GenesisAI

Genesis AI Hackathon

Architecture

Backend Structure

The application uses a FastAPI backend server that:

1. Connects to the Supabase database
2. Provides RESTful endpoints for accessing news articles
3. Handles analysis and processing of queries

API Modules

The API modules (e.g., natural_language_understanding.py) now communicate with the FastAPI backend via HTTP requests:

• They do not directly access the database
• They use the FastAPI backend as a gateway to the database
• This ensures a clean separation of concerns

Architecture Changes

Previous Implementation:

• API modules directly accessed the Supabase database through a Database class.
• This created tight coupling between the API modules and the database.

New Implementation:

• API modules now communicate with the FastAPI backend server via HTTP requests.
• The FastAPI backend acts as a gateway to the database.
• This creates a clean separation of concerns and follows RESTful principles.

Benefits:

1. Separation of concerns - API modules don't need to know database details
2. Single source of truth - Only the FastAPI backend connects to the database
3. Testability - API modules can be tested with mock responses
4. Scalability - FastAPI backend can be scaled independently

Added Functionality:

• Robust error handling when the backend is not available
• Fallback to mock data when the database can't provide real articles
• Detailed error logging to help diagnose issues

Setup

1. Start the FastAPI backend server:

python backend/main.py

2. The API modules will automatically connect to the backend server running at the URL specified in the BACKEND_URL environment variable (defaults to http://localhost:8000)

3. To test the Natural Language Understanding module:

python test_nlu_api.py

Environment Variables

Set these in your .env file:

• SUPABASE_URL: Your Supabase project URL
• SUPABASE_KEY: Your Supabase API key
• COHERE_API_KEY: Your Cohere API key
• BACKEND_URL: URL of the FastAPI backend (default: http://localhost:8000)

Testing

To test this implementation, make sure the FastAPI backend is running:

cd backend
python main.py

Then run the test script in another terminal:

python test_nlu_api.py

Database Search Implementation

The application now uses a dedicated approach to search articles in the database:

1. Table Structure:

   • The Supabase database contains a table named "news"
   • Key columns: id, source_name, source_link, news_information

2. Search Functionality:

   • A new /search endpoint that specifically searches the news_information column
   • Enhanced keyword matching in the NLU module
   • Fallback mechanisms when search fails

3. Improvements:

   • Direct searching in the news_information column for better relevance
   • Multiple search strategies with progressive fallbacks
   • Robust error handling and logging
   • Mock data fallback for testing when the database is unavailable

Testing The Implementation

To test the search functionality:

1. Start the FastAPI backend:

   cd backend
   python main.py

2. Use the test script to search for specific terms:

   python test_search.py "search term"

3. Or use the full NLU pipeline:

   python test_nlu_api.py