Text2SQL

Description

📔 Text ➡️ SQL 🧑‍💻

Tools

• LangChain SQL Q&A Tutorial 🦜🔗
• Vanna.AI 🔮
• LlamaIndex 🦙

Tutorials

• How to Use LangChain to Build a Text-to-SQL Solution
• Text2SQL GitHub Repository
• Text2SQL Workshop GitHub Repository

Papers

• A Survey on Employing Large Language Models for Text-to-SQL Tasks
• PET-SQL: A Prompt-enhanced Two-stage Text-to-SQL Framework with Cross-consistency
• SeaD: End-to-end Text-to-SQL Generation with Schema-aware Denoising
• Next-Generation Database Interfaces:A Survey of LLM-based Text-to-SQL

Running

1 - install requirements

pip install requirements.txt

2- navigate to /data and run sqlite-synthetic.py to create a toy dataset

cd data
python sqlite-synthetic.py

after this step you should see a synthetic_data.db in /src

3- navigate to src and run either main.py or the UI:

For command line interface:

cd ../src
# Use OpenAI (default)
python main.py

# Or use Google's Gemini model
python main.py -gemini

For the web interface:

cd ../src
# Run the Streamlit UI
python run_ui.py

Optional run visualize_workflows.py to show workflow graphs

API Keys

Create a keys.env file in the src/agents directory with your API keys:

OPENAI_API_KEY=your_openai_key_here
GOOGLE_API_KEY=your_gemini_key_here # Optional, only if using Gemini

Development

See TODO.md for planned features and improvements.

Implementaions

Langgraph Workflow 🦜

• Master Workflow

• Python Workflow

• SQL Workflow