Skip to content
Kitbay
Skill

ci-metrics

NOASSERTIONUpdated on Nov 11, 2026105,390GitHub starsIncludes scripts

A contributor skill in the PyTorch repository for answering questions about PyTorch's CI: job durations, failures, queue times, workflow trends and runner health.

What the skill teaches

A contributor skill in the PyTorch repository for answering questions about PyTorch's CI: job durations, failures, queue times, workflow trends and runner health. It tries PyTorch's HUD tests API first, passing the user's GitHub token, and falls back to ClickHouse queries for test and webhook data or Prometheus queries for infrastructure metrics, all through a bundled wrapper that downloads a pinned, checksum-checked Grafana CLI and signs in with the GitHub CLI session. It sets rules for which databases and tables to use and points to PyTorch's test-infra and ci-infra repositories for saved queries and metric names. Access needs write permission on pytorch/pytorch or a place on HUD's allow list.

Capabilities

  • Query PyTorch CI test results through the HUD API
  • Run ClickHouse queries on CI and webhook data
  • Query Prometheus for runner and queue metrics