🤖 fix: keep thinking preference across model switches #1758

ammar-agent · 2026-01-18T23:04:05Z

Summary

preserve thinking preference when switching models by storing the chosen level and clamping only for display/sends
clamp above-policy requests to the highest allowed level instead of defaulting to medium
add a UI integration test covering xhigh → Opus 4.5 → back to codex

make static-check
bun test src/common/utils/thinking/policy.test.ts
Not run (requires TEST_INTEGRATION=1): bun test tests/ui/thinkingPersistence.integration.test.ts

Generated with mux • Model: openai:gpt-5.2-codex • Thinking: xhigh • Cost: $2.69

github-actions bot added the enhancement New feature or functionality label Jan 18, 2026

ammar-agent force-pushed the fix/thinking-preference branch 2 times, most recently from 6f094ae to 233ac98 Compare January 18, 2026 23:19

🤖 fix: keep thinking preference across model switches

1460a6a

ammar-agent force-pushed the fix/thinking-preference branch from 233ac98 to 1460a6a Compare January 18, 2026 23:28

ammar-agent added 2 commits January 18, 2026 17:41

fix: pick nearest allowed thinking level

7d22a7a

tests: stabilize thinking persistence UI test

2879505

ammario merged commit a40aba5 into main Jan 19, 2026
21 checks passed

ammario deleted the fix/thinking-preference branch January 19, 2026 00:14