Optimizing Vercel Functions With Upstash Workflow

This demo shows the cost-saving benefits of using Upstash Workflow for Vercel functions. It compares two methods of calling an image generation API:

- Method 1: Calling the API in a standard Vercel function
- Method 2: Calling the API using Upstash Workflow

Both methods start at the same time and take about the same time to finish. The key difference is the estimated cost per 1M requests (hover for details).

Prompt: A futuristic city skyline at dusk, with towering skyscrapers and flying vehicles in the style of retro sci-fi art. Colors should feature deep purples, bright neon pinks, and glowing electric blues.

Traditional Serverless Function

Total Duration0ms

Vercel Function Duration0ms

Cost for 1M Requests~$0

Vercel Function

import { NextRequest, NextResponse } from 'next/server'
import { ImageResponse } from 'utils/types'

export const POST = async (request: NextRequest) => {
  // get prompt from request
  const params = await request.json()
  const prompt = params.prompt as string

  // make the fetch request
  const response = await fetch(
    "https://api.ideogram.ai/generate",
    {
      method: "POST",
      body: JSON.stringify({
        image_request: {
          model: 'V_2',
          prompt,
          aspect_ratio: 'ASPECT_1_1',
          magic_prompt_option: 'AUTO',
        },
      }),
      headers: {
        'Content-Type': 'application/json',
        'Api-Key': process.env.IDEOGRAM_API_KEY!
      },
    }
  )

  // get the image url
  const payload = await response.json() as ImageResponse
  const url = payload.data[0].url

  return new NextResponse(
    JSON.stringify({ url }),
    { status: 200 }
  )
}

Serverless Function with Upstash Workflow

Total Duration0ms

Vercel Function Duration0ms

Cost for 1M Requests~$0

Workflow Function

import { serve } from "@upstash/qstash/nextjs"
import { Redis } from "@upstash/redis"
import { ImageResponse } from "utils/types"

const redis = Redis.fromEnv()

export const POST = serve<{ prompt: string }>(
  async (context) => {
    // get prompt from request
    const { prompt } = context.requestPayload

    // make the call to Idogram through QStash
    const result = await context.call<ImageResponse>(
      'call Ideogram',
      "https://api.ideogram.ai/generate",
      "POST",
      {
        image_request: {
          model: 'V_2',
          prompt,
          aspect_ratio: 'ASPECT_1_1',
          magic_prompt_option: 'AUTO',
        },
      },
      {
        'Content-Type': 'application/json',
        'Api-Key': process.env.IDEOGRAM_API_KEY!,
      },
    )

    // save the image url in redis
    // so that UI can access it
    await context.run(
      'save results in redis',
      async () => {
        await redis.set<string>(
          context.headers.get('callKey')!,
          result.data[0].url,
          { ex: 120 }, // expire in 120 seconds
        )
      }
    )
  }
)