Webhook

EP152 Async AI APIs Make Webhooks Reliable for Long-Running Jobs Welcome back to AI DevTools, the Crazy Router Podcast.

Some AI tasks take longer than an HTTP request should remain open.

Video generation, document processing, large batch jobs, and multi-step workflows need a durable asynchronous interface.

Webhooks can notify a client when work finishes, but only when delivery, authentication, duplication, and state transitions are designed deliberately.

Today we will build a reliable webhook workflow for long-running AI jobs.

Start with a durable job state.

So, I described briefly what they are, what we do there, and how to start with them.

And also, these buildpacks are available on Docker Hub, so you can browse our Webhooks repository on Docker Hub and read more about them.

And choose the hardened images if you're there, if you're still thinking about build packs or Docker files, right? So you can choose something.

And yeah, I guess that's it.

Fourteen, use webhooks instead of constant polling.

Polling asks Faucet Pay every few seconds.

Webhooks invert the relationship.

Faucet Pay tells your server when something happens.

The current V2 API supports payout events including and signs webhook deliveries using HMAC SHA-256 in the header.

Instead, only after cryptographic verification should the event be trusted.

Sixteen, make webhook processing idempotent.

Webhooks can be delivered more than once.

GitHub restores services after nearly an eight-hour outage, disrupting actions of APIs, PRs, and Copilot.

GitHub has restored services after nearly eight-hour outage disrupted several of its core development tools, including actions, pull requests, APIs, Git operations, webhooks, and Copilot, impacting software development workflows across its platform.

This incident has been resolved.

Thank you for your patience and understanding as we address the issue the company wrote on its status page.

The disruption was first reported at 1.40 p.m. on August 17th, with GitHub initially flagging degraded performance across the platform.

Within minutes, the disruption had spread to API requests, actions, webhooks, issues, and pull requests or PRs.

Microsoft

finally patches critical one-click copilot vulnerability almost eight weeks after learning of it.

Among the comments we can read, this is HWDR, human-written, did read, and every single JIRA ticket is a novel worth of useless text.

GitHub's status page led this story with a broad outage that disrupted Git operations, API requests, pull requests, issues, actions, pages, webhooks, and some co-pilot authentication before the incident was marked resolved with a root cause analysis promised later.

The hacker news reaction in this thread was basically procedural because the post was quickly closed as a duplicate instead of becoming a real discussion.

In the comments, the only visible activity was redirecting readers to the main thread rather than debating the outage itself.