YouTube Video Summarizer & Q&A

A powerful tool that downloads YouTube videos, transcribes the audio, generates concise summaries, and allows you to ask questions about the video content using RAG (Retrieval-Augmented Generation).

🚀 Features

• Video Ingestion: Downloads audio from YouTube videos using yt-dlp.
• Intelligent Transcription:
  • Auto-Switching: Automatically chooses between Single-threaded (short videos) and Parallel Processing (long videos > 5 mins) for optimal speed.
  • Whisper Integration: Uses OpenAI's Whisper model for high-accuracy speech-to-text.
• Advanced RAG Pipeline:
  • Hybrid Search: Retrieves top candidates using Vector Search (FAISS).
  • Re-ranking: Refines results using a Cross-Encoder (ms-marco-MiniLM-L-6-v2) to ensure high relevance.
• Background Processing:
  • Asynchronous Architecture: Uses Redis and RQ (Redis Queue) to handle long-running tasks without blocking the UI.
  • Real-time ETA: Calculates and displays estimated time remaining based on video duration and current stage.
• Summarization & Q&A: Generates concise summaries and answers questions using Llama 3.3 70B via Groq.

🛠️ Tech Stack

• Backend: FastAPI
• Frontend: Streamlit
• LLM / Inference: Groq API (Llama 3.3-70b-versatile)
• Transcription: OpenAI Whisper (Local) with Parallel Execution
• Vector DB: FAISS
• Re-ranking: SentenceTransformers (cross-encoder/ms-marco-MiniLM-L-6-v2)
• Queue System: Redis + RQ
• Video Processing: yt-dlp, ffmpeg-python, pydub

📋 Prerequisites

• Python 3.8+
• FFmpeg installed and added to PATH.
• Redis installed and running locally.
• A Groq API Key.

⚙️ Installation

1. Clone the repository:

   git clone <repository-url>
   cd video-summarizer

2. Create and activate a virtual environment:

   python -m venv .venv
   # Windows
   .venv\Scripts\activate
   # Mac/Linux
   source .venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Configure Environment: Create a .env file in the root directory:

   GROQ_API_KEY=your_groq_api_key_here

🏃‍♂️ Running the Application

This app uses an async worker architecture. You need 3 terminals:

1. Start Redis Server

Ensure your Redis server is running.

docker start redis-stack

2. Start the Worker (Background Processor)

This handles the heavy lifting (transcription, embedding).

rq worker ingestion --worker-class rq.worker.SimpleWorker

3. Start the Backend API

uvicorn app.main:app --reload

4. Start the Frontend UI

streamlit run frontend/streamlit_app.py

📂 Project Structure

├── app/
│   ├── api/            # FastAPI Routes
│   ├── db/             # Vector Database (FAISS)
│   ├── rag/            # Chunking & Prompting
│   ├── services/       # Core Logic (YouTube, STT, QA, Re-ranker)
│   ├── utils/          # Redis connection, Task Queue, Audio utils
│   ├── workers/        # RQ Worker logic (ingest_worker.py)
│   └── main.py         # App Entry Point
├── data/               # Persistent storage for indices and summaries
├── frontend/           # Streamlit Interface
├── scripts/            # Testing scripts
├── .env                # Secrets
└── requirements.txt    # Dependencies