Support for reasoning in Ollama 0.9.0+ #24
Description
Ollama added API support for reasoning in their 0.9.0 release.
https://github.com/ollama/ollama/releases/tag/v0.9.0
TODOs:
- Parse the thinking attribute in assistant message JSON like we do for e.g. Gemini.
- Probably also reconstruct previous messages in the same way when sending the chat for the next turn.
- Enable/disable reasoning according to the flag in chat settings.
- What about backwards compatibility?
- Is this model-dependent? E.g. are there reasoning models that still send tags even with newer Ollama installs?