quarkus-langchain4j
quarkus-langchain4j copied to clipboard
[Question] How to get the WatsonxChatModel instance
I need to access the WatsonxChatModel class to use some specific functionality that watsonx.ai provides to the developer.
Specifically, I need to call a rest api that returns a prompt stored in watsonx.ai that I would like to use as input of an `AiService'. This would allow me to use a prompt that can be updated and tested by an AI engineer without touching the backend during the dev phase.
Initially my idea was to find a way to integrate this feature into the library with a specific annotation, but this functionality is something that would only be used by watsonx.ai and would require a lot of changes touching different parts of the core module, so the quickest way could be to inject ChatLanguageModel and use ClientProxy.unwrap to have access to the WatsonxChatModel instance where I will add the methods I need.
This change will allow me to avoid rewriting the rest client. Is this the right way to go? Or is it better to use another approach?