Introduction

Generate precise SQL queries through simple natural language descriptions.

Support Status

• Interface languages: Chinese, English (more to come)
• Supported databases: Any type
  • For query import scripts, currently supports: SQLite, MySQL, PostgreSQL, SQLServer
• Supported vector databases: Chroma (more to come)
• Supported LLMs: OpenAI (more to come)
• Supported translation (optional): Azure Translator (more to come)

SQL Generation Process

Similar to a workflow concept, executed step by step:

1. Match business documents
2. Match generation records
3. AI generates potentially relevant fields
4. Match most similar tables and fields based on AI-generated fields
5. AI generates SQL
6. Learning: Learn from results to improve table comments, field comments, and field relationships

Features

• Progressive improvement, becomes more accurate with use
• Convenient and easy to understand
• Rich and refined functionality

Usage

Project Structure

• backend: Backend project, using Python and Flask framework
• frontend: Frontend project, using React and Tailwind CSS framework

Dependencies

Backend project depends on the following:

• Database: Stores application data. Default configuration uses SQLite (created in backend/instance directory). Can be modified to connect to other databases. Tables are automatically created on first startup
• Vector database: Stores application data. Can start a Chroma docker container and modify configuration to connect to it
• LLM: Uses OpenAI by default. Can modify apikey and other configurations
• Translation: Uses Azure Translator by default. Can modify configuration. If you're using English, this is optional. However, it's highly recommended for other languages as it greatly improves vector database matching accuracy

Backend Deployment

• Install dependencies: pip install -r requirements.txt
• Local development: Enter backend directory, run python app.py
• Production deployment: Deploy using gunicorn, refer to ./start.sh
• Docker deployment:
  • Refer to ./docker_build.sh to build Docker image
  • Refer to ./docker_run.sh to run Docker image

Frontend Deployment

• Local development: Enter frontend directory, run npm run dev for development debugging, or npm run start to fetch backend API and regenerate API definitions before starting debug
• Production deployment: Use vite to package, run npm run build

Development URLs

In local development mode, frontend URL: http://localhost:5173

Screenshots

Projects page

Project page

DDL page

Document page

Rule page

Settings page