DocsAI ProvidersThinking Modes
Back to Docs
AI Providers

Thinking Modes

Control how much reasoning time YeePilot requests before it answers, plans, or verifies work.

Last updated: May 15, 2026

What Thinking Mode Does

Thinking mode tells YeePilot how much reasoning effort to request before it answers or plans.

This matters most for:

  • debugging
  • migration planning
  • multi-step troubleshooting
  • high-risk changes
  • code or config review

Available Modes

Current supported values are:

  • off
  • auto
  • best
  • low
  • medium
  • high

In the TUI, /think on is accepted as a shortcut and maps to medium.

Change It During A Session

text
/think off
/think auto
/think best
/think low
/think medium
/think high

Set A Default In Config

yaml
ai:
  think_mode: medium

Which Mode Should You Use?

ModeBest for
offvery simple checks, fast answers
lowroutine multi-step work
mediumgeneral troubleshooting and planning
highdeep troubleshooting, high-stakes review
autolet the runtime choose a reasonable level
bestask for the strongest available reasoning behavior

Interaction With Other Controls

Thinking mode works best when combined with:

  • /plan <task> for structured planning
  • /autonomy guarded or /autonomy strict_review for careful execution
  • /tokenmode saver if stronger reasoning starts using too much context or cost

End-User Recommendation

For most users:

  • keep medium as the default
  • switch to off for quick one-step questions
  • switch to high or best for difficult troubleshooting
Previous

Model Selection & Token Modes

AI Providers

Next

Command Safety & Risk Classification

Security