Orchestrator/Agent Structure
How BB delegates complex tasks to specialized sub-agents for more efficient processing and parallel execution.
Understanding the Orchestrator/Agent Model
The Orchestrator/Agent structure is a powerful feature in BB that allows complex tasks to be broken down and delegated to specialized sub-agents. This approach enables:
- More efficient processing of large or complex projects
- Parallel execution of independent tasks
- Specialized handling of domain-specific requirements
- Better token utilization and resource management
- Enhanced problem decomposition for complicated workflows
Key Components:
- Orchestrator: The main BB conversation that coordinates tasks and manages sub-agents
- Sub-agents: Specialized BB instances that handle specific delegated tasks
- Task definitions: Structured descriptions of work to be delegated
- Results integration: How completed tasks are collected and synthesized by the orchestrator
When to Use the Orchestrator/Agent Structure
The Orchestrator/Agent approach is ideal for specific use cases:
Token Efficiency
When working with very large files or multiple heavy resources that would exceed token limits in a single conversation:
- Processing multiple large documents
- Analyzing extensive codebases
- Working with lengthy historical conversations
- Handling high-volume data analysis
Problem Decomposition
For complex problems that benefit from being broken down into distinct, manageable parts:
- Multi-stage research projects
- Software architecture with several components
- Content creation across multiple channels
- Data processing pipelines with distinct steps
Specialization
When different parts of a project require specialized focus or domain knowledge:
- Tasks requiring different research methodologies
- Projects with varying technical requirements
- Content for different audience segments
- Cross-functional workflows
Parallel Processing
For tasks that can be executed simultaneously rather than sequentially:
- Independent research on multiple topics
- Processing multiple data sources
- Generating various content pieces
- Running multiple independent analyses
Focused Execution
When quality is enhanced by having dedicated focus on specific subtasks:
- Detailed technical documentation
- In-depth literature reviews
- Meticulous data analysis
- Complex code implementation
Using the Orchestrator/Agent Structure
You can explicitly request BB to use the Orchestrator/Agent approach for complex tasks:
Requesting Delegation
You can explicitly ask BB to use sub-agents in your request:
"Please use sub-agents to analyze these three research papers and identify the key methodological differences between them."
"This is a complex project with multiple components. Could you use the Orchestrator/Agent structure to break this down efficiently?"
BB-Initiated Delegation
BB may also proactively suggest using sub-agents when it recognizes a task would benefit from delegation:
"This task involves processing several large files and would benefit from parallel processing. Would you like me to use sub-agents to handle this more efficiently?"
How It Works
The Orchestrator/Agent structure follows a systematic workflow:
- Task Analysis: The orchestrator analyzes the overall task to identify components that can be delegated
- Task Definition: Each sub-task is precisely defined with clear instructions, requirements, and resources
- Sub-agent Creation: Specialized sub-agents are created to handle each task, with appropriate capabilities
- Parallel Execution: Sub-agents work on their assigned tasks simultaneously or in sequence as needed
- Result Collection: The orchestrator collects the results from each sub-agent as tasks are completed
- Integration: The orchestrator synthesizes the individual results into a cohesive whole
- Final Output: The combined results are presented back to you in a unified format
Task Definition Components
Each delegated task includes essential information:
- Title: A clear, descriptive name for the task
- Background: Contextual information needed for the task
- Instructions: Detailed directions for what to accomplish
- Requirements: Specifications for the expected output format and quality
- Resources: Files, URLs, or other materials needed to complete the task
- Capabilities: Special abilities or tools needed by the sub-agent
Example Use Cases
Research Example
Main Task: Literature Review on Climate Change Adaptation
Delegation approach:
- Sub-agent 1: Review papers on urban planning approaches
- Sub-agent 2: Analyze studies on agricultural adaptations
- Sub-agent 3: Examine economic impact assessments
- Orchestrator: Synthesize findings into a comprehensive review
Content Creation Example
Main Task: Multi-Channel Content Strategy
Delegation approach:
- Sub-agent 1: Develop detailed blog content plan
- Sub-agent 2: Create social media content calendar
- Sub-agent 3: Outline email newsletter series
- Sub-agent 4: Develop video content scripts
- Orchestrator: Ensure consistent messaging and optimal timing
Data Analysis Example
Main Task: Customer Behavior Analysis
Delegation approach:
- Sub-agent 1: Analyze purchase history patterns
- Sub-agent 2: Process customer demographic data
- Sub-agent 3: Review customer support interactions
- Sub-agent 4: Examine website behavior analytics
- Orchestrator: Create integrated customer behavior profiles
Development Example
Main Task: System Integration Project
Delegation approach:
- Sub-agent 1: Design API architecture
- Sub-agent 2: Develop data transformation layer
- Sub-agent 3: Create authentication system
- Sub-agent 4: Build error handling and logging
- Orchestrator: Ensure components work together cohesively
Best Practices
To get the most out of the Orchestrator/Agent structure:
- Clear Task Boundaries: Define sub-tasks with minimal dependencies between them
- Comprehensive Background: Provide each sub-agent with all necessary context
- Specific Instructions: Be precise about what each sub-agent should accomplish
- Output Requirements: Specify exactly what format and content the results should contain
- Resource Management: Ensure relevant files and URLs are included with each task
- Synchronization: Choose between synchronous (sequential) and asynchronous (parallel) execution based on task dependencies