LLM native returns

[newzhiMing]

Hi！Does SpringAI dynamically support users to directly get the native information returned by the large model? For example, if I add a switch or set a feature into it, it will put back a large model and encapsulate the information directly returned by the class, and now this ChatRseponse does not feel like the openai paradigm？

thanks！

[markpollack]

I don't understand the question, we map back the response data from openai .

[sunyuhan1998]

I don't understand the question, we map back the response data from openai .

I guess what he means is wanting to directly obtain the raw JSON returned by the LLM, rather than the objects currently mapped by Spring AI.

I recall that a while back there were some issues discussing this topic—specifically around observability in Spring AI. Later on, Spring AI did introduce some enhanced observability capabilities, but those seem to be limited to retrieving metadata information (such as the number of tokens used). As of now, there still doesn't appear to be an elegant way to access the LLM's raw JSON response. The only available approach seems to be creating a custom WebClientCustomizer to intercept the results returned by the RestClient.

[newzhiMing]

@sunyuhan1998 Yes，That's what I mean, and I'm going to try what you said！