How to use different env variable name for API Key in openai-compatible endpoint?

[awagner-mainz]

Dear all,

I want to set up a workflow comparing different LLMs for the spancat/ner task(s). Besides the current OpenAI GPT-4o, I also have access to models hosted by my infrastructure provider via a "OpenAI compatible" API. I have tried this with insomnium and it works, as long as I provide my infrastructure provider's API key in the Bearer Token HTTP Request Header.

I have my OpenAI API key in an env variable $OPENAI_API_KEY and my provider's API key in another env variable $PROV_API_KEY. How can I make spacy-llm use the latter env variable to populate the Bearer Token Header?

I suppose I cannot use something like

[components.llm.model]
@llm_models = "spacy.GPT-4.v3"
name = "meta-llama-3-70b-instruct"
endpoint = "https://my.provider.ai/v1/chat/completions"
key = ${os.environ['PROV_API_KEY']}

, can I?

Can I somehow temporarily shuffle/rename my env variables in an automated way from the spacy setup?

TIA