Speech-to-text providers

A read-only catalog of every speech-to-text vendor the runtime knows about — what's usable right now, at what price, and whether it supports streaming. This mirrors the equivalent catalog endpoints for TTS (/guides/voices) and LLM providers: same shape, same purpose — it drives the agent builder's provider picker, and lets you make an informed choice without guessing at pricing or availability.

curl https://convostack.ai/api/stt/providers \
  -H "Authorization: Bearer cak_live_..."

[
  {
    "id": "deepgram",
    "displayName": "Deepgram",
    "description": "Real-time speech-to-text with sub-300ms latency, native VAD events (SpeechStarted + UtteranceEnd), and word-level confidence. The platform default — every existing agent uses this provider for live telephony.",
    "availability": "available",
    "supportsStreaming": true,
    "pricing": {
      "default": [{ "unit": "per_minute_audio", "usd": 0.0043 }],
      "byModel": {
        "nova-2": [{ "unit": "per_minute_audio", "usd": 0.0043, "tierLabel": "nova-2 streaming" }]
      }
    }
  }
]

availability is available, not_configured (the platform hasn't set up credentials for it), or not_implemented (listed for completeness, not yet wired into the runtime). There's no BYOK layer for STT today — unlike TTS, every org shares the platform's configured provider.

To choose a provider for an agent, set model.stt.provider in its definition to one of the ids returned here — see Agents.