coverify

What problem does it solve?

Semantic commitment can drift or be leaked when signals are transformed; CoVerify provides a lightweight instrument to extract and compare the commitment kernel.

Core Features & Use Cases

• Kernel extraction: Pulls the irreducible commitment kernel from a text signal.
• Comparison & scoring: Uses Jaccard similarity to quantify preservation and detect leakage.
• Ghost-token analysis: Identifies cascade-risk leakage patterns and cross-model structural holes.
• Audit-ready: Supports an input-output ledger and cross-model comparisons for governance validation.
• Use Case: Validate model swaps and transformation pipelines where preserving intent matters.

Quick Start

Run commitment_verify.py extract "Agents must always verify lineage before executing." to extract the commitment kernel from a sample signal.