• Overview
• Features
• Installation
• Usage
• Codebase
  • Dependencies
  • File Structure
• Contributing
• License

This CLI tool is a TypeScript application that leverages the OpenAI API, HNSWLib, and Langchain to analyze YouTube videos and engage in an AI-driven conversation based on the video's content. Aimed at TypeScript engineers, this README provides an exhaustive guide for understanding, installing, and using this tool.

• Prompt for YouTube URL: Asks the user for a YouTube video URL with a default option.
• Transcript Download: Downloads the transcript of the given YouTube video and splits it into documents.
• OpenAI Embeddings: Utilizes OpenAI embeddings to create vector embeddings for the downloaded documents.
• Local Vector Store: Utilizes the HNSWLib library to store embeddings in a local vector store on the filesystem.
• AI-Driven Chat: Uses the Langchain library to build a chat interface that uses the embeddings as context.
• Summary Display: Evaluates and displays a summary of the video based on its transcripts.
• Interactive Q&A: Allows the user to ask follow-up questions, which are answered using the embeddings until the user types "exit".

To install the CLI tool, you need to have Node.js and npm installed on your system. Then run:

# Clone the repository
git clone https://github.com/djheru/gpt-cli.git

# Navigate into the directory
cd gpt-cli

# Install dependencies
npm install

To execute the application, run:

npm run youtube

Follow the on-screen prompts to input a YouTube URL (or use the default) and engage in a conversation based on the video's content.

Here are the main libraries and their roles in this project as per package.json:

• openai: For generating embeddings and AI-related functionalities.
• hnswlib: Used for efficient similarity search and storage of embeddings.
• langchain: For building the chat interface.

• index.ts: Entry point for the application.
• youtube.ts: Contains the logic for handling YouTube video analysis and chat.
• utils.ts: Utility functions used across the project.

// File: index.ts
// Entry point for the CLI application

// File: youtube.ts
// Main logic for YouTube video analysis and chat functionality

// File: utils.ts
// Contains utility functions for embedding creation, summary, etc.

If you're interested in contributing, please follow the standard fork, feature-branch, pull request workflow.

1. Fork the repository.
2. Create a new feature branch.
3. Make your changes.
4. Create a pull request.

This project is licensed under the MIT License.

This README aims to be a thorough guide for TypeScript engineers looking to understand and contribute to this tool. Feel free to submit issues or pull requests to improve it.