Local Knowledge Graph
This application uses a local Llama model to answer queries, build embeddings, and create a knowledge graph for exploring related questions and answers.
Description
The Local Knowledge Graph is a Flask-based web application that leverages a local Llama language model to process user queries, generate step-by-step reasoning, and visualize the thought process as an interactive knowledge graph. It also finds and displays related questions and answers based on semantic similarity.
Features
- Interactive web interface for submitting queries
- Step-by-step reasoning process displayed in real-time
- Dynamic knowledge graph visualization of the reasoning steps
- Calculation and display of the strongest reasoning path
- Related questions and answers based on semantic similarity
- Local processing using a Llama language model
Usage
- Ensure you have all the required dependencies installed.
- Start the Flask application by running
app.py. - Open a web browser and navigate to
http://localhost:5100(or the appropriate port if modified). - Enter your query in the input field and click "Submit".
- Watch as the application generates a step-by-step reasoning process, updating the knowledge graph in real-time.
- Review the final answer and the strongest reasoning path.
- Explore related questions and answers displayed below the main response.
Requirements
- Python 3.7+
- Flask
- NumPy
- scikit-learn
- Annoy
- NetworkX
- A local Llama language model (e.g., llama3.1:8b) running on
http://localhost:11434
Installation
- Clone this repository.
- Install the required Python packages using the requirements.txt file:
pip install -r requirements.txt - Ensure you have a local Llama model running and accessible.
- Run the Flask application:
Note
This application requires a local Llama language model to be running and accessible. Make sure you have the appropriate model set up and running before using this application.