oclaw/openclaw/extensions/groq/media-understanding-provider.ts

import {
  transcribeOpenAiCompatibleAudio,
  type MediaUnderstandingProvider,
} from "openclaw/plugin-sdk/media-understanding";

const DEFAULT_GROQ_AUDIO_BASE_URL = "https://api.groq.com/openai/v1";
const DEFAULT_GROQ_AUDIO_MODEL = "whisper-large-v3-turbo";

export const groqMediaUnderstandingProvider: MediaUnderstandingProvider = {
  id: "groq",
  capabilities: ["audio"],
  defaultModels: { audio: DEFAULT_GROQ_AUDIO_MODEL },
  autoPriority: { audio: 20 },
  transcribeAudio: (req) =>
    transcribeOpenAiCompatibleAudio({
      ...req,
      baseUrl: req.baseUrl ?? DEFAULT_GROQ_AUDIO_BASE_URL,
      defaultBaseUrl: DEFAULT_GROQ_AUDIO_BASE_URL,
      defaultModel: DEFAULT_GROQ_AUDIO_MODEL,
    }),
};