Generative-AI-Module-Dauphine-Twitter_Bot

Repository Overview

This project, part of the "Generative AI" course for the Master IASD Executive program at Dauphine-PSL University, demonstrates the use of generative AI in a tool that helps agents in customer service to answer customer's tweets faster.

This project can be reuse for any other use cases, that requires to use company data to generate a text from a user query (chatbot, email, calls, ...).

Key Components

• app.py: The Flask app, containing the backend.
• notebooks/:
  • embed_data.ipynb : This notebook embeds tweet data into a chromadb vector database using the OpenAI API.
  • evaluate_tool.ipynb : This notebook sends data from twitter_data_clean_eval.csv through the Flask API to generate responses to customer tweets, comparing them to original tweets for analysis.

Pre-requisites

Before running this project, ensure you have completed the environment setup as outlined in the Setup Guide from the Generative AI module repository.

Configuration and Setup

1. OpenAI API Credentials

   • Create a config.ini file in the project root with the following content, replacing key with your actual OpenAI API key:

     [OPENAI_API]
     OPENAI_KEY = key

2. Data Preparation

   • Clone the Generative-AI-Module-Dauphine repository.
   • Create a data/ directory in the Twitter Bot project folder.
   • Copy all CSV files from the Generative-AI-Module-Dauphine repository into the data/ directory.

3. Embedding Tweet Data

   • Navigate to the notebooks/ directory.
   • Open and run embed_data.ipynb in a Jupyter environment. This notebook embeds tweet data into a chromadb vector database using the OpenAI API.

4. Running the Flask Application

   • Execute the Flask app to start the server:

     python app.py

5. Evaluation and Interaction

   • Use the notebooks/evaluate_tool.ipynb notebook for evaluating the system. This notebook sends data from twitter_data_clean_eval.csv through the Flask API to generate responses to customer tweets, comparing them to original tweets for analysis.

Running the Project

After completing the setup and configuration steps, you can start experimenting with interface, and generate answers for customer's tweets. This project offers a hands-on opportunity to showcase the practical applications and challenges of generative AI technologies.