createOpenAITranscription (src/transcription.ts)

* transcription.ts — speech-to-text (2.0). `transcribe()` is the STT twin of `generateImage`: one request, one response, no streaming, no loop. It resolves the key through the G1 chain (`deps.keyProvider` → factory config → `ClientConfig.apiKeys`), sends the audio through the inj

// Source: https://github.com/Deuz-AI/Deuz-SDK | License: MIT | Copyright 2026 Umutcan Edizaslan (U-C4N / Deuz)
// Provided via justclaw.it.com; keep this notice with the code.
/**
 * transcription.ts — speech-to-text (2.0).
 *
 * `transcribe()` is the STT twin of `generateImage`: one request, one response,
 * no streaming, no loop. It resolves the key through the G1 chain
 * (`deps.keyProvider` → factory config → `ClientConfig.apiKeys`), sends the
 * audio through the injected `deps.fetch`, and wraps the whole call in
 * `operation.*` observation events under the `'transcription'` subsystem.
 *
 * Two wires ship (`src/adapters/transcription.ts`): OpenAI's multipart
 * `/audio/transcriptions` and Deepgram's raw-body `/listen`. They differ enough
 * that the differences are worth knowing be

... (truncated -- full source via MCP)