Label Prometheus metrics by workflow name

[cnicolov]

Summary

• Adds a backend aggregation for workflow run counts grouped by workflow name and status.
• Emits workflow_name alongside status on openworkflow_workflow_runs Prometheus samples.
• Updates dashboard/backend tests and Prometheus docs for the new label.

Verification

• npx npm@11.8.0 run format
• npm_config_engine_strict=false npx npm@11.8.0 run build
• npm_config_engine_strict=false npx npm@11.8.0 run typecheck
• npm_config_engine_strict=false npx npm@11.8.0 run lint
• npm_config_engine_strict=false npx npm@11.8.0 run knip
• npm_config_engine_strict=false npx npm@11.8.0 run lint:duplication
• npm_config_engine_strict=false npx npm@11.8.0 run test:coverage

Notes

• npm_config_engine_strict=false was needed locally because cspell declares Node >=22.18.0 while this machine has Node 22.16.0.
• npm run lint:spell could not run locally for the same Node version reason.

[jamescmartinez]

The query itself looks performant and supported by the existing indexes, but adding workflow_name as a Prom label creates a cardinality risk. The Prom metrics are already pretty heavy as-is, and the full counts we're doing for the current metrics are causing problems in prod.

I'm going to do some more tests on a bigger deployment this week to see how this works in prod.

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/openworkflowdev/openworkflow/@openworkflow/cli@517

npm i https://pkg.pr.new/openworkflowdev/openworkflow/@openworkflow/dashboard@517

npm i https://pkg.pr.new/openworkflowdev/openworkflow@517

commit: 1f853cc

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[cnicolov]

The query itself looks performant and supported by the existing indexes, but adding workflow_name as a Prom label creates a cardinality risk. The Prom metrics are already pretty heavy as-is, and the full counts we're doing for the current metrics are causing problems in prod.

I'm going to do some more tests on a bigger deployment this week to see how this works in prod.

Hey @jamescmartinez what do you think to make it configurable and off by default?