Telnyx Hermes STT Provider
Use this skill when integrating or validating the Telnyx Speech-to-Text provider for Hermes.
Architecture
This is not a standalone plugin. Hermes handles STT through built-in
providers dispatched in tools/transcription_tools.py. This repo contains the
Telnyx provider function and tests, ready to be contributed upstream.
Provider details
| Field | Value |
|---|---|
| Provider ID | telnyx |
| Endpoint | https://api.telnyx.com/v2/ai/audio/transcriptions |
| Default model | openai/whisper-large-v3-turbo |
| Protocol | OpenAI-compatible (multipart/form-data) |
| Auth | TELNYX_API_KEY (Bearer) |
| Base URL override | TELNYX_STT_BASE_URL env var |
| Language override | TELNYX_STT_LANGUAGE env var (ISO-639-1, default en) |
Integration
See README.md for step-by-step instructions on adding the Telnyx STT provider
to tools/transcription_tools.py in hermes-agent. No new dependencies are
required — the existing openai package is reused with the Telnyx base URL.
Configure
export TELNYX_API_KEY="***"
Optional:
export TELNYX_STT_BASE_URL="https://api.telnyx.com/v2/ai"
export TELNYX_STT_LANGUAGE="en"
Running tests
# No credentials needed
python -m pytest tests/test_telnyx_stt_static.py tests/test_telnyx_stt_runtime.py -q
# Live test (requires TELNYX_API_KEY)
export TELNYX_API_KEY=***
python -m pytest tests/test_telnyx_stt_live.py -q
Notes
- No
pip installneeded — the provider function is copy-pasted into hermes-agent. - Local tests use a mock OpenAI client; live tests hit the real Telnyx API.
- The endpoint override (
TELNYX_STT_BASE_URL) is covered by tests.