Run yarn vercel to have a local development server at http://localhost:3001. The web app can then call the API at http://localhost:3000/api/pdf with the expected JSON body in a POST request.
Using a service called Browserless that will run chrome used for puppeteer in order to stay under the vercel function size limit of 50 mb.
Update: Now hosting my own version of Browserless on Railway using the browserless/chrome docker image. This is because the free tier of Browserless only allows 1000 requests per month, and I was hitting that limit somehow (I don't know how since my soccer league doesn't have many PDF download requests.). Railway has a free hosting plan so I can use that for free.