hidai25/EvalView - AI Agent Testing

Catch agent regressions in CI. Detects tool changes, output drift, cost spikes. Works with LangGraph, CrewAI, Claude, GPT.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Activelast commit Jul 3, 2026
License
Apache 2.0

Pinned Snippet

workflow.ymlSHA-pinned
uses: hidai25/eval-view@c996a89324800c0d68b6d9629dd46adb0637c91b # v0.8.0

tags can be moved; commit SHAs can't. why a SHA?

namedescriptionrequireddefault
modeTesting mode: "check" for regression detection (default), "run" for full evaluationnocheck
commandAlias for "mode" — accepts "check" or "run". If both are set, "command" takes precedence.no
openai-api-keyOpenAI API key for LLM-as-judge evaluationno
anthropic-api-keyAnthropic API key (optional alternative to OpenAI)no
test-pathPath to test cases directorynotests
config-pathPath to EvalView config fileno.evalview/config.yaml
filterFilter tests by name patternno
max-workersMaximum parallel workersno4
max-retriesMaximum retries for failed testsno2
fail-on-errorFail the workflow if any test fails or regressions detectednotrue
fail-onComma-separated statuses that fail CI (REGRESSION, TOOLS_CHANGED, OUTPUT_CHANGED, CONTRACT_DRIFT)noREGRESSION
strictFail on any change (REGRESSION, TOOLS_CHANGED, OUTPUT_CHANGED)nofalse
diffCompare against golden baselines in "run" mode (ignored in "check" mode which always diffs)nofalse
contractsCheck MCP contracts for interface drift before running testsnofalse
generate-reportGenerate HTML reportnotrue
post-commentAutomatically post results as PR comment (requires GITHUB_TOKEN)notrue
upload-artifactUpload results and report as workflow artifactsnotrue
evalview-versionEvalView version to install (e.g. "0.5.3", default: latest)no""
python-versionPython version to useno3.11
namedescription
results-filePath to JSON results file
report-filePath to HTML report file
total-testsTotal number of tests run
passed-testsNumber of passed/unchanged tests
failed-testsNumber of failed/regressed tests
regressionsNumber of regressions detected (check mode)
tools-changedNumber of tests with tool changes (check mode)
pass-ratePass rate percentage
statusOverall status: passed, regression, tools_changed, output_changed