Multi-Agent YouTube Research & Blog Writer

A simple multi-agent system built with CrewAI that researches YouTube videos and automatically generates blog posts based on video content.

Overview

This project uses two AI agents working in sequence:

1. Blog Researcher - Finds and analyzes YouTube videos on specified topics
2. Blog Writer - Creates engaging blog content based on the research

Features

• 🎥 Search and analyze YouTube channel content
• 🤖 Two specialized AI agents with distinct roles
• 📝 Automatic blog post generation in Markdown format
• 🔄 Sequential workflow with memory and caching
• ⚡ Built on CrewAI framework with GPT-4o-mini

Setup

1. Clone the repository

   git clone https://github.com/Tusharbecoding/multi-youtube-agent
   cd multi-youtube-agent

2. Create virtual environment

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

   pip install -r requirements.txt

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

   OPENAI_API_KEY=your_openai_api_key_here
   

5. Update YouTube channel Edit tools.py to specify the target YouTube channel:

   yt_tool = YoutubeChannelSearchTool(youtube_channel_handle="@yourchannel")

Usage

Run the multi-agent system:

python crew.py

The system will:

1. Research the specified video topic on the configured YouTube channel
2. Generate a comprehensive blog post
3. Save the output to blog_post.md

Project Structure

multi-youtube-agent/
├── agents.py      # AI agent definitions (researcher & writer)
├── tasks.py       # Task definitions for each agent
├── tools.py       # YouTube search tool configuration
├── crew.py        # Main orchestration and execution
└── requirements.txt

Customization

• Change the topic: Modify the topic parameter in crew.py
• Adjust agent behavior: Edit agent backstories and goals in agents.py
• Modify output format: Update task descriptions in tasks.py
• Switch YouTube channel: Update the channel handle in tools.py

Dependencies

• crewai - Multi-agent framework
• crewai-tools - YouTube integration tools
• python-dotenv - Environment variable management
• openai - GPT model access

Output

The system generates a blog_post.md file containing a well-structured blog post based on the YouTube video research.