🎥 Advanced YouTube RAG Chatbot

---

Deployed App Link: https://chatwithvideos.streamlit.app/

---

📌 Project Overview

This application allows users to interact conversationally with YouTube videos.

The system extracts video transcripts, converts them into semantic embeddings, stores them in a FAISS vector database, retrieves relevant transcript chunks, reranks results semantically, and generates contextual answers using Gemini AI.

The project demonstrates:

• RAG pipeline engineering
• NLP workflows
• conversational memory
• semantic retrieval
• vector databases
• production AI application architecture

---

---

🚀 Live Features

✅ YouTube transcript extraction
✅ Conversational AI chatbot
✅ Multi-video support
✅ Semantic search with FAISS
✅ Gemini-powered responses
✅ Semantic reranking
✅ Timestamp citations
✅ Transcript download
✅ Streaming responses
✅ Video thumbnail preview
✅ Persistent vector database
✅ Production-style modular architecture

---

System Architecture

---

⚙️ Tech Stack

Frontend

• Streamlit

AI / NLP

• LangChain
• Google Gemini AI
• Sentence Transformers

Vector Database

• FAISS

Data Processing

• YouTube Transcript API
• Recursive Text Chunking

---

📂 Project Structure

youtube-rag-chatbot/
│
├── app.py
├── requirements.txt
├── README.md
├── .gitignore
├── .env
│
├── utils/
│   ├── transcript.py
│   ├── chunking.py
│   ├── embeddings.py
│   ├── vectorstore.py
│   ├── rag_chain.py
│   ├── reranker.py
│
├── data/
│   └── faiss_index/
│
└── images/
    └── screenshots/

---

🚀 Installation

Clone Repository

git clone github.com/maw-khan/youtube-rag-chatbot.git

Install Dependencies

pip install -r requirements.txt

Run Application

streamlit run app.py

---

🔑 API Key Setup You need a Google Gemini API Key. Get it from: 👉 https://ai.google.dev/ No need for .env file — the app accepts it directly via Streamlit sidebar.

---

▶️ Run the Deployed App (Link):

https://chatwithvideos.streamlit.app/

---

💡 How to Use

1. Enter your Gemini API Key in the sidebar
2. Paste youtube videos links/URL
3. Click “Process Video”
4. Wait for processing to complete
5. Start asking questions in the chat box
6. View answers + expandable source references

📚 Example Query “What is the main conclusion of the video?” The chatbot will:

• Retrieve relevant chunks
• Generate an answer using Gemini
• Show source excerpts used for reasoning

---

---

🔥 Key Features Explained

🎥 Transcript Extraction

Automatically extracts transcripts from YouTube videos.

📊 Retrieval-Augmented Generation (RAG)

Relevant transcript chunks are retrieved before generating responses.

📚 Semantic Reranking

Retrieved chunks are reranked using transformer-based semantic similarity models.

📌 Timestamp Citations

Responses include transcript timestamps for source verification.

⚡ Streaming Responses

Answers stream in real time for better conversational UX.

📥 Transcript Download

Users can download processed transcripts directly.

---

🚀 Future Improvements

• YouTube playlist support
• Multi-user authentication
• Chat export
• PDF summary generation
• Voice interaction
• Hybrid BM25 + vector retrieval
• LangGraph agent workflows

---

📌 Learning Outcomes

This project helped strengthen understanding of:

• AI application engineering
• RAG pipelines
• vector search systems
• semantic retrieval
• conversational memory
• Streamlit deployment
• modular software architecture

---

👨‍💻 Author

Muhammad Ali Waris Khan

AI Developer | RAG Systems | Streamlit AI Apps | Python