38 lines
1.3 KiB
Markdown
38 lines
1.3 KiB
Markdown
---
|
||
title: "Chained Agents"
|
||
type: concept
|
||
tags: [multi-agent, workflow, automation]
|
||
sources: [content-factory]
|
||
last_updated: 2026-04-22
|
||
---
|
||
|
||
## Definition
|
||
|
||
Chained Agents(链式 Agent)是一种多 Agent 协作模式,其中上游 Agent 的输出直接作为下游 Agent 的输入,无需人工逐步干预。核心价值在于将复杂任务拆解为多个专业化子任务,通过流水线式编排实现全自动化执行。
|
||
|
||
## Key Characteristics
|
||
|
||
- **顺序依赖**:Agent A 输出 → Agent B 输入 → Agent B 输出 → Agent C 输入
|
||
- **无需人工中转**:每个环节由 Agent 自动触发和推进
|
||
- **可插拔**:可替换任意环节 Agent(如将本地模型替换为 API)
|
||
- **透明可查**:通过 Discord 频道等隔离机制,每个 Agent 的工作独立可审查
|
||
|
||
## Example: Content Factory
|
||
|
||
```
|
||
Research Agent (#research)
|
||
→ 输出:Top 5 内容机会(含来源)
|
||
↓
|
||
Writing Agent (#scripts)
|
||
→ 输出:完整脚本/推文串/Newsletter 草稿
|
||
↓
|
||
Thumbnail Agent (#thumbnails)
|
||
→ 输出:AI 生成的缩略图/封面
|
||
```
|
||
|
||
## Related Concepts
|
||
|
||
- [[Multi-Agent Coordination]]:多 Agent 协作的整体框架
|
||
- [[Workflow Architecture]]:工作流架构设计原则
|
||
- [[Content Automation]]:内容创作自动化
|