CascadeFlowInstall

Summary

Validates drafter AI output quality and cascades to verifier when below threshold, or routes hard/expert queries directly to verifier.

Inputs

• qualityThreshold — Minimum quality score (0-1) to accept drafter response. Lower = more cost savings, higher = better quality.
• useSemanticValidation — Whether to use ML-based semantic similarity validation for better quality detection.
• useAlignmentScoring — Whether to score query-response alignment for improved validation accuracy.
• useComplexityRouting — Whether to route queries directly to the verifier based on detected complexity.

Output shape

ai_languageModel

Returns a CascadeChatModel instance implementing drafter-first cascading with quality validation and optional verifier escalation. Response metadata includes cascadeflow flow tags (type, confidence, latency, cost, reason) for workflow visibility.