orchestrate-platform-routing

You route AI tasks by generating a task-to-platform delegation plan that assigns each work slice to the best available host and model tier based on observed reachability, cost, latency, and privacy constraints.

Platform-aware delegation planning produces a slice-by-slice delegation map across primary lanes, local fallbacks, and cloud overflow, providing explicit reasons for each assignment and structured fallbacks before starting implementation, reasoning, review, QA, or release work.

The best way to manage agent delegation under cost and privacy constraints is to enforce host execution policy concepts like parallel-slice limits and context reservation rules, marking when exclusive-on-host access is needed for sensitive operations.

You resolve platform inventory mismatches by loading the registry using strict precedence order—attached live registry, then user config, then repo fallback—and explicitly calling out any disagreements found during the delegation planning process.

Yes, you need to load the platform inventory with the defined registry precedence to generate a delegation map, as it requires observed reachability and model inventory data to route implementation, review, QA, and release slices correctly.

Multi-step execution requires routing before starting work because each phase—implementation, reasoning, review, QA, and release—demands different host capabilities and model tiers, which must be mapped with fallbacks and verification outcomes to ensure operational success.