arifOS — Constitutional AI Governance

A constitutional AI governance layer that treats every LLM interaction as a stateful, auditable session. Exposes 12 tools across a 13-floor verification pipeline: init_anchor bootstraps sessions with entropy baselines and Gödel locks, agi_mind runs three-phase Ollama reasoning (fast/reflect/decide) with constitutional prefixes, apex_soul validates against sovereignty rules, and vault_ledger maintains a Merkle-chained immutable log. The physics_reality tool grounds queries in real-world data to prevent hallucination, while math_estimator tracks thermodynamic health metrics. Human-in-loop by design means every decision returns uncertainty ranges and requires explicit verdict confirmation. Reach for this when you need provable AI governance with cryptographic auditability rather than just prompt engineering.