I got the same result with the OpenAI module.

I have found that this issue is not related to Spring AI.

doOnEach handles multiple events.
In the provided code, it calls the onComplete event as the final step after all tasks are completed.
Since there is no data in the onComplete event, it results in null.

Using doOnNext instead of doOnEach will resolve the issue.

I have found that this issue is not related to Spring AI.

doOnEach handles multiple events. In the provided code, it calls the onComplete event as the final step after all tasks are completed. Since there is no data in the onComplete event, it results in null.

Using doOnNext instead of doOnEach will resolve the issue.

I am a beginner, please forgive me if there are any mistakes in what I said. Here is my opinion:
I think the null pointer caused by the Azure OpenAiChatModel not processing ChatResponse properly cannot be avoided when obtaining results, whether it is doOnNext or doOnEach

I have found that this issue is not related to Spring AI.
doOnEach handles multiple events. In the provided code, it calls the onComplete event as the final step after all tasks are completed. Since there is no data in the onComplete event, it results in null.
Using doOnNext instead of doOnEach will resolve the issue.

I am a beginner, please forgive me if there are any mistakes in what I said. Here is my opinion: I think the null pointer caused by the Azure OpenAiChatModel not processing ChatResponse properly cannot be avoided when obtaining results, whether it is doOnNext or doOnEach

I agree. Created a PR to make the processing of chatReponses more robust

thanks so much! will review.

I am testing with M7. The snippit to reproduce this error

chatClient.prompt().user("How are you?").stream().content().doOnEach(data -> System.out.println(data.get())); 

Passes with M7.

Also the test in the PR - #2789 passes without the fix in the code.

I'm not sure what is going on here, though I suppose the extra checks in the PR don't hurt.

Thoughts?

Thanks for testing @markpollack!
I did some testing with different setups in Azure and I found the cause. It is because of the Content Filtering option: asynchronous-filter
As you can see in the example below. WIth the asynchronous-filter enabled the last data message is null/empty and the content-filter result is send at a later moment.

data: {"id":"","object":"","created":0,"model":"","prompt_annotations":[{"prompt_index":0,"content_filter_results":{"hate":{"filtered":false,"severity":"safe"},"self_harm":{"filtered":false,"severity":"safe"},"sexual":{"filtered":false,"severity":"safe"},"violence":{"filtered":false,"severity":"safe"}}}],"choices":[],"usage":null} 

data: {"id":"chatcmpl-7rCNsVeZy0PGnX3H6jK8STps5nZUY","object":"chat.completion.chunk","created":1692913344,"model":"gpt-35-turbo","choices":[{"index":0,"finish_reason":null,"delta":{"role":"assistant"}}],"usage":null} 

data: {"id":"chatcmpl-7rCNsVeZy0PGnX3H6jK8STps5nZUY","object":"chat.completion.chunk","created":1692913344,"model":"gpt-35-turbo","choices":[{"index":0,"finish_reason":null,"delta":{"content":"Color"}}],"usage":null} 

data: {"id":"chatcmpl-7rCNsVeZy0PGnX3H6jK8STps5nZUY","object":"chat.completion.chunk","created":1692913344,"model":"gpt-35-turbo","choices":[{"index":0,"finish_reason":null,"delta":{"content":" is"}}],"usage":null} 

data: {"id":"chatcmpl-7rCNsVeZy0PGnX3H6jK8STps5nZUY","object":"chat.completion.chunk","created":1692913344,"model":"gpt-35-turbo","choices":[{"index":0,"finish_reason":null,"delta":{"content":" a"}}],"usage":null} 

data: {"id":"chatcmpl-7rCNsVeZy0PGnX3H6jK8STps5nZUY","object":"chat.completion.chunk","created":1692913344,"model":"gpt-35-turbo","choices":[{"index":0,"finish_reason":"stop","delta":{}}],"usage":null} 

data: {"id":"","object":"","created":0,"model":"","choices":[{"index":0,"finish_reason":null,"content_filter_results":{"hate":{"filtered":false,"severity":"safe"},"self_harm":{"filtered":false,"severity":"safe"},"sexual":{"filtered":false,"severity":"safe"},"violence":{"filtered":false,"severity":"safe"}},"content_filter_offsets":{"check_offset":506,"start_offset":44,"end_offset":571}}],"usage":null} 

data: [DONE]

You can resproduce this if you set a custom content-filer on your model: Azure AI Foundry->Safety+Security->Create content filter->Output Filter->Streaming mode (Preview)