Gemini webhooks for 1M-token contexts.

The timing problem

Long-context Gemini in a webhook handler

AnyHook ack-and-forwards. You run the Gemini call against your full corpus. No polling loop, no Redis queue, no Cloud Tasks.

import { verifyWebhook } from "anyhook-verify";
import { GoogleGenerativeAI } from "@google/generative-ai";

const gemini = new GoogleGenerativeAI(process.env.GEMINI_API_KEY!);

export async function POST(req: Request) {
  if (!(await verifyWebhook(req, process.env.ANYHOOK_SIGNING_SECRET!))) {
    return new Response("invalid", { status: 401 });
  }

  const event = await req.json();

  // Run RAG over the full corpus. 45 seconds. Stripe doesn't care.
  const model = gemini.getGenerativeModel({ model: "gemini-1.5-pro" });
  const result = await model.generateContent([
    fullCorpusFor(tenantOf(event)),
    `Process: ${JSON.stringify(event)}`,
  ]);

  await persistDecision(result.response.text());
  return new Response("ok", { status: 200 });
}

Questions teams ask

Does AnyHook support Gemini's webhook signature scheme?

Yes. Gemini signs with the Standard Webhooks scheme. Set the app's source to Standard Webhooks and paste the signing secret: AnyHook verifies each delivery at the edge before forwarding it.

What if my Gemini call hits a rate limit?

Return a 429 from your handler. AnyHook treats it as retriable, applies exponential backoff, and surfaces the retry chain in the dashboard. You can replay the original event once your quota resets.

Is this only for Gemini API webhooks, or also Vertex AI?

Both. Anything that sends HTTP webhooks can go through AnyHook. Signatures are checked at the edge for the schemes in the docs, Standard Webhooks included; a sender outside them runs as Generic, with its original headers forwarded for your own check.