invokeRemote throws Error: Resolver has no definition for 'undefined'

I am trying to call invokeRemote from a Custom UI module, which keeps throwing.

The error from the rejected promise in the Custom UI looks as follows:

Error: There was an error invoking the function - Resolver has no definition for 'undefined'.

There is also an error in the app log console that looks like this:

Error: Resolver has no definition for 'undefined'.
at Resolver.getFunction (webpack://my-app/node_modules/@forge/resolver/out/index.js:44:1)
at Object.resolve (webpack://my-app/node_modules/@forge/resolver/out/index.js:52:1)
at r.<computed> [as handler] (/var/task/index.cjs:2:139447)
at Runtime.handleOnceNonStreaming (file:///var/runtime/index.mjs:1173:29)

The fact that the @forge/resolver package is looking for a resolver named undefined does not seem right. I cannot tell if this is a user or library error.

I tried to update my @forge/resolver package to the latest version but that did not help.

The request never hits my remote server.

My manifest contains these configurations (module details omitted):

app:
  id: ari:cloud:ecosystem::app/123
  runtime:
    name: nodejs18.x
  connect:
    key: my.app.key
    remote: connect
    authentication: jwt
remotes:
  - key: connect
    baseUrl: https://my-remote.example.com
modules:
  endpoint:
    - key: connect-endpoint
      remote: connect
  jira:adminPage:
    - key: my-admin-page
      resource: r-NeVU2u
      title: My admin page
      layout: blank
      resolver:
        function: my-function-key
  # UI modules, permissions, etc.

Happy to share more details and appreciate if anyone has some input. Thanks!

Hi @tbinna, I am not too sure where the @forge/resolver package comes into play here, that package is to help with native Forge functions. invokeRemote lives in the @forge/bridge package, are you able to provide a code snippet?

I found the issue. The problem is in my app manifest.yml where the resolver property of the UI module has to point to an endpoint and not to a function:

app:
  id: ari:cloud:ecosystem::app/123
  runtime:
    name: nodejs18.x
  connect:
    key: my.app.key
    remote: connect
    authentication: jwt
remotes:
  - key: connect
    baseUrl: https://my-remote.example.com
modules:
  endpoint:
    - key: connect-endpoint
      remote: connect
  jira:adminPage:
    - key: my-admin-page
      resource: r-NeVU2u
      title: My admin page
      layout: blank
      resolver:
        endpoint: connect-endpoint
  # other UI modules, permissions, etc.

Ahh, that would do it. While we are on the subject of debugging, we also emit some very basic logs for remote invocations that can be viewed through the developer console (if you haven’t come across it already). It might come in handy for debugging.

1 Like

@BoZhang, is my understanding correct that once a module opts into resolving from an endpoint, the same module cannot also call invoke to call into a Forge backend (i.e., call a resolver function)?

Yeah, unfortunately, you can’t mix and match. However, there is a feature coming soon that will allow you to call out to your remote from a Forge resolver (“Forge function to remote” under this board). It’s not as good as being able to configure your module to call both directly, but it will still allow a module to use Forge resolver functions and remotes.

1 Like

Thanks! It is unfortunate that we cannot mix, but it sounds like the “Forge function to remote” feature will solve that problem. That may be quite a good solution because the Custom UI can go back to a regular resolver and let the Forge function handle the interaction with the remote—transparently to the front end.

1 Like

Hey @tbinna it is an enhancement we’re thinking about for future. I’ve created a FRGE ticket to track it.

3 Likes