Joe Haddad

@consolelogwill @rauchg @CaseyGowrie You could leverage Queues to roll resumable streams, a few of our internal demos actually do this. Have you seen github.com/vercel/resumab…?

@rauchg @timer150 @CaseyGowrie If I just want resumable llm streams, is this overkill or the right thing to use?

Vercel Queues learns extensively from its predecessors and peer primitives in the cloud ecosystem. The brand is "Queues" but the implementation is durable event streaming. We settled on an API that's very simple (2 key methods!), and yet insanely powerful… You can use it as-is, but Queues also serve as a foundation for higher-level DX tools. For instance, Workflow (useworkflow.dev) on Vercel is implemented via Queues. For Python folk for example, Celery could be supported transparently with our serverless queues as its backend 🤤

@neoromantic @rauchg @CaseyGowrie We are going to continue investment into the Workflow SDK! cc @pranaygp

Tried to use workflows several times but it’s sadly every time felt unstable and full of issues, many of th reported. I suppose queues are simpler and absolutely stable by now. Any thoughts about workflows future? Is it more like a small side project or like something that you would invest?

@tobiastornros @rauchg @CaseyGowrie Yes! You can o11y in your project's o11y tab. We'll update the docs for this!

@rauchg @timer150 @CaseyGowrie I see the power in it. Do you offer any observability around it? I scanned docs but didn't find anything

@marclou @rauchg Let us know what we can improve! Looking forward to feedback 😁

@rauchg Qsatsh is giving me headaches Trying next week

Queues are one of the most requested services since I started Vercel. They're now here. It's just two APIs: 𝚜𝚎𝚗𝚍 and 𝚑𝚊𝚗𝚍𝚕𝚎𝙲𝚊𝚕𝚕𝚋𝚊𝚌𝚔 😌. The use-cases are basically infinite. Notably: queues can make agents and AI apps reliable. Quality and reliability are top of mind for everyone now, including our own team. We went through 3 iterations of the infrastructure while in private beta, and we're excited for you to build 'unbreakable software' with it.

@BenjDicken @rauchg It is hard for me to imagine maintaining Vercel’s DX for arbitrary compute where we do not have an event binding interface. What does this look like to you? 👀 Have you looked at Sandbox for your use case? Can you share the specific type of work you want to do in this VPS?

@timer150 @rauchg In my case, yes I genuinely want indefinite. More broadly, would just be amazing to have a vps with vercel dx.

What service should we build next, with deep care and investment into its security, availability, and durability?

Is indefinitely actually desirable here? For long running tasks, we typically see someone implement checkpointing so that a transient failure (e.g. upstream API) does not cause you to lose all progress. This can generalize to Queues or Workflow where a task might take 8 hours, but you carve it up into 10 minute chunks. I would love to learn more specifics!

@rauchg Server-side ability to run background tasks indefinitely. I dont think cron is a good fit - not enough timing granularity. Queues may have helped. But what I'd just love is a way to manage my own simple dedicated hosts/vms with the DX of vercel.

@k_firsov @cramforce @rauchg Mind sharing more details (here or in DM)?

Also bypassing Vercel WAF @cramforce @rauchg

A little morning research. New React/Next.js CVE-2025-55183 Source code leakage PoC Bypasses Cloudflare! @FearsOff #react #bypass #cloudflare

@maxleiter Do you have thoughts on srcml.org?

Formatting code should be unnecessary, and we solved it back in the 80s maxleiter.com/blog/formatting

@mxkaske Sent you a PR to replace your middleware with routes instead! This will guarantee your site is always fast github.com/mxkaske/mxkask…

resolved: middleware + hobby plan

just reduced the ping frequency from 1min to 10min for a static content on @vercel - is this normal?

@karthikkalyan90 very aligned- nothing’s worse than a ton of nits!

@timer150 - catch regressions - catch security vuln That’s all, nothing more or less. Anything more is too much noise. Anything less is not useful enough.

I would switch to a new AI code review tool tomorrow if it _______. We'll build the best ideas and give you early access!

@huozhi I can imagine many layers to this, for starters the bot should auto resolve comments it previously found and are now fixed. Automated approvals sound great if it can follow a repo policy (only for docs, bug fixes that do not introduce new features, etc)!

@timer150 Approves my PR when I resolve all their comments

@hp_arora 🎙️ we're waiting

@timer150 👀

@dobroslav_dev @vercel_changes @elysiaJS @anthonysheww think we can get this working? We have our bun builder that could be used automatically: github.com/vercel-communi…

@vercel_changes What about @elysiaJS ? 👀

You can now deploy Hono backends with zero-configuration. Hono apps deployed to Vercel benefit from Fluid compute, with Active CPU pricing, automatic cold start optimizations, background processing, and more. vercel.com/changelog/depl…

@preetsuthar17 @vercel what is the first feature you’d ship?

one day I wish to work at @vercel I got this

@ScriptedAlchemy @rauchg @kdy1dev @tomlienard Real Rust with Active CPU pricing. vercel.com/blog/introduci…

@rauchg @kdy1dev @tomlienard Should promote this more. Is it wasm rust or real rust?

What’s Vercel for backend? I need to deploy my Rust servers. Railway? Is there less costly alternatives?

@Tofugrass @vercel_changes All plans, including Hobby (free)

@vercel_changes Pro plan only?

Vercel’s WAF rules can now be managed directly in vercel.json. Both developers and agents can declaratively write and push rules that are versioned with code and tied to git commits. vercel.com/changelog/web-…

@WebReflection @rauchg @tomasz_ducin @WebReflection I think we're talking past one another - the receive handler in initial tweet handles the entire lifecycle of message handling. Without a callback, you'd need: const { message, ack, reject, extend } = await receive("topic", "consumer"); gist.github.com/Timer/d5952ec5…

@timer150 @rauchg @tomasz_ducin so you asked, I wrote this ... hopefully *something* is useful in there or at least explains why I didn't like the initial tweet or the blog post example 👋 gist.github.com/WebReflection/…

Queues. The most requested feature since I started Vercel 🥹 1️⃣ Send 𝚊𝚠𝚊𝚒𝚝 𝚜𝚎𝚗𝚍("𝚝𝚘𝚙𝚒𝚌", { 𝚖𝚎𝚜𝚜𝚊𝚐𝚎: "🔥" }); 2️⃣ Receive 𝚊𝚠𝚊𝚒𝚝 𝚛𝚎𝚌𝚎𝚒𝚟𝚎("𝚝𝚘𝚙𝚒𝚌", "𝚌𝚘𝚗𝚜𝚞𝚖𝚎𝚛", 𝚖 => { 𝚌𝚘𝚗𝚜𝚘𝚕𝚎.𝚕𝚘𝚐(𝚖.𝚖𝚎𝚜𝚜𝚊𝚐𝚎); // 𝙻𝚘𝚐𝚜 "🔥" }); Combine it with Active CPU pricing on Fluid, and you have a cost-efficient, performant, and now reliable backend system on Vercel.

@WebReflection @rauchg @tomasz_ducin send() resolves once the message is durably persisted to the queue, before processing starts.

@timer150 @rauchg @tomasz_ducin had a look here (not even sure if relevant) npmjs.com/package/vercel… but I still think something is off … is the send() resolving *after* the processing? That might explain the API choice to confine/trap completion through the attached callback reducing try/catch orchestration 👍

@rauchg @WebReflection @tomasz_ducin @WebReflection given this insight, how would you design the API to receive and process a single message? 👀

@rauchg @WebReflection @tomasz_ducin It's semantically equivalent to await new Promise((resolve, reject) => { .. }), where the handler is invoked immediately.