Auto thinking mode: enable extended reasoning only for complex prompts

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Thinking mode consume tokens extras mesmo em perguntas simples. Seria útil ter um modo automático que ative o raciocínio estendido apenas quando a complexidade da pergunta justificar, ou permitir
▎ configurar um threshold.

Proposed Solution

Add a "Smart" or "Auto" option to the Thinking mode setting, alongside the current On/Off toggle.
▎
▎ In Auto mode, the model would evaluate prompt complexity before deciding whether to engage extended reasoning. Simple prompts (factual questions, short explanations, quick lookups) would skip the
▎ thinking step, while complex prompts (code generation, debugging, architecture planning, multi-step analysis) would trigger it automatically.
▎
▎ Alternatively, expose a user-configurable complexity threshold (e.g., Low / Medium / High) so users can tune the behavior to their workflow.
▎
▎ This would reduce unnecessary token consumption without requiring users to manually toggle the setting between tasks.

Alternative Solutions

A simpler alternative would be to allow users to set Thinking mode per session via a slash command, such as /think on and /think off, making it faster to toggle without opening the settings
▎ dialog.
▎
▎ Another alternative would be a keyboard shortcut to toggle Thinking mode on/off instantly during a conversation, without interrupting the workflow.

Priority

High - Significant impact on productivity

Feature Category

Configuration and settings

Use Case Example

_No response_

Additional Context

_No response_