Configuration Guide

Mastering Crew.ai’s multi_agent_config.yaml: Enabling and Disabling Agents

The multi_agent_config.yaml file is the brain of the Crew.ai Ollama Multi-Agent System. It lets you easily control which agents (like virtual helpers) are active, which are disabled, and how they collaborate to perform tasks. Whether you’re new to tech or customizing for professional workflows, this guide will help you set up and manage your system step by step.

---

What is multi_agent_config.yaml?

Think of multi_agent_config.yaml as your system’s instruction manual. It helps you:

1. Define how the system works as a whole.
2. List all available agents (helpers) in your system.
3. Turn specific agents on or off to match your needs.

Here’s an example of what the file looks like:

orchestration:
  multi_agent: true
  memory_store: vector_db

agents:
  - name: ollama_agent
    enabled: true

  - name: household_chores_agent
    enabled: true

  - name: diet_meal_planning_agent
    enabled: false

  - name: photo_management_agent
    enabled: false

---

Breaking Down the File

1. Orchestration Section

This part defines how your system operates:

• multi_agent: If true, multiple agents can collaborate to solve tasks. If false, only one agent will handle everything.
• memory_store: Defines how your system stores information. Options include:
  • vector_db: Uses advanced storage for quick recall (great for complex tasks).
  • in_memory: Stores data temporarily while the system is running.

2. Agents Section

This section lists all your agents. Each agent has:

• name: A unique name for the agent (e.g., ollama_agent).
• enabled: A true or false value to control whether the agent is active.

---

How to Enable or Disable Agents

Enabling an Agent

To turn on an agent, set its enabled field to true:

- name: diet_meal_planning_agent
  enabled: true

Once enabled, this agent will activate the next time you restart the system.

Disabling an Agent

To turn off an agent, set its enabled field to false:

- name: photo_management_agent
  enabled: false

Disabled agents won’t run, which saves system resources.

---

Why Would You Enable or Disable Agents?

1. Save Resources

Some agents require more memory or processing power. Disabling non-essential agents can improve performance, especially on less powerful machines.

2. Focus on What You Need

If you only care about meal planning and tracking expenses, you can disable agents like photo management or streaming recommendations to keep things simple.

3. Debugging or Testing

When creating or fixing a new agent, disable the others to isolate and test its functionality.

4. Privacy

Certain agents might use the internet (e.g., for weather or streaming data). If you prefer everything offline, disable agents that require external connections.

---

How the System Reads multi_agent_config.yaml

The AgentManager in Crew.ai reads your configuration file to decide which agents to activate. Here’s a simplified version of how it works:

import yaml
from agents.ollama_agent import OllamaAgent
from agents.household_chores_agent import HouseholdChoresAgent
from agents.diet_meal_planning_agent import DietMealPlanningAgent

class AgentManager:
    def __init__(self):
        # Load the configuration file
        with open("src/config/multi_agent_config.yaml", "r") as f:
            self.config = yaml.safe_load(f)
        self.active_agents = self.load_agents()

    def load_agents(self):
        agents = []
        for agent_def in self.config.get("agents", []):
            if agent_def["enabled"]:
                if agent_def["name"] == "ollama_agent":
                    agents.append(OllamaAgent())
                elif agent_def["name"] == "household_chores_agent":
                    agents.append(HouseholdChoresAgent())
                elif agent_def["name"] == "diet_meal_planning_agent":
                    agents.append(DietMealPlanningAgent())
        return agents

Key Takeaways:

• The system looks for agents marked enabled: true in the multi_agent_config.yaml file.
• Only enabled agents are added to the system and ready to handle tasks.

---

Example Configurations

Minimal Setup

For a basic system that handles general tasks and household chores:

orchestration:
  multi_agent: true
  memory_store: in_memory

agents:
  - name: ollama_agent
    enabled: true

  - name: household_chores_agent
    enabled: true

  - name: diet_meal_planning_agent
    enabled: false

Full Feature Setup

For a system with all available agents enabled:

orchestration:
  multi_agent: true
  memory_store: vector_db

agents:
  - name: ollama_agent
    enabled: true

  - name: household_chores_agent
    enabled: true

  - name: diet_meal_planning_agent
    enabled: true

  - name: photo_management_agent
    enabled: true

---

Tips for Success

1. Keep a Backup: Always save a copy of your working configuration file in case you need to revert.

2. Test Changes: When enabling a new agent, test it individually to make sure it works as expected.

3. Monitor Performance: If the system feels slow, try disabling resource-heavy agents.

4. Document Your Setup: Add comments to your multi_agent_config.yaml file to track why certain agents are enabled or disabled.

Example:

# Enabling photo management for organizing holiday pictures
- name: photo_management_agent
  enabled: true

---

Conclusion

The multi_agent_config.yaml file gives you full control over how Crew.ai operates. By turning agents on or off, you can:

• Customize the system to match your goals.
• Optimize its performance for your hardware.
• Keep things simple or expand as needed.

Whether you’re a tech enthusiast or just looking for a practical assistant, mastering this configuration file is the key to unlocking Crew.ai’s full potential. Start experimenting today and create your perfect AI setup!