Interface: ISpeechToTextProvider

Defined in: packages/agentos/src/cognition/rag/multimodal/types.ts:344

Minimal interface for a speech-to-text provider.

This is kept intentionally narrow to avoid coupling the multimodal indexer to a specific STT service. Any service that can transcribe audio buffers satisfies this contract.

Example

const sttProvider: ISpeechToTextProvider = {
  transcribe: async (audio, language) => {
    const response = await openai.audio.transcriptions.create({
      model: 'whisper-1',
      file: audio,
      language,
    });
    return response.text;
  },
};

Methods

transcribe()

transcribe(audio, language?): Promise<string>

Defined in: packages/agentos/src/cognition/rag/multimodal/types.ts:352

Transcribe audio data to text.

Parameters

audio

Buffer

Raw audio data as a Buffer.

language?

string

Optional BCP-47 language code hint.

Returns

Promise<string>

The transcribed text.