ai-agents-ollama-model-integra-using-deepseek-r1-model-with-ollama-via-langchain--b2e1ea58

THEN Explicitly set the format option to 'json' when initializing the ChatOllama model for deepseek-r1 to bypass the thinking section. Alternatively, wait for a parser update in LangChain that correctly strips the thinking section from the response.

const model = new ChatOllama({ model: 'deepseek-r1:14b', format: 'json' });