PR #59602: fix: send TTS audio as voice message in Feishu

• Repository: openclaw/openclaw
• Author: KChow-ctrl
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/59602

Description (problem / solution / changelog)

Summary

Add .mp3 and audio/mpeg detection in resolveFeishuOutboundMediaKind() so that TTS-generated audio files (typically .mp3) are sent as voice messages instead of file attachments in Feishu.

Changes

• extensions/feishu/src/media.ts: Added .mp3 extension and audio/mpeg contentType to audio detection in resolveFeishuOutboundMediaKind()

Testing

Verified the fix locally by reviewing the updated code - the function now correctly identifies .mp3 and audio/mpeg as audio type.

Fixes openclaw/openclaw#59588

Changed files

• extensions/feishu/src/media.ts (modified, +3/-1)

PR #59619: fix(feishu): send TTS audio as voice message instead of file attachment (#59588)

• Repository: openclaw/openclaw
• Author: MoerAI
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/59619

Description (problem / solution / changelog)

Summary

TTS audio is sent as a file attachment instead of a clickable voice message bubble in Feishu because the audioAsVoice flag from the outbound context is not propagated to the Feishu media routing logic.

Root Cause

The Feishu sendMedia adapter (outbound.ts) did not destructure audioAsVoice from the ChannelOutboundContext, so sendMediaFeishu and resolveFeishuOutboundMediaKind never saw the flag. When TTS generates MP3 audio with audioAsVoice: true, resolveFeishuOutboundMediaKind classified it as msgType: "file" (generic file attachment) instead of msgType: "audio" (voice message bubble).

Changes

• extensions/feishu/src/outbound.ts: destructure audioAsVoice from outbound context, pass it to sendMediaFeishu
• extensions/feishu/src/media.ts: add audioAsVoice parameter to sendMediaFeishu and resolveFeishuOutboundMediaKind; when audioAsVoice is true and file is audio (mp3/wav/m4a/aac), force msgType: "audio"
• extensions/feishu/src/media.test.ts: add regression tests for mp3 with/without audioAsVoice

Test

pnpm test -- extensions/feishu/src/media.test.ts -t "msg_type"  # 20 passed
pnpm test -- extensions/feishu/src/outbound.test.ts             # 18 passed

Closes #59588

Changed files

• extensions/feishu/src/media.test.ts (modified, +31/-0)
• extensions/feishu/src/media.ts (modified, +18/-3)
• extensions/feishu/src/outbound.ts (modified, +2/-0)

PR #59652: fix(microsoft-tts): default to Opus output format for voice message compatibility

• Repository: openclaw/openclaw
• Author: w-sss
• State: closed | merged: False
• Link: https://github.com/openclaw/openclaw/pull/59652

Description (problem / solution / changelog)

Fixes #59588

Problem

TTS audio from Microsoft provider is delivered as MP3 file attachment instead of clickable voice message in Feishu (and likely Telegram/WhatsApp/Matrix). Users must download the file and play it manually.

Root Cause

The default Edge TTS output format is audio-24khz-48kbitrate-mono-mp3, which produces .mp3 files. The Feishu channel's media classification (resolveFeishuOutboundMediaKind) only treats .opus and .ogg as audio voice messages; MP3 files are sent as generic file attachments.

Additionally, isVoiceCompatibleAudio() in src/media/audio.ts checks for Telegram voice formats (ogg, opus, m4a) but not mp3.

Fix

Change DEFAULT_EDGE_OUTPUT_FORMAT from audio-24khz-48kbitrate-mono-mp3 to audio-24khz-48kbitrate-mono-opus. This ensures TTS audio is sent as Opus format, which is compatible with voice message features on all channels in OPUS_CHANNELS (Telegram, Feishu, WhatsApp, Matrix).

Test Plan

• Microsoft TTS tests pass (11/11)
• Manual testing: enable Microsoft TTS with Feishu, verify audio is sent as voice message (clickable bubble) not file attachment

Changed files

• src/agents/tools/sessions-send-tool.ts (modified, +15/-12)
• src/infra/exec-approvals-persist.test.ts (added, +142/-0)
• src/infra/exec-approvals.ts (modified, +30/-3)