Vercel

Vercel is the default provider when hosting is not Cloudflare. Set it explicitly when you want local builds to match Vercel deployment output.

import { defineConfig } from 'vite'
import { hubWorkflow } from '@vitehub/workflow/vite'

export default defineConfig({
  plugins: [hubWorkflow()],
  workflow: {
    provider: 'vercel',
  },
})

import { defineNitroConfig } from 'nitro/config'

export default defineNitroConfig({
  modules: ['@vitehub/workflow/nitro'],
  workflow: {
    provider: 'vercel',
  },
})

Generated output

Vite builds emit Vercel Build Output with a server function that installs the generated workflow registry before app code runs.

The application keeps the same calls:

await runWorkflow('welcome', payload)
await getWorkflowRun('welcome', run.id)
deferWorkflow('welcome', payload)

The generated workflow name uses the same stable encoding as Cloudflare:

workflow--77656c636f6d65

Provider resolution

Provider resolution matches the other ViteHub packages:

1. workflow.provider wins when set.
2. Cloudflare hosting selects cloudflare.
3. All other hosting targets select vercel.

This means Vercel can be the local default:

workflow: {}

Use an explicit provider when test output should be deterministic:

workflow: {
  provider: 'vercel',
}

Status behavior

Vercel runs return normalized metadata from the generated runtime:

{
  "id": "wrun_lvn4hx4f_x8k2p9s1",
  "provider": "vercel",
  "status": "queued",
  "metadata": {
    "workflow": "workflow--77656c636f6d65"
  }
}

getWorkflowRun() can report completed, failed, or unknown for runs visible to the current deployment process.

Deferred starts

deferWorkflow() uses the current request waitUntil() hook when Vercel exposes one through the runtime adapter:

export default defineEventHandler(() => {
  deferWorkflow('welcome', {
    email: 'ava@example.com',
  })

  return { ok: true }
})

The route returns immediately while the generated runtime starts the workflow asynchronously.