Problem

For some instances it is important to use different API keys from different Azure environments for different clients. Right now an instance can only use 1 for all clients.

Solution

Similar to the DeepL solution, multiple API keys for Azure should be possible on client level. For language resources shared between clients, this is only possible if the clients use the system API key.

In the resource creation window when Azure is selected, then we introduce

1. a mandatory selection "use system Azure config/API key" or "use client level Azure config/API key"
2. the next selection must be the client. If system key was selected, list all clients and allow multi selection, when client key was selected, list all clients with client overwrite and only one client selection is allowed
3. after selection of client/clients we know from the settings if we need the model dropdown or the text field where to add the exact deployment name (according to rights for Azure)

Only if system level API key is used the resource may have multiple clients, if client level API key is used, it will be 1 resource : 1 client. The client for client overwrite keys cannot be changed afterwards. Client(s) for system key can be added/deleted.

P.S. It was decided to make client overridable keys for all AI resources (OpenAI,Azure,Translate5AI), not only for Azure

Testing

Create a client and create a config overrides for him for all 3 type of AI resources.
Create a language resource with the system level configs and check it uses system level (pretranslate with this resource)
Create a language resource with the client level configs and check it uses client level configs (pretranslate with this resource)
Try different combinations

The easiest way to check if client configs are used is to reset system level to empty values or to fake values that if used will trigger an error and then pretranslate with the client level language resource and vice versa.