Bug Description

After a fresh install using the setup wizard, I configured faster-whisper as my local STT provider. The wizard left stt.model: "whisper-1" in my config.yaml. When I sent a voice message via Telegram, transcription crashed silently — the voice message was just ignored with no response.

*Diagnosed and generated by Slopus 4.6

Steps to Reproduce

1. Run hermes setup, configure STT with local faster-whisper provider
2. Note that config.yaml contains stt.model: "whisper-1" (the OpenAI API name)
3. Start the gateway with Telegram
4. Send a voice message to the bot

Expected Behavior

The bot should either map "whisper-1" to a valid local model size (e.g. "base") or show a clear error/warning in logs and fall back gracefully.

Actual Behavior

Transcription crashes with:
Invalid model size 'whisper-1', expected one of: tiny, base, small, medium, large...The voice message is silently ignored.

Affected Component

Gateway (Telegram/Discord/Slack/WhatsApp)

Messaging Platform (if gateway-related)

Telegram

Operating System

Debian GNU/Linux 13 (trixie)

Python Version

3.13.5

Hermes Version

0.4.0

Relevant Logs / Traceback

Root Cause Analysis (optional)

tools/transcription_tools.py passes the model name from config directly to faster-whisper without validating it. Cloud-only model names like "whisper-1" are invalid for the local provider but there's no guard or mapping.

Proposed Fix (optional)

Add a validation guard in transcription_tools.py that detects cloud-only model names (e.g. "whisper-1") when using a local provider and either maps them to a sensible default like "base" or ignores the override entirely.

Are you willing to submit a PR for this?

• I'd like to fix this myself and submit a PR