Skip to content

Latest commit

 

History

History
221 lines (167 loc) · 7.78 KB

README-en.md

File metadata and controls

221 lines (167 loc) · 7.78 KB

Large Model Data Assistant

中文文档

🌟 Project Introduction

A lightweight, full - link supported, and easily customizable large model application project.

Compatible with large models such as DeepSeek and Qwen2.5

This is a one - stop large model application development project built on technologies like Dify, Ollama & Vllm, Sanic, and Text2SQL 📊. It features a modern UI crafted with Vue3, TypeScript, and Vite 5. It supports data graphical Q&A based on large models through ECharts 📈 and can handle tabular Q&A for CSV files 📂. Additionally, it can be easily integrated with third - party open - source RAG systems and retrieval systems 🌐 to support a wide range of general knowledge Q&A.

As a lightweight large model application development project, Sanic - Web 🛠️ supports rapid iteration and expansion, facilitating the quick implementation of large model projects. 🚀

Architecture Diagram

image

🎉 Features

  • Core Technology Stack: Dify + Ollama + RAG + (Qwen2.5/DeepSeek) + Text2SQL
  • UI Framework: Vue 3 + TypeScript + Vite 5
  • Data Q&A: Integrates ECharts and large models to achieve lightweight graphical data Q&A display via Text2SQL.
  • Table Q&A: Supports uploading CSV files and provides table data Q&A based on large model summarization, preprocessing, and Text2SQL.
  • General Q&A: Supports general data Q&A through integration with third - party RAG systems and public network retrieval.
  • Application Architecture: Serves as a lightweight, full - link, one - stop large model application development framework for easy expansion and implementation.
  • Flexible Deployment: Supports one - click deployment of all dependent components for large model application development using docker - compose, with zero configuration required.

Demo

image image image image

💡 Environment Requirements

Before you start, make sure your development environment meets the following minimum requirements:

  • Operating System: Windows 10/11, macOS M series, Centos/Ubuntu
  • GPU: For local deployment with Ollama, an NVIDIA GPU is recommended. Alternatively, you can use the CPU mode or purchase an API key from the public network.
  • Memory: 8GB+

🔧 Prerequisites

  • Python 3.8+
  • Poetry 1.8.3+
  • Dify 0.7.1+
  • Mysql 8.0+
  • Node.js 18.12.x+
  • Pnpm 9.x

📚 Large Model Deployment

⚙️ Dify Environment Configuration

  1. Install Dify
    • Official Documentation
    • To assist those new to large model applications, this project provides a one - click solution to start the Dify service for a quick experience.
    • Local access address for Dify: http://localhost:18000. You need to register your own account and password.
    # Start the built - in Dify service
cd docker/dify/docker
docker - compose up -d
  2. Configure Dify
  • Add the Ollama large model provider in Dify and configure the Qwen2.5 and DeepSeek R1 models.
  • Import the docker/dify/数据问答_v1.1.2_deepseek.yml canvas from the project root directory.
  • Copy the API key corresponding to the canvas for use in the following steps.
  • After importing the canvas, manually select the locally configured large model and save the settings.

image image image image

🚀 Quick Start

The specific steps are as follows:

  1. Clone the repository
git clone https://github.com/apconw/sanic - web.git
  1. Start the service
  • Modify the environment variables starting with DIFY_ in the chat - service of docker - compose.
  • Set the DIFY_DATABASE_QA_API_KEY to the API key obtained from the Dify canvas.
# Start the front - end, back - end services, and middleware
cd docker
docker compose up -d
  1. Configure Minio
  • Access the MinIO service at http://localhost:19001/. The default account is admin, and the password is 12345678.
  • Create a bucket named filedata and configure the Access Key.
  • Modify the environment variables starting with MINIO_ in the chat - service of docker - compose and restart the service.
# Restart the front - end, back - end services, and middleware
cd docker
docker compose up -d
  1. Initialize the data
# Install the dependency package
pip install pymysql

# For Mac or Linux users
cd docker
./init.sh

# For Windows users
cd common
python initialize_mysql.py
  1. Access the application

🛠️ Local Development

  • Clone the repository
  • Deploy large models: Refer to the above Large Model Deployment section to install Ollama and deploy the Qwen2.5 and DeepSeek R1 models.
  • Configure Dify: Refer to the above Dify Environment Configuration section to obtain the API key from the Dify canvas and modify the DIFY_DATABASE_QA_API_KEY in the .env.dev file.
  • Configure Minio: Modify the Minio - related keys in the .env.dev file.
  • Install dependencies and start the services
  1. Back - end dependencies
  • Install Poetry: [Refer to the official Poetry documentation](https://python - poetry.org/docs/)
# Install Poetry
pip install poetry

# Install dependencies at the root directory
# Set up the domestic mirror
poetry source add --priority = primary mirrors https://pypi.tuna.tsinghua.edu.cn/simple/
poetry install --no - root
  1. Install middleware
cd docker
docker compose up -d mysql minio
  1. Configure Minio
  • Access the MinIO service at http://localhost:19001/. The default account is admin, and the password is 12345678.
  • Create a bucket named filedata and configure the Access Key.
  • Modify the Minio - related keys in the .env.dev file.
  1. Initialize the data
  • If you are using a local MySQL environment, you need to modify the database connection information in the initialize_mysql source code.
# For Mac or Linux users
cd docker
./init.sh

# For Windows users
cd common
python initialize_mysql.py
  1. Front - end dependencies
# Install front - end dependencies and start the service
cd web

# Install pnpm globally
npm install -g pnpm

# Install dependencies
pnpm i

# Start the service
pnpm dev
  1. Start the back - end service
# Start the back - end service
python serv.py
  1. Access the service

🐳 Build Images

  • Execute the following commands to build the images:
# Build the front - end image
make web - build

# Build the back - end image
make server - build

🌹 Support

If you find this project useful, please give it a star on GitHub by clicking the Star button. Your support is our motivation to keep improving. Thank you! ^_^

⭐ Star History

Star History Chart

QA Community

  • Join our large model application community to discuss and share experiences.
  • Follow the official WeChat account and click the "WeChat Group" menu to join.
微信群
image

License

MIT License | Copyright © 2024 - PRESENT AiAdventurer