Bug

The message tool requires components for Discord sends, but Discord voice messages (asVoice: true) cannot have components. This makes it impossible to send pre-recorded audio as inline playable voice messages in Discord.

Steps to Reproduce

1. Try sending a voice message with pre-recorded audio:

   • asVoice: true + media: /path/to/audio.ogg (no components) → Validation error: components: must have required property components
   • asVoice: true + media + components: {} → API error: Discord components cannot be sent as voice messages.

2. Try sending audio without asVoice:

   • media + components: { blocks: [{ type: "file", file: "attachment://file.ogg" }] } → File attaches as download card, NOT inline playable
   • media + components: { text: "..." } → Attachment silently stripped (attachments: [])

Expected Behavior

Sending asVoice: true with media and no components should work. Discord voice messages need duration_secs and waveform metadata on the attachment — the tool should either:

1. Skip components validation when asVoice: true
2. Auto-generate waveform data from the audio file (like Discord clients do)

Actual Behavior

Caught between two constraints: components is required but voice messages cannot have components.

Workaround

The tts tool can send inline playable audio, but it uses the bot's own TTS voice — not custom/cloned audio. There is no way to send pre-recorded custom audio as a playable voice message through the current tool.

Environment

• OpenClaw 2026.3.14
• Discord channel
• macOS arm64