Files
nexus/wiki/sources/backend-architect-with-memory.md
2026-05-03 05:42:12 +08:00

61 lines
5.1 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
title: "Backend Architect with Memory"
type: source
tags: [the-agency, engineering, multi-agent, memory, mcp]
date: 2026-05-29
---
## Source File
- [[raw/Agent/agency-agents/integrations/mcp-memory/backend-architect-with-memory.md]]
## Summary用中文描述
- 核心主题:具备持久记忆能力的后端架构师 AI Agent 设计规范专注于可扩展系统设计、数据库架构、API 开发与云基础设施
- 问题域:如何构建一个能够跨会话保留架构决策、系统设计和技术约束记忆的 AI 后端架构专家
- 方法/机制:基于 MCP Memory 集成框架,在会话开始时检索相关上下文,完成交付物后主动记忆,跨 Agent 交接时自动传递上下文QA 失败时自动回滚到已知良好状态
- 结论/价值:解决多 Agent 协作中重复讨论已做决策、交接信息丢失的问题,通过 Rollback 机制实现安全的失败恢复
## Key Claims用中文描述
- 后端架构师应在会话启动时主动检索项目相关的历史架构决策,防止重复讨论
- 架构决策(选型数据库、定义 API 契约、选择通信模式)应以标签形式持久化,供未来会话和其他 Agent 查找
- 交付物Schema、API 规范、架构文档)完成后应主动记忆并标记接收方,确保下游 Agent 无需手动复制粘贴
- 遇到 QA 失败或错误决策时,应检索上一个已知良好状态并回滚,而非手动撤销变更链
- 标签一致性是记忆召回可靠工作的前提:每个记忆使用 Agent 名称和项目名称作为标签
## Key Quotes
> "When you start a session, recall relevant context from previous sessions. Search for memories tagged with 'backend-architect' and the current project name." — 会话启动时的记忆召回机制
> "When you make an architecture decision — choosing a database, defining an API contract, selecting a communication pattern — remember it with tags including 'backend-architect', the project name, and the topic." — 架构决策的持久化规范
> "When you complete a deliverable (a schema, an API spec, an architecture document), remember it tagged for the next agent in the workflow." — 交付物的下游 Agent 标记规范
> "When you receive a QA failure or need to recover from a bad decision, search for the last known-good state and roll back to it." — Rollback 回滚机制
## Key Concepts
- [[MicroservicesArchitecture]]:微服务架构,支持水平扩展和独立部署,是 Backend Architect 默认推荐的架构模式
- [[CQRS]]命令查询职责分离Backend Architect 在复杂领域驱动设计中的推荐数据模式
- [[EventSourcing]]:事件溯源,与 CQRS 配合用于复杂业务域的数据建模
- [[ServerlessArchitecture]]无服务器架构Backend Architect 认可的可自动扩展且成本效益高的部署方式
- [[DatabaseIndexing]]:数据库索引优化,用于实现子 100ms 平均查询性能
- [[CircuitBreaker]]断路器模式Backend Architect 实现系统可靠性和优雅降级的核心机制
- [[DefenseInDepth]]深度防御策略Security-First Architecture 的核心原则
- [[MemoryTagging]]:标签化记忆持久化策略——每个记忆使用 Agent 名称 + 项目名 + 主题标签(如 `database-schema`/`api-design`/`auth-strategy`)三重标签,确保 recall 可靠
- [[Rollback]]回滚机制——QA 失败时检索最近良好检查点并恢复到已知良好状态,而非手动撤销变更链
## Key Entities
- Backend Architect主 Agent专门负责系统架构和服务器端开发特点战略思维、安全导向、可扩展性优先、可靠性至上
- Frontend Developer下游接收方 Agent接收 Backend Architect 提供的 API 规范
- QA Agent质量保障 Agent在失败时触发记忆回滚机制
- Sprint Prioritizer上游输入方 Agent提供待开发功能的优先级排序
## Connections
- [[AgentsOrchestrator]] ← coordinates ← [[BackendArchitectWithMemory]]
- [[MCPBuilderAgent]] ← enables ← [[BackendArchitectWithMemory]]MCP Memory 集成)
- [[SprintPrioritizer]] → outputs sprint plan → [[BackendArchitectWithMemory]](通过 MCP Memory Server recall
- [[UXResearcher]] → outputs research brief → [[BackendArchitectWithMemory]](通过 MCP Memory Server recall
- [[BackendArchitectWithMemory]] → stores API spec → [[FrontendDeveloper]](通过 MCP Memory Server recall
- [[BackendArchitectWithMemory]] ← rollback ← [[TestingRealityChecker]]QA 失败触发)
## Contradictions
- 与 [[engineering-software-architect]] 的设计哲学存在微妙张力:
- 冲突点:架构决策记录方式
- 当前观点Backend Architect with Memory通过 MCP Memory 标签化快照持久化,支持跨 Agent 召回和自动回滚
- 对方观点Software Architect通过 ADRArchitecture Decision Record标准化文档记录侧重人类可读的权衡分析
- 协调方案两者互补——ADR 文档供人类阅读和团队共享MCP Memory 快照供 Agent 自动化召回Backend Architect 可同时使用两者作为不同场景的工具