Results and exports

Read a run's terminal status and coverage before comparing its score:

evalrouter status RUN_ID --json
evalrouter results RUN_ID --json

The result records the evaluation identity, completed and missing work, native scores, execution errors and billing state. Sampled coverage is not a full-benchmark result. A failed or partial run must remain visible in comparisons; do not report it as a completed evaluation.

Save a versioned report

Use the result's version when you need a repeatable export:

evalrouter export RUN_ID --version RESULT_VERSION --format json --output result.json
evalrouter export RUN_ID --version RESULT_VERSION --format csv --output result.csv
evalrouter export RUN_ID --version RESULT_VERSION --format html --output result.html

Replace RESULT_VERSION with the returned integer. Exports require an explicit destination and refuse overwrite by default. Keep the version and checksum with downstream reports. Export permissions and evidence availability remain subject to your workspace and the selected evaluation's rights.

A local download failure does not start another run. Retry the export for the same run and version. For an evaluation still in progress, use evalrouter wait RUN_ID --wait-timeout 3600 --json.