whylabs-monitorSkill whylabs-monitor

WhyLabs监控技能是一个专注于机器学习(ML)模型生产环境可观测性的集成工具。它通过自动化的数据画像生成、实时异常检测和性能监控,帮助数据科学家和工程师监控模型漂移、保障数据质量、快速响应生产事故。核心功能包括数据画像记录、模型漂移检测、细分分析、约束验证和仪表板可视化,适用于金融风控、推荐系统、自动驾驶等AI应用场景。

机器学习 0 次安装 0 次浏览 更新于 2/23/2026

name: whylabs-monitor description: 用于机器学习可观测性、数据画像记录和生产环境异常检测的WhyLabs集成技能。 allowed-tools:

  • Read
  • Write
  • Bash
  • Glob
  • Grep

whylabs-monitor

概述

用于生产环境机器学习系统的可观测性、数据画像记录和异常检测的WhyLabs集成技能。

功能

  • 数据画像生成(whylogs)
  • 画像上传至WhyLabs平台
  • 异常检测与告警
  • 数据子集的细分分析
  • 性能监控仪表板
  • 与机器学习流水线集成
  • 历史画像对比
  • 自定义约束验证

目标流程

  • 模型性能监控与漂移检测
  • 机器学习系统可观测性与事件响应

工具与库

  • whylogs
  • WhyLabs平台
  • pandas

输入模式

{
  "type": "object",
  "required": ["action"],
  "properties": {
    "action": {
      "type": "string",
      "enum": ["profile", "upload", "compare", "validate", "alert-config"],
      "description": "要执行的WhyLabs操作"
    },
    "profileConfig": {
      "type": "object",
      "properties": {
        "dataPath": { "type": "string" },
        "datasetId": { "type": "string" },
        "segments": { "type": "array", "items": { "type": "string" } },
        "timestamp": { "type": "string" }
      }
    },
    "uploadConfig": {
      "type": "object",
      "properties": {
        "orgId": { "type": "string" },
        "modelId": { "type": "string" },
        "profilePath": { "type": "string" }
      }
    },
    "compareConfig": {
      "type": "object",
      "properties": {
        "baselineProfile": { "type": "string" },
        "targetProfile": { "type": "string" },
        "metrics": { "type": "array", "items": { "type": "string" } }
      }
    },
    "validationConfig": {
      "type": "object",
      "properties": {
        "constraints": {
          "type": "array",
          "items": {
            "type": "object",
            "properties": {
              "column": { "type": "string" },
              "constraint": { "type": "string" },
              "value": { "type": "number" }
            }
          }
        }
      }
    }
  }
}

输出模式

{
  "type": "object",
  "required": ["status", "action"],
  "properties": {
    "status": {
      "type": "string",
      "enum": ["success", "error", "warning"]
    },
    "action": {
      "type": "string"
    },
    "profilePath": {
      "type": "string"
    },
    "uploadId": {
      "type": "string"
    },
    "dashboardUrl": {
      "type": "string"
    },
    "comparison": {
      "type": "object",
      "properties": {
        "driftScore": { "type": "number" },
        "driftedFeatures": { "type": "array", "items": { "type": "string" } },
        "alerts": { "type": "array" }
      }
    },
    "validation": {
      "type": "object",
      "properties": {
        "passed": { "type": "boolean" },
        "failures": { "type": "array" }
      }
    }
  }
}

使用示例

{
  kind: 'skill',
  title: '分析并上传生产环境数据',
  skill: {
    name: 'whylabs-monitor',
    context: {
      action: 'profile',
      profileConfig: {
        dataPath: 'data/production_batch.parquet',
        datasetId: 'fraud-detection',
        segments: ['region', 'customer_type'],
        timestamp: '2024-01-15T00:00:00Z'
      },
      uploadConfig: {
        orgId: 'org-123',
        modelId: 'model-fraud-v2'
      }
    }
  }
}