sarcascoder/parakh — extraction eval gate

Field-level OCR/extraction accuracy evaluation. Fails the PR if accuracy regresses past a threshold.

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Stale
License
None
namedescriptionrequireddefault
configPath to parakh config YAML (relative to the workspace root).yeseval/config.yaml
modelModel adapter to use for this run (must be defined in your parakh config or available as a built-in adapter).nodefault
thresholdMinimum overall field-level accuracy (0.0–1.0) required to pass the gate.no0.90
reading-order-thresholdMinimum reading-order accuracy (0.0–1.0). Optional. Set to 0 to disable.no0.85
goldenPath to the golden dataset directory (overrides what is set in the config).no""
python-versionPython version used to install and run parakh.no3.11
upload-reportIf true, upload the parakh HTML report as a GitHub artifact (default: true).notrue
fail-on-regressionIf true, exit non-zero when accuracy is below threshold (default: true).notrue
namedescription
accuracyOverall field-level accuracy achieved on this run.
reading-orderReading-order accuracy on this run.
failed-fieldsNumber of fields below the per-field threshold.
report-pathPath to the HTML report produced by parakh.