TAG ARCHIVE
action-router
5 MARIA OS blog articles tagged action-router, organized as a Bonginkan topic archive for search engines and LLM retrieval.
Judgment OS / Decision Intelligence OS
Core MARIA OS research on turning organizational judgment into executable decision systems.
Agentic Company Architecture
Research on human-agent organizations, delegation boundaries, role topology, and governed autonomy.
Responsibility Gates and AI Governance
Safety, accountability, fail-closed gates, auditability, and human-in-the-loop control for AI agents.
Multi-Agent Mathematics
Formal models for convergence, stability, game theory, graph dynamics, and multi-agent evaluation.
Evidence, RAG, and Knowledge Governance
Evidence bundles, retrieval architecture, Graph RAG, knowledge trust, and auditable reasoning pipelines.
Agentic R&D and Judgment Science
Research operations, simulation labs, judgment science, recursive improvement, and experimental AI governance.
Action Router Intelligence Theory: Why Routing Must Control Actions, Not Classify Words
From keyword detection to action-level control: a formal shift that recasts AI routing from text classification to governance-aware execution control
Traditional AI routers treat routing as text classification: extract keywords, map to categories, and dispatch handlers. For enterprise-grade agentic systems, this approach is often insufficient. We formalize the Action Router as a function R: (Context × Intent × State) → Action, replacing the naive R: Input → Category mapping. The Action Router integrates with the MARIA OS Gate Engine so responsibility is enforced at routing time, not retrofitted afterward. We formalize the action space, define precondition-effect semantics for routable actions, derive routing cost over feasible actions, and show in simulation that action-level routing reduces misrouting by 67%, cuts responsibility-attribution failures by 94%, and achieves 3.2x lower latency than semantic-similarity routing on enterprise decision workloads.
The Complete Action Router: From Theory to Implementation to Scaling in MARIA OS
End-to-end architecture of the three-layer Action Router stack (Intent Parser, Action Resolver, Gate Controller), with recursive optimization and scaling patterns for 100+ agent deployments
The Action Router Intelligence Theory established that routing must control actions, not classify words. This paper presents the full implementation architecture: a three-layer stack of Intent Parser (context-aware goal extraction), Action Resolver (state-dependent action selection with precondition-effect semantics), and Gate Controller (risk-tiered execution envelopes integrated with MARIA OS governance). We detail a recursive optimization loop in which routing policies learn from execution outcomes, formalized as an online convex optimization problem with O(√T) regret. We then present a scaling architecture for 100+ concurrent agents using coordinate-based sharding, hierarchical action caches, and zone-local resolution. Integration with the MARIA OS Decision Pipeline state machine is formalized as a product automaton. Production benchmarks show sub-30ms P99 latency at 10,000 routing decisions per second, with first-attempt accuracy improving from 93.4% to 97.8% after 30 days of recursive learning.
Action Router × Gate Engine Composition: Formal Theory of Responsibility-Aware Routing
How action routing and gate control compose into a provably safe routing system where each routed action carries complete responsibility provenance
Enterprise AI systems face a core tension: routers must maximize throughput and decision quality, while gate engines must enforce safety constraints and responsibility boundaries. When these subsystems are implemented independently and stacked in sequence, interface failures emerge: routed actions can satisfy routing criteria but violate gate invariants, and gate rules can block optimal routes without considering alternatives. This paper presents a formal composition theory that unifies Gate operator G and Router operator R into a composite operator G ∘ R that preserves safety invariants by construction. We prove a Safety Preservation Theorem showing the composed system maintains gate invariants while maximizing routing quality inside the feasible safety envelope. Using Lagrangian optimization, we derive the constrained-optimal routing policy and show a 31.4% routing-quality improvement over sequential stacking, with zero safety violations across 18 production MARIA OS deployments (1,247 agents, 180 days).
Recursive Adaptation in Action Routing: How MARIA OS Routes Learn from Execution Outcomes
How self-improving routing uses recursive execution feedback to converge toward high-quality policies while preserving Lyapunov stability guarantees
Static action routing — where rules are configured once and applied uniformly — is inadequate for enterprise AI governance. Agent capabilities evolve, workloads shift, and routing quality depends on context that is only observed after execution. This paper introduces a recursive adaptation framework for MARIA OS action routing in which execution outcomes update routing parameters through a formal learning rule. We define θ_{t+1} = θ_t + η∇J(θ_t), where J(θ) is expected routing quality and gradients are estimated from outcome signals. We prove convergence under standard stochastic-approximation assumptions and establish Lyapunov stability guarantees, showing the adaptation process remains bounded while converging toward locally optimal routing policies. Thompson sampling provides principled exploration, and a multi-agent coordination protocol prevents oscillatory conflicts under concurrent adaptation. The quantitative figures in this article should be read as replay and simulation outputs over 14 operating contexts, not as audited production metrics of the current shipping router.
Sentence-Level Streaming VUI Architecture: From Cognitive Theory to Production Implementation in MARIA OS
How sentence-boundary detection, sequential TTS chaining, and rolling conversation summaries create a natural-feeling voice interface with long-session stability
Voice user interfaces face a core tradeoff: stream tokens immediately for low latency, or wait for larger semantic units to improve naturalness. MARIA OS resolves this with sentence-level streaming: detect sentence boundaries from Gemini token streams in real time, queue each sentence for sequential ElevenLabs TTS playback, and coordinate full-duplex interaction through barge-in control, speech debouncing, and heartbeat-based recovery. This paper presents the cognitive basis for sentence-level granularity, the production `useGeminiLive` architecture, a 29-tool action router across 4 teams with confidence-weighted team inference, and the rolling-summary mechanism for long voice sessions. In 2,400+ production sessions, the system achieved sub-800ms first-sentence latency with zero sentence-ordering violations, including compatibility handling for 9 in-app browser environments.