Hey, quick question, I just deployed a FastAPI as a separate k8s deployment that will be exposed to the outside and will be used to transform non-graphql requests to graphql requests to be used by Dagit. However, I'm having issues calling dagit from this new deployment, all the requests to execute pipelines return the following:

{
  "data": {
    "launchPipelineExecution": {
      "__typename": "PipelineNotFoundError"
    }
  }
}

However, this does not happen when using docker-compose. For docker-compose im just using the simple approach where dagster and dagit are in the same container. My pipelines repository location is defined as

pipelines_repository

and my workspace.yaml looks as follows:

load_from:
  - python_file:
      relative_path: app/pipelines_repository.py
      location_name: pipelines_repository

I believe that the problem is in my graphql with repositoryLocationName that somehow is different when using the kubernetes deployment once dagit, dagster daemon and user code is separated. Do I have to prepend anything to the

repositoryLocationName

in my graphql request to tell dagit where to look at? Edit: These are the parameters passed to my graphql query:

{
  "repositoryLocationName": "pipelines_repository",
  "repositoryName": "my_repo",
  "pipelineName": "my_pipeline",
  "runConfigData": {
    "solids": {
      "my_solid": {
        "inputs": {
          "company_name": "my_company"
        }
      }
    },
    "execution": {
      "multiprocess": {
        "config": {
          "max_concurrent": 6
        }
      }
    },
    "storage": {
      "filesystem": {}
    }
  },
  "mode": "default"
}

Glad you were able to sort it out - you’re right that the helm chart constructs a workspace.yaml file for you if you’re using user deployments, so it wouldn’t know to look at the one that you created manually.