Shining Forest v6.29: Model Versioning & Archival Protocol

Overview Shining Forest v6.29 provides a model versioning and archival protocol for computational chemistry ML pipelines. It addresses model serialization fragility across library versions (e.g., sklearn 1.8.0 → 1.2.2), architecture tracking, scientific provenance for deprecated models, and instant rollback via symlink management. Context & Rationale This protocol emerged from real-world incidents during the Sentinel Sovereign project where sklearn version incompatibilities and architecture mismatches caused silent failures in production pipelines. Key Components ADR-012 (Architectural Decision Record): Defines the versioning standards, including the bundled model format and "NEVER-DELETE" policy. Bundled Model Format: Enforces architecture metadata (e.g., config.architecture == 'composite_pbml') to prevent deployment of mismatched models. Comparision Tool (compare_model_versions.py): Automated generation of COMPARISON_REPORT.md to validate metric stability before promotion. Protocol Lineage v6.26: Cryptographic Provenance (DOI: 10.5281/zenodo.17716208) v6.27.1: Chemistry Data Calibration (DOI: 10.5281/zenodo.18084420) v6.28.1: ML Lifecycle Metrology (DOI: 10.5281/zenodo.18084246)