HTTP SSE API

端点列表

方法	路径	描述
GET	/api/v1/agents	列出所有已加载的 Agent
POST	/api/v1/chat/stream	流式对话（SSE）
POST	/api/v1/chat/resume	恢复中断的对话
GET	/api/v1/chat/interrupt_state	查询中断状态
GET	/api/v1/admin/evolution/stats	Evolution 引擎状态
GET	/api/v1/admin/evolution/genes	基因列表查询
GET	/api/v1/admin/evolution/federated	联邦搜索
GET	/metrics	Prometheus 指标

流式对话

请求

POST /api/v1/chat/stream
Content-Type: application/json

{
  "agent_id": "research-agent",
  "session_id": "session-123",
  "message": "介绍一下 Eino 框架"
}

响应（Legacy 模式）

data: Eino
data:  是一个
data:  基于
data:  Go
data:  的 LLM 框架
data: [DONE]

响应（A2UI 模式）

添加 ?a2ui=true 或 X-A2UI: true 头：

event: text
data: {"type":"text","timestamp":1234567890,"data":{"delta":"Eino "}}

event: tool_call
data: {"type":"tool_call","timestamp":1234567891,"data":{"id":"tc1","name":"web_search","status":"calling"}}

event: tool_result
data: {"type":"tool_result","timestamp":1234567892,"data":{"id":"tc1","name":"web_search","result":"...","is_error":false}}

event: text
data: {"type":"text","timestamp":1234567893,"data":{"delta":"根据搜索结果..."}}

event: done
data: {"type":"done","timestamp":1234567900,"data":null}

恢复中断

POST /api/v1/chat/resume
Content-Type: application/json

{
  "agent_id": "approval-agent",
  "session_id": "session-123",
  "input": {
    "confirm": true,
    "reason": "用户已确认执行"
  }
}

列出 Agent

GET /api/v1/agents

{
  "agents": [
    {"name": "research-agent", "description": "研究助手"},
    {"name": "code-review-agent", "description": "代码审查"}
  ]
}