@genkit-ai/compat-oai/lib/openai/whisper.mjs

Genkit AI framework plugin for OpenAI APIs.

import { modelRef, z } from "genkit";
import { model } from "genkit/plugin";
import {
  TranscriptionConfigSchema,
  toSttRequest,
  transcriptionToGenerateResponse
} from "../audio.mjs";
import {
  toTranslationRequest,
  translationToGenerateResponse
} from "../translate.mjs";
import { maybeCreateRequestScopedOpenAIClient, toModelName } from "../utils.mjs";
const WHISPER_MODEL_INFO = {
  supports: {
    media: true,
    output: ["text", "json"],
    multiturn: false,
    systemRole: false,
    tools: false
  }
};
const WhisperConfigSchema = TranscriptionConfigSchema.extend({
  /** When true, uses Translation API instead of Transcription. Default: false */
  translate: z.boolean().optional().default(false)
});
function defineOpenAIWhisperModel(params) {
  const { name, client: defaultClient, pluginOptions, modelRef: modelRef2 } = params;
  const modelName = toModelName(name, pluginOptions?.name);
  const actionName = modelRef2?.name ?? `${pluginOptions?.name ?? "openai"}/${modelName}`;
  return model(
    {
      name: actionName,
      ...modelRef2?.info,
      configSchema: modelRef2?.configSchema
    },
    async (request, { abortSignal }) => {
      const { translate, ...cleanConfig } = request.config ?? {};
      const cleanRequest = { ...request, config: cleanConfig };
      const client = maybeCreateRequestScopedOpenAIClient(
        pluginOptions,
        request,
        defaultClient
      );
      if (translate === true) {
        const params2 = toTranslationRequest(modelName, cleanRequest);
        const result = await client.audio.translations.create(params2, {
          signal: abortSignal
        });
        return translationToGenerateResponse(result);
      } else {
        const params2 = toSttRequest(modelName, cleanRequest);
        const result = await client.audio.transcriptions.create(
          {
            ...params2,
            stream: false
          },
          { signal: abortSignal }
        );
        return transcriptionToGenerateResponse(result);
      }
    }
  );
}
function openAIWhisperModelRef(params) {
  const {
    name,
    info = WHISPER_MODEL_INFO,
    configSchema,
    config = void 0
  } = params;
  return modelRef({
    name,
    configSchema: configSchema || WhisperConfigSchema,
    info,
    config,
    namespace: "openai"
  });
}
const SUPPORTED_WHISPER_MODELS = {
  "whisper-1": openAIWhisperModelRef({
    name: "whisper-1"
  })
};
export {
  SUPPORTED_WHISPER_MODELS,
  WHISPER_MODEL_INFO,
  WhisperConfigSchema,
  defineOpenAIWhisperModel,
  openAIWhisperModelRef
};
//# sourceMappingURL=whisper.mjs.map