AgentListenConfig
STT (speech-to-text) configuration for the Agent API listen stage.
Defined in: src/providers/agent/deepgram/types.ts:404
STT (speech-to-text) configuration for the Agent API listen stage.
Remarks
The listen stage is always powered by Deepgram’s own STT models (Nova / Flux). This config controls model selection, language, and formatting options.
Properties
| Property | Type | Description | Defined in |
|---|---|---|---|
provider? | { keyterms?: string[]; language?: string; model?: string; smart_format?: boolean; type: "deepgram"; version?: "v1" | "v2"; } | Deepgram STT provider settings. | src/providers/agent/deepgram/types.ts:406 |
provider.keyterms? | string[] | Key terms to boost during recognition. Remarks Improves accuracy for domain-specific vocabulary, proper nouns, or uncommon words by biasing the model towards these terms. | src/providers/agent/deepgram/types.ts:422 |
provider.language? | string | BCP-47 language code (e.g. 'en', 'es'). | src/providers/agent/deepgram/types.ts:414 |
provider.model? | string | STT model ID (e.g. 'nova-3'). | src/providers/agent/deepgram/types.ts:412 |
provider.smart_format? | boolean | Enable smart formatting of transcription output. Remarks When true, Deepgram applies automatic formatting such as punctuation, numerals, and dates. | src/providers/agent/deepgram/types.ts:430 |
provider.type | "deepgram" | Provider identifier, must be 'deepgram'. | src/providers/agent/deepgram/types.ts:408 |
provider.version? | "v1" | "v2" | STT API version. | src/providers/agent/deepgram/types.ts:410 |