---

📌 Project Overview

YouTube AI Analyzer is an AI-powered agent that takes any YouTube video URL and returns a structured, timestamped content breakdown — including video type identification, segment summaries, topic progression, and key learning points.

Built with the Agno agent framework, powered by Groq's Qwen3-32B LLM, and wrapped in a fully branded Streamlit UI — no OpenAI API needed, entirely free to run.

🎯 Use Case: Students, content creators, and researchers who want instant structured summaries of long YouTube videos without watching the full content.

---

🧠 Agent Architecture

YouTube URL → Agno Agent → Groq (Qwen3-32B) → YouTubeTools (transcript fetch) → Structured Analysis → Streamlit UI

How It Works

1️⃣ URL Intake

• User pastes any YouTube video URL into the Streamlit UI
• Input validated for correct YouTube domain format

2️⃣ Transcript Extraction

• YouTubeTools from Agno fetches the auto-generated or manual caption transcript
• Handles both youtube.com and youtu.be short URLs

3️⃣ LLM Analysis via Groq

• Transcript sent to qwen/qwen3-32b running on Groq's ultra-fast inference
• Agent follows structured instructions: Video Overview → Timestamp Creation → Content Organization

4️⃣ Streaming Output

• Response streamed token-by-token back to the UI in real time
• Final output downloadable as .md file

---

📊 Output Format

The agent produces a structured markdown report containing:

Section	Description
🎬 Video Overview	Title, type, length, structure
⏱️ Timestamps	[start, end] with detailed segment summary
📚 Key Learning Points	Bullet list of main takeaways
🔗 References	Notable links or resources mentioned

---

🔍 Key Insights

• Groq inference is 10–20× faster than standard OpenAI API — ideal for long transcripts
• Qwen3-32B handles multilingual transcripts better than smaller models
• YouTubeTools works on any video with captions (auto-generated or manual)
• Streaming output keeps UX snappy even for 1-hour long videos
• CSS-file separation from Streamlit logic prevents raw HTML leaking issues

---

🗂️ Repository Structure

youtube-ai-analyzer/
├── ui.py                  # Streamlit frontend
├── style.css              # Custom UI styles (instapost brand theme)
├── .env                   # API keys (not committed)
├── requirements.txt       # Dependencies
├── .streamlit/
│   └── config.toml        # Streamlit server config
└── README.md

---

🚀 Quick Start

# 1. Clone the repo
git clone https://github.com/ronakrajput8882/Youtube-Video-Analyzer.git
cd Youtube-Video-Analyzer

# 2. Install dependencies
pip install streamlit agno groq python-dotenv youtube-transcript-api

# 3. Set up API key
echo "GROQ_API_KEY=your_groq_key_here" > .env

# 4. Run the app
streamlit run ui.py

Get your free Groq API key at → console.groq.com

---

🧠 Key Learnings

• Agno's Agent abstraction simplifies tool-augmented LLM calls to a few lines
• Groq's free tier supports Qwen3-32B with generous rate limits — no cost to run
• Streamlit CSS injection must use hardcoded hex values — CSS variables from :root don't resolve inside Streamlit's shadow DOM
• Separating CSS into style.css and loading via open() prevents raw HTML render bugs
• @st.cache_resource is critical — without it the agent re-initializes on every button click

---

🛠️ Tech Stack

Tool	Purpose
Python 3.10+	Core language
Streamlit 1.57	Web UI framework
Agno	AI agent framework
Groq API	LLM inference (Qwen3-32B)
YouTubeTools	Transcript extraction
python-dotenv	Environment variable management
youtube-transcript-api	YouTube caption fetching

---

Connect with me

If you found this useful, please ⭐ the repo!