Airline Data Query Assistant 🛫

A natural language interface for querying airline/flight data using LLMs

Workflow Diagram

Features

• ✈️ Natural language to SQL conversion using LLMs
• 🔍 SQL query validation and execution
• 📊 Results conversion to human-readable format
• 🌐 Web interface for easy interaction
• ✅ Comprehensive testing suite
• 📁 Structured project layout with Poetry dependency management

Installation

Prerequisites

• Python 3.9+

Setup

1. Clone the repository

git clone https://github.com/gtmray/smart-air
cd smart-air

2. Install poetry

pip install poetry

3. Install dependencies

poetry install

4. Configure environment variables for LLM access and API configurations (in .env file)

HOSTNAME=localhost
PORT=<YOUR_AVAILABLE_PORT>

API_TYPE=<enter azure if using azure else comment this out>
OPENAI_API_VERSION=<YOUR_API_VERSION_IF_USING_AZURE_API_TYPE>

API_BASE_URL=<YOUR_API_BASE_URL>
API_KEY=<YOUR_API_KEY>
MODEL_NAME=<YOUR_MODEL_NAME>

5. Store raw files downloaded from Kaggle using KaggleHub to src/sqlite_db/raw_data

See: https://github.com/Kaggle/kagglehub or download from kaggle UI

6. Create database

poetry run python src/sqlite_db/create.py

7. Start the application

poetry run src/app/main.py

8. Access the web interface

http://localhost:<YOUR_ASSIGNED_PORT>

Future Improvements

• Integrate industry standard database
• Implement caching for frequently asked queries
• Improve UI/UX for better user experience
• Optimize query execution performance
• Consider LLM’s token limit and output result consider it
• Add support for real-time streaming data
• Explore multi-model orchestration for complex queries
• Expand testing coverage to include performance benchmarks