急语箱MCP服务器 - 智能场景与人物识别服务

基于MCP协议的沟通场景深度分析服务

🎯 项目定位

急语箱MCP服务器是一个专注于沟通场景深度分析的智能服务，

3. 第一个API调用

# 综合分析沟通场景
curl -X POST http://localhost:8000/api/v1/analyze \
  -H "Content-Type: application/json" \
  -H "X-API-Key: your_api_key" \
  -d '{
    "text": "领导批评我业绩不好，说需要改进",
    "analysis_options": {
      "include_participants": true,
      "include_relationships": true,
      "include_attitudes": true,
      "include_decision_factors": true
    }
  }'

返回示例:

{
  "analysis_id": "analysis_20260316_103000",
  "scenario": {
    "type": "workplace",
    "confidence": 0.98,
    "uncertain": false
  },
  "participants": [
    {
      "id": "person_1",
      "role": "领导",
      "position": "发送方",
      "confidence": 0.95
    },
    {
      "id": "person_2",
      "role": "下属/我",
      "position": "接收方",
      "confidence": 0.90
    }
  ],
  "relationships": [
    {
      "type": "上下级",
      "power_balance": "领导主导",
      "confidence": 0.95
    }
  ],
  "attitudes": [
    {
      "person_id": "person_1",
      "emotion": "不满",
      "stance": "批评",
      "tone": "严肃",
      "confidence": 0.90
    }
  ],
  "overall_confidence": 0.94,
  "uncertain_items": []
}

🔔 实时事件推送（SSE）

服务器支持 Server-Sent Events (SSE) 实时事件推送，可以订阅服务器事件流，实时接收场景识别结果和系统通知。

快速体验

// JavaScript 客户端
const eventSource = new EventSource('http://localhost:8000/events');

// 监听场景识别完成事件
eventSource.addEventListener('scenario.recognized', (e) => {
    const data = JSON.parse(e.data);
    console.log('场景:', data.scenario);
    console.log('置信度:', data.confidence);
    console.log('文本:', data.text);
});

// 监听系统通知
eventSource.addEventListener('system.notification', (e) => {
    const data = JSON.parse(e.data);
    console.log('通知:', data.title, '-', data.message);
});

主要端点

测试SSE

# 测试SSE连接
curl -X POST http://localhost:8000/events/test

# 触发场景识别事件
curl -X POST http://localhost:8000/events/scenario/recognized \
  -H "Content-Type: application/json" \
  -d '{
    "scenario": "workplace",
    "confidence": 0.95,
    "text": "老板让我加班到很晚"
  }'

# 发送系统通知
curl -X POST http://localhost:8000/events/notify \
  -H "Content-Type: application/json" \
  -d '{
    "level": "info",
    "title": "系统通知",
    "message": "场景识别服务已启动"
  }'

详细文档

查看 SSE_USAGE.md 获取完整的SSE使用文档，包括：

• 所有事件类型说明
• 客户端集成示例（JavaScript、Python）
• 使用场景和最佳实践
• 故障排除指南

📡 MCP工具

工具1: analyze_communication_scenario

综合分析沟通场景（推荐使用）

POST /mcp/tools/analyze_communication_scenario
{
  "text": "领导批评我业绩不好，说需要改进",
  "image_data": "base64_image...",  // 可选
  "analysis_options": {
    "include_participants": true,
    "include_relationships": true,
    "include_attitudes": true,
    "include_decision_factors": true
  }
}

工具2: identify_scenario

单独识别场景类型

POST /mcp/tools/identify_scenario
{
  "text": "对象生气了，怎么哄？"
}

工具3: identify_participants

识别参与人和角色

POST /mcp/tools/identify_participants
{
  "text": "妈妈说太晚回家不安全",
  "scenario_hint": "family"
}

工具4: analyze_relationships

分析人物关系

POST /mcp/tools/analyze_relationships
{
  "text": "领导找我和小王谈项目",
  "participants": [
    {"id": "p1", "role": "领导"},
    {"id": "p2", "role": "我"},
    {"id": "p3", "role": "小王"}
  ]
}

工具5: analyze_attitudes

分析态度和情绪

POST /mcp/tools/analyze_attitudes
{
  "text": "老板说这次做得不错，继续努力",
  "focus_person": "p1"
}

工具6: extract_decision_factors

提取决策要素

POST /mcp/tools/extract_decision_factors
{
  "text": "客户说价格太高，需要考虑一下",
  "scenario": "workplace"
}