Is Engine required?

[jedrichards]

Is it possible to use these tools without using the Engine service at all? I can successfully run commands on my client project with a valid ENGINE_API_KEY, but as soon as I remove it I run into problems.

Trying to run apollo client:check without an api key env var, against a live server on a url with the following config:

module.exports = {
  client: {
    service: {
      url: 'http://...',
    },
    includes: ['src/**/*.graphql'],
  },
}

Results in an error: No service found to link to Engine. If I add an arbitrary name,

module.exports = {
  client: {
    service: {
      name: 'foo',
      url: 'http://...',
    },
    includes: ['src/**/*.graphql'],
  },
}

I get an error: Unable to find ENGINE_API_KEY.

Using a localSchemaFile results in no joy either.

[JakeDawkins]

Hi @jedrichards! For future reference, the apollo spectrum community would be the place to go to get questions like this answered more quickly (since there are more than a few people watching). We'd like to keep these issues reserved for bug reports or feature requests.

But to answer your question here, No, Engine is not required to do client:check 😄

The instructions for setting up a client to use a local endpoint or schema file are here. Make sure you're up to date with the CLI and that there isn't a .env file (even empty) or a key in your env or else the CLI may parse the env for the key instead of using the local endpoint.

I'll close this since I can't reproduce locally, but if it turns out to be a bug, I can reopen :)

[jedrichards]

@JakeDawkins Thanks! I believe this to be a bug then. I've created a minimal repro repository here: https://github.com/jedrichards/apollo-tooling-1075 (private repo, sent you an invite to collaborate).

[JakeDawkins]

@jedrichards oops! Sorry, I was confused. Yes engine is required for the client:check command. The reason is that operation validation isn't something done locally by the language server. It's handled on the server-side of engine.

The CLI fires a mutation containing the client operations and those operations are validated against the most recently published schema. There's no easy way to make this a client-run command, as it would require a lot of duplicated work all across the board. So I don't believe this will be a feature than can be supported without Engine anytime soon.

[jedrichards]

@JakeDawkins Ah ok, thanks for checking and updating the messaging 👍