This project provides a lightweight, full-stack framework for developing large model applications, targeting developers and researchers looking for a rapid deployment solution. It integrates various LLM technologies, data visualization, and RAG capabilities for versatile data interaction and general knowledge Q&A.

How It Works

The framework is built upon Dify for orchestration, Ollama/vLLM for local model serving (supporting models like Qwen2.5 and DeepSeek), and Sanic for the backend API. It features a Vue3/TypeScript/Vite frontend for a modern UI. Key functionalities include Text2SQL for data visualization via ECharts and direct CSV file analysis, alongside RAG integration for broader knowledge retrieval.

Quick Start & Requirements

• Installation: Primarily uses docker-compose up -d for a full stack with Dify, Ollama, and the application services.
• Prerequisites: Python 3.8+, Poetry 1.8.3+, Dify 0.7.1+, MySQL 8.0+, Node.js 18.12.x+, pnpm 9.x. Local LLM deployment via Ollama is recommended, or public API keys can be used.
• Setup: One-click deployment for Dify is provided. Full setup involves cloning, deploying models, configuring Dify, and running Docker Compose.
• Links: Dify Official Docs, Ollama

Highlighted Details

• Supports data visualization through LLM-powered Text2SQL with ECharts.
• Enables tabular data Q&A via CSV file uploads and summarization.
• Facilitates general knowledge Q&A by integrating with third-party RAG systems.
• Offers a flexible, full-stack architecture for easy extension and deployment.

Maintenance & Community

The project is maintained by apconw. Community interaction is encouraged via WeChat groups for technical support and discussion.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

The project requires significant setup with multiple dependencies (Dify, MySQL, Ollama, Node.js). While offering one-click deployment for Dify, other components require manual configuration and deployment. The README indicates potential for paid one-on-one technical support, suggesting limited free support bandwidth.