Request-scoped cache.
Aborts when the client goes away or the request times out.
Runs another executor, or a model id, to completion.
Executor value or model id to run.
Program to send.
The response program.
Runs another executor, or a model id, and streams its chunks.
Executor value or model id to run.
Program to send.
The response chunks.
The runtime services available while a request is being served.
Executors and transforms never talk to providers directly: they call RuntimeContext.invoke, which lets the runtime handle routing, auditing and cancellation.