[Question]: When I want to use a deep thinking model, such as R1, how can I print the deep thinking process?

[tiebingice]

Question Validation

• I have searched both the documentation and discord for an answer.

Question

I make code for llama-index-llms-deepseek，When I want to use a deep thinking model R1, how can I print the deep thinking process?

[dosubot]

Hello @tiebingice! I'm Dosu, here to help you solve bugs, answer questions, and navigate your way to becoming a contributor. While you're waiting for a human maintainer, feel free to ask me anything you need assistance with!

To print the deep thinking process when using the deep thinking model R1 in the llama-index-llms-deepseek codebase, you can use the SimpleLLMHandler class. This class includes a function called _print_llm_event that prints the prompt and completion or messages and response based on the payload provided. This function is called within the on_event_end method when the event type is CBEventType.LLM ^[1].

Additionally, you can configure the SimpleLLMHandler class to provide more detailed output by utilizing the thinking_dict parameter when initializing the model. This allows the model to generate a chain-of-thought response. You can also access the exact thinking process by accessing the thinking attribute in the response's additional_kwargs ^[2].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}