多 Agent 协作通信协议设计模板
为多 Agent 系统设计标准化的通信协议和消息格式，支持任务分发、状态同步、错误传播和结果聚合。
You are a distributed systems architect designing communication protocols for multi-agent AI systems. Create a complete protocol specification.
System Requirements
- Number of agent roles: [e.g., planner, researcher, coder, reviewer]
- Communication pattern: [hub-spoke / peer-to-peer / hierarchical / pub-sub]
- Execution model: [synchronous / asynchronous / event-driven]
- Failure tolerance: [at-least-once / exactly-once / best-effort]
Design the Protocol
1. Message Format
Define a standard message envelope with fields: msg_id (uuid), type (task_assign/status_update/result/error/heartbeat), from (agent_id), to (agent_id or broadcast), correlation_id, timestamp (ISO-8601), payload, metadata (priority 0-9, ttl_ms, retry_count).
2. Task Lifecycle
Define state machine: pending -> assigned -> in_progress -> review -> completed/failed. Specify transitions, required fields, timeout handling and escalation rules per state.
3. Error Propagation
- Error categories: transient (retry) vs permanent (escalate)
- Circuit breaker pattern for failing agents
- Graceful degradation when agents are unavailable
- Dead letter queue for unprocessable messages
4. Context Sharing
- Shared memory/blackboard pattern vs message passing trade-offs
- Context window management and compression strategies
- Conflict resolution when agents produce contradictory results
5. Implementation Skeleton
Provide Python code scaffolding for: Base agent class with protocol support, Message router/dispatcher, Simple orchestrator managing the task lifecycle.
Focus on practical, implementable designs rather than theoretical frameworks.