Problem
The yellow background highlight on thinking/reasoning blocks is visually loud, especially when the model runs in high reasoning mode. The entire transcript page accumulates multiple highlighted blocks (thinking, tool outputs, etc.), and combined they create a noisy reading experience.
Current state
There is show_thinking = true/false in settings.toml which fully hides thinking output. But this is binary — you either see nothing or you see the yellow-highlighted blocks. There is no middle ground.
Request
Add a setting to keep thinking content visible but without the intrusive background highlight. For example:
thinking_highlight = true/false — controls whether the yellow bg is applied- Or a more subtle visual treatment (border/indent only, no filled background)
The core ask: let me read the reasoning without the highlighter pen effect.
Problem
The yellow background highlight on thinking/reasoning blocks is visually loud, especially when the model runs in high reasoning mode. The entire transcript page accumulates multiple highlighted blocks (thinking, tool outputs, etc.), and combined they create a noisy reading experience.
Current state
There is
show_thinking = true/falseinsettings.tomlwhich fully hides thinking output. But this is binary — you either see nothing or you see the yellow-highlighted blocks. There is no middle ground.Request
Add a setting to keep thinking content visible but without the intrusive background highlight. For example:
thinking_highlight = true/false— controls whether the yellow bg is appliedThe core ask: let me read the reasoning without the highlighter pen effect.