dispatch

To delegate tasks to background AI workers, you describe a task to dispatch and a specialized worker executes it using a plan-driven checklist. It orchestrates multi-model backends for delegation, tracking, and isolated work contexts.

Orchestrating multi-model AI workflows is best handled through config-driven model routing and dynamic worker spawning. This approach maintains isolated work contexts and tracks progress via a plan-based checklist under the .dispatch/tasks directory.

Inter-process communication for AI task planning uses a dedicated IPC protocol that allows workers to ask questions and enables dynamic reconfiguration. This ensures smooth coordination and progress reporting across multiple specialized workers.

Yes, you can use it for dynamic worker spawning and model routing. It supports config-driven routing to multi-model backends, applies frontmatter validation, and manages optional resource directories to reconfigure tasks dynamically.

No external dependencies are required to run worker orchestration tasks. The system operates independently by implementing its own frontmatter validation, plan creation, IPC protocol, and worker spawning logic without relying on external packages.

Dispatched AI workers maintain progress tracking through checklist-based plans created in the .dispatch/tasks directory. If progress is lost, it likely indicates a failure in the frontmatter validation or a disruption in the IPC protocol used for status reporting.