Per-Call Options

The channel-level timeout applies to every call. When a few calls need a different timeout — or need to be cancelable — derive a proxy with withCallOptions() instead of changing the whole channel.

Overriding timeout

import { wrap, withCallOptions } from "kkrpc"

const api = wrap<RemoteAPI>(transport) // channel timeout, e.g. 30s
const quick = withCallOptions(api, { timeout: 2000 })

await api.normalCall() // uses the channel timeout
await quick.slowCall() // rejects after 2s with an RPCTimeoutError

The options apply to every call made through the returned proxy and any nested property proxies derived from it. The original proxy is unchanged. Pass timeout: 0 to disable the timeout for a particular proxy.

This replaces the pattern of setting one large channel-wide timeout to accommodate a few long-running methods: keep the channel default tight, and widen only the calls that need it.

Canceling a call

Pass an AbortSignal to cancel an in-flight call. The returned promise rejects with an AbortError, and if the signal is already aborted the call rejects immediately without sending anything.

const controller = new AbortController()
const cancelable = withCallOptions(api, { signal: controller.signal })

const promise = cancelable.longRunning()
controller.abort() // promise rejects with AbortError

// Combine with a timeout signal:
withCallOptions(api, { signal: AbortSignal.timeout(5000) })

Canceling stops the caller from waiting; it does not abort work already running on the remote side. Both kkrpc/streaming and kkrpc/remote-refs proxies support withCallOptions.