Decoupling AI Agent Execution from HTTP Requests

Decoupling AI agent execution from HTTP requests lets tasks outlive client connections, preserve intermediate work, and stream progress via subscriptions.