- Introduction - Presenter: David Andre - Topic: Building AI agents with Llama fre model - Tools: AMA, VS Code, Gro - Performance: 216 tokens per second with big model, instant with Llama fre 8 billion - LLM Arena - Llama 370b is open source and outperforms GPT-4 - Workshop Offer - Step-by-step AI agent building workshop for non-programmers - Available in David's community - Building from Scratch - Download AMA and VS Code - Choose Llama 3 model (8 billion recommended) - Install model via terminal in VS Code - Download times: 20 minutes for small model, 3 hours for large model - Quick Tip - Use 'SL bu' to end chat without killing the terminal - Setting Up in VS Code - Create new Python file (main.py) - Import AMA and install necessary packages (e.g., PIP install crew AI) - Import from crew AI (Agent, Task, Crew, Process) - Defining Agents and Tasks - Create variables for models and agents - Example: Email classifier and responder agents - Define tasks with descriptions and expected outputs - Assembling the Crew - Create a crew with a list of agents and tasks - Set verbosity and process type (sequential) - Kick off the crew and print output - Troubleshooting - Issues with Llama 3 performance through crew AI - Llama 3 works well in terminal but not in crew AI - Connecting to Gro API - For users with bad computers - Create API key on Gro cloud - Set environment variables for Gro API - Import OS and set API base, model name, and API key - Test and observe improved speed with API - Conclusion - AI Revolution is happening - Join David's community to stay ahead in AI - Community link provided in the video description