llm-d-infra

Generate weekly nightly E2E trend report with failure patterns and success rates

Analyze nightly E2E job failures across llm-d repos. Smart router to sub-skills based on failure type.

Diagnose deployment failures in nightly E2E runs (helmfile, helm, install.sh)

Diagnose GPU contention failures in nightly E2E runs

Diagnose infrastructure and authentication failures in nightly E2E runs (GCP, AWS, runner issues)

Diagnose pod readiness timeout failures in nightly E2E runs

Full root cause analysis for nightly E2E failures that don't fit simple categories

Diagnose E2E test failures in nightly runs (e2e-validate.sh, make test-e2e)