Deepgram

Deepgram is a speech-to-text API. OpenClaw uses it for inbound audio/voice-note transcription through tools.media.audio and for Voice Call streaming STT through plugins.entries.voice-call.config.streaming.

Batch transcription uploads the complete audio file to Deepgram. Flux models use Deepgram's one-shot WebSocket instead. Both paths inject the transcript into the reply pipeline ({{Transcript}} + [Audio] block). Voice Call streaming forwards live G.711 u-law frames over Deepgram's WebSocket listen endpoint and emits partial/final transcripts as Deepgram returns them.

Getting started

DEEPGRAM_API_KEY=dg_...

{  tools: {    media: {      models: [{ provider: "deepgram", model: "nova-3", capabilities: ["audio"] }],      audio: {        enabled: true,      },    },  },}

Configuration options

providerOptions.deepgram merges extra query params directly into the Deepgram /listen request, so any Deepgram-supported param name works (for example detect_language, punctuate, smart_format):

{  tools: {    media: {      models: [        { provider: "deepgram", model: "nova-3", language: "en", capabilities: ["audio"] },      ],      audio: {        enabled: true,      },    },  },}

{  tools: {    media: {      models: [{ provider: "deepgram", model: "nova-3", capabilities: ["audio"] }],      audio: {        enabled: true,        providerOptions: {          deepgram: {            detect_language: true,            punctuate: true,            smart_format: true,          },        },      },    },  },}

Flux models

Use flux-general-en or flux-general-multi for Deepgram Flux. OpenClaw converts the voice note to 16 kHz mono linear16 audio with ffmpeg, then sends it to Deepgram's /v2/listen WebSocket endpoint. Set tools.media.models[].model to either Flux model in the getting-started configuration above.

Flux supports eager_eot_threshold, eot_threshold, eot_timeout_ms, keyterm, language_hint, mip_opt_out, numerals, profanity_filter, redact, and tag in providerOptions.deepgram. OpenClaw ignores batch-only options such as detect_language, punctuate, and smart_format on Flux. Language hints apply only to flux-general-multi: OpenClaw maps the model entry's language setting to language_hint, with an explicit provider option taking precedence. Both settings are ignored for the English-only flux-general-en model.

Voice Call streaming STT

The bundled deepgram plugin also registers a realtime transcription provider for the Voice Call plugin.

{  plugins: {    entries: {      "voice-call": {        config: {          streaming: {            enabled: true,            provider: "deepgram",            providers: {              deepgram: {                apiKey: "${DEEPGRAM_API_KEY}",                model: "nova-3",                endpointingMs: 800,                language: "en-US",              },            },          },        },      },    },  },}

For a Deepgram custom endpoint, set baseUrl to the endpoint root, including any base path but not /listen. Realtime endpoints accept http://, https://, ws://, and wss://. HTTP maps to WS, HTTPS maps to WSS, and explicit WebSocket schemes stay unchanged. Malformed URLs and other schemes fail during session setup.

Notes

Related