Player not loading? Watch on YouTube
This beginner tutorial explains CrewAI through a research and script-writing workflow backed by Ollama. The speaker describes flows as the layer that controls execution and crews as teams of agents with assigned roles and goals. The demonstration builds a sequential crew.
Setup covers Windows Command Prompt or PowerShell and the terminal in Visual Studio Code. The presenter specifies Python 3.10 or higher, below 3.14, then explains virtual environment creation and CrewAI installation with pip. Ollama provides the local LLM runtime. The setup section downloads and tests llama3, while the later demonstration uses TinyLlama. The speaker says downloaded models can work offline; the example leaves the tools file empty and does not demonstrate web research.
The project separates agent definitions, tasks, optional tools and the entry point into Python files. A researcher produces three bullet points about a supplied topic, and a script writer uses that output to draft a short YouTube script. Both agents have memory disabled. The crew runs the research task first, then the writing task through crew.kickoff, and prints the result. The demonstrated topic is CrewAI itself.
