The cloud-native orchestrator for the whole development lifecycle, with integrated lineage and observability.

dagster

Hello! I was wondering if there was any way to create a configurable, third party API resource with the new Pythonic resources ? Doesn’t look so looking at the documentation, which is kind of a regression compared to the @resource. Thank you by advance.

this is the answer to your previous question (same question really though, right?) as well <https://docs.dagster.io/concepts/resources#managing-state-in-resources>

make the `setup_for_execution` function set your private attribute, which should be your client

like
```class BespokeResource(ConfigurableResource):
    config_value: str|None

    _client: YourClientType

    def setup_for_execution(self, context):
        if self.config_value:
            self._client = YourClient(config_value)
        else:
            self._client = YourClient()```

then other functions can either use the client directly, or you can have another function return the client directly

```def get_client(self):
    return self._client```
or
```def use_client(self, *args):
    self._client(*args)```

with an asset looking like
```@asset
def my_asset(bespoke_resource: BespokeResource):
    your_client = bespoke_resource.get_client()```

Indeed, seems like a good option, but this implies to change our entire code-base. I am try to do it directly with inheritance, hopefully I can make it work.

in that case, `IAttachDifferentObjectToOpContext` may be helpful - check out how the dagster_aws integration handles the s3 resource for both styles <https://github.com/dagster-io/dagster/blob/master/python_modules/libraries/dagster-aws/dagster_aws/s3/resources.py|here>

one returns the other, so the resource inside the op/asset is the client itself without an additional call

at least future code can be written to use the pythonic resources natively

Oh nice ! That was exactly what I needed, thanks!