When I run prisma deploy, the api endpoints are automatically generated. WHen I deploy to a server (zeit.now) they are not. What do I need to do to have those endpoints generated when deploying to now?

what do you mean with api endpoints are generated?

The prisma endpoints for my data models

It's unexpected that the endpoint is not updated after deployment. Are you sure you are deploying the changes to the same endpoint?

For example see the difference in what queries are available in the palyground

The second screenshot you sent is not a Prisma endpoint.

This is the boilerplate I used for it.. https://github.com/graphql-boilerplates/node-graphql-server/tree/master/advanced

Ok. Do you understand why this doesn't work after what I just wrote?

Kind of. I’m a little confused about what the difference is between the prisma endpoint and my graphql server.

You can read this to get a better understanding here: https://www.prisma.io/docs/understand-prisma/prisma-introduction-what-why-how-j9ff/

So say I wanted to use the prisma endpoint in production. How do I do that? Do I need to deploy a graphql server?

see this as well: https://prisma.slack.com/archives/CA491RJH0/p1536282891000100?thread_ts=1536282378.000100&cid=CA491RJH0

Okay thanks @patrickdevivo. I don’t know why this has been so difficult to wrap my head around but I appreciate the help! So basically the graphql server’s primary responsibility is to communicate with the database, and also can handle some other logic like user authentication tokens for example. But there’s also the prisma endpoint which is created when I run prisma deploy. And that is what I will use to actually communicate with the database from the front end. Am I understanding correctly?

it depends on your use case…but im going to guess in most cases you dont want your front-end directly accessing prisma/your db. if this were the case, it would mean that anyone could run arbitrary commands in your database without proper (application level) checks. most real deployments are going to require another graphql server in between your prisma and your front end. this let’s you build your application logic with prisma (see

prisma-bindings

)

So are you saying that in most production scenarios, you still need to write out all the resolvers?

it does…do you want all those CRUD operations exposed to your frontend with no authentication? if so, you understand that that means that anyone can make a request to your prisma to delete or modify anything they want

Okay gotcha, no I don’t want that. Is that where the authorization token comes in to play?

yes, but again, probably not from your frontend. the token is generated using the secret you specify in your prisma config. you should use

prisma-binding

in a new graphql server you set-up (pass in the secret) and write the resolvers you need for your application using calls to prisma

Okay. Also does prisma client change any of this? I was reading this (https://www.prisma.io/blog/prisma-client-beta-ahph4o1umail/) it looks like prisma client accomplishes what I’m looking to do? (Not have to write a resolver function for every CRUD operation for every type in my datamodel)

i mean…you still have to define resolvers for your application level graphql calls, even in that example: https://github.com/prisma/prisma-examples/blob/master/node-graphql-schema-delegation/src/resolvers/Mutation.js#L2

Okay thanks for your help