🎥 YouTube RAG (Retrieval-Augmented Generation)

An AI-powered Retrieval-Augmented Generation (RAG) application that allows users to ask questions about a YouTube video. The application extracts the video's transcript, converts it into vector embeddings, stores them in a FAISS vector database, and generates context-aware answers using a Large Language Model (LLM).

---

📌 Features

• Extract transcript from any YouTube video
• Automatic text preprocessing
• Intelligent text chunking
• Generate vector embeddings
• Store embeddings in FAISS
• Semantic similarity search
• Context-aware question answering
• Simple Streamlit user interface
• Supports AI, Data Science, Python, SAP, and educational videos

---

🚀 Project Workflow

           User
             │
             ▼
      Enter YouTube URL
             │
             ▼
    Extract Video Transcript
             │
             ▼
       Clean & Split Text
             │
             ▼
     Generate Embeddings
             │
             ▼
    Store Vectors in FAISS
             │
             ▼
      User Asks Question
             │
             ▼
 Retrieve Relevant Chunks
             │
             ▼
 Large Language Model (LLM)
             │
             ▼
      Generate Response

---

🛠 Technology Stack

• Python
• LangChain
• FAISS
• OpenAI
• HuggingFace Transformers
• Sentence Transformers
• YouTube Transcript API
• Streamlit
• Dotenv

---

📂 Project Structure

YouTube-RAG/
│
├── app.py
├── rag.py
├── transcript.py
├── embeddings.py
├── requirements.txt
├── README.md
├── .env
├── vectorstore/
├── data/
├── screenshots/
└── assets/

---

📦 Installation

Clone the repository

git clone https://github.com/yourusername/youtube-rag.git

Go to project folder

cd youtube-rag

Install dependencies

pip install -r requirements.txt

---

▶️ Run the Project

Using Streamlit

streamlit run app.py

---

📚 Required Libraries

langchain
langchain-community
faiss-cpu
transformers
sentence-transformers
youtube-transcript-api
streamlit
python-dotenv
openai
numpy
pandas
tiktoken

---

💡 How It Works

Step 1

Enter a YouTube video URL.

Step 2

The transcript is extracted automatically.

Step 3

The transcript is cleaned and divided into smaller chunks.

Step 4

Each chunk is converted into vector embeddings.

Step 5

Embeddings are stored in a FAISS vector database.

Step 6

The user asks a question.

Step 7

The retriever searches the most relevant transcript chunks.

Step 8

The LLM generates a final answer using the retrieved context.

---

🧠 RAG Architecture

YouTube Video
      │
      ▼
Transcript Extraction
      │
      ▼
Text Chunking
      │
      ▼
Embedding Model
      │
      ▼
FAISS Vector Store
      │
      ▼
Similarity Search
      │
      ▼
Retrieved Context
      │
      ▼
Large Language Model
      │
      ▼
Final Answer

---

📈 Advantages

• Fast semantic search
• Accurate context-based answers
• No need to watch long videos
• Easy to scale
• Better than keyword search
• Supports educational content
• Easy integration with Streamlit

---

⚠️ Limitations

• Transcript must be available
• Performance depends on embedding quality
• Internet connection required
• Long videos require additional processing time

---

🎯 Applications

• Education
• AI Tutorials
• Data Science Learning
• Interview Preparation
• Online Courses
• Research
• Lecture Summarization
• Corporate Training

---

🔮 Future Scope

• Multiple YouTube videos
• Chat history
• Voice assistant
• PDF support
• Audio upload
• Image understanding
• Multilingual support
• Cloud deployment
• User authentication

---

📊 Sample Output

Question:
What is Retrieval-Augmented Generation?

Answer:
Retrieval-Augmented Generation (RAG) combines information retrieval with a Large Language Model. It retrieves relevant information from a vector database before generating an accurate and context-aware response.

---

📸 Screenshots

Add screenshots here.

screenshots/

home.png

output.png

workflow.png

---

👨‍💻 Author

Ashwin Kumar

MBA (Data Analytics)

Python | Machine Learning | Deep Learning | Generative AI | RAG | SQL | Power BI

---

⭐ Acknowledgement

Special thanks to the open-source community for providing tools such as LangChain, FAISS, Hugging Face Transformers, Streamlit, and the YouTube Transcript API, which made this project possible.

---

📄 License

This project is intended for educational and learning purposes.