jarjarbeatyourattitude/Evalt performance gate

Find or verify the lowest-cost AI route that clears your frozen quality contract.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Stale
License
None
namedescriptionrequireddefault
suitePath to the versioned Evalt suite.noevalt.json
resultPath where the machine-readable Evalt result is written or read.noevalt-result.json
optimizeRun the budget-capped tournament before applying the gate. Set false to gate an existing result without provider calls.notrue
evalt-versionExact PyPI version to install. Mutable aliases such as latest are rejected.no0.9.5
python-versionPython version used by the action.no3.12
min-pass-rateRequired final-test pass rate from 0 through 1.no0.95
max-cost-per-successOptional maximum USD cost per successful production call.no""
require-complete-coverageFail when the tournament did not finish all decision-relevant coverage.notrue
fixed-promptDisable prompt rewrites and few-shot search while still comparing model routes.nofalse
max-parallel-modelsOptional override for parallel model lanes.no""
max-parallel-scenariosOptional override for parallel scenario executions per model.no""
request-timeout-secondsOptional wall-clock deadline for one provider response, not the tournament.no""
html-reportPath for the self-contained HTML evidence report.noevalt-report.html
junit-reportPath for the case-level JUnit report.noevalt-junit.xml
namedescription
statusPASS, FAIL, or ERROR.
selected-modelModel selected in the saved result, if one qualified.
final-test-pass-rateMeasured final-test pass rate.
cost-per-1000-successful-calls-usdMeasured production cost per 1,000 successful calls.
resultPath to the machine-readable result.
html-reportPath to the self-contained HTML report.
junit-reportPath to the case-level JUnit report.