Hierarchical Crew

A component that allows you to organize and orchestrate multiple AI agents into a collaborative team. The Hierarchical Crew component enables complex, multi-step workflows with specialized agents working together toward a common goal.

Hierarchical Crew component interface and configuration

Advanced Configuration: This component requires careful planning of agent roles and responsibilities. Proper configuration of memory sharing and agent interactions is crucial for effective collaboration.

Component Inputs

Verbose: Level of detail in crew operation logs
Example: 0 (minimal logging) to 2 (detailed logging of all agent interactions)
Memory: Shared memory configuration for the crew
Example: "Shared" (all agents access the same memory) or "Individual" (each agent has private memory)
Cache: Whether to cache agent responses
Example: Enabled (for faster repeated operations) or Disabled (for always fresh responses)
Max RPM: Maximum requests per minute for the crew
Example: 100 (limits the rate of API calls to prevent throttling)
Share Crew: Whether agents can reference other agents in the crew
Example: Enabled (agents can collaborate directly) or Disabled (agents work independently)

Component Outputs

Toolset: The result of the crew's collaborative work
Example: The final output from the orchestrated workflow, which could be a document, analysis, or other collaborative result

Agent Types and Roles

Function Calling LLM

An agent that can execute specific functions and interact with external tools

Role: Primary interface for tool interactions and function execution
Example: API caller, data processor, or query executor

Agents

Specialized agents with specific skills or knowledge domains

Role: Domain-specific experts that handle specialized tasks
Example: Research agent, creative writer, code reviewer

Tasks

Discrete work items that agents can perform

Role: Structured units of work assigned to agents
Example: Data analysis task, content generation task, evaluation task

Manager LLM

Oversees the workflow and coordinates agent activities

Role: Orchestrator that assigns tasks and monitors progress
Example: Project manager, workflow coordinator

Manager Agent

Assistant to the manager that handles specific coordination aspects

Role: Task dispatcher and progress tracker
Example: Quality assurance, timeline manager

Implementation Example

// Example configuration for a content creation crew
const contentCreationCrew = {
  verbose: 1,
  memory: "Shared",
  cache: true,
  maxRpm: 60,
  shareCrew: true,
  agents: [
    {
      role: "Manager",
      type: "ManagerLLM",
      model: "gpt-4",
      description: "Coordinates the content creation workflow and ensures quality"
    },
    {
      role: "Researcher",
      type: "Agent",
      model: "gpt-4",
      description: "Gathers relevant information on the topic"
    },
    {
      role: "Writer",
      type: "Agent",
      model: "gpt-4",
      description: "Creates engaging content based on research"
    },
    {
      role: "Editor",
      type: "Agent",
      model: "gpt-4",
      description: "Reviews and refines the content for clarity and accuracy"
    }
  ],
  workflow: [
    {
      name: "Research",
      type: "Task",
      assignedTo: "Researcher",
      description: "Collect key information and data points on the topic"
    },
    {
      name: "Draft",
      type: "Task",
      assignedTo: "Writer",
      description: "Create a compelling first draft based on research findings"
    },
    {
      name: "Edit",
      type: "Task",
      assignedTo: "Editor",
      description: "Review and refine the draft for quality and accuracy"
    },
    {
      name: "Finalize",
      type: "Task",
      assignedTo: "Manager",
      description: "Approve the final content and prepare for delivery"
    }
  ]
};

// Example integration in an application
async function createContent(topic, audience, format) {
  // Initialize the crew component
  const crewResult = await hierarchicalCrewComponent.execute({
    verbose: contentCreationCrew.verbose,
    memory: contentCreationCrew.memory,
    cache: contentCreationCrew.cache,
    maxRpm: contentCreationCrew.maxRpm,
    shareCrew: contentCreationCrew.shareCrew,
    context: {
      topic: topic,
      audience: audience,
      format: format,
      deadline: new Date().addDays(1).toISOString()
    }
  });
  
  // Extract and use the crew's output
  return {
    content: crewResult.toolset,
    metadata: {
      created: new Date().toISOString(),
      topic: topic,
      audience: audience
    }
  };
}

Use Cases

Content Creation: Orchestrate research, writing, and editing agents
Product Development: Coordinate design, development, and testing workflows
Customer Support: Manage ticket classification, research, and response generation
Market Analysis: Combine data collection, processing, and insight generation
Research Projects: Coordinate literature review, analysis, and reporting

Useful Resources

Best Practices

Clearly define each agent's role and responsibilities
Start with a simpler crew structure and expand gradually
Use shared memory for tasks requiring context preservation
Set appropriate verbosity levels for debugging and monitoring
Design workflows with clear dependencies and handoffs
Implement error handling and recovery mechanisms
Monitor resource usage to prevent excessive API consumption
Test crews with various inputs to ensure robust performance

Documentation