Skip to main content

Interface: OpenAIWhisperSpeechToTextProviderConfig

Defined in: packages/agentos/src/io/hearing/providers/OpenAIWhisperSpeechToTextProvider.ts:17

Configuration for the OpenAIWhisperSpeechToTextProvider.

See​

Properties​

apiKey​

apiKey: string

Defined in: packages/agentos/src/io/hearing/providers/OpenAIWhisperSpeechToTextProvider.ts:22

OpenAI API key used for authentication. Sent as Authorization: Bearer <apiKey> in the request header.

baseUrl?​

optional baseUrl: string

Defined in: packages/agentos/src/io/hearing/providers/OpenAIWhisperSpeechToTextProvider.ts:29

Base URL for the OpenAI API. Override for proxies, Azure OpenAI, or compatible third-party endpoints.

Default​

'https://api.openai.com/v1'

fetchImpl()?​

optional fetchImpl: {(input, init?): Promise<Response>; (input, init?): Promise<Response>; }

Defined in: packages/agentos/src/io/hearing/providers/OpenAIWhisperSpeechToTextProvider.ts:51

Custom fetch implementation for dependency injection in tests.

Call Signature​

(input, init?): Promise<Response>

MDN Reference

Parameters​
input​

RequestInfo | URL

init?​

RequestInit

Returns​

Promise<Response>

Call Signature​

(input, init?): Promise<Response>

MDN Reference

Parameters​
input​

string | Request | URL

init?​

RequestInit

Returns​

Promise<Response>

Default​

globalThis.fetch

model?​

optional model: string

Defined in: packages/agentos/src/io/hearing/providers/OpenAIWhisperSpeechToTextProvider.ts:45

Default transcription model.

gpt-transcribe is the model OpenAI recommends for recorded speech. OpenAI removes whisper-1 from the API on 2027-02-26 and names gpt-live-transcribe or gpt-transcribe as the replacement, while its guide still sends callers who need word or segment timestamps to whisper-1. A call that asks for verbose_json, srt or vtt runs on whisper-1, which serves those formats, when no model is configured or the configured model is one of OpenAI's gpt- transcription models. A configured model of a whisper-compatible server keeps those calls, and a model named on the call is always used as given.

Default​

'gpt-transcribe'