TextReasoningContent implementation?

[flintlouis]

Hi,
When using the Microsoft.Extensions.AI implementation of OllamaSharp, using the GetStreamingResponseAsync and using a thinking model like qwen3, the response I get is something like:

User: Hello
Assistant: <think>
Okay, the user said "Hello" after I greeted them with "Hi, how can I help you today?" Now they're responding with just "Hello." I need to acknowledge their greeting and keep the conversation going. Let me make sure to be friendly and open-ended. Maybe ask them how I can assist or what they need help with. I should keep it simple and welcoming. Let me check if there's anything specific they might need, but since they just said hello, I'll stick to a general offer of help. Alright, time to respond.
</think>

Hello! ?? How can I assist you today? Let me know if you need help with anything!

The <think>...</think> part is added to the response message as a TextContent. I saw that there is also a TextReasoningContent. I was wondering if the <think> part will be implemented as a TextReasoningContent in the future?

[awaescher]

Older Ollama version or older model definitions may cause that the think tags are included in the model response text directly. Can you make sure that both, the Ollama version and the model definition are new? When in doubt, just re-pull your models.

[flintlouis]

My ollama version is 0.10.1
The qwen3 model version is qwen3:latest

It's not about if they are included in the response for me (in the Ollama CLI the thinking is also included but in a different colour), I was just wondering if the TextReasoningContent will be used to seperate the thinking in the future? For now I can just filter it out myself I was just curious :)

[awaescher]

Should be, yes. See the ExtensionAiToolConsole.

[flintlouis]

Ah yes I see that they also filter for TextReasoningContent but then I am guessing it is not implemented yet. Or am I missing something

Thankyou for your quick responses.

[flintlouis]

@awaescher I take it back, everything is working. I had a small bug in my code :$ sorry