{
  "schemaVersion": "1.0",
  "updates": [],
  "companyNews": [],
  "id": "b701",
  "language": "zh",
  "name": "Opik",
  "originalName": "Opik",
  "library": "tools",
  "category": "模型开发",
  "categoryKey": "模型开发",
  "categoryKeys": [
    "模型开发"
  ],
  "categories": [
    "模型开发"
  ],
  "entryPoints": [],
  "summary": "记录多步智能体调用，构建评估数据和实验，并以线上评分与 CI 检查跟踪应用质量变化。",
  "introduction": "Opik 将模型调用、对话和工具步骤组成追踪树，帮助查看智能体如何得到结果。数据集与实验可比较输出，评价指标涵盖 RAG 等应用任务，也可加入人工反馈。Agent Optimizer 用于迭代提示和智能体配置，在线规则及 PyTest 集成把评估接入运行监控和提交检查。模型评分仍需结合人工核查与任务样本解释。",
  "capabilities": [],
  "company": {
    "name": "Comet",
    "originalName": "Comet",
    "country": {
      "codes": [
        "US"
      ],
      "names": [
        "美国"
      ],
      "display": "美国",
      "basis": "headquarters",
      "basisLabel": "公司总部"
    }
  },
  "officialUrl": "https://www.comet.com/site/products/opik/",
  "profileUpdatedOn": "2026-10-20",
  "sources": [
    {
      "kind": "product",
      "url": "https://www.comet.com/site/products/opik/"
    },
    {
      "kind": "documentation",
      "url": "https://www.comet.com/docs/opik/quickstart/"
    }
  ],
  "news": [],
  "links": {
    "interactive": "products/opik--b701.html",
    "reading": "read/zh/products/opik--b701.html",
    "markdown": "read/zh/products/opik--b701.md",
    "json": "read/zh/products/opik--b701.json"
  }
}