OMOS Runtime
Health, manifest, provider discovery, Council orchestration, governed run records, human disposition, and persistence status are exposed through the shared runtime API.
Run one difficult question through governed input, Layer 1 distillation, alignment, multi-model Council Review, synthesis, human authorization, a persistent Decision Record, and retrievable history.
Health, manifest, provider discovery, Council orchestration, governed run records, human disposition, and persistence status are exposed through the shared runtime API.
OMOS identifies the objective, constraints, ambiguity, evidence and quarantined instructions before model orchestration begins.
Test Layer 1 →Truth, clarity, coherence, dignity, constructive unity, evidence support, verifiability and execution readiness remain visible as separate signals.
Run Alignment →Independent outputs are followed by cross-review. Agreement, contradiction, missing evidence and novel insight are preserved separately from factual verification.
Run Council →OMOS prepares a structured result without erasing uncertainty or dissent. Human approval or rejection remains required before final record disposition.
Start Run →The record retains normalized input, provider outputs, reviews, synthesis state, verification state, hashes, timestamps, persistence metadata, and human disposition.
The human gate is enforced inside the Record transition. Durable restart-safe persistence depends on the production PostgreSQL backend reporting durable=true.
Recent server run records can be reopened through the authenticated Council APIs and directly inside Ask OMOS.
GET /api/v1/council/runs
The visible product rail is intentionally simpler than the internal governance machinery. Human approval/rejection remains mandatory inside the Record transition.