# multi-agent

**Repository Path**: qingyitianxia/multi-agent

## Basic Information

- **Project Name**: multi-agent
- **Description**: 通用agent平台
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-01-04
- **Last Updated**: 2026-01-04

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# NL2SQL System

This project demonstrates a Natural Language to SQL (NL2SQL) system built with LangGraph.

## Overview

The system translates natural language queries into SQL statements and executes them against a database. It uses a multi-agent approach with LangGraph to orchestrate the various steps involved in this process.

## Architecture

The system consists of several components organized as nodes in a LangGraph workflow:

1. **Intent Recognition**: Parses the natural language query to understand user intent
2. **Schema Recall**: Identifies relevant database tables based on the query
3. **SQL Generation**: Generates SQL based on the query and schema information
4. **SQL Validation**: Validates the generated SQL for safety and correctness
5. **SQL Execution**: Executes the SQL query against the database
6. **Result Analysis**: Analyzes the results and determines the best output format

## Setup

1. Install the required packages:
   ```
   pip install -r requirements.txt
   ```

2. Set up your API key in the `.env` file:
   
   For OpenAI:
   ```
   OPENAI_API_KEY=your_openai_api_key_here
   ```
   
   For SiliconFlow:
   ```
   LLM_PROVIDER=siliconflow
   SILICONFLOW_API_KEY=your_siliconflow_api_key_here
   SILICONFLOW_MODEL=Qwen/Qwen2.5-7B-Instruct
   SILICONFLOW_API_BASE=https://api.siliconflow.cn/v1
   ```

## Usage

To run the NL2SQL agent:

```python
from src.nl2sql_agent import create_nl2sql_workflow

# Create the workflow
app = create_nl2sql_workflow()

# Run with a query
result = app.invoke({
    "user_query": "Show me all employees",
    "retry_count": 0
})

print(result)
```

## Database

The system uses the Chinook database, a sample database that represents a digital media store. The database includes tables for tracks, albums, artists, customers, and more.

The database will be automatically downloaded when you run the system for the first time.

## Supported Databases

The system supports multiple database types through a modular database manager:

1. **SQLite** - Default Chinook database
2. **PostgreSQL** - Using psycopg2 driver
3. **MySQL** - Using PyMySQL driver

To add a new database, use the DatabaseManager class in `src/db_utils.py`:

```python
from src.db_utils import db_manager, DatabaseConfig

# Register a PostgreSQL database
config = DatabaseConfig(
    db_type="postgresql",
    uri="postgresql://user:password@localhost:5432/mydb",
    name="my_postgres_db"
)
db_manager.register_database("my_postgres_db", config)
```

## Supported LLM Providers

The system supports multiple LLM providers:

1. **OpenAI** - Default provider
2. **SiliconFlow** - Chinese LLM provider with OpenAI-compatible API

To switch providers, update the environment variables in `.env`.

## Security Features

- Only allows SELECT statements
- Blocks dangerous SQL keywords (DROP, DELETE, INSERT, etc.)
- Uses parameterized queries where possible