🎥 YouTube Transcript AI Assistant

Ask Questions About Any YouTube Video Using AI

FastAPI • LangChain • Ollama • FAISS • RAG

Turn any YouTube video into an intelligent knowledge base and chat with its content.

---

🚀 Overview

YouTube videos contain valuable information, but finding specific insights often requires watching hours of content.

This project solves that problem by combining:

• YouTube Transcript Extraction
• Retrieval-Augmented Generation (RAG)
• FAISS Vector Search
• Ollama Local LLMs
• FastAPI Backend

Users can load a YouTube video's transcript into a vector database and ask natural language questions about the video's content.

The system retrieves relevant transcript segments and generates context-aware answers using a local LLM.

---

✨ Features

📺 YouTube Transcript Processing

• Automatic transcript extraction
• Support for English transcripts
• Transcript chunking for efficient retrieval

🧠 AI-Powered Question Answering

• Retrieval-Augmented Generation (RAG)
• Context-aware responses
• Grounded answers from transcript content
• Hallucination reduction through retrieval

⚡ Fast Semantic Search

• FAISS Vector Database
• Embedding-based similarity search
• Relevant transcript retrieval

🤖 Local AI Inference

• Powered by Ollama
• No OpenAI API required
• Fully local execution

🌐 Modern Web Interface

• Clean responsive UI
• Real-time interaction
• FastAPI REST endpoints

---

🏗️ System Architecture

YouTube Video
      │
      ▼
Transcript Extraction
      │
      ▼
Text Chunking
      │
      ▼
Ollama Embeddings
      │
      ▼
FAISS Vector Store
      │
      ▼
Similarity Search
      │
      ▼
Relevant Context
      │
      ▼
ChatOllama (LLM)
      │
      ▼
Generated Answer

---

⚙️ Tech Stack

Backend

• FastAPI
• Python
• Pydantic

AI & RAG

• LangChain
• Ollama
• ChatOllama
• Ollama Embeddings
• FAISS

Data Processing

• YouTube Transcript API
• Recursive Character Text Splitter

Frontend

• HTML
• CSS
• JavaScript

---

📂 Project Structure

YT-Transcript-AI-Assistant/

├── main.py
├── requirements.txt
├── index.html
├── README.md
└── assets/

---

🔌 API Endpoints

Load Video Transcript

POST /load-video

Request

{
  "videolink": "https://www.youtube.com/watch?v=VIDEO_ID"
}

Response

{
  "message": "Video transcript loaded into FAISS.",
  "video_id": "VIDEO_ID",
  "cached": false,
  "chunks": 32
}

---

Ask Assistant

POST /ask-assistant

Request

{
  "videolink": "https://www.youtube.com/watch?v=VIDEO_ID",
  "target_question": "What is the video about?"
}

Response

{
  "answer": "Generated answer from transcript context"
}

---

🛠 Installation

Clone Repository

git clone https://github.com/your-username/yt-transcript-ai-assistant.git

cd yt-transcript-ai-assistant

Install Dependencies

pip install -r requirements.txt

---

🤖 Install Ollama

Download:

https://ollama.com

Pull the model:

ollama pull qwen2.5-coder:3b

Verify:

ollama list

---

▶️ Run the Backend

uvicorn main:app --reload

Server starts at:

http://127.0.0.1:8000

---

🌐 Run the Frontend

Open:

index.html

Or use:

• VS Code Live Server
• Python HTTP Server

python -m http.server 5500

---

💡 Example Questions

Summarize this video.

What are the key points discussed?

Explain the main concept.

What technologies were mentioned?

Give me a short overview.

What are the speaker's conclusions?

---

🎯 Use Cases

• Educational video summarization
• Technical tutorial analysis
• Lecture understanding
• Research assistance
• Knowledge extraction
• Content review
• Personal learning assistant

---

🔮 Future Improvements

• Multi-video knowledge base
• Playlist ingestion
• Persistent vector storage
• Chat memory
• Streaming responses
• PDF summary export
• User authentication
• Docker deployment
• Multi-language support

---

👨‍💻 Author

Nishant Rajora

---

⭐ Support

If you found this project useful:

• Star the repository
• Fork the project
• Share feedback
• Contribute improvements

---

Built with ❤️ using FastAPI, LangChain, Ollama and FAISS