PR #68107: fix(feishu): respect reasoningLevel=off so thinking text stops rendering in Feishu cards

• Repository: openclaw/openclaw
• Author: mvanhorn
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/68107

Description (problem / solution / changelog)

Summary

• Problem: Feishu cards rendered the model's reasoning text even when a session had reasoningLevel: "off". /reasoning off, verboseDefault: off, and blockStreaming: false had no effect because the Feishu dispatcher never read the session's reasoning level.
• Why it matters: Webchat (and Telegram, as of extensions/telegram/src/bot-message-dispatch.ts:303) already suppress thinking output when the user turned reasoning off. Feishu users had no way to hide it short of disabling reasoning at the agent level.
• What changed: Added a Feishu-local resolveFeishuReasoningLevel helper that reads reasoningLevel from the session store and threaded the resolved level into buildCombinedStreamText at all three call sites in extensions/feishu/src/reply-dispatcher.ts. When the level is the exact string "off", the reasoning branch is skipped.
• What did NOT change (scope boundary): Telegram, webchat, and core reply pipelines are untouched. No changes to the session-store schema or the /reasoning command. Default behavior for any level other than "off" is unchanged.

Change Type (select all)

• Bug fix
• Feature
• Refactor required for the fix
• Docs
• Security hardening
• Chore/infra

Scope (select all touched areas)

• Gateway / orchestration
• Skills / tool execution
• Auth / tokens
• Memory / storage
• Integrations
• API / contracts
• UI / DX
• CI/CD / infra

Linked Issue/PR

• Closes #68052
• Related #
• This PR fixes a bug or regression

Root Cause (if applicable)

• Root cause: The Feishu dispatcher constructed buildCombinedStreamText(thinking, answer) at three sites without consulting the session's reasoningLevel. Telegram already does this via resolveTelegramReasoningLevel at extensions/telegram/src/bot-message-dispatch.ts:132-156 and gates output with suppressReasoning = resolvedReasoningLevel === "off" at :303. Feishu never had the equivalent plumbing.
• Missing detection / guardrail: No Feishu regression test covered the reasoningLevel=off path. This PR adds one.
• Contributing context (if known): Traced by @hclsys on the issue thread - they identified the exact call sites in reply-dispatcher.ts and the Telegram reference pattern to mirror.

Regression Test Plan (if applicable)

• Coverage level that should have caught this:
  • Unit test
  • Seam / integration test
  • End-to-end test
  • Existing coverage already sufficient
• Target test or file: extensions/feishu/src/reply-dispatcher.test.ts
• Scenario the test should lock in: Dispatcher handling a reply on a session whose store entry has reasoningLevel: "off" must produce card text that contains only the answer, with no reasoning prefix.
• Why this is the smallest reliable guardrail: The bug lives entirely inside the per-dispatch helper that assembles card text. A unit-level test that drives the dispatcher with a fake session store exercises the exact suppression branch without needing a live Feishu bot.
• Existing test that already covers this (if any): None. extensions/feishu/src/reply-dispatcher.test.ts previously had 30 tests; this PR adds 1.
• If no new test is added, why not: N/A. New test added (31 passing).

This contribution was developed with AI assistance (Codex).

Changed files

• extensions/feishu/src/bot.ts (modified, +2/-0)
• extensions/feishu/src/reply-dispatcher.test.ts (modified, +90/-0)
• extensions/feishu/src/reply-dispatcher.ts (modified, +56/-7)