Copilot 用量 API 为 AI 采用阶段新增审查轮次与审查耗时指标

Copilot 用量指标 API 现在为每个 AI 采用阶段报告两个额外的代码审查速度指标，扩展了企业和组织报告中可用的采用阶段群组字段。除了已有的按阶段统计的合并时间和合并次数之外，您现在可以比较不同采用群组中拉取请求的审查速度以及它们所经历的审查循环次数。

totals_by_ai_adoption_phase 分解中的每个阶段条目现在包含两个新字段：

• avg_pull_requests_minutes_to_review：从拉取请求创建到首次审查的中位时间（以分钟计）。
• avg_pull_requests_review_cycles：拉取请求在合并前收到的审查提交次数的中位数。

这两个指标均限定于已合并的拉取请求，并按每个拉取请求的合并日期进行归属，因此每个拉取请求仅被计数一次。仅被审查但从未合并的拉取请求不会计入任一指标。

审查延迟和审查循环次数是工程吞吐量的先行指标。通过按 AI 采用阶段分解，您可以了解 Copilot 采用程度更深的团队是否拉取请求审查速度更快、经历的审查轮次更少。这有助于量化 Copilot 对审查流程的下游影响，并帮助您将赋能工作聚焦于影响最显著之处。

• 这些指标会同时出现在企业和组织的 1 天与 28 天报告中。
• 由于仅限于已合并的拉取请求，审查延迟和循环次数反映的是实际落地的工作。
• 阶段定义和群组分配保持不变。详情请参阅采用阶段群组公告。

请访问 Copilot 用量指标 API 文档 开始使用。

The Copilot usage metrics API now reports two additional code-review velocity metrics for each AI adoption phase, extending the adoption phase cohorts fields available in the enterprise and organization reports. Alongside the existing per-phase merge time and merge counts, you can now compare how quickly pull requests are reviewed and how many review cycles they take across your adoption cohorts.

Each phase entry in the totals_by_ai_adoption_phase breakdown now includes two new fields:

• avg_pull_requests_minutes_to_review : The median time, in minutes, from when a pull request is created to its first review.
• avg_pull_requests_review_cycles : The median number of review submissions a pull request receives before it merges.

Both metrics are scoped to merged pull requests and attributed to each pull request’s merge day, so every pull request is counted exactly once. Pull requests that are reviewed but never merged do not contribute to either metric.

Review latency and review-cycle counts are leading indicators of engineering throughput. By breaking them out by AI adoption phase, you can see whether teams with deeper Copilot adoption get their pull requests reviewed faster and iterate through fewer review cycles. This helps quantify the downstream impact of Copilot on your review process and helps you target enablement where it moves the needle most.

• These metrics appear in both the enterprise and organization 1-day and 28-day reports.
• Because they are scoped to merged pull requests, review latency and cycle counts reflect work that actually landed.
• Phase definitions and cohort assignment are unchanged. For details, see the adoption phase cohorts announcement.

Visit the Copilot usage metrics API documentation to get started.