how does ChatHistory.messagesSupplier() got update?

[nxa06464]

Hello,
I run the code on my pc, and want to know how chat history got update.
I monitor when call this.chatService.stream, chatHistory.messagesSupplier() size is 0, when go subscribe the size is 1, complete do finally size is 2. 1 is USER request message, 2 is THINK and ASSISTANT. but I dont know where or which class update chatHistory.messagesSupplier(). could explain me thanks? is it control by spring or Vaadin?

Regards
zwz

[JM-Lab]

Hi @nxa06464,

Currently, Ollama models like Qwen3 and DeepSeek-R1 send thinking content as <think> tags in the assistant message, which gets rendered in the Vaadin UI during streaming.
For gpt-oss, thinking comes through assistant metadata fields, but I haven’t integrated this yet since Spring AI is still developing unified support for thinking across providers.
Once Spring AI exposes a consistent interface, I’ll add framework-level parsing/persistence support alongside the existing UI handling. For now, check ChatContentView for the current streaming logic.

Thanks for the question!

[nxa06464]

thanks for your reply. but I still not understand how chat history got update.
eg. before click submit button
this.chatHistory.messagesSupplier().get().size() is 0
after I click the button it call method inputEvent, and process this.chatService.stream, get response and refresh page.
then this.chatHistory.messagesSupplier().get().size() change to 1 and 2. which code put the chat message to chatHistory.messagesSupplier()? would like to understand this, thanks.

[JM-Lab]

The chatHistory.messagesSupplier().get() retrieves messages from Spring AI’s ChatMemory. The chatClient has ChatMemory registered as an Advisor, so all interactions with the LLM are automatically recorded by the Spring AI framework.
When ChatHistory is created, it’s configured to pull from the associated memory. The implementation details are in ChatHistoryService.
So the flow is:
1. User submits → chatService.stream() calls the LLM
2. Spring AI framework automatically stores both user message and AI response via ChatMemory Advisor
3. UI refreshes → messagesSupplier() pulls the updated messages from ChatMemory
That’s why you see the size go from 0 → 1 (user message) → 2 (+ AI response) without explicit message adding in the UI code!

[nxa06464]

thanks for your help, very helpful, thanks.