Allow Dynamic Configuration of Model Thinking, Temperature, and Reasoning Parameters

[Kaspazza]

To upvote this issue, give it a thumbs up. See this list for the most upvoted issues.

Based on this slack conversation: https://clojurians.slack.com/archives/C093426FPUG/p1770901436896429
Please look there for all the details of current discussion in the subject.

Problem description:
Currently, adjusting model parameters such as reasoning effort, thinking mode, temperature, or verbosity requires manual configuration and is not easily changeable at runtime. This makes it inconvenient to experiment with different generation behaviors, especially across providers like OpenAI and Anthropic that expose different controls.

For example, using claude code it's usefull to stay on opus 4.6 and just change it's thinking level to low for easy tasks that don't need much compute and are better to be done quickly.

We need a simple, consistent way to configure and switch between predefined parameter sets per model, while keeping the system flexible and provider-agnostic.

[rschmukler]

Just thinking that the header of the agent.md could be a good way to forward on these parameters

[ericdallo]

I already have something working in variants branch but there are some rough edges to be able to release that, it should be done in a week or so