🎥 YouTube Transcript Chat App

A Streamlit-based web application that lets users ask questions about the content of a YouTube video using its English transcript. It extracts the transcript, breaks it into chunks, embeds it using HuggingFace models, and answers questions via the Perplexity LLM.

🚀 Features

• 🔗 Enter any YouTube video URL with English captions
• 🤖 Ask natural language questions about the video
• 🧠 Powered by Perplexity LLM & HuggingFace sentence embeddings
• 📚 Context-aware QA using FAISS vector store
• 🎨 Clean UI with custom background and styled answer box

The core Streamlit app that:

• Accepts a YouTube URL and user question
• Fetches and processes transcript using youtube-transcript-api
• Splits transcript into chunks with RecursiveCharacterTextSplitter
• Generates embeddings using sentence-transformers/all-MiniLM-L6-v2
• Stores vectors in FAISS and retrieves top k similar chunks
• Feeds context and question into a prompt chain using Perplexity
• Displays the generated answer in a styled output box

🛠️ Technologies Used

• Python
• Streamlit
• LangChain
• HuggingFace Sentence Transformers
• Perplexity Chat API
• FAISS
• YouTube Transcript API

📦 Installation

1. Clone the repository

   git clone https://github.com/TahaCoder563/Youtube-Transcript-Chat.git
   cd Youtube-Transcript-Chat

2. Install dependencies

   pip install -r requirements.txt

3. Set up API keys

   🔑 Get a Perplexity API Key:

   • Go to https://docs.perplexity.ai
   • Sign in or create an account
   • Generate a new API key from the dashboard

   🔑 Get a HuggingFace API Token:

   • Go to https://huggingface.co/settings/tokens
   • Create a token with read access

   Create a .env file in the root of the project and paste your keys:

   PERPLEXITY_API_KEY="your_perplexity_api_key_here"
   HUGGINGFACEHUB_API_TOKEN="your_huggingface_token_here"

   Replace the your_perplexity_api_key_here and your_huggingface_token_here with the respective api keys.

4. Run the app

   streamlit run main.py