Build Anything with Local Agents, Here’s How

---

---

AI Summary

Summary: Building and Running AI Agents Locally with OLLAMA and CREWAI

Setting Up OLLAMA

• Download OAMA from ama.com for your OS.
• Unzip and run the application.
• Ensure OAMA icon appears, indicating it’s running.

Installing CREI

• Use Visual Studio Code (VS Code) to create a new file agents.py.
• Open a new terminal in VS Code.
• Install CREI using pip install crei or, if using Conda, create a new environment and install CREI there.
• Restart VS Code if necessary to see the new environment.

Importing Models with Lang Chain

• Import Ola using from lang_chain import olama.
• Install Lang Chain Community with pip install -U lang_chain_community.

Choosing a Language Model

• Select a model based on hardware capability and internet speed.
• Run the model using olama run <model_name>.

Building Agents with CREI

• Define agents with roles, goals, backstories, verbosity, delegation, and model.
• Create a variable topic to assign different topics to agents.
• Define tasks with descriptions, assigned agents, and expected outputs.
• Build a crew by assigning agents and tasks, setting verbosity, and using process.sequential.
• Execute with crew.execute_kickoff() and print the output.

Running and Testing Agents

• Change the topic variable to test different prompts.
• Run the agents and observe the memory usage.
• Review the generated prompts for quality and relevance.

Additional Information

• The video is part of a workshop on AI agents.
• Join the community for access to the full workshop and code.

Notes

• Ensure correct syntax in code (tasks and agents plural, llm not model).
• Use smaller models for faster results but with less intelligence.
• The next video will optimize the agents further, available exclusively in the community.