niceeval view --out <directory>, or explicit niceeval view --run <id> --out <directory>, selects published Runs.
The internal host forms a pure AnalysisSample, projects the facts declared by the Report, and exports one completed ReportExecution.
This is the export path for sharing results.
Select the scope before exporting
Without a locator or--run, the command selects every matching result by current-project identity.
--run selects named historical Runs.
Both forms create a Sample first.
A Sample preserves the selected slots’ denominator and included, not-recorded, core-invalid, and excluded states.
A Report reads only the RecordProjection it declares.
An unrequested bad Attachment does not affect the export.
A requested recorded-data problem appears in the built-in problems surface instead of disappearing from the denominator.
What prevents export
unavailable, migration-required, migration-unavailable, unsupported, invalid, and partial collection are recorded-data problems.
They can be exported, but pages must show the problem.
A projector or author-callback defect, an invalid semantic document, or a route conflict is an execution problem.
Any execution problem stops static export, so NiceEval never publishes a half-successful report.
What the export directory contains
A successful directory includes:- Every fixed page and PageFamily-expanded page.
- The host data each page needs.
- Already-produced downloads.
- The manifest and the exporter’s built-in runtime.
- A final
completemarker.
Generate it in CI
Run evals, then export in the same job:runIds and use view --run <runId> instead.
Publish the directory
Give./report-site to your hosting provider as the static-site directory.
Do not point the platform build at a source Record; it only needs to deploy the already generated directory.