PR #84085: fix(ui): de-duplicate Off option in chat thinking-level picker [AI-assisted]

• Repository: openclaw/openclaw
• Author: luyao618
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/84085

Description (problem / solution / changelog)

🤖 AI-assisted (built with Codex via Hermes orchestration). Test level: lightly tested (new co-located unit tests for the affected helper). Prompt summary available on request.

Summary

• Problem: The chat thinking-level dropdown shows two indistinguishable "Off" rows for models whose level list includes an explicit off entry, and stays interactive even when no real thinking options exist.
• Why it matters: Users can't tell the two "Off" rows apart and may pick the "wrong" one; for non-thinking models the picker invites a meaningless choice.
• What changed: buildThinkingOptions now skips any entry whose normalized value is off (the inherited <option value=""> already represents that choice), and renderChatThinkingSelect disables the select when no real options remain so it stays visible but inert.
• What did NOT change (scope boundary): No changes to thinking-labels.ts, thinking.ts, gateway-side level resolution, or the sessions view picker. The inherited-default label logic is untouched.

Change Type (select all)

• Bug fix

Scope (select all touched areas)

• UI

Linked Issue/PR

• Closes #84069
• This PR fixes a bug or regression

Root Cause

• Root cause: The renderer always emits an <option value=""> whose label is defaultLabel ("Off" when the effective level is off), then concatenates every entry from buildThinkingOptions. When the model's level list itself contains off, the picker ends up with two "Off" rows: one empty-value, one value="off". Separately, models with no thinking levels still rendered an interactive picker because disabled did not consider an empty options list.
• Missing detection / guardrail: No unit test asserted that the inherited default row and the dynamic option list are disjoint, and there was no test covering the empty-options case.
• Contributing context (if known): buildThinkingOptions was added to dedupe by lowercase key but it didn't filter the inherited-default sentinel, since the inherited row is emitted by the renderer rather than by buildThinkingOptions.

Regression Test Plan

• Coverage level that should have caught this:
  • Unit test
• Target test or file: ui/src/ui/chat/session-controls.test.ts (new)
• Scenario the test should lock in: buildThinkingOptions strips any off/none/OFF entry from the levels list and from the current override, while preserving other levels and non-off overrides.
• Why this is the smallest reliable guardrail: The duplicate-Off bug lives entirely in the option-building helper; locking down its output is sufficient and avoids a heavy lit/DOM test.
• Existing test that already covers this (if any): N/A
• If no new test is added, why not: N/A

User-visible / Behavior Changes

• Chat thinking-level picker no longer shows two "Off" rows for models that expose an off level.
• For models with no real thinking levels, the picker is rendered but disabled (no selectable choices beyond the inherited "Off").

Security Impact (required)

• New permissions/capabilities? No
• Pure UI-side option filtering and a disabled-state tweak; no auth, network, or persistence changes.

Changed files

• ui/src/ui/chat/session-controls.test.ts (added, +49/-0)
• ui/src/ui/chat/session-controls.ts (modified, +14/-3)

PR #84087: fix(chat): remove duplicate Off option in thinking level dropdown (#84069)

• Repository: openclaw/openclaw
• Author: 1027866388
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/84087

Description (problem / solution / changelog)

Summary

Fixes #84069 - Thinking Selection dropdown shows two identical "Off" options.

Problem

When using the thinking level dropdown with an inherited default (value=""), the backend thinking levels list also includes an "off" option. This results in two identical "Off" entries in the select menu.

Solution

In buildThinkingOptions(), skip adding "off" from thinking levels when currentOverride is empty (inherited default), since the default option already represents that state.

Changes

• ui/src/ui/chat/session-controls.ts: Skip "off" from thinking levels when using inherited default
• ui/src/ui/views/chat.test.ts: Updated tests to reflect correct (non-duplicate) behavior

Testing

All 38 chat UI tests pass.

Real behavior proof

Behavior or issue addressed: Thinking level dropdown in chat session controls renders duplicate "Off" entries when using inherited default (currentOverride="").

Real environment tested: Local development setup with pnpm dev, running openclaw web UI.

Exact steps or command run after this patch:

1. Open chat session with thinking level dropdown
2. Observe the dropdown options when currentOverride is empty (inherited default)
3. Before fix: Two "Off" entries visible
4. After fix: One "Off" entry visible

Evidence after fix: Screenshot of the fixed UI showing the thinking level dropdown with correct non-duplicate options.

Observed result after fix: The thinking level dropdown now shows a single "Off" option instead of duplicating it. The fix in session-controls.ts line 297 adds a condition to skip adding "off" from backend thinking levels when using the inherited default.

What was not tested: Non-default thinking level selections, other session control scenarios.

Changed files

• ui/src/ui/chat/session-controls.ts (modified, +6/-0)
• ui/src/ui/views/chat.test.ts (modified, +1/-3)