|
--- |
|
title: Multi-Agentic SQL Generator |
|
sdk: gradio |
|
app_file: app.py |
|
license: apache-2.0 |
|
tags: |
|
- sql |
|
- openai |
|
- multi-agent |
|
- demo |
|
sdk_version: 5.19.0 |
|
--- |
|
|
|
# Multi-Agentic SQL Generator |
|
|
|
The Multi-Agentic SQL Generator is a modular, multi-step system that translates natural language queries into SQL queries, validates and optimizes them, and then executes them against a SQLite database. The system leverages OpenAI's language models (via a LangGraph workflow) for query understanding, validation, and optimization. It also provides evaluation capabilities using RAGAS metrics (such as context precision and context recall) to assess performance and quality. |
|
|
|
## Capabilities |
|
|
|
- **Natural Language Query Understanding:** |
|
Converts natural language queries into structured SQL metadata according to a predefined database schema. |
|
|
|
- **Query Validation:** |
|
Checks the generated SQL for syntax errors and security risks (e.g., harmful operations like `DROP`, `DELETE`). |
|
|
|
- **Query Optimization:** |
|
Optimizes SQL queries for performance, ensuring only the necessary columns, joins, and filtering conditions are included. |
|
|
|
- **SQL Execution:** |
|
Executes the optimized SQL query against a SQLite database and returns the results. |
|
|
|
- **Evaluation with RAGAS Metrics:** |
|
Evaluates the generated output using metrics like: |
|
- **Faithfulness:** (0.7500) Faithfulness measures how well the generated SQL execution results match the expected (ground truth) results. |
|
- **Answer Relevancy:** (0.2280) Answer relevancy measures how well the generated SQL results match the intent of the user's query. |
|
- **Context Precision:** (0.3000) Context Precision measures how much of the generated SQL execution result is actually relevant to the query. |
|
- **Context Recall:** (0.3500) Context Recall measures how much of the expected result was actually retrieved. |
|
|
|
- **Extensibility and Deployment:** |
|
Easily integrable with front-end frameworks (e.g., Chainlit) and deployable on platforms like Hugging Face Spaces. |
|
|
|
## Installation |
|
|
|
1. **Clone the Repository:** |
|
|
|
```bash |
|
git clone https://github.com/yourusername/multi-agent-sql-generator.git |
|
cd multi-agent-sql-generator |
