[Feature]: Support `think` parameter for Ollama models

[saattrupdan]

The Feature

Ollama now supports thinking with the new think parameter in the ollama.generate and ollama.chat functions. This doesn't seem to be implemented in LiteLLM yet, however.

Motivation, pitch

Support for reasoning Ollama models.

LiteLLM is hiring a founding backend engineer, are you interested in joining us and shipping to all our users?

No

Twitter / LinkedIn details

No response

[enzomich]

I support the proposal, and it should be easy to do: just add another boolean kwarg ("think") in the call to completion() to be passed to the Ollama API.

[enzomich]

@saattrupdan Actually LiteLLM already supports thinking for a number of providers, but they don't include Ollama:
https://docs.litellm.ai/docs/reasoning_content

[saattrupdan]

@saattrupdan Actually LiteLLM already supports thinking for a number of providers, but they don't include Ollama:
https://docs.litellm.ai/docs/reasoning_content

Indeed, so it hopefully shouldn't be too difficult to add Ollama support as well 🙂

[jamie-dit]

Just updating that Ollama for gpt-oss models are not supported in LiteLLM

[jvsteiner]

I just arrived here to report the same thing ;)