OpenClaw Gladia

Gladia batch speech-to-text provider for OpenClaw media understanding.

This plugin registers the gladia media-understanding provider for audio transcription. It uses Gladia v2 pre-recorded transcription:

1. POST /v2/upload
2. POST /v2/pre-recorded
3. poll GET /v2/pre-recorded/{id} until done

Install Locally

cd openclaw-gladia
openclaw plugins install .
openclaw gateway restart

Configure

Expose the Gladia API key to the Gateway:

GLADIA_API_KEY=gladia_xxx

Then configure audio media understanding in ~/.openclaw/openclaw.json:

{
  tools: {
    media: {
      audio: {
        enabled: true,
        maxBytes: 20971520,
        models: [
          {
            provider: "gladia",
            model: "solaria-1",
            language: "fr"
          }
        ]
      }
    }
  }
}

Use

openclaw infer audio transcribe --file ./memo.m4a --model gladia/solaria-1 --json

Provider options can be passed through OpenClaw tools.media.audio provider query/options. Supported primitive query fields:

• languages: comma-separated language hints, for example fr,en
• code_switching: boolean
• diarization: boolean
• sentences: boolean
• punctuation_enhanced: boolean
• poll_interval_ms: polling interval, default 3000

Development

npm install
npm test
npm run pack:check

This package is intentionally small: the runtime entry registers one OpenClaw media-understanding provider named gladia, and the provider implementation contains the Gladia v2 upload/job/polling flow.

Publishing Notes

Before publishing:

• confirm the package name and npm ownership;
• add repository metadata to package.json once the GitHub URL exists;
• publish to npm;
• register the plugin in the OpenClaw plugin catalog or ClawHub flow with the manifest id gladia.