeaFlow, GeaLearn, and GeaMaker.

  12. Overview of DB-GPT capabilities

    main

    DB-GPT is an open-source agentic AI data assistant designed for AI + Data products. It provides a platform for building AI-native data agents, workflows, and applications using agents, AWEL, RAG, and multi-model support.

    Core Capabilities:

    • Agentic Analysis: Task planning, step-by-step execution, tool use, and iterative reasoning.
    • SQL + Code Execution: Natural language to SQL, Python-based analysis, metric calculation, and chart generation.
    • Multi-Source Data Access: Support for relational databases, CSV/Excel, documents, and knowledge bases.
    • Skills and Agents: Reusable skills, domain workflows, agent orchestration, and customizable execution flows.
    • Reporting: Generation of database profiling reports, financial analysis reports, dashboards, and summaries.
    • Safe Execution: Sandboxed code execution and controlled tool use for reliable and reproducible outputs.