Datasette was designed for Vercel from the start. It's a really great way of turning data into a read-only API, with both JSON and GraphQL endpoints.

https://datasette.io/

Demo here: https://datasette.simonwillison.net/simonwillisonblog

GraphQL demo here: https://datasette.simonwillison.net/graphql?query=%7B%0A%20%...

To deploy a SQLite database to Vercel with a GraphQL API:

    brew install datasette
    datasette install datasette-publish-vercel
    datasette publish vercel mydatabase.db \
      --project my-new-vercel-project-name \
      --install datasette-graphql

Here's a tutorial on how to get data into that SQLite database in the first place: https://datasette.io/tutorials/clean-data

You might like what we're building with Seafowl [0], which is an open-source, single binary database you can deploy to the edge for executing cache-friendly queries (so e.g. you can deploy to fly.io free tier, load your data into it, and put Cloudflare in front of it to cache query results like any other HTTP response).

Here's an example [1] multi-page site with Next.js and Seafowl.

[0] https://seafowl.io/docs/getting-started/introduction

[1] https://github.com/splitgraph/madatdata/tree/main/examples/r...

Devil's advocate: why can't you deploy Dockerfile of roapi to Vercel (or fly.io)?

What is the defacto "hobbyist tinker-er" standard for "I want to deploy this Docker thing into the cloud and play with it on the Internet"? There's got to be some one-click deploy solution where it goes to a Github repo, picks up a Helm chart, deploys it to a k8s cluster or something?

You can deploy a docker container to heroku, digitalocean app platform etc. no helm config required.

I think fly.io is intending to be a good choice for that use case.

I say 'intending' because I've not tried it rather than to imply any informed opinion about where they are in the process of achieving said goal.