Fix voice transcription

[dim]

📝 Description

Addresses #945

🗣️ Type of Change

• 🐞 Bug fix (non-breaking change which fixes an issue)
• ✨ New feature (non-breaking change which adds functionality)
• 📖 Documentation update
• ⚡ Code refactoring (no functional changes, no api changes)

🤖 AI Code Generation

• 🤖 Fully AI-generated (100% AI, 0% Human)
• 🛠️ Mostly AI-generated (AI draft, Human verified/modified)
• 👨‍💻 Mostly Human-written (Human lead, AI assisted or none)

🔗 Related Issue

Fixes #945

📚 Technical Context (Skip for Docs)

• Reference URL:
• Reasoning:

🧪 Test Environment

• Hardware: Raspberry Pi 5
• OS: Debian 13
• Model/Provider: MiniMax M25 + Groq Whisper
• Channels: Telegram

📸 Evidence (Optional)

2026/03/01 08:29:59 [2026-03-01T08:29:59Z] [INFO] agent: Processing message from telegram:telegram:REDACTED: [voice] {channel=telegram, chat_id=REDACTED, sender_id=telegram:REDACTED, session_key=}
2026/03/01 08:29:59 [2026-03-01T08:29:59Z] [INFO] voice: Starting transcription {audio_file=/tmp/picoclaw_media/452372f1_file_6.oga.ogg}
2026/03/01 08:29:59 [2026-03-01T08:29:59Z] [INFO] voice: Transcription completed successfully {text_length=45, language=, duration_seconds=0, transcription_preview= Let me check if voice recognition now works.}
2026/03/01 08:29:59 [2026-03-01T08:29:59Z] [INFO] agent: Routed message {agent_id=main, session_key=agent:main:telegram:direct:REDACTED, matched_by=default}
2026/03/01 08:30:05 [2026-03-01T08:30:05Z] [INFO] agent: LLM response without tool calls (direct answer) {agent_id=main, iteration=1, content_chars=294}
2026/03/01 08:30:05 [2026-03-01T08:30:05Z] [INFO] agent: Response: Hey, I can read that! 🎉

☑️ Checklist

• My code/docs follow the style of this project.
• I have performed a self-review of my own changes.
• I have updated the documentation accordingly.

[xiaket]

Longer term, we may want to consider multiple transcribe providers and how can we prioritise them. For now this works and address an issue, I'm happy with it.

[dim]

@xiaket let me address this then slightly more elegantly

[xiaket]

LGTM

[nikolasdehor]

LGTM. This is a well-structured fix that moves voice transcription from channel-specific (Telegram only) to agent-level, making it work across all channels.

Key improvements:

1. Transcriber interface with Name() and Transcribe() methods -- clean abstraction
2. DetectTranscriber auto-detects Groq from either direct provider config or model_list entries
3. Agent-level transcription via transcribeAudioInMessage replaces audio annotations [voice] with transcribed text
4. Falls back gracefully: if transcription fails, the annotation is left as-is (empty string appended)
5. README updates across all translations correctly reflect the change

The regex-based annotation replacement is clean: audioAnnotationRe matches [voice] and [audio:*] patterns, and transcriptions are applied in order. Remaining transcriptions (more audio than annotations) are appended with newlines.

Test coverage is thorough: interface satisfaction, DetectTranscriber with various configs, actual transcription with mocked HTTP server, API errors, and missing files. Good work.

[afjcjsbx]

LGTM! @dim could you fix the lint?

[Orgmar]

@dim Thanks for fixing the voice transcription issue! Audio handling bugs can be tough to track down, glad this one got sorted out.

We have a PicoClaw Dev Group on Discord where contributors connect and share ideas. If you'd like to join, send an email to support@sipeed.com with the subject [Join PicoClaw Dev Group] dim and we'll send the invite!

[dim]

@Orgmar thanks, that's kind. I am a bit busy at the moment but would be very happy to join in a few weeks time. Will email support@sipeed.com as suggested. Thanks again!