This project demonstrates a Retrieval-Augmented Generation (RAG) application using a Knowledge Graph to provide context-aware responses from a language model. The application is built using Python, NetworkX for the knowledge graph, and Streamlit for the web interface.
-
Clone the repository:
git clone https://github.com/GloryKO/RAG-TEST.git cd RAG-TEST
-
Create and activate a virtual environment:
python -m venv venv source venv/bin/activate # On Windows, use
venv\Scripts\activate -
Install dependencies:
pip install -r requirements.txt
-
Environment Variables: Create a
.envfile in the root directory of the project and add your OpenAI API key:OPENAI_API_KEY=your_openai_api_key
-
Project Structure:
├── main.py ├── knowledge_graph.py ├── llm_graph_transformer.py ├── requirements.txt ├── .env _.gitignore ├── README.md - screnshots
The project is structured into three main modules:
-
Knowledge Graph Module (
knowledge_graph.py):- Handles the creation and management of a knowledge graph using NetworkX.
- Adds data to the graph and retrieves context relevant to a query.
-
LLM Graph Transformer Module (
llm_graph_transformer.py):- Integrates with OpenAI's GPT-3.5-turbo model.
- Uses the context from the knowledge graph to generate context-aware responses.
-
Streamlit Application (
main.py):- Provides a user interface for interacting with the knowledge graph and the LLM.
- Takes user queries and displays the generated responses.
-
Start the Streamlit application:
streamlit run main.py
-
Access the application: Open your web browser and go to
http://localhost:8501.
- Enter your query: Type a question or a prompt into the text input field.
- Submit the query: Click the "Submit" button to send the query.
- View the response: The application will display a response generated by the language model using the context from the knowledge graph.
Here are some example queries you can try:
-
Basic Entity Queries:
- "Tell me about Inception."
- "Who directed Interstellar?"
- "Who starred in Mad Max: Fury Road?"
- "who directed the movie "New One"?" #this is testing with sample data not online but in the KG
- "Who starred in the Movie "New One" ? #also testing with sample data not availabe online but in the KG
-
Relation-Based Queries:
- "Who directed Inception?"
- "What genre is The Matrix?"
- "Who starred in Interstellar?"
-
Complex Queries:
- "Which movies were directed by Christopher Nolan?"
- "List all actors in Science Fiction movies."
- "Tell me about movies in the Action genre."
-
Edge Case Queries:
- "What is the genre of Inception?"
- "Who is the director of a movie starring Leonardo DiCaprio?"
- "who is the director of a movie starring Boluwatife Rolland?"