Reporting

Qontinuum renders results for wherever your pipeline needs them — from a one-line shell summary to a full engineering report. All formats come from the qont report group and share the same versioned suite-result JSON as their source of truth.

Formats

Command	Output	Use
qont report junit	JUnit XML	Jenkins/GitLab/Buildkite render quantum checks natively
qont report badge	SVG badge	README / dashboards
qont report md	Markdown	docs, wikis, PR bodies
qont report pr	Markdown + cost table	the sticky PR comment, rendered locally
qont report json	Versioned suite JSON	machine consumption, archival
qont report html	Self-contained HTML dashboard	trends & observability
qont report summary	One greppable line	shell pipelines
qont report engineering	Full engineering report	the "where should this run, and why?" decision doc

Every format can read a prior run (--from suite.json) or run the suite fresh, and write to a file (--out) or stdout.

The engineering report

qont report engineering is the platform's decision document. It pulls the whole intelligence layer into one markdown file:

• Execution plan — the selected device with expected runtime, queue, cost, and fidelity, plus its risk level and confidence, the reasons it was chosen, risks, fallbacks, assumptions, and missing evidence.
• Ranked recommendations — the feasible devices under your chosen strategy.
• Provider analysis — per-device reliability from calibration, local history, and (if synced) community intelligence, with the evidence behind each.
• Cost analysis — estimated hardware cost per device.
• Historical context — pass rate, hardware success rate, and spend from your local run history.

$ qont report engineering --path . --strategy balanced --out report.md
$ qont report engineering --strategy value --budget 20

It reuses the exact engine behind qont recommend, qont plan, and qont health, so the report never disagrees with the individual commands. The renderer is pure (structured objects in, markdown out), so it is easy to test and to embed in CI.

Confidence and honesty

Reports never present a number without context. Fidelity is labelled as a per-shot survival estimate (not a measured value); queue time is blank when it is not modelled offline; and every plan carries a confidence score reflecting how much evidence backed it. Missing evidence is stated, not hidden.