Ollama Chat - parse tool calls on streaming

[krrishdholakia]

• fix(user_api_key_auth.py): fix else block

Fixes #11170

• refactor(ollama/chat): refactor to base config pattern

easier to maintain fixes

• fix(ollama/chat): support tool call parsing on streaming

Closes #11104

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
litellm	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	May 27, 2025 11:15pm

[lsorber]

@krrishdholakia Thanks for working on this! May I ask to also verify that litellm.stream_chunk_builder successfully recovers the tool calls from the chunks? Because in debugging my unit test for this issue, I found that stream_chunk_builder does not include the tool calls because Ollama does not return tool call ids as part of its response.

[krrishdholakia]

Hey @lsorber this is solved by handling the tool call translation in the ollama/chat.py - this follows are newer pattern of keeping all the llm translation work inside the llm folder.

This then returns a standard ModelResponseStream object which is returned by the stream chunk builder