lukecarr/Litmus Test Runner

Run Litmus specification tests against LLM models and annotate failures inline.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Activelast commit Jun 5, 2026
License
MIT

Pinned Snippet

workflow.ymlSHA-pinned
uses: lukecarr/litmus@e59ae24a2a03cf61a60075400a9fe3e15b4468da # v0.3.3

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

namedescriptionrequireddefault
testsPath to the test cases JSON file.yes
schemaPath to the JSON schema file.yes
modelModel(s) to test against, one per line or comma-separated (e.g. openai/gpt-4o).yes
promptSystem prompt for the LLM. Mutually exclusive with prompt-file.no""
prompt-filePath to a file containing the system prompt.no""
parallelNumber of parallel requests per model.no1
outputOutput format: terminal, json, html, or github.nogithub
providerLLM provider: openrouter or cloudflare.noopenrouter
api-keyProvider API key. For openrouter this is the OpenRouter key; for cloudflare it is the downstream provider key. Passed to litmus through the environment, never on the command line. Leave empty to use a key already set in the job's environment.no""
cf-account-idCloudflare account ID (used with provider=cloudflare).no""
cf-gatewayCloudflare AI Gateway ID (used with provider=cloudflare).no""
cf-tokenCloudflare AI Gateway token for authenticated gateways.no""
versionLitmus release to download, e.g. v0.3.0. Defaults to the release matching the action ref you pinned (an exact vX.Y.Z tag), or the latest release when the ref is a branch, SHA, or major-version tag.no""
working-directoryDirectory to run litmus from.no.

no outputs