Project Description

This project is a learning agent that uses various APIs and tools to gather and process information, store it in a shared memory, and use it to answer user queries. The agent is designed to be neutral and objective, providing accurate responses based on reliable information.

Features

• Uses WikipediaAPIWrapper, ArxivAPIWrapper, and BraveSearch to gather information.
• Stores information in a shared memory using a custom SharedMemory class.
• Uses GraphQAChain and NetworkxEntityGraph to process and analyze the information.
• Uses OpenAI's ChatOpenAI model for natural language processing.
• Implements a Graph Agent and a Cognitive Agent to handle user queries.

Classes

• learning_agent: An empty class for future development.
• SharedMemory: A class to handle storing and retrieving information in a shared memory.
• Graph_Agent: A class that uses various tools to gather and process information, and answer user queries.
• CognitiveAgent: A class that interacts with the Graph Agent to handle user queries.

Tools

• write_memory: A tool to write information into the shared memory.
• read_memory: A tool to read information from the shared memory.
• node_information: A tool to retrieve metadata related to a node in the graph.
• Noeuds_Proches: A tool to search for nodes connected to a specific node.
• list_of_nodes: A tool to return all nodes in the graph.

Requirements

• json
• langchain
• networkx
• pickle
• GrandCypher
• OpenAI API key
• BraveSearch API key

Usage

Initialize a CognitiveAgent with a data path and an API key, then use the search method to ask questions. The agent will use the tools at its disposal to find the answer and respond with a detailed response, citing its sources.

Running the Application

The project also includes a Flask application that allows users to interact with the agent through a web interface. To run the application, set the OPENAI_API_KEY environment variable to your OpenAI API key, then run flask run. The application will be accessible at http://localhost:5000.

File Structure

• main.py: The main file containing the classes and tools.
• app.py: The Flask application file.
• templates/interface_user.html: The HTML template for the Flask application.
• Data directory: Contains the data used by the agent, including the graph data and node information.

Future Work

• Implement the learning_agent class.
• Add more tools for information gathering and processing.
• Improve the agent's ability to understand and respond to complex queries.
• Enhance the shared memory to store and retrieve information more efficiently.