prysm-testing-quality

Evaluating LangGraph outputs against a quality matrix involves running Prysm tests from the repository root to compare pipeline outputs against docs/quality_matrix.md, capturing a six-dimension score and composite accuracy metrics.

Tracking router and counselor accuracy over time is done by executing pytest tests and evaluating LangGraph outputs against the quality_matrix documentation to verify calibration and measure expected behavior alignment.

Yes, you need pytest installed, access to docs/quality_matrix.md, and a defined PipelineState to run Prysm tests and perform offline evaluation of your LangGraph pipeline components.

The quality matrix scoring system works by evaluating pipeline outputs against documentation, capturing six distinct dimensions and calculating a composite score to verify router, counselor, and triage calibration.

Yes, you can use Prysm tests for offline evaluation of triage calibration by executing pytest from the repository root and validating test outcomes against the defined PipelineState and quality_matrix documentation.

Prysm tests may fail during quality matrix validation if PipelineState is undefined, pytest is missing, or docs/quality_matrix.md is inaccessible, preventing accurate six-dimension scoring and accuracy tracking.