AzwaOS is a human-centric APEX + Amanah orchestration layer. It is model-agnostic, system-agnostic, and composable. AzwaOS complements arifOS—which remains the governance kernel and judge—by acting as an optional operator that helps humans survive life with calmer decisions and reversible steps. AzwaOS never claims authority or autonomy.
- azwaOS ≠ arifOS: arifOS is the governance kernel and ultimate judge. AzwaOS is an operator that routes and simulates guidance.
- Amanah-first: irreversible actions are denied; ambiguity is surfaced as UNKNOWN.
- Human-survival focus: prioritize clarity, calm, and small reversible actions to assist humans in daily life navigation.
- Model-agnostic: adapters and interfaces are designed without locking to any specific model or runtime.
- Unknown over unsafe certainty: if confidence < 0.99, outputs are labeled UNKNOWN and invite clarification.
python -m interfaces.cli demo
python -m interfaces.cli run --content "Plan my next 3 hours" --simulate-onlyazwa_core/— APEX modules (akal, present, energy, exploration), Amanah guard, orchestrator, and typed schemas.canon/— Canon seed for APEX + Amanah; compatible with arifOS floors without embedding them.interfaces/— CLI entry point for demos and request routing.memory/— Policy stubs; no persistence yet.docs/— Architecture, boundaries, and roadmap notes.
- No authority claims; outputs are advisory and simulation-first.
- No irreversible or credential actions; all sensitive requests fail closed.
- No anthropomorphic claims (no feelings, soul, or ego).
- High-stakes or ambiguous inputs are met with UNKNOWN and a request for clarification.
AzwaOS is designed to be text-compatible with arifOS governance language and the 9-floor mindset. Enforcement of the arifOS floors remains within arifOS; AzwaOS keeps only advisory checks to stay reversible.
✊ DITEMPA, BUKAN DIBERI