fix: persist reasoningLevel 'off' instead of deleting it (#24406) by echoVic · Pull Request #24559 · openclaw/openclaw

echoVic · 2026-02-23T14:49:53Z

Problem

When a user runs /reasoning off, reasoning content continues to leak into channel messages (Telegram, Matrix, etc.). The WebUI correctly hides reasoning, but channel delivery paths still include it.

Reported in #24406 (Telegram, thinking=low) and #24411 (Matrix, /reasoning off).

Root Cause

In sessions-patch.ts (the WebSocket/API session patch path), when reasoningLevel is set to "off", the code deletes the field from the session entry instead of persisting it:

if (normalized === 'off') {
  delete next.reasoningLevel;  // ← field removed entirely
}

Later, in get-reply-directives.ts, the code checks whether reasoning was explicitly set:

const reasoningExplicitlySet =
  directives.reasoningLevel !== undefined ||
  (sessionEntry?.reasoningLevel !== undefined && sessionEntry?.reasoningLevel !== null);

Since the field was deleted (undefined), reasoningExplicitlySet is false. This triggers the auto-enable path:

if (!reasoningExplicitlySet && resolvedReasoningLevel === 'off' && !thinkingActive) {
  resolvedReasoningLevel = await modelState.resolveDefaultReasoningLevel();
}

For reasoning-capable models (Claude Opus, etc.), this re-enables reasoning — undoing the user's explicit /reasoning off.

Fix

Persist "off" explicitly instead of deleting the field, matching how directive-handling.persist.ts (the inline command path) already handles it correctly:

// Before (broken):
if (normalized === 'off') { delete next.reasoningLevel; }

// After (fixed):
next.reasoningLevel = normalized;  // persists 'off' explicitly

Testing

/reasoning off → sessionEntry.reasoningLevel === 'off' → reasoningExplicitlySet === true → no auto-enable
/reasoning on and /reasoning stream continue to work as before
null (reset) still deletes the field, allowing model defaults to apply

Fixes #24406
Fixes #24411

Greptile Summary

Fixed reasoning content leak in channel messages (Telegram, Matrix) by persisting reasoningLevel: "off" explicitly instead of deleting it. The previous implementation deleted the field when set to "off", causing get-reply-directives.ts:396-398 to treat it as unset and auto-enable reasoning for capable models (Claude Opus, etc.). The fix aligns with the existing pattern used by elevatedLevel (line 222) and directive-handling.persist.ts:93-95, ensuring the user's explicit /reasoning off command is respected across all delivery paths.

Confidence Score: 5/5

This PR is safe to merge with minimal risk
The fix is a simple 3-line change that correctly aligns the implementation with existing patterns for similar fields (elevatedLevel on line 222 of the same file). The change matches the pattern already used in directive-handling.persist.ts:93-95, and the PR description provides clear root cause analysis with references to the specific logic that was broken. The fix ensures consistency across two code paths (WebSocket/API vs inline command) and prevents unintended re-enabling of reasoning.
No files require special attention

_{Last reviewed commit: d385867}

) When a user runs /reasoning off, the session patch handler deleted the reasoningLevel field from the session entry. This caused get-reply-directives to treat reasoning as 'not explicitly set', which triggered resolveDefaultReasoningLevel() to re-enable reasoning for capable models (e.g. Claude Opus). The fix persists 'off' explicitly, matching how directive-handling.persist.ts already handles the inline /reasoning off command. Fixes openclaw#24406 Fixes openclaw#24411

steipete

Reviewed for correctness and simpler alternatives; this is the best scoped change for the issue.

) (openclaw#24559) When a user runs /reasoning off, the session patch handler deleted the reasoningLevel field from the session entry. This caused get-reply-directives to treat reasoning as 'not explicitly set', which triggered resolveDefaultReasoningLevel() to re-enable reasoning for capable models (e.g. Claude Opus). The fix persists 'off' explicitly, matching how directive-handling.persist.ts already handles the inline /reasoning off command. Fixes openclaw#24406 Fixes openclaw#24411 Co-authored-by: echoVic <AkiraVic@outlook.com>

) (openclaw#24559) When a user runs /reasoning off, the session patch handler deleted the reasoningLevel field from the session entry. This caused get-reply-directives to treat reasoning as 'not explicitly set', which triggered resolveDefaultReasoningLevel() to re-enable reasoning for capable models (e.g. Claude Opus). The fix persists 'off' explicitly, matching how directive-handling.persist.ts already handles the inline /reasoning off command. Fixes openclaw#24406 Fixes openclaw#24411 Co-authored-by: echoVic <AkiraVic@outlook.com> (cherry picked from commit 52ac763)

) (openclaw#24559) When a user runs /reasoning off, the session patch handler deleted the reasoningLevel field from the session entry. This caused get-reply-directives to treat reasoning as 'not explicitly set', which triggered resolveDefaultReasoningLevel() to re-enable reasoning for capable models (e.g. Claude Opus). The fix persists 'off' explicitly, matching how directive-handling.persist.ts already handles the inline /reasoning off command. Fixes openclaw#24406 Fixes openclaw#24411 Co-authored-by: echoVic <AkiraVic@outlook.com> (cherry picked from commit 52ac763) # Conflicts: # src/gateway/sessions-patch.ts

) (openclaw#24559) When a user runs /reasoning off, the session patch handler deleted the reasoningLevel field from the session entry. This caused get-reply-directives to treat reasoning as 'not explicitly set', which triggered resolveDefaultReasoningLevel() to re-enable reasoning for capable models (e.g. Claude Opus). The fix persists 'off' explicitly, matching how directive-handling.persist.ts already handles the inline /reasoning off command. Fixes openclaw#24406 Fixes openclaw#24411 Co-authored-by: echoVic <AkiraVic@outlook.com>

openclaw-barnacle bot added gateway Gateway runtime size: XS trusted-contributor labels Feb 23, 2026

steipete approved these changes Feb 24, 2026

View reviewed changes

steipete merged commit 52ac763 into openclaw:main Feb 24, 2026
29 checks passed

abdelkadermeflahi6-web referenced this pull request Feb 24, 2026

docs(changelog): trim 2026.2.24 unreleased entries

e3ac491

abdelkadermeflahi6-web referenced this pull request Feb 24, 2026

chore(release): bump version to 2026.2.24

2bad30b

arjunaskykok mentioned this pull request Feb 26, 2026

fix/test a2ui bundle preflight #27345

Open

2 tasks

github-actions bot mentioned this pull request Mar 1, 2026

cherry-pick: upstream bugfix commits (2026-03-01-0443) hughdidit/DAISy-Agency#140

Closed

6 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix: persist reasoningLevel 'off' instead of deleting it (#24406)#24559

fix: persist reasoningLevel 'off' instead of deleting it (#24406)#24559
steipete merged 1 commit intoopenclaw:mainfrom
echoVic:fix/reasoning-off-persisted

echoVic commented Feb 23, 2026 •

edited by greptile-apps bot

Loading

Uh oh!

steipete left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

echoVic commented Feb 23, 2026 • edited by greptile-apps bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Root Cause

Fix

Testing

Greptile Summary

Confidence Score: 5/5

Uh oh!

steipete left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

echoVic commented Feb 23, 2026 •

edited by greptile-apps bot

Loading