QaCockpit

DocumentationPipeline Intelligence

Pipeline Intelligence in QaCockpit for Azure DevOps

See where pipeline time goes, where queues add delay, and where test evidence is missing without separating the signal from its exact Azure run.

QaCockpit for Azure DevOps documentation

These guides cover the public Azure DevOps Marketplace extension, its current permissions, personal storage, evidence boundaries, and supported environment.

Looking for QaCockpit Cloud documentation?

01

Start with the available pipeline history

Choose all accessible projects or one project, then select 7, 14, or 30 days. Overview uses every accessible run queued in that window across all branches. It shows active definitions, outcomes, run-time median, report completeness, current operational attention, and daily activity without inventing a previous-period comparison.

Signals to review keep shared task failures, queue pressure, incomplete latest evidence, and inactive definitions separate. Each available action opens the affected pipeline, evidence view, or exact run.

Pipeline Intelligence Overview showing six KPI cards in one row, actionable cross-project signals, and daily run and queue trends.
Overview uses all accessible branches queued in the selected window and keeps outcomes, evidence, and operational attention separate.

02

Separate run time from queue wait and feedback time

Performance uses completed default-branch runs queued in the selected window. Its run count can therefore be lower than Overview. Daily medians keep run time and queue wait separate, while feedback time covers the interval from queue to finish.

Highest queue share and slowest pipeline medians show at most five facts and state the total when more are available. A longer run is a measured fact, not an automatic quality verdict. Failed and canceled runtime remains factual wall-clock time.

Pipeline Intelligence Performance showing five KPI cards in one row, daily run and queue trends, queue-share ranking, pipeline medians, and aggregate agent-pool facts.
Performance compares timing without hiding unavailable Analytics observations or treating them as zero.

03

Open the exact run behind a timing signal

Exact-run detail shows queue wait, run time, end-to-end feedback time, test-report state, anonymized worker count, peak concurrency, agent job time, and the available stage, job, and task timeline. Rankings stay bounded to five rows while stating the complete observed total.

The Azure link opens the exact build. Test evidence identifies complete, partial, missing, and unavailable reports independently of the pipeline result.

Pipeline Intelligence exact-run drawer for QC 08 Multi-stage Shards showing queue, run and feedback time, 312 tests, anonymized concurrency, and bounded stage, job, and task rankings.
The exact-run drawer keeps timing, 312 reported tests, concurrency, and the 62-record timeline connected to Azure run 225.

04

Unavailable data remains explicit

Pipeline Intelligence follows both the extension scopes and the signed-in user's native Azure permissions. Inaccessible projects and pools are not disclosed. If Azure Analytics does not return agent-request observations, observed queue p95 and agent-request wait show Not available instead of zero.

  • vso.code reads the exact run commit to locate one unique YAML task after you open an affected run.
  • vso.analytics reads bounded pipeline, task, and agent-request timing facts.
  • vso.agentpools reads pools, queues, agents, and recent jobs for aggregate and anonymized capacity views.

QaCockpit does not change source code, display worker identities or capabilities, or persist raw agent assignments. Existing installations require an Azure DevOps organization administrator to authorize these scopes before upgrading to 1.1.0.

Next: Permissions & data.