#Prevent <think> section from being returned to user

1 messages · Page 1 of 1 (latest)

azure aspen
#

I'm using deepseek-r1 via ollama as the conversation agent for an ai assistant
The model outputs a <think> ... </think> section before giving an answer, which is not very useful when using the model as a chat agent
Is it possible to put configuration somewhere to stop the whole think section being fed back to the speaker (wyoming satellite)?

#

Prevent <think> section from being returned to user