@neondatabase/functions

Runtime helpers for Neon Functions. Currently provides a waitUntil primitive for deferring background work past a response.

Install

npm install @neondatabase/functions

Usage

The API mirrors @vercel/functions: import waitUntil and call it directly with the promise you want to keep alive.

import { waitUntil } from "@neondatabase/functions/v1";

export default {
	async fetch(req: Request): Promise<Response> {
		// Fire-and-forget background work that should outlive the response.
		waitUntil(logRequest(req));
		return new Response("ok");
	},
};

waitUntil(promise) forwards the promise to the Neon Functions runtime, which keeps the invocation alive until the promise settles (up to the 15-minute waitUntil limit). When no invocation context is in scope — local dev, tests, or any non-Neon host — it is a no-op: the promise is accepted and ignored, so the same code runs everywhere without branching. Passing a non-Promise throws a TypeError.

Runtime integration

The runtime carries the per-invocation context in an AsyncLocalStorage and publishes an accessor at globalThis[NEON_FUNCTIONS_CONTEXT] (a Symbol.for("@neondatabase/functions/request-context")) shaped like the providers used by Vercel and Next.js — a stable object exposing get().

To make waitUntil resolve to a given invocation, wrap the handler with runWithRequestContext. This is intended for the Neon Functions runtime; application code should not need it.

import { runWithRequestContext } from "@neondatabase/functions/v1";

runWithRequestContext({ waitUntil: realWaitUntil }, () => handler(req));

Because the context lives in AsyncLocalStorage, concurrent invocations in the same isolate each see their own context and never clobber one another.