unshdee/proofrag

Evaluate a RAG/LLM app and gate CI on answer quality + retrieval regressions.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Stale
License
None
namedescriptionrequireddefault
goldensetPath to the golden set JSONL.yes
predictionsPath to predictions JSONL ({id, answer, retrieved_contexts}) from your RAG.yes
resultsWhere to write results.json.results.json
scorecardWhere to write the HTML scorecard.scorecard.html
fail-underFail if overall generation score drops below this (0-1). Empty = no absolute gate.""
baselineBaseline results.json to diff against. Empty = no regression check.""
toleranceAllowed metric drop vs baseline before it counts as a regression.0.02
kCutoff k for retrieval metrics (Recall@k, NDCG@k, ...).5
semanticUse embedding cosine for chunk relevance (needs the openai extra + key).false
extraproofrag extra to install: anthropic or openai.anthropic
versionPin a proofrag version (e.g. 0.3.0). Empty = latest.""
summaryWrite a markdown score summary to the GitHub Actions job summary.true
upload-artifactUpload the results JSON and HTML scorecard as a workflow artifact.true
artifact-nameName for the uploaded scorecard/results artifact.proofrag-scorecard
namedescription
resultsPath to the written results.json.
scorecardPath to the written HTML scorecard.