vitepress-template

Appearance

Agent Communication and Coordination ​

**Referenced Files in This Document** - [[bi_chat/src/agents/base_agent.py]](file/bi-chat/bi-chat/src/agents/base-agent.py) - [[bi_chat/src/agents/bi_agent.py]](file/bi-chat/bi-chat/src/agents/bi-agent.py) - [[bi_chat/src/agents/data_agent.py]](file/bi-chat/bi-chat/src/agents/data-agent.py) - [[bi-chat/bi_chat/src/agents/manager_agent.py]](file/bi-chat/bi-chat/src/agents/manager-agent.py) - [[bi-chat/bi_chat/src/agents/conversation_manager.py]](file/bi-chat/bi-chat/src/agents/conversation-manager.py) - [[bi-chat/bi_chat/src/agents/model_wrapper.py]](file/bi-chat/bi-chat/src/agents/model-wrapper.py) - [[bi-chat/bi_chat/src/agents/prompts/data_analysis_prompts.py]](file/bi-chat/bi-chat/src/agents/prompts/data-analysis-prompts.py) - [[_bmad/_config/agent-manifest.csv]](file/bmad/config/agent-manifest.csv) - [[_bmad/bmb/workflows/agent/data/agent-compilation.md]](file/bmad/bmb/workflows/agent/data/agent-compilation.md) - [[_bmad/bmb/workflows/agent/data/agent-menu-patterns.md]](file/bmad/bmb/workflows/agent/data/agent-menu-patterns.md) - [[_bmad/bmb/workflows/agent/data/agent-metadata.md]](file/bmad/bmb/workflows/agent/data/agent-metadata.md) - [[_bmad/bmb/workflows/agent/data/expert-agent-architecture.md]](file/bmad/bmb/workflows/agent/data/expert-agent-architecture.md) - [[_bmad/bmb/workflows/agent/data/module-agent-validation.md]](file/bmad/bmb/workflows/agent/data/module-agent-validation.md) - [[_bmad/bmb/workflows/agent/data/simple-agent-architecture.md]](file/bmad/bmb/workflows/agent/data/simple-agent-architecture.md) - [[_bmad/bmb/workflows/agent/data/simple-agent-validation.md]](file/bmad/bmb/workflows/agent/data/simple-agent-validation.md) - [[_bmad/bmb/workflows/agent/data/understanding-agent-types.md]](file/bmad/bmb/workflows/agent/data/understanding-agent-types.md) - [[_bmad/bmb/workflows/agent/templates/agent-plan.template.md]](file/bmad/bmb/workflows/agent/templates/agent-plan.template.md) - [[_bmad/bmb/workflows/agent/templates/expert-agent-template/expert-agent.template.md]](file/bmad/bmb/workflows/agent/templates/expert-agent-template/expert-agent.template.md) - [[_bmad/bmb/workflows/agent/templates/simple-agent.template.md]](file/bmad/bmb/workflows/agent/templates/simple-agent.template.md) - [[bi-chat/bi_chat/src/agents/tool_registry.py]](file/bi-chat/bi-chat/src/agents/tool-registry.py) - [[bi-chat/bi_chat/src/agents/tool_executor.py]](file/bi-chat/bi-chat/src/agents/tool-executor.py) - [[bi-chat/bi_chat/src/agents/tool_base.py]](file/bi-chat/bi-chat/src/agents/tool-base.py) - [[bi-chat/bi_chat/src/agents/complex_task_handler.py]](file/bi-chat/bi-chat/src/agents/complex-task-handler.py) - [[bi-chat/bi_chat/src/agents/clarification_guard.py]](file/bi-chat/bi-chat/src/agents/clarification-guard.py) - [[bi-chat/bi_chat/src/agents/report_agent.py]](file/bi-chat/bi-chat/src/agents/report-agent.py) - [[bi-chat/bi_chat/src/agents/summary_agent.py]](file/bi-chat/bi-chat/src/agents/summary-agent.py) - [[bi-chat/bi_chat/src/agents/knowledge_agent.py]](file/bi-chat/bi-chat/src/agents/knowledge-agent.py) - [[bi-chat/bi_chat/src/agents/calculation_agent.py]](file/bi-chat/bi-chat/src/agents/calculation-agent.py) - [[bi-chat/bi_chat/src/agents/general_agent.py]](file/bi-chat/bi-chat/src/agents/general-agent.py) - [[bi-chat/bi_chat/src/agents/optimization_agent.py]](file/bi-chat/bi-chat/src/agents/optimization-agent.py)

Table of Contents ​

  1. Introduction
  2. Project Structure
  3. Core Components
  4. Architecture Overview
  5. Detailed Component Analysis
  6. Dependency Analysis
  7. Performance Considerations
  8. Troubleshooting Guide
  9. Conclusion
  10. Appendices

Introduction ​

This document describes the agent communication patterns and coordination mechanisms in the multi-agent system. It focuses on:

  • BaseAgentHelper common functionality for message formatting, response parsing, and error handling
  • Tool registry system enabling dynamic tool injection and agent capability expansion
  • Inter-agent messaging protocols: request-response, broadcast, and collaborative workflows
  • Coordination strategies for concurrent execution and resource sharing
  • Examples of collaboration patterns, conflict resolution, and workflow branching
  • Integration with external tools and APIs via standardized interfaces

Project Structure ​

The multi-agent system spans two primary areas:

  • A Python-based chat agent framework under bi-chat with specialized agents and orchestration utilities
  • A B-MAD agent configuration and workflow system under _bmad for agent composition and lifecycle

Diagram sources

Section sources

Core Components ​

  • Base Agent: Provides foundational capabilities for message formatting, response parsing, and error handling
  • Manager Agent: Coordinates agent interactions, orchestrates workflows, and manages shared resources
  • Conversation Manager: Handles conversational context, turn-taking, and session persistence
  • Specialized Agents: Data, General, Knowledge, Report, Summary, Calculation, Optimization agents tailored for domain tasks
  • Tool Registry and Executor: Dynamic tool injection, capability expansion, and standardized API integration
  • B-MAD Agent Manifest and Workflows: Structured agent definitions, templates, and compilation/validation artifacts

Section sources

Architecture Overview ​

The system employs a layered architecture:

  • Orchestration Layer: Manager Agent and Conversation Manager coordinate agent activities
  • Execution Layer: Specialized agents perform domain-specific tasks
  • Tool Layer: Tool Registry and Executor enable dynamic capability expansion
  • Configuration Layer: B-MAD manifests and workflows define agent composition and lifecycle

Diagram sources

Detailed Component Analysis ​

Base Agent Helper ​

The Base Agent encapsulates common functionality for:

  • Message formatting: Ensures consistent input/output structures for downstream processing
  • Response parsing: Extracts structured data from model outputs
  • Error handling: Centralizes error propagation and recovery strategies

Diagram sources

Section sources

Tool Registry System ​

The tool registry enables dynamic tool injection and standardized integration:

  • Tool registration: Adds tools with metadata and capabilities
  • Tool resolution: Selects appropriate tools per task
  • Tool execution: Executes tools with validated parameters and captures results
  • Standardized interfaces: Ensures compatibility across external APIs

Diagram sources

Section sources

Inter-Agent Messaging Protocols ​

  • Request-Response: Manager Agent dispatches structured tasks; agents return formatted responses
  • Broadcast: Manager Agent can broadcast updates to specialized agents for awareness
  • Collaborative Workflows: Multi-step orchestration with intermediate context updates

Diagram sources

Section sources

Coordination Strategies for Concurrent Execution ​

  • Resource Sharing: Shared conversation context and tool registries
  • Conflict Resolution: Manager Agent resolves conflicting requests and prioritizes tasks
  • Workflow Branching: Decision points route tasks to appropriate agents based on intent and context

Diagram sources

Section sources

Agent Collaboration Patterns and Examples ​

  • Data Analysis Collaboration: Data Agent prepares insights; Report Agent composes reports; Summary Agent extracts key takeaways
  • Knowledge Retrieval: Knowledge Agent cross-references documents; General Agent synthesizes answers
  • Calculation and Optimization: Calculation Agent performs computations; Optimization Agent refines solutions iteratively

Diagram sources

Section sources

B-MAD Agent Configuration and Workflows ​

  • Agent Manifest: Defines agent capabilities, roles, and dependencies
  • Workflows and Templates: Provide structured patterns for agent creation, validation, and compilation
  • Validation Artifacts: Ensure agent correctness and consistency across environments

Diagram sources

Section sources

Dependency Analysis ​

The system exhibits clear separation of concerns:

  • Manager Agent depends on Conversation Manager, Tool Registry, and specialized agents
  • Tool Registry depends on Tool Base and Tool Executor
  • Specialized agents depend on Base Agent and optional tools
  • B-MAD configuration influences agent composition and capabilities

Diagram sources

Section sources

Performance Considerations ​

  • Asynchronous tool execution reduces latency for long-running operations
  • Caching of conversation context minimizes repeated computation
  • Batch tool execution consolidates network calls and improves throughput
  • Modular agent design allows selective activation to reduce overhead

Troubleshooting Guide ​

Common issues and resolutions:

  • Tool execution failures: Validate tool schema and parameters; check tool availability in registry
  • Conversation drift: Ensure Conversation Manager updates context consistently after each agent turn
  • Agent conflicts: Use Manager Agent conflict resolution to prioritize and sequence tasks
  • Parsing errors: Verify Base Agent response parsing logic aligns with model output format

Section sources

Conclusion ​

The multi-agent system integrates orchestration, specialized execution, and dynamic tooling through a cohesive architecture. The Base Agent provides foundational capabilities, the Manager Agent coordinates interactions, and the Tool Registry expands agent capabilities. B-MAD configuration ensures scalable composition and validation. Together, these components enable robust collaboration, conflict resolution, and efficient resource sharing across concurrent agent execution.

Appendices ​

  • Prompt templates for data analysis tasks are available for agent customization and reuse
  • Agent templates and workflows provide reusable patterns for rapid agent development

Section sources