telnyx-stt

Telnyx speech-to-text provider for OpenClaw audio transcription.

Registers Telnyx as a first-class mediaUnderstandingProviders backend so inbound voice notes, audio uploads, and recordings can be transcribed as an alternative media-understanding backend.

Features

• Provider ID: telnyx
• Plugin ID: telnyx-stt
• Default model: openai/whisper-large-v3-turbo
• Auto-select priority: 25
• Auth: TELNYX_API_KEY env var or provider config
• Transport: multipart upload to POST /v2/ai/audio/transcriptions

Prerequisites

• Node.js >=22.14.0
• OpenClaw >=2026.4.0
• A valid TELNYX_API_KEY

Install

From a local checkout:

cd /path/to/telnyx-openclaw-stt
npm install && npm run build
openclaw plugins install /path/to/telnyx-openclaw-stt

Or from your current directory:

ocplatform plugins install .

Configuration

Minimal (env var)

export TELNYX_API_KEY=KEY...

Enable Telnyx for audio transcription

In openclaw.json:

{
  "tools": {
    "media": {
      "audio": {
        "enabled": true,
        "models": [
          {
            "provider": "telnyx",
            "model": "openai/whisper-large-v3-turbo"
          }
        ]
      }
    }
  }
}

Optional language hint and provider flags

{
  "tools": {
    "media": {
      "audio": {
        "prompt": "Expect OpenClaw and Telnyx product names",
        "providerOptions": {
          "telnyx": {
            "temperature": 0,
            "timestamps": true
          }
        },
        "models": [
          {
            "provider": "telnyx",
            "model": "openai/whisper-large-v3-turbo",
            "language": "en"
          }
        ]
      }
    }
  }
}

Local development

git clone https://github.com/team-telnyx/telnyx-openclaw-stt.git
cd telnyx-openclaw-stt
npm install
npm run build
npm test
npm run lint

Live smoke test

The repo includes a small WAV fixture (fixtures/sample.wav) for testing.

export TELNYX_API_KEY=KEY...
node --input-type=module <<'EOF'
import { readFileSync } from 'node:fs';
const sample = readFileSync('./fixtures/sample.wav');
const form = new FormData();
form.set('model', 'openai/whisper-large-v3-turbo');
form.set('file', new Blob([sample], { type: 'audio/wav' }), 'sample.wav');
const res = await fetch('https://api.telnyx.com/v2/ai/audio/transcriptions', {
  method: 'POST',
  headers: { Authorization: `Bearer ${process.env.TELNYX_API_KEY}` },
  body: form,
});
if (!res.ok) throw new Error(`Telnyx API error ${res.status}: ${await res.text()}`);
const data = await res.json();
console.log(data.text);
EOF

Unit Tests

npm install
npm test
npm run lint

How it works

• Builds multipart form data with model, language, prompt, and extra provider flags
• Uploads audio to Telnyx STT
• Reads text from the response payload
• Returns the transcript through OpenClaw's media-understanding contract

Files

• index.ts — plugin entry
• src/audio.ts — transcription transport logic
• src/media-understanding-provider.ts — provider registration object
• src/audio.test.ts — unit tests
• fixtures/sample.wav — test audio fixture (440 Hz sine, 0.5s)

License

MIT