Summary

• add agent.reasoning_language with visible options auto, zh, and en
• apply zh/en as transient per-turn context while auto injects no additional language instruction
• expose the setting in Desktop Settings and refine the Agent runtime copy, tooltip behavior, model usage layout, and step-limit order
• document the new config option and add focused config, render, input, and desktop settings coverage

Cache behavior

• auto keeps the existing stable language policy and does not add dynamic content
• explicit language preferences are composed into the user turn, not the stable system prompt or tool schema

Validation

• pnpm --dir desktop/frontend check:css
• pnpm --dir desktop/frontend typecheck
• pnpm --dir desktop/frontend test:all
• go test ./internal/config ./internal/control ./internal/boot ./internal/i18n
• go test ./... in desktop/
• verified the Settings > Models > Usage UI in the in-app browser