@genkit-ai/compat-oai 1.28.0 → 1.29.0

package/lib/{audio-CUuCwm-y.d.mts → audio-CJ8rzf35.d.mts}

@@ -1,7 +1,8 @@
 import { EmbedderReference, EmbedderAction, GenerateRequest, z, ModelReference, StreamingCallback, GenerateResponseChunkData, GenerateResponseData, Role, Part, MessageData, ToolRequestPart, ActionMetadata } from 'genkit';
 import { ModelInfo, ModelAction, ToolDefinition } from 'genkit/model';
 import OpenAI, { ClientOptions } from 'openai';
-import { SpeechCreateParams, TranscriptionCreateParams } from 'openai/resources/audio/index.mjs';
+import { Response } from 'openai/core.mjs';
+import { SpeechCreateParams, TranscriptionCreateParams, Transcription } from 'openai/resources/audio/index.mjs';
 import * as genkit_plugin from 'genkit/plugin';
 import { ResolvableAction } from 'genkit/plugin';
 import { ActionType } from 'genkit/registry';
@@ -476,6 +477,8 @@ declare const RESPONSE_FORMAT_MEDIA_TYPES: {
     wav: string;
     pcm: string;
 };
+declare function toTTSRequest(modelName: string, request: GenerateRequest, requestBuilder?: SpeechRequestBuilder): SpeechCreateParams;
+declare function speechToGenerateResponse(response: Response, responseFormat?: 'mp3' | 'opus' | 'aac' | 'flac' | 'wav' | 'pcm'): Promise<GenerateResponseData>;
 /**
  * Method to define a new Genkit Model that is compatible with the Open AI Audio
  * API.
@@ -507,6 +510,8 @@ declare function compatOaiSpeechModelRef<CustomOptions extends z.ZodTypeAny = z.
     config?: any;
     namespace?: string;
 }): ModelReference<any>;
+declare function toSttRequest(modelName: string, request: GenerateRequest, requestBuilder?: TranscriptionRequestBuilder): TranscriptionCreateParams;
+declare function transcriptionToGenerateResponse(result: Transcription | string): GenerateResponseData;
 /**
  * Method to define a new Genkit Model that is compatible with Open AI
  * Transcriptions API.
@@ -540,4 +545,4 @@ declare function compatOaiTranscriptionModelRef<CustomOptions extends z.ZodTypeA
     namespace?: string;
 }): ModelReference<any>;
 
-export { ChatCompletionCommonConfigSchema as C, ImageGenerationCommonConfigSchema as I, type ModelRequestBuilder as M, type PluginOptions as P, RESPONSE_FORMAT_MEDIA_TYPES as R, SpeechConfigSchema as S, TranscriptionConfigSchema as T, type ImageRequestBuilder as a, IMAGE_GENERATION_MODEL_INFO as b, defineCompatOpenAIImageModel as c, defineCompatOpenAIEmbedder as d, compatOaiImageModelRef as e, toOpenAITool as f, toOpenAITextAndMedia as g, toOpenAIMessages as h, fromOpenAIToolCall as i, fromOpenAIChoice as j, fromOpenAIChunkChoice as k, toOpenAIRequestBody as l, defineCompatOpenAIModel as m, compatOaiModelRef as n, openAIModelRunner as o, openAICompatible as p, compatOaiSpeechModelRef as q, compatOaiTranscriptionModelRef as r, defineCompatOpenAISpeechModel as s, toOpenAIRole as t, defineCompatOpenAITranscriptionModel as u, type SpeechRequestBuilder as v, type TranscriptionRequestBuilder as w, TRANSCRIPTION_MODEL_INFO as x, SPEECH_MODEL_INFO as y };
+export { speechToGenerateResponse as A, toSttRequest as B, ChatCompletionCommonConfigSchema as C, transcriptionToGenerateResponse as D, ImageGenerationCommonConfigSchema as I, type ModelRequestBuilder as M, type PluginOptions as P, RESPONSE_FORMAT_MEDIA_TYPES as R, SpeechConfigSchema as S, TranscriptionConfigSchema as T, type ImageRequestBuilder as a, IMAGE_GENERATION_MODEL_INFO as b, defineCompatOpenAIImageModel as c, defineCompatOpenAIEmbedder as d, compatOaiImageModelRef as e, toOpenAITool as f, toOpenAITextAndMedia as g, toOpenAIMessages as h, fromOpenAIToolCall as i, fromOpenAIChoice as j, fromOpenAIChunkChoice as k, toOpenAIRequestBody as l, defineCompatOpenAIModel as m, compatOaiModelRef as n, openAIModelRunner as o, openAICompatible as p, compatOaiSpeechModelRef as q, compatOaiTranscriptionModelRef as r, defineCompatOpenAISpeechModel as s, toOpenAIRole as t, defineCompatOpenAITranscriptionModel as u, type SpeechRequestBuilder as v, type TranscriptionRequestBuilder as w, TRANSCRIPTION_MODEL_INFO as x, SPEECH_MODEL_INFO as y, toTTSRequest as z };

package/lib/{audio-CUuCwm-y.d.ts → audio-CJ8rzf35.d.ts}

@@ -1,7 +1,8 @@
 import { EmbedderReference, EmbedderAction, GenerateRequest, z, ModelReference, StreamingCallback, GenerateResponseChunkData, GenerateResponseData, Role, Part, MessageData, ToolRequestPart, ActionMetadata } from 'genkit';
 import { ModelInfo, ModelAction, ToolDefinition } from 'genkit/model';
 import OpenAI, { ClientOptions } from 'openai';
-import { SpeechCreateParams, TranscriptionCreateParams } from 'openai/resources/audio/index.mjs';
+import { Response } from 'openai/core.mjs';
+import { SpeechCreateParams, TranscriptionCreateParams, Transcription } from 'openai/resources/audio/index.mjs';
 import * as genkit_plugin from 'genkit/plugin';
 import { ResolvableAction } from 'genkit/plugin';
 import { ActionType } from 'genkit/registry';
@@ -476,6 +477,8 @@ declare const RESPONSE_FORMAT_MEDIA_TYPES: {
     wav: string;
     pcm: string;
 };
+declare function toTTSRequest(modelName: string, request: GenerateRequest, requestBuilder?: SpeechRequestBuilder): SpeechCreateParams;
+declare function speechToGenerateResponse(response: Response, responseFormat?: 'mp3' | 'opus' | 'aac' | 'flac' | 'wav' | 'pcm'): Promise<GenerateResponseData>;
 /**
  * Method to define a new Genkit Model that is compatible with the Open AI Audio
  * API.
@@ -507,6 +510,8 @@ declare function compatOaiSpeechModelRef<CustomOptions extends z.ZodTypeAny = z.
     config?: any;
     namespace?: string;
 }): ModelReference<any>;
+declare function toSttRequest(modelName: string, request: GenerateRequest, requestBuilder?: TranscriptionRequestBuilder): TranscriptionCreateParams;
+declare function transcriptionToGenerateResponse(result: Transcription | string): GenerateResponseData;
 /**
  * Method to define a new Genkit Model that is compatible with Open AI
  * Transcriptions API.
@@ -540,4 +545,4 @@ declare function compatOaiTranscriptionModelRef<CustomOptions extends z.ZodTypeA
     namespace?: string;
 }): ModelReference<any>;
 
-export { ChatCompletionCommonConfigSchema as C, ImageGenerationCommonConfigSchema as I, type ModelRequestBuilder as M, type PluginOptions as P, RESPONSE_FORMAT_MEDIA_TYPES as R, SpeechConfigSchema as S, TranscriptionConfigSchema as T, type ImageRequestBuilder as a, IMAGE_GENERATION_MODEL_INFO as b, defineCompatOpenAIImageModel as c, defineCompatOpenAIEmbedder as d, compatOaiImageModelRef as e, toOpenAITool as f, toOpenAITextAndMedia as g, toOpenAIMessages as h, fromOpenAIToolCall as i, fromOpenAIChoice as j, fromOpenAIChunkChoice as k, toOpenAIRequestBody as l, defineCompatOpenAIModel as m, compatOaiModelRef as n, openAIModelRunner as o, openAICompatible as p, compatOaiSpeechModelRef as q, compatOaiTranscriptionModelRef as r, defineCompatOpenAISpeechModel as s, toOpenAIRole as t, defineCompatOpenAITranscriptionModel as u, type SpeechRequestBuilder as v, type TranscriptionRequestBuilder as w, TRANSCRIPTION_MODEL_INFO as x, SPEECH_MODEL_INFO as y };
+export { speechToGenerateResponse as A, toSttRequest as B, ChatCompletionCommonConfigSchema as C, transcriptionToGenerateResponse as D, ImageGenerationCommonConfigSchema as I, type ModelRequestBuilder as M, type PluginOptions as P, RESPONSE_FORMAT_MEDIA_TYPES as R, SpeechConfigSchema as S, TranscriptionConfigSchema as T, type ImageRequestBuilder as a, IMAGE_GENERATION_MODEL_INFO as b, defineCompatOpenAIImageModel as c, defineCompatOpenAIEmbedder as d, compatOaiImageModelRef as e, toOpenAITool as f, toOpenAITextAndMedia as g, toOpenAIMessages as h, fromOpenAIToolCall as i, fromOpenAIChoice as j, fromOpenAIChunkChoice as k, toOpenAIRequestBody as l, defineCompatOpenAIModel as m, compatOaiModelRef as n, openAIModelRunner as o, openAICompatible as p, compatOaiSpeechModelRef as q, compatOaiTranscriptionModelRef as r, defineCompatOpenAISpeechModel as s, toOpenAIRole as t, defineCompatOpenAITranscriptionModel as u, type SpeechRequestBuilder as v, type TranscriptionRequestBuilder as w, TRANSCRIPTION_MODEL_INFO as x, SPEECH_MODEL_INFO as y, toTTSRequest as z };

package/lib/audio.d.mts

@@ -1,8 +1,9 @@
 import 'genkit';
 import 'genkit/model';
 import 'openai';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
-export { R as RESPONSE_FORMAT_MEDIA_TYPES, y as SPEECH_MODEL_INFO, S as SpeechConfigSchema, v as SpeechRequestBuilder, x as TRANSCRIPTION_MODEL_INFO, T as TranscriptionConfigSchema, w as TranscriptionRequestBuilder, q as compatOaiSpeechModelRef, r as compatOaiTranscriptionModelRef, s as defineCompatOpenAISpeechModel, u as defineCompatOpenAITranscriptionModel } from './audio-CUuCwm-y.mjs';
+export { R as RESPONSE_FORMAT_MEDIA_TYPES, y as SPEECH_MODEL_INFO, S as SpeechConfigSchema, v as SpeechRequestBuilder, x as TRANSCRIPTION_MODEL_INFO, T as TranscriptionConfigSchema, w as TranscriptionRequestBuilder, q as compatOaiSpeechModelRef, r as compatOaiTranscriptionModelRef, s as defineCompatOpenAISpeechModel, u as defineCompatOpenAITranscriptionModel, A as speechToGenerateResponse, B as toSttRequest, z as toTTSRequest, D as transcriptionToGenerateResponse } from './audio-CJ8rzf35.mjs';
 import 'genkit/plugin';
 import 'genkit/registry';
 import 'openai/resources/images.mjs';

package/lib/audio.d.ts

@@ -1,8 +1,9 @@
 import 'genkit';
 import 'genkit/model';
 import 'openai';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
-export { R as RESPONSE_FORMAT_MEDIA_TYPES, y as SPEECH_MODEL_INFO, S as SpeechConfigSchema, v as SpeechRequestBuilder, x as TRANSCRIPTION_MODEL_INFO, T as TranscriptionConfigSchema, w as TranscriptionRequestBuilder, q as compatOaiSpeechModelRef, r as compatOaiTranscriptionModelRef, s as defineCompatOpenAISpeechModel, u as defineCompatOpenAITranscriptionModel } from './audio-CUuCwm-y.js';
+export { R as RESPONSE_FORMAT_MEDIA_TYPES, y as SPEECH_MODEL_INFO, S as SpeechConfigSchema, v as SpeechRequestBuilder, x as TRANSCRIPTION_MODEL_INFO, T as TranscriptionConfigSchema, w as TranscriptionRequestBuilder, q as compatOaiSpeechModelRef, r as compatOaiTranscriptionModelRef, s as defineCompatOpenAISpeechModel, u as defineCompatOpenAITranscriptionModel, A as speechToGenerateResponse, B as toSttRequest, z as toTTSRequest, D as transcriptionToGenerateResponse } from './audio-CJ8rzf35.js';
 import 'genkit/plugin';
 import 'genkit/registry';
 import 'openai/resources/images.mjs';

package/lib/audio.js

@@ -26,7 +26,11 @@ __export(audio_exports, {
   compatOaiSpeechModelRef: () => compatOaiSpeechModelRef,
   compatOaiTranscriptionModelRef: () => compatOaiTranscriptionModelRef,
   defineCompatOpenAISpeechModel: () => defineCompatOpenAISpeechModel,
-  defineCompatOpenAITranscriptionModel: () => defineCompatOpenAITranscriptionModel
+  defineCompatOpenAITranscriptionModel: () => defineCompatOpenAITranscriptionModel,
+  speechToGenerateResponse: () => speechToGenerateResponse,
+  toSttRequest: () => toSttRequest,
+  toTTSRequest: () => toTTSRequest,
+  transcriptionToGenerateResponse: () => transcriptionToGenerateResponse
 });
 module.exports = __toCommonJS(audio_exports);
 var import_genkit = require("genkit");
@@ -111,7 +115,7 @@ function toTTSRequest(modelName, request, requestBuilder) {
   }
   return options;
 }
-async function toGenerateResponse(response, responseFormat = "mp3") {
+async function speechToGenerateResponse(response, responseFormat = "mp3") {
   const resultArrayBuffer = await response.arrayBuffer();
   const resultBuffer = Buffer.from(new Uint8Array(resultArrayBuffer));
   const mediaType = RESPONSE_FORMAT_MEDIA_TYPES[responseFormat];
@@ -157,7 +161,7 @@ function defineCompatOpenAISpeechModel(params) {
       const result = await client.audio.speech.create(ttsRequest, {
         signal: abortSignal
       });
-      return await toGenerateResponse(result, ttsRequest.response_format);
+      return await speechToGenerateResponse(result, ttsRequest.response_format);
     }
   );
 }
@@ -250,12 +254,13 @@ function transcriptionToGenerateResponse(result) {
 }
 function defineCompatOpenAITranscriptionModel(params) {
   const {
-    name: modelName,
+    name,
     pluginOptions,
     client: defaultClient,
     modelRef: modelRef2,
     requestBuilder
   } = params;
+  const modelName = (0, import_utils.toModelName)(name, pluginOptions?.name);
   const actionName = modelRef2?.name ?? `${pluginOptions?.name ?? "compat-oai"}/${modelName}`;
   return (0, import_plugin.model)(
     {
@@ -307,6 +312,10 @@ function compatOaiTranscriptionModelRef(params) {
   compatOaiSpeechModelRef,
   compatOaiTranscriptionModelRef,
   defineCompatOpenAISpeechModel,
-  defineCompatOpenAITranscriptionModel
+  defineCompatOpenAITranscriptionModel,
+  speechToGenerateResponse,
+  toSttRequest,
+  toTTSRequest,
+  transcriptionToGenerateResponse
 });
 //# sourceMappingURL=audio.js.map

package/lib/audio.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/audio.ts"],"sourcesContent":["/**\n * Copyright 2024 The Fire Company\n * Copyright 2024 Google LLC\n *\n * Licensed under the Apache License, Version 2.0 (the \"License\");\n * you may not use this file except in compliance with the License.\n * You may obtain a copy of the License at\n *\n *     http://www.apache.org/licenses/LICENSE-2.0\n *\n * Unless required by applicable law or agreed to in writing, software\n * distributed under the License is distributed on an \"AS IS\" BASIS,\n * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n * See the License for the specific language governing permissions and\n * limitations under the License.\n */\nimport type {\n  GenerateRequest,\n  GenerateResponseData,\n  ModelReference,\n} from 'genkit';\nimport { GenerationCommonConfigSchema, Message, modelRef, z } from 'genkit';\nimport type { ModelAction, ModelInfo } from 'genkit/model';\nimport { model } from 'genkit/plugin';\nimport OpenAI from 'openai';\nimport { Response } from 'openai/core.mjs';\nimport type {\n  SpeechCreateParams,\n  Transcription,\n  TranscriptionCreateParams,\n} from 'openai/resources/audio/index.mjs';\nimport { PluginOptions } from './index.js';\nimport { maybeCreateRequestScopedOpenAIClient, toModelName } from './utils.js';\n\nexport type SpeechRequestBuilder = (\n  req: GenerateRequest,\n  params: SpeechCreateParams\n) => void;\nexport type TranscriptionRequestBuilder = (\n  req: GenerateRequest,\n  params: TranscriptionCreateParams\n) => void;\n\nexport const TRANSCRIPTION_MODEL_INFO = {\n  supports: {\n    media: true,\n    output: ['text', 'json'],\n    multiturn: false,\n    systemRole: false,\n    tools: false,\n  },\n};\n\nexport const SPEECH_MODEL_INFO: ModelInfo = {\n  supports: {\n    media: false,\n    output: ['media'],\n    multiturn: false,\n    systemRole: false,\n    tools: false,\n  },\n};\n\nconst ChunkingStrategySchema = z.object({\n  type: z.string(),\n  prefix_padding_ms: z.number().int().optional(),\n  silence_duration_ms: z.number().int().optional(),\n  threshold: z.number().min(0).max(1.0).optional(),\n});\nexport const TranscriptionConfigSchema = GenerationCommonConfigSchema.pick({\n  temperature: true,\n}).extend({\n  chunking_strategy: z\n    .union([z.literal('auto'), ChunkingStrategySchema])\n    .optional(),\n  include: z.array(z.any()).optional(),\n  language: z.string().optional(),\n  timestamp_granularities: z.array(z.enum(['word', 'segment'])).optional(),\n  response_format: z\n    .enum(['json', 'text', 'srt', 'verbose_json', 'vtt'])\n    .optional(),\n  // TODO stream support\n});\n\nexport const SpeechConfigSchema = z.object({\n  voice: z\n    .enum(['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'])\n    .default('alloy'),\n  speed: z.number().min(0.25).max(4.0).optional(),\n  response_format: z\n    .enum(['mp3', 'opus', 'aac', 'flac', 'wav', 'pcm'])\n    .optional(),\n});\n\n/**\n * Supported media formats for Audio generation\n */\nexport const RESPONSE_FORMAT_MEDIA_TYPES = {\n  mp3: 'audio/mpeg',\n  opus: 'audio/opus',\n  aac: 'audio/aac',\n  flac: 'audio/flac',\n  wav: 'audio/wav',\n  pcm: 'audio/L16',\n};\n\nfunction toTTSRequest(\n  modelName: string,\n  request: GenerateRequest,\n  requestBuilder?: SpeechRequestBuilder\n): SpeechCreateParams {\n  const {\n    voice,\n    version: modelVersion,\n    temperature,\n    maxOutputTokens,\n    stopSequences,\n    topK,\n    topP,\n    ...restOfConfig\n  } = request.config ?? {};\n\n  let options: SpeechCreateParams = {\n    model: modelVersion ?? modelName,\n    input: new Message(request.messages[0]).text,\n    voice: voice ?? 'alloy',\n  };\n  if (requestBuilder) {\n    requestBuilder(request, options);\n  } else {\n    options = {\n      ...options,\n      ...restOfConfig, // passthorugh rest of the config\n    };\n  }\n  for (const k in options) {\n    if (options[k] === undefined) {\n      delete options[k];\n    }\n  }\n  return options;\n}\n\nasync function toGenerateResponse(\n  response: Response,\n  responseFormat: 'mp3' | 'opus' | 'aac' | 'flac' | 'wav' | 'pcm' = 'mp3'\n): Promise<GenerateResponseData> {\n  const resultArrayBuffer = await response.arrayBuffer();\n  const resultBuffer = Buffer.from(new Uint8Array(resultArrayBuffer));\n  const mediaType = RESPONSE_FORMAT_MEDIA_TYPES[responseFormat];\n  return {\n    message: {\n      role: 'model',\n      content: [\n        {\n          media: {\n            contentType: mediaType,\n            url: `data:${mediaType};base64,${resultBuffer.toString('base64')}`,\n          },\n        },\n      ],\n    },\n    finishReason: 'stop',\n    raw: response,\n  };\n}\n\n/**\n * Method to define a new Genkit Model that is compatible with the Open AI Audio\n * API. \n *\n * These models are to be used to create audio speech from a given request.\n * @param params An object containing parameters for defining the OpenAI speech\n * model.\n * @param params.ai The Genkit AI instance.\n * @param params.name The name of the model.\n * @param params.client The OpenAI client instance.\n * @param params.modelRef Optional reference to the model's configuration and\n * custom options.\n\n * @returns the created {@link ModelAction}\n */\nexport function defineCompatOpenAISpeechModel<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  client: OpenAI;\n  modelRef?: ModelReference<CustomOptions>;\n  requestBuilder?: SpeechRequestBuilder;\n  pluginOptions: PluginOptions;\n}): ModelAction {\n  const {\n    name,\n    client: defaultClient,\n    pluginOptions,\n    modelRef,\n    requestBuilder,\n  } = params;\n  const modelName = toModelName(name, pluginOptions?.name);\n  const actionName = `${pluginOptions?.name ?? 'compat-oai'}/${modelName}`;\n\n  return model(\n    {\n      name: actionName,\n      ...modelRef?.info,\n      configSchema: modelRef?.configSchema,\n    },\n    async (request, { abortSignal }) => {\n      const ttsRequest = toTTSRequest(modelName, request, requestBuilder);\n      const client = maybeCreateRequestScopedOpenAIClient(\n        pluginOptions,\n        request,\n        defaultClient\n      );\n      const result = await client.audio.speech.create(ttsRequest, {\n        signal: abortSignal,\n      });\n      return await toGenerateResponse(result, ttsRequest.response_format);\n    }\n  );\n}\n\n/** Speech generation ModelRef helper, with reasonable defaults for\n * OpenAI-compatible providers */\nexport function compatOaiSpeechModelRef<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  info?: ModelInfo;\n  configSchema?: CustomOptions;\n  config?: any;\n  namespace?: string;\n}) {\n  const {\n    name,\n    info = SPEECH_MODEL_INFO,\n    configSchema,\n    config = undefined,\n    namespace,\n  } = params;\n  return modelRef({\n    name,\n    configSchema: configSchema || (SpeechConfigSchema as any),\n    info,\n    config,\n    namespace,\n  });\n}\n\nfunction toSttRequest(\n  modelName: string,\n  request: GenerateRequest,\n  requestBuilder?: TranscriptionRequestBuilder\n): TranscriptionCreateParams {\n  const message = new Message(request.messages[0]);\n  const media = message.media;\n  if (!media?.url) {\n    throw new Error('No media found in the request');\n  }\n  const mediaBuffer = Buffer.from(\n    media.url.slice(media.url.indexOf(',') + 1),\n    'base64'\n  );\n  const mediaFile = new File([mediaBuffer], 'input', {\n    type:\n      media.contentType ??\n      media.url.slice('data:'.length, media.url.indexOf(';')),\n  });\n  const {\n    temperature,\n    version: modelVersion,\n    maxOutputTokens,\n    stopSequences,\n    topK,\n    topP,\n    ...restOfConfig\n  } = request.config ?? {};\n\n  let options: TranscriptionCreateParams = {\n    model: modelVersion ?? modelName,\n    file: mediaFile,\n    prompt: message.text,\n    temperature,\n  };\n  if (requestBuilder) {\n    requestBuilder(request, options);\n  } else {\n    options = {\n      ...options,\n      ...restOfConfig, // passthrough rest of the config\n    };\n  }\n  const outputFormat = request.output?.format as 'json' | 'text' | 'media';\n  const customFormat = request.config?.response_format;\n  if (outputFormat && customFormat) {\n    if (\n      outputFormat === 'json' &&\n      customFormat !== 'json' &&\n      customFormat !== 'verbose_json'\n    ) {\n      throw new Error(\n        `Custom response format ${customFormat} is not compatible with output format ${outputFormat}`\n      );\n    }\n  }\n  if (outputFormat === 'media') {\n    throw new Error(`Output format ${outputFormat} is not supported.`);\n  }\n  options.response_format = customFormat || outputFormat || 'text';\n  for (const k in options) {\n    if (options[k] === undefined) {\n      delete options[k];\n    }\n  }\n  return options;\n}\n\nfunction transcriptionToGenerateResponse(\n  result: Transcription | string\n): GenerateResponseData {\n  return {\n    message: {\n      role: 'model',\n      content: [\n        {\n          text: typeof result === 'string' ? result : result.text,\n        },\n      ],\n    },\n    finishReason: 'stop',\n    raw: result,\n  };\n}\n\n/**\n * Method to define a new Genkit Model that is compatible with Open AI\n * Transcriptions API. \n *\n * These models are to be used to transcribe audio to text.\n *\n * @param params An object containing parameters for defining the OpenAI\n * transcription model.\n * @param params.ai The Genkit AI instance.\n * @param params.name The name of the model.\n * @param params.client The OpenAI client instance.\n * @param params.modelRef Optional reference to the model's configuration and\n * custom options.\n\n * @returns the created {@link ModelAction}\n */\nexport function defineCompatOpenAITranscriptionModel<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  client: OpenAI;\n  pluginOptions?: PluginOptions;\n  modelRef?: ModelReference<CustomOptions>;\n  requestBuilder?: TranscriptionRequestBuilder;\n}): ModelAction {\n  const {\n    name: modelName,\n    pluginOptions,\n    client: defaultClient,\n    modelRef,\n    requestBuilder,\n  } = params;\n\n  const actionName =\n    modelRef?.name ?? `${pluginOptions?.name ?? 'compat-oai'}/${modelName}`;\n  return model(\n    {\n      name: actionName,\n      ...modelRef?.info,\n      configSchema: modelRef?.configSchema,\n    },\n    async (request, { abortSignal }) => {\n      const params = toSttRequest(modelName, request, requestBuilder);\n      const client = maybeCreateRequestScopedOpenAIClient(\n        pluginOptions,\n        request,\n        defaultClient\n      );\n      // Explicitly setting stream to false ensures we use the non-streaming overload\n      const result = await client.audio.transcriptions.create(\n        {\n          ...params,\n          stream: false,\n        },\n        { signal: abortSignal }\n      );\n      return transcriptionToGenerateResponse(result);\n    }\n  );\n}\n\n/** Transcription ModelRef helper, with reasonable defaults for\n * OpenAI-compatible providers */\nexport function compatOaiTranscriptionModelRef<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  info?: ModelInfo;\n  configSchema?: CustomOptions;\n  config?: any;\n  namespace?: string;\n}) {\n  const {\n    name,\n    info = TRANSCRIPTION_MODEL_INFO,\n    configSchema,\n    config = undefined,\n    namespace,\n  } = params;\n  return modelRef({\n    name,\n    configSchema: configSchema || (TranscriptionConfigSchema as any),\n    info,\n    config,\n    namespace,\n  });\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAqBA,oBAAmE;AAEnE,oBAAsB;AAStB,mBAAkE;AAW3D,MAAM,2BAA2B;AAAA,EACtC,UAAU;AAAA,IACR,OAAO;AAAA,IACP,QAAQ,CAAC,QAAQ,MAAM;AAAA,IACvB,WAAW;AAAA,IACX,YAAY;AAAA,IACZ,OAAO;AAAA,EACT;AACF;AAEO,MAAM,oBAA+B;AAAA,EAC1C,UAAU;AAAA,IACR,OAAO;AAAA,IACP,QAAQ,CAAC,OAAO;AAAA,IAChB,WAAW;AAAA,IACX,YAAY;AAAA,IACZ,OAAO;AAAA,EACT;AACF;AAEA,MAAM,yBAAyB,gBAAE,OAAO;AAAA,EACtC,MAAM,gBAAE,OAAO;AAAA,EACf,mBAAmB,gBAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EAC7C,qBAAqB,gBAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EAC/C,WAAW,gBAAE,OAAO,EAAE,IAAI,CAAC,EAAE,IAAI,CAAG,EAAE,SAAS;AACjD,CAAC;AACM,MAAM,4BAA4B,2CAA6B,KAAK;AAAA,EACzE,aAAa;AACf,CAAC,EAAE,OAAO;AAAA,EACR,mBAAmB,gBAChB,MAAM,CAAC,gBAAE,QAAQ,MAAM,GAAG,sBAAsB,CAAC,EACjD,SAAS;AAAA,EACZ,SAAS,gBAAE,MAAM,gBAAE,IAAI,CAAC,EAAE,SAAS;AAAA,EACnC,UAAU,gBAAE,OAAO,EAAE,SAAS;AAAA,EAC9B,yBAAyB,gBAAE,MAAM,gBAAE,KAAK,CAAC,QAAQ,SAAS,CAAC,CAAC,EAAE,SAAS;AAAA,EACvE,iBAAiB,gBACd,KAAK,CAAC,QAAQ,QAAQ,OAAO,gBAAgB,KAAK,CAAC,EACnD,SAAS;AAAA;AAEd,CAAC;AAEM,MAAM,qBAAqB,gBAAE,OAAO;AAAA,EACzC,OAAO,gBACJ,KAAK,CAAC,SAAS,QAAQ,SAAS,QAAQ,QAAQ,SAAS,CAAC,EAC1D,QAAQ,OAAO;AAAA,EAClB,OAAO,gBAAE,OAAO,EAAE,IAAI,IAAI,EAAE,IAAI,CAAG,EAAE,SAAS;AAAA,EAC9C,iBAAiB,gBACd,KAAK,CAAC,OAAO,QAAQ,OAAO,QAAQ,OAAO,KAAK,CAAC,EACjD,SAAS;AACd,CAAC;AAKM,MAAM,8BAA8B;AAAA,EACzC,KAAK;AAAA,EACL,MAAM;AAAA,EACN,KAAK;AAAA,EACL,MAAM;AAAA,EACN,KAAK;AAAA,EACL,KAAK;AACP;AAEA,SAAS,aACP,WACA,SACA,gBACoB;AACpB,QAAM;AAAA,IACJ;AAAA,IACA,SAAS;AAAA,IACT;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,GAAG;AAAA,EACL,IAAI,QAAQ,UAAU,CAAC;AAEvB,MAAI,UAA8B;AAAA,IAChC,OAAO,gBAAgB;AAAA,IACvB,OAAO,IAAI,sBAAQ,QAAQ,SAAS,CAAC,CAAC,EAAE;AAAA,IACxC,OAAO,SAAS;AAAA,EAClB;AACA,MAAI,gBAAgB;AAClB,mBAAe,SAAS,OAAO;AAAA,EACjC,OAAO;AACL,cAAU;AAAA,MACR,GAAG;AAAA,MACH,GAAG;AAAA;AAAA,IACL;AAAA,EACF;AACA,aAAW,KAAK,SAAS;AACvB,QAAI,QAAQ,CAAC,MAAM,QAAW;AAC5B,aAAO,QAAQ,CAAC;AAAA,IAClB;AAAA,EACF;AACA,SAAO;AACT;AAEA,eAAe,mBACb,UACA,iBAAkE,OACnC;AAC/B,QAAM,oBAAoB,MAAM,SAAS,YAAY;AACrD,QAAM,eAAe,OAAO,KAAK,IAAI,WAAW,iBAAiB,CAAC;AAClE,QAAM,YAAY,4BAA4B,cAAc;AAC5D,SAAO;AAAA,IACL,SAAS;AAAA,MACP,MAAM;AAAA,MACN,SAAS;AAAA,QACP;AAAA,UACE,OAAO;AAAA,YACL,aAAa;AAAA,YACb,KAAK,QAAQ,SAAS,WAAW,aAAa,SAAS,QAAQ,CAAC;AAAA,UAClE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,IACA,cAAc;AAAA,IACd,KAAK;AAAA,EACP;AACF;AAiBO,SAAS,8BAEd,QAMc;AACd,QAAM;AAAA,IACJ;AAAA,IACA,QAAQ;AAAA,IACR;AAAA,IACA,UAAAA;AAAA,IACA;AAAA,EACF,IAAI;AACJ,QAAM,gBAAY,0BAAY,MAAM,eAAe,IAAI;AACvD,QAAM,aAAa,GAAG,eAAe,QAAQ,YAAY,IAAI,SAAS;AAEtE,aAAO;AAAA,IACL;AAAA,MACE,MAAM;AAAA,MACN,GAAGA,WAAU;AAAA,MACb,cAAcA,WAAU;AAAA,IAC1B;AAAA,IACA,OAAO,SAAS,EAAE,YAAY,MAAM;AAClC,YAAM,aAAa,aAAa,WAAW,SAAS,cAAc;AAClE,YAAM,aAAS;AAAA,QACb;AAAA,QACA;AAAA,QACA;AAAA,MACF;AACA,YAAM,SAAS,MAAM,OAAO,MAAM,OAAO,OAAO,YAAY;AAAA,QAC1D,QAAQ;AAAA,MACV,CAAC;AACD,aAAO,MAAM,mBAAmB,QAAQ,WAAW,eAAe;AAAA,IACpE;AAAA,EACF;AACF;AAIO,SAAS,wBAEd,QAMC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,OAAO;AAAA,IACP;AAAA,IACA,SAAS;AAAA,IACT;AAAA,EACF,IAAI;AACJ,aAAO,wBAAS;AAAA,IACd;AAAA,IACA,cAAc,gBAAiB;AAAA,IAC/B;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AAEA,SAAS,aACP,WACA,SACA,gBAC2B;AAC3B,QAAM,UAAU,IAAI,sBAAQ,QAAQ,SAAS,CAAC,CAAC;AAC/C,QAAM,QAAQ,QAAQ;AACtB,MAAI,CAAC,OAAO,KAAK;AACf,UAAM,IAAI,MAAM,+BAA+B;AAAA,EACjD;AACA,QAAM,cAAc,OAAO;AAAA,IACzB,MAAM,IAAI,MAAM,MAAM,IAAI,QAAQ,GAAG,IAAI,CAAC;AAAA,IAC1C;AAAA,EACF;AACA,QAAM,YAAY,IAAI,KAAK,CAAC,WAAW,GAAG,SAAS;AAAA,IACjD,MACE,MAAM,eACN,MAAM,IAAI,MAAM,QAAQ,QAAQ,MAAM,IAAI,QAAQ,GAAG,CAAC;AAAA,EAC1D,CAAC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,SAAS;AAAA,IACT;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,GAAG;AAAA,EACL,IAAI,QAAQ,UAAU,CAAC;AAEvB,MAAI,UAAqC;AAAA,IACvC,OAAO,gBAAgB;AAAA,IACvB,MAAM;AAAA,IACN,QAAQ,QAAQ;AAAA,IAChB;AAAA,EACF;AACA,MAAI,gBAAgB;AAClB,mBAAe,SAAS,OAAO;AAAA,EACjC,OAAO;AACL,cAAU;AAAA,MACR,GAAG;AAAA,MACH,GAAG;AAAA;AAAA,IACL;AAAA,EACF;AACA,QAAM,eAAe,QAAQ,QAAQ;AACrC,QAAM,eAAe,QAAQ,QAAQ;AACrC,MAAI,gBAAgB,cAAc;AAChC,QACE,iBAAiB,UACjB,iBAAiB,UACjB,iBAAiB,gBACjB;AACA,YAAM,IAAI;AAAA,QACR,0BAA0B,YAAY,yCAAyC,YAAY;AAAA,MAC7F;AAAA,IACF;AAAA,EACF;AACA,MAAI,iBAAiB,SAAS;AAC5B,UAAM,IAAI,MAAM,iBAAiB,YAAY,oBAAoB;AAAA,EACnE;AACA,UAAQ,kBAAkB,gBAAgB,gBAAgB;AAC1D,aAAW,KAAK,SAAS;AACvB,QAAI,QAAQ,CAAC,MAAM,QAAW;AAC5B,aAAO,QAAQ,CAAC;AAAA,IAClB;AAAA,EACF;AACA,SAAO;AACT;AAEA,SAAS,gCACP,QACsB;AACtB,SAAO;AAAA,IACL,SAAS;AAAA,MACP,MAAM;AAAA,MACN,SAAS;AAAA,QACP;AAAA,UACE,MAAM,OAAO,WAAW,WAAW,SAAS,OAAO;AAAA,QACrD;AAAA,MACF;AAAA,IACF;AAAA,IACA,cAAc;AAAA,IACd,KAAK;AAAA,EACP;AACF;AAkBO,SAAS,qCAEd,QAMc;AACd,QAAM;AAAA,IACJ,MAAM;AAAA,IACN;AAAA,IACA,QAAQ;AAAA,IACR,UAAAA;AAAA,IACA;AAAA,EACF,IAAI;AAEJ,QAAM,aACJA,WAAU,QAAQ,GAAG,eAAe,QAAQ,YAAY,IAAI,SAAS;AACvE,aAAO;AAAA,IACL;AAAA,MACE,MAAM;AAAA,MACN,GAAGA,WAAU;AAAA,MACb,cAAcA,WAAU;AAAA,IAC1B;AAAA,IACA,OAAO,SAAS,EAAE,YAAY,MAAM;AAClC,YAAMC,UAAS,aAAa,WAAW,SAAS,cAAc;AAC9D,YAAM,aAAS;AAAA,QACb;AAAA,QACA;AAAA,QACA;AAAA,MACF;AAEA,YAAM,SAAS,MAAM,OAAO,MAAM,eAAe;AAAA,QAC/C;AAAA,UACE,GAAGA;AAAA,UACH,QAAQ;AAAA,QACV;AAAA,QACA,EAAE,QAAQ,YAAY;AAAA,MACxB;AACA,aAAO,gCAAgC,MAAM;AAAA,IAC/C;AAAA,EACF;AACF;AAIO,SAAS,+BAEd,QAMC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,OAAO;AAAA,IACP;AAAA,IACA,SAAS;AAAA,IACT;AAAA,EACF,IAAI;AACJ,aAAO,wBAAS;AAAA,IACd;AAAA,IACA,cAAc,gBAAiB;AAAA,IAC/B;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;","names":["modelRef","params"]}
+{"version":3,"sources":["../src/audio.ts"],"sourcesContent":["/**\n * Copyright 2024 The Fire Company\n * Copyright 2024 Google LLC\n *\n * Licensed under the Apache License, Version 2.0 (the \"License\");\n * you may not use this file except in compliance with the License.\n * You may obtain a copy of the License at\n *\n *     http://www.apache.org/licenses/LICENSE-2.0\n *\n * Unless required by applicable law or agreed to in writing, software\n * distributed under the License is distributed on an \"AS IS\" BASIS,\n * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n * See the License for the specific language governing permissions and\n * limitations under the License.\n */\nimport type {\n  GenerateRequest,\n  GenerateResponseData,\n  ModelReference,\n} from 'genkit';\nimport { GenerationCommonConfigSchema, Message, modelRef, z } from 'genkit';\nimport type { ModelAction, ModelInfo } from 'genkit/model';\nimport { model } from 'genkit/plugin';\nimport OpenAI from 'openai';\nimport { Response } from 'openai/core.mjs';\nimport type {\n  SpeechCreateParams,\n  Transcription,\n  TranscriptionCreateParams,\n} from 'openai/resources/audio/index.mjs';\nimport { PluginOptions } from './index.js';\nimport { maybeCreateRequestScopedOpenAIClient, toModelName } from './utils.js';\n\nexport type SpeechRequestBuilder = (\n  req: GenerateRequest,\n  params: SpeechCreateParams\n) => void;\nexport type TranscriptionRequestBuilder = (\n  req: GenerateRequest,\n  params: TranscriptionCreateParams\n) => void;\n\nexport const TRANSCRIPTION_MODEL_INFO = {\n  supports: {\n    media: true,\n    output: ['text', 'json'],\n    multiturn: false,\n    systemRole: false,\n    tools: false,\n  },\n};\n\nexport const SPEECH_MODEL_INFO: ModelInfo = {\n  supports: {\n    media: false,\n    output: ['media'],\n    multiturn: false,\n    systemRole: false,\n    tools: false,\n  },\n};\n\nconst ChunkingStrategySchema = z.object({\n  type: z.string(),\n  prefix_padding_ms: z.number().int().optional(),\n  silence_duration_ms: z.number().int().optional(),\n  threshold: z.number().min(0).max(1.0).optional(),\n});\nexport const TranscriptionConfigSchema = GenerationCommonConfigSchema.pick({\n  temperature: true,\n}).extend({\n  chunking_strategy: z\n    .union([z.literal('auto'), ChunkingStrategySchema])\n    .optional(),\n  include: z.array(z.any()).optional(),\n  language: z.string().optional(),\n  timestamp_granularities: z.array(z.enum(['word', 'segment'])).optional(),\n  response_format: z\n    .enum(['json', 'text', 'srt', 'verbose_json', 'vtt'])\n    .optional(),\n  // TODO stream support\n});\n\nexport const SpeechConfigSchema = z.object({\n  voice: z\n    .enum(['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'])\n    .default('alloy'),\n  speed: z.number().min(0.25).max(4.0).optional(),\n  response_format: z\n    .enum(['mp3', 'opus', 'aac', 'flac', 'wav', 'pcm'])\n    .optional(),\n});\n\n/**\n * Supported media formats for Audio generation\n */\nexport const RESPONSE_FORMAT_MEDIA_TYPES = {\n  mp3: 'audio/mpeg',\n  opus: 'audio/opus',\n  aac: 'audio/aac',\n  flac: 'audio/flac',\n  wav: 'audio/wav',\n  pcm: 'audio/L16',\n};\n\nexport function toTTSRequest(\n  modelName: string,\n  request: GenerateRequest,\n  requestBuilder?: SpeechRequestBuilder\n): SpeechCreateParams {\n  const {\n    voice,\n    version: modelVersion,\n    temperature,\n    maxOutputTokens,\n    stopSequences,\n    topK,\n    topP,\n    ...restOfConfig\n  } = request.config ?? {};\n\n  let options: SpeechCreateParams = {\n    model: modelVersion ?? modelName,\n    input: new Message(request.messages[0]).text,\n    voice: voice ?? 'alloy',\n  };\n  if (requestBuilder) {\n    requestBuilder(request, options);\n  } else {\n    options = {\n      ...options,\n      ...restOfConfig, // passthorugh rest of the config\n    };\n  }\n  for (const k in options) {\n    if (options[k] === undefined) {\n      delete options[k];\n    }\n  }\n  return options;\n}\n\nexport async function speechToGenerateResponse(\n  response: Response,\n  responseFormat: 'mp3' | 'opus' | 'aac' | 'flac' | 'wav' | 'pcm' = 'mp3'\n): Promise<GenerateResponseData> {\n  const resultArrayBuffer = await response.arrayBuffer();\n  const resultBuffer = Buffer.from(new Uint8Array(resultArrayBuffer));\n  const mediaType = RESPONSE_FORMAT_MEDIA_TYPES[responseFormat];\n  return {\n    message: {\n      role: 'model',\n      content: [\n        {\n          media: {\n            contentType: mediaType,\n            url: `data:${mediaType};base64,${resultBuffer.toString('base64')}`,\n          },\n        },\n      ],\n    },\n    finishReason: 'stop',\n    raw: response,\n  };\n}\n\n/**\n * Method to define a new Genkit Model that is compatible with the Open AI Audio\n * API. \n *\n * These models are to be used to create audio speech from a given request.\n * @param params An object containing parameters for defining the OpenAI speech\n * model.\n * @param params.ai The Genkit AI instance.\n * @param params.name The name of the model.\n * @param params.client The OpenAI client instance.\n * @param params.modelRef Optional reference to the model's configuration and\n * custom options.\n\n * @returns the created {@link ModelAction}\n */\nexport function defineCompatOpenAISpeechModel<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  client: OpenAI;\n  modelRef?: ModelReference<CustomOptions>;\n  requestBuilder?: SpeechRequestBuilder;\n  pluginOptions: PluginOptions;\n}): ModelAction {\n  const {\n    name,\n    client: defaultClient,\n    pluginOptions,\n    modelRef,\n    requestBuilder,\n  } = params;\n  const modelName = toModelName(name, pluginOptions?.name);\n  const actionName = `${pluginOptions?.name ?? 'compat-oai'}/${modelName}`;\n\n  return model(\n    {\n      name: actionName,\n      ...modelRef?.info,\n      configSchema: modelRef?.configSchema,\n    },\n    async (request, { abortSignal }) => {\n      const ttsRequest = toTTSRequest(modelName, request, requestBuilder);\n      const client = maybeCreateRequestScopedOpenAIClient(\n        pluginOptions,\n        request,\n        defaultClient\n      );\n      const result = await client.audio.speech.create(ttsRequest, {\n        signal: abortSignal,\n      });\n      return await speechToGenerateResponse(result, ttsRequest.response_format);\n    }\n  );\n}\n\n/** Speech generation ModelRef helper, with reasonable defaults for\n * OpenAI-compatible providers */\nexport function compatOaiSpeechModelRef<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  info?: ModelInfo;\n  configSchema?: CustomOptions;\n  config?: any;\n  namespace?: string;\n}) {\n  const {\n    name,\n    info = SPEECH_MODEL_INFO,\n    configSchema,\n    config = undefined,\n    namespace,\n  } = params;\n  return modelRef({\n    name,\n    configSchema: configSchema || (SpeechConfigSchema as any),\n    info,\n    config,\n    namespace,\n  });\n}\n\nexport function toSttRequest(\n  modelName: string,\n  request: GenerateRequest,\n  requestBuilder?: TranscriptionRequestBuilder\n): TranscriptionCreateParams {\n  const message = new Message(request.messages[0]);\n  const media = message.media;\n  if (!media?.url) {\n    throw new Error('No media found in the request');\n  }\n  const mediaBuffer = Buffer.from(\n    media.url.slice(media.url.indexOf(',') + 1),\n    'base64'\n  );\n  const mediaFile = new File([mediaBuffer], 'input', {\n    type:\n      media.contentType ??\n      media.url.slice('data:'.length, media.url.indexOf(';')),\n  });\n  const {\n    temperature,\n    version: modelVersion,\n    maxOutputTokens,\n    stopSequences,\n    topK,\n    topP,\n    ...restOfConfig\n  } = request.config ?? {};\n\n  let options: TranscriptionCreateParams = {\n    model: modelVersion ?? modelName,\n    file: mediaFile,\n    prompt: message.text,\n    temperature,\n  };\n  if (requestBuilder) {\n    requestBuilder(request, options);\n  } else {\n    options = {\n      ...options,\n      ...restOfConfig, // passthrough rest of the config\n    };\n  }\n  const outputFormat = request.output?.format as 'json' | 'text' | 'media';\n  const customFormat = request.config?.response_format;\n  if (outputFormat && customFormat) {\n    if (\n      outputFormat === 'json' &&\n      customFormat !== 'json' &&\n      customFormat !== 'verbose_json'\n    ) {\n      throw new Error(\n        `Custom response format ${customFormat} is not compatible with output format ${outputFormat}`\n      );\n    }\n  }\n  if (outputFormat === 'media') {\n    throw new Error(`Output format ${outputFormat} is not supported.`);\n  }\n  options.response_format = customFormat || outputFormat || 'text';\n  for (const k in options) {\n    if (options[k] === undefined) {\n      delete options[k];\n    }\n  }\n  return options;\n}\n\nexport function transcriptionToGenerateResponse(\n  result: Transcription | string\n): GenerateResponseData {\n  return {\n    message: {\n      role: 'model',\n      content: [\n        {\n          text: typeof result === 'string' ? result : result.text,\n        },\n      ],\n    },\n    finishReason: 'stop',\n    raw: result,\n  };\n}\n\n/**\n * Method to define a new Genkit Model that is compatible with Open AI\n * Transcriptions API. \n *\n * These models are to be used to transcribe audio to text.\n *\n * @param params An object containing parameters for defining the OpenAI\n * transcription model.\n * @param params.ai The Genkit AI instance.\n * @param params.name The name of the model.\n * @param params.client The OpenAI client instance.\n * @param params.modelRef Optional reference to the model's configuration and\n * custom options.\n\n * @returns the created {@link ModelAction}\n */\nexport function defineCompatOpenAITranscriptionModel<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  client: OpenAI;\n  pluginOptions?: PluginOptions;\n  modelRef?: ModelReference<CustomOptions>;\n  requestBuilder?: TranscriptionRequestBuilder;\n}): ModelAction {\n  const {\n    name,\n    pluginOptions,\n    client: defaultClient,\n    modelRef,\n    requestBuilder,\n  } = params;\n  const modelName = toModelName(name, pluginOptions?.name);\n  const actionName =\n    modelRef?.name ?? `${pluginOptions?.name ?? 'compat-oai'}/${modelName}`;\n\n  return model(\n    {\n      name: actionName,\n      ...modelRef?.info,\n      configSchema: modelRef?.configSchema,\n    },\n    async (request, { abortSignal }) => {\n      const params = toSttRequest(modelName, request, requestBuilder);\n      const client = maybeCreateRequestScopedOpenAIClient(\n        pluginOptions,\n        request,\n        defaultClient\n      );\n      // Explicitly setting stream to false ensures we use the non-streaming overload\n      const result = await client.audio.transcriptions.create(\n        {\n          ...params,\n          stream: false,\n        },\n        { signal: abortSignal }\n      );\n      return transcriptionToGenerateResponse(result);\n    }\n  );\n}\n\n/** Transcription ModelRef helper, with reasonable defaults for\n * OpenAI-compatible providers */\nexport function compatOaiTranscriptionModelRef<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  info?: ModelInfo;\n  configSchema?: CustomOptions;\n  config?: any;\n  namespace?: string;\n}) {\n  const {\n    name,\n    info = TRANSCRIPTION_MODEL_INFO,\n    configSchema,\n    config = undefined,\n    namespace,\n  } = params;\n  return modelRef({\n    name,\n    configSchema: configSchema || (TranscriptionConfigSchema as any),\n    info,\n    config,\n    namespace,\n  });\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAqBA,oBAAmE;AAEnE,oBAAsB;AAStB,mBAAkE;AAW3D,MAAM,2BAA2B;AAAA,EACtC,UAAU;AAAA,IACR,OAAO;AAAA,IACP,QAAQ,CAAC,QAAQ,MAAM;AAAA,IACvB,WAAW;AAAA,IACX,YAAY;AAAA,IACZ,OAAO;AAAA,EACT;AACF;AAEO,MAAM,oBAA+B;AAAA,EAC1C,UAAU;AAAA,IACR,OAAO;AAAA,IACP,QAAQ,CAAC,OAAO;AAAA,IAChB,WAAW;AAAA,IACX,YAAY;AAAA,IACZ,OAAO;AAAA,EACT;AACF;AAEA,MAAM,yBAAyB,gBAAE,OAAO;AAAA,EACtC,MAAM,gBAAE,OAAO;AAAA,EACf,mBAAmB,gBAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EAC7C,qBAAqB,gBAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EAC/C,WAAW,gBAAE,OAAO,EAAE,IAAI,CAAC,EAAE,IAAI,CAAG,EAAE,SAAS;AACjD,CAAC;AACM,MAAM,4BAA4B,2CAA6B,KAAK;AAAA,EACzE,aAAa;AACf,CAAC,EAAE,OAAO;AAAA,EACR,mBAAmB,gBAChB,MAAM,CAAC,gBAAE,QAAQ,MAAM,GAAG,sBAAsB,CAAC,EACjD,SAAS;AAAA,EACZ,SAAS,gBAAE,MAAM,gBAAE,IAAI,CAAC,EAAE,SAAS;AAAA,EACnC,UAAU,gBAAE,OAAO,EAAE,SAAS;AAAA,EAC9B,yBAAyB,gBAAE,MAAM,gBAAE,KAAK,CAAC,QAAQ,SAAS,CAAC,CAAC,EAAE,SAAS;AAAA,EACvE,iBAAiB,gBACd,KAAK,CAAC,QAAQ,QAAQ,OAAO,gBAAgB,KAAK,CAAC,EACnD,SAAS;AAAA;AAEd,CAAC;AAEM,MAAM,qBAAqB,gBAAE,OAAO;AAAA,EACzC,OAAO,gBACJ,KAAK,CAAC,SAAS,QAAQ,SAAS,QAAQ,QAAQ,SAAS,CAAC,EAC1D,QAAQ,OAAO;AAAA,EAClB,OAAO,gBAAE,OAAO,EAAE,IAAI,IAAI,EAAE,IAAI,CAAG,EAAE,SAAS;AAAA,EAC9C,iBAAiB,gBACd,KAAK,CAAC,OAAO,QAAQ,OAAO,QAAQ,OAAO,KAAK,CAAC,EACjD,SAAS;AACd,CAAC;AAKM,MAAM,8BAA8B;AAAA,EACzC,KAAK;AAAA,EACL,MAAM;AAAA,EACN,KAAK;AAAA,EACL,MAAM;AAAA,EACN,KAAK;AAAA,EACL,KAAK;AACP;AAEO,SAAS,aACd,WACA,SACA,gBACoB;AACpB,QAAM;AAAA,IACJ;AAAA,IACA,SAAS;AAAA,IACT;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,GAAG;AAAA,EACL,IAAI,QAAQ,UAAU,CAAC;AAEvB,MAAI,UAA8B;AAAA,IAChC,OAAO,gBAAgB;AAAA,IACvB,OAAO,IAAI,sBAAQ,QAAQ,SAAS,CAAC,CAAC,EAAE;AAAA,IACxC,OAAO,SAAS;AAAA,EAClB;AACA,MAAI,gBAAgB;AAClB,mBAAe,SAAS,OAAO;AAAA,EACjC,OAAO;AACL,cAAU;AAAA,MACR,GAAG;AAAA,MACH,GAAG;AAAA;AAAA,IACL;AAAA,EACF;AACA,aAAW,KAAK,SAAS;AACvB,QAAI,QAAQ,CAAC,MAAM,QAAW;AAC5B,aAAO,QAAQ,CAAC;AAAA,IAClB;AAAA,EACF;AACA,SAAO;AACT;AAEA,eAAsB,yBACpB,UACA,iBAAkE,OACnC;AAC/B,QAAM,oBAAoB,MAAM,SAAS,YAAY;AACrD,QAAM,eAAe,OAAO,KAAK,IAAI,WAAW,iBAAiB,CAAC;AAClE,QAAM,YAAY,4BAA4B,cAAc;AAC5D,SAAO;AAAA,IACL,SAAS;AAAA,MACP,MAAM;AAAA,MACN,SAAS;AAAA,QACP;AAAA,UACE,OAAO;AAAA,YACL,aAAa;AAAA,YACb,KAAK,QAAQ,SAAS,WAAW,aAAa,SAAS,QAAQ,CAAC;AAAA,UAClE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,IACA,cAAc;AAAA,IACd,KAAK;AAAA,EACP;AACF;AAiBO,SAAS,8BAEd,QAMc;AACd,QAAM;AAAA,IACJ;AAAA,IACA,QAAQ;AAAA,IACR;AAAA,IACA,UAAAA;AAAA,IACA;AAAA,EACF,IAAI;AACJ,QAAM,gBAAY,0BAAY,MAAM,eAAe,IAAI;AACvD,QAAM,aAAa,GAAG,eAAe,QAAQ,YAAY,IAAI,SAAS;AAEtE,aAAO;AAAA,IACL;AAAA,MACE,MAAM;AAAA,MACN,GAAGA,WAAU;AAAA,MACb,cAAcA,WAAU;AAAA,IAC1B;AAAA,IACA,OAAO,SAAS,EAAE,YAAY,MAAM;AAClC,YAAM,aAAa,aAAa,WAAW,SAAS,cAAc;AAClE,YAAM,aAAS;AAAA,QACb;AAAA,QACA;AAAA,QACA;AAAA,MACF;AACA,YAAM,SAAS,MAAM,OAAO,MAAM,OAAO,OAAO,YAAY;AAAA,QAC1D,QAAQ;AAAA,MACV,CAAC;AACD,aAAO,MAAM,yBAAyB,QAAQ,WAAW,eAAe;AAAA,IAC1E;AAAA,EACF;AACF;AAIO,SAAS,wBAEd,QAMC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,OAAO;AAAA,IACP;AAAA,IACA,SAAS;AAAA,IACT;AAAA,EACF,IAAI;AACJ,aAAO,wBAAS;AAAA,IACd;AAAA,IACA,cAAc,gBAAiB;AAAA,IAC/B;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AAEO,SAAS,aACd,WACA,SACA,gBAC2B;AAC3B,QAAM,UAAU,IAAI,sBAAQ,QAAQ,SAAS,CAAC,CAAC;AAC/C,QAAM,QAAQ,QAAQ;AACtB,MAAI,CAAC,OAAO,KAAK;AACf,UAAM,IAAI,MAAM,+BAA+B;AAAA,EACjD;AACA,QAAM,cAAc,OAAO;AAAA,IACzB,MAAM,IAAI,MAAM,MAAM,IAAI,QAAQ,GAAG,IAAI,CAAC;AAAA,IAC1C;AAAA,EACF;AACA,QAAM,YAAY,IAAI,KAAK,CAAC,WAAW,GAAG,SAAS;AAAA,IACjD,MACE,MAAM,eACN,MAAM,IAAI,MAAM,QAAQ,QAAQ,MAAM,IAAI,QAAQ,GAAG,CAAC;AAAA,EAC1D,CAAC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,SAAS;AAAA,IACT;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,GAAG;AAAA,EACL,IAAI,QAAQ,UAAU,CAAC;AAEvB,MAAI,UAAqC;AAAA,IACvC,OAAO,gBAAgB;AAAA,IACvB,MAAM;AAAA,IACN,QAAQ,QAAQ;AAAA,IAChB;AAAA,EACF;AACA,MAAI,gBAAgB;AAClB,mBAAe,SAAS,OAAO;AAAA,EACjC,OAAO;AACL,cAAU;AAAA,MACR,GAAG;AAAA,MACH,GAAG;AAAA;AAAA,IACL;AAAA,EACF;AACA,QAAM,eAAe,QAAQ,QAAQ;AACrC,QAAM,eAAe,QAAQ,QAAQ;AACrC,MAAI,gBAAgB,cAAc;AAChC,QACE,iBAAiB,UACjB,iBAAiB,UACjB,iBAAiB,gBACjB;AACA,YAAM,IAAI;AAAA,QACR,0BAA0B,YAAY,yCAAyC,YAAY;AAAA,MAC7F;AAAA,IACF;AAAA,EACF;AACA,MAAI,iBAAiB,SAAS;AAC5B,UAAM,IAAI,MAAM,iBAAiB,YAAY,oBAAoB;AAAA,EACnE;AACA,UAAQ,kBAAkB,gBAAgB,gBAAgB;AAC1D,aAAW,KAAK,SAAS;AACvB,QAAI,QAAQ,CAAC,MAAM,QAAW;AAC5B,aAAO,QAAQ,CAAC;AAAA,IAClB;AAAA,EACF;AACA,SAAO;AACT;AAEO,SAAS,gCACd,QACsB;AACtB,SAAO;AAAA,IACL,SAAS;AAAA,MACP,MAAM;AAAA,MACN,SAAS;AAAA,QACP;AAAA,UACE,MAAM,OAAO,WAAW,WAAW,SAAS,OAAO;AAAA,QACrD;AAAA,MACF;AAAA,IACF;AAAA,IACA,cAAc;AAAA,IACd,KAAK;AAAA,EACP;AACF;AAkBO,SAAS,qCAEd,QAMc;AACd,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,IACA,QAAQ;AAAA,IACR,UAAAA;AAAA,IACA;AAAA,EACF,IAAI;AACJ,QAAM,gBAAY,0BAAY,MAAM,eAAe,IAAI;AACvD,QAAM,aACJA,WAAU,QAAQ,GAAG,eAAe,QAAQ,YAAY,IAAI,SAAS;AAEvE,aAAO;AAAA,IACL;AAAA,MACE,MAAM;AAAA,MACN,GAAGA,WAAU;AAAA,MACb,cAAcA,WAAU;AAAA,IAC1B;AAAA,IACA,OAAO,SAAS,EAAE,YAAY,MAAM;AAClC,YAAMC,UAAS,aAAa,WAAW,SAAS,cAAc;AAC9D,YAAM,aAAS;AAAA,QACb;AAAA,QACA;AAAA,QACA;AAAA,MACF;AAEA,YAAM,SAAS,MAAM,OAAO,MAAM,eAAe;AAAA,QAC/C;AAAA,UACE,GAAGA;AAAA,UACH,QAAQ;AAAA,QACV;AAAA,QACA,EAAE,QAAQ,YAAY;AAAA,MACxB;AACA,aAAO,gCAAgC,MAAM;AAAA,IAC/C;AAAA,EACF;AACF;AAIO,SAAS,+BAEd,QAMC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,OAAO;AAAA,IACP;AAAA,IACA,SAAS;AAAA,IACT;AAAA,EACF,IAAI;AACJ,aAAO,wBAAS;AAAA,IACd;AAAA,IACA,cAAc,gBAAiB;AAAA,IAC/B;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;","names":["modelRef","params"]}

package/lib/audio.mjs

@@ -80,7 +80,7 @@ function toTTSRequest(modelName, request, requestBuilder) {
   }
   return options;
 }
-async function toGenerateResponse(response, responseFormat = "mp3") {
+async function speechToGenerateResponse(response, responseFormat = "mp3") {
   const resultArrayBuffer = await response.arrayBuffer();
   const resultBuffer = Buffer.from(new Uint8Array(resultArrayBuffer));
   const mediaType = RESPONSE_FORMAT_MEDIA_TYPES[responseFormat];
@@ -126,7 +126,7 @@ function defineCompatOpenAISpeechModel(params) {
       const result = await client.audio.speech.create(ttsRequest, {
         signal: abortSignal
       });
-      return await toGenerateResponse(result, ttsRequest.response_format);
+      return await speechToGenerateResponse(result, ttsRequest.response_format);
     }
   );
 }
@@ -219,12 +219,13 @@ function transcriptionToGenerateResponse(result) {
 }
 function defineCompatOpenAITranscriptionModel(params) {
   const {
-    name: modelName,
+    name,
     pluginOptions,
     client: defaultClient,
     modelRef: modelRef2,
     requestBuilder
   } = params;
+  const modelName = toModelName(name, pluginOptions?.name);
   const actionName = modelRef2?.name ?? `${pluginOptions?.name ?? "compat-oai"}/${modelName}`;
   return model(
     {
@@ -275,6 +276,10 @@ export {
   compatOaiSpeechModelRef,
   compatOaiTranscriptionModelRef,
   defineCompatOpenAISpeechModel,
-  defineCompatOpenAITranscriptionModel
+  defineCompatOpenAITranscriptionModel,
+  speechToGenerateResponse,
+  toSttRequest,
+  toTTSRequest,
+  transcriptionToGenerateResponse
 };
 //# sourceMappingURL=audio.mjs.map

package/lib/audio.mjs.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/audio.ts"],"sourcesContent":["/**\n * Copyright 2024 The Fire Company\n * Copyright 2024 Google LLC\n *\n * Licensed under the Apache License, Version 2.0 (the \"License\");\n * you may not use this file except in compliance with the License.\n * You may obtain a copy of the License at\n *\n *     http://www.apache.org/licenses/LICENSE-2.0\n *\n * Unless required by applicable law or agreed to in writing, software\n * distributed under the License is distributed on an \"AS IS\" BASIS,\n * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n * See the License for the specific language governing permissions and\n * limitations under the License.\n */\nimport type {\n  GenerateRequest,\n  GenerateResponseData,\n  ModelReference,\n} from 'genkit';\nimport { GenerationCommonConfigSchema, Message, modelRef, z } from 'genkit';\nimport type { ModelAction, ModelInfo } from 'genkit/model';\nimport { model } from 'genkit/plugin';\nimport OpenAI from 'openai';\nimport { Response } from 'openai/core.mjs';\nimport type {\n  SpeechCreateParams,\n  Transcription,\n  TranscriptionCreateParams,\n} from 'openai/resources/audio/index.mjs';\nimport { PluginOptions } from './index.js';\nimport { maybeCreateRequestScopedOpenAIClient, toModelName } from './utils.js';\n\nexport type SpeechRequestBuilder = (\n  req: GenerateRequest,\n  params: SpeechCreateParams\n) => void;\nexport type TranscriptionRequestBuilder = (\n  req: GenerateRequest,\n  params: TranscriptionCreateParams\n) => void;\n\nexport const TRANSCRIPTION_MODEL_INFO = {\n  supports: {\n    media: true,\n    output: ['text', 'json'],\n    multiturn: false,\n    systemRole: false,\n    tools: false,\n  },\n};\n\nexport const SPEECH_MODEL_INFO: ModelInfo = {\n  supports: {\n    media: false,\n    output: ['media'],\n    multiturn: false,\n    systemRole: false,\n    tools: false,\n  },\n};\n\nconst ChunkingStrategySchema = z.object({\n  type: z.string(),\n  prefix_padding_ms: z.number().int().optional(),\n  silence_duration_ms: z.number().int().optional(),\n  threshold: z.number().min(0).max(1.0).optional(),\n});\nexport const TranscriptionConfigSchema = GenerationCommonConfigSchema.pick({\n  temperature: true,\n}).extend({\n  chunking_strategy: z\n    .union([z.literal('auto'), ChunkingStrategySchema])\n    .optional(),\n  include: z.array(z.any()).optional(),\n  language: z.string().optional(),\n  timestamp_granularities: z.array(z.enum(['word', 'segment'])).optional(),\n  response_format: z\n    .enum(['json', 'text', 'srt', 'verbose_json', 'vtt'])\n    .optional(),\n  // TODO stream support\n});\n\nexport const SpeechConfigSchema = z.object({\n  voice: z\n    .enum(['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'])\n    .default('alloy'),\n  speed: z.number().min(0.25).max(4.0).optional(),\n  response_format: z\n    .enum(['mp3', 'opus', 'aac', 'flac', 'wav', 'pcm'])\n    .optional(),\n});\n\n/**\n * Supported media formats for Audio generation\n */\nexport const RESPONSE_FORMAT_MEDIA_TYPES = {\n  mp3: 'audio/mpeg',\n  opus: 'audio/opus',\n  aac: 'audio/aac',\n  flac: 'audio/flac',\n  wav: 'audio/wav',\n  pcm: 'audio/L16',\n};\n\nfunction toTTSRequest(\n  modelName: string,\n  request: GenerateRequest,\n  requestBuilder?: SpeechRequestBuilder\n): SpeechCreateParams {\n  const {\n    voice,\n    version: modelVersion,\n    temperature,\n    maxOutputTokens,\n    stopSequences,\n    topK,\n    topP,\n    ...restOfConfig\n  } = request.config ?? {};\n\n  let options: SpeechCreateParams = {\n    model: modelVersion ?? modelName,\n    input: new Message(request.messages[0]).text,\n    voice: voice ?? 'alloy',\n  };\n  if (requestBuilder) {\n    requestBuilder(request, options);\n  } else {\n    options = {\n      ...options,\n      ...restOfConfig, // passthorugh rest of the config\n    };\n  }\n  for (const k in options) {\n    if (options[k] === undefined) {\n      delete options[k];\n    }\n  }\n  return options;\n}\n\nasync function toGenerateResponse(\n  response: Response,\n  responseFormat: 'mp3' | 'opus' | 'aac' | 'flac' | 'wav' | 'pcm' = 'mp3'\n): Promise<GenerateResponseData> {\n  const resultArrayBuffer = await response.arrayBuffer();\n  const resultBuffer = Buffer.from(new Uint8Array(resultArrayBuffer));\n  const mediaType = RESPONSE_FORMAT_MEDIA_TYPES[responseFormat];\n  return {\n    message: {\n      role: 'model',\n      content: [\n        {\n          media: {\n            contentType: mediaType,\n            url: `data:${mediaType};base64,${resultBuffer.toString('base64')}`,\n          },\n        },\n      ],\n    },\n    finishReason: 'stop',\n    raw: response,\n  };\n}\n\n/**\n * Method to define a new Genkit Model that is compatible with the Open AI Audio\n * API. \n *\n * These models are to be used to create audio speech from a given request.\n * @param params An object containing parameters for defining the OpenAI speech\n * model.\n * @param params.ai The Genkit AI instance.\n * @param params.name The name of the model.\n * @param params.client The OpenAI client instance.\n * @param params.modelRef Optional reference to the model's configuration and\n * custom options.\n\n * @returns the created {@link ModelAction}\n */\nexport function defineCompatOpenAISpeechModel<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  client: OpenAI;\n  modelRef?: ModelReference<CustomOptions>;\n  requestBuilder?: SpeechRequestBuilder;\n  pluginOptions: PluginOptions;\n}): ModelAction {\n  const {\n    name,\n    client: defaultClient,\n    pluginOptions,\n    modelRef,\n    requestBuilder,\n  } = params;\n  const modelName = toModelName(name, pluginOptions?.name);\n  const actionName = `${pluginOptions?.name ?? 'compat-oai'}/${modelName}`;\n\n  return model(\n    {\n      name: actionName,\n      ...modelRef?.info,\n      configSchema: modelRef?.configSchema,\n    },\n    async (request, { abortSignal }) => {\n      const ttsRequest = toTTSRequest(modelName, request, requestBuilder);\n      const client = maybeCreateRequestScopedOpenAIClient(\n        pluginOptions,\n        request,\n        defaultClient\n      );\n      const result = await client.audio.speech.create(ttsRequest, {\n        signal: abortSignal,\n      });\n      return await toGenerateResponse(result, ttsRequest.response_format);\n    }\n  );\n}\n\n/** Speech generation ModelRef helper, with reasonable defaults for\n * OpenAI-compatible providers */\nexport function compatOaiSpeechModelRef<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  info?: ModelInfo;\n  configSchema?: CustomOptions;\n  config?: any;\n  namespace?: string;\n}) {\n  const {\n    name,\n    info = SPEECH_MODEL_INFO,\n    configSchema,\n    config = undefined,\n    namespace,\n  } = params;\n  return modelRef({\n    name,\n    configSchema: configSchema || (SpeechConfigSchema as any),\n    info,\n    config,\n    namespace,\n  });\n}\n\nfunction toSttRequest(\n  modelName: string,\n  request: GenerateRequest,\n  requestBuilder?: TranscriptionRequestBuilder\n): TranscriptionCreateParams {\n  const message = new Message(request.messages[0]);\n  const media = message.media;\n  if (!media?.url) {\n    throw new Error('No media found in the request');\n  }\n  const mediaBuffer = Buffer.from(\n    media.url.slice(media.url.indexOf(',') + 1),\n    'base64'\n  );\n  const mediaFile = new File([mediaBuffer], 'input', {\n    type:\n      media.contentType ??\n      media.url.slice('data:'.length, media.url.indexOf(';')),\n  });\n  const {\n    temperature,\n    version: modelVersion,\n    maxOutputTokens,\n    stopSequences,\n    topK,\n    topP,\n    ...restOfConfig\n  } = request.config ?? {};\n\n  let options: TranscriptionCreateParams = {\n    model: modelVersion ?? modelName,\n    file: mediaFile,\n    prompt: message.text,\n    temperature,\n  };\n  if (requestBuilder) {\n    requestBuilder(request, options);\n  } else {\n    options = {\n      ...options,\n      ...restOfConfig, // passthrough rest of the config\n    };\n  }\n  const outputFormat = request.output?.format as 'json' | 'text' | 'media';\n  const customFormat = request.config?.response_format;\n  if (outputFormat && customFormat) {\n    if (\n      outputFormat === 'json' &&\n      customFormat !== 'json' &&\n      customFormat !== 'verbose_json'\n    ) {\n      throw new Error(\n        `Custom response format ${customFormat} is not compatible with output format ${outputFormat}`\n      );\n    }\n  }\n  if (outputFormat === 'media') {\n    throw new Error(`Output format ${outputFormat} is not supported.`);\n  }\n  options.response_format = customFormat || outputFormat || 'text';\n  for (const k in options) {\n    if (options[k] === undefined) {\n      delete options[k];\n    }\n  }\n  return options;\n}\n\nfunction transcriptionToGenerateResponse(\n  result: Transcription | string\n): GenerateResponseData {\n  return {\n    message: {\n      role: 'model',\n      content: [\n        {\n          text: typeof result === 'string' ? result : result.text,\n        },\n      ],\n    },\n    finishReason: 'stop',\n    raw: result,\n  };\n}\n\n/**\n * Method to define a new Genkit Model that is compatible with Open AI\n * Transcriptions API. \n *\n * These models are to be used to transcribe audio to text.\n *\n * @param params An object containing parameters for defining the OpenAI\n * transcription model.\n * @param params.ai The Genkit AI instance.\n * @param params.name The name of the model.\n * @param params.client The OpenAI client instance.\n * @param params.modelRef Optional reference to the model's configuration and\n * custom options.\n\n * @returns the created {@link ModelAction}\n */\nexport function defineCompatOpenAITranscriptionModel<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  client: OpenAI;\n  pluginOptions?: PluginOptions;\n  modelRef?: ModelReference<CustomOptions>;\n  requestBuilder?: TranscriptionRequestBuilder;\n}): ModelAction {\n  const {\n    name: modelName,\n    pluginOptions,\n    client: defaultClient,\n    modelRef,\n    requestBuilder,\n  } = params;\n\n  const actionName =\n    modelRef?.name ?? `${pluginOptions?.name ?? 'compat-oai'}/${modelName}`;\n  return model(\n    {\n      name: actionName,\n      ...modelRef?.info,\n      configSchema: modelRef?.configSchema,\n    },\n    async (request, { abortSignal }) => {\n      const params = toSttRequest(modelName, request, requestBuilder);\n      const client = maybeCreateRequestScopedOpenAIClient(\n        pluginOptions,\n        request,\n        defaultClient\n      );\n      // Explicitly setting stream to false ensures we use the non-streaming overload\n      const result = await client.audio.transcriptions.create(\n        {\n          ...params,\n          stream: false,\n        },\n        { signal: abortSignal }\n      );\n      return transcriptionToGenerateResponse(result);\n    }\n  );\n}\n\n/** Transcription ModelRef helper, with reasonable defaults for\n * OpenAI-compatible providers */\nexport function compatOaiTranscriptionModelRef<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  info?: ModelInfo;\n  configSchema?: CustomOptions;\n  config?: any;\n  namespace?: string;\n}) {\n  const {\n    name,\n    info = TRANSCRIPTION_MODEL_INFO,\n    configSchema,\n    config = undefined,\n    namespace,\n  } = params;\n  return modelRef({\n    name,\n    configSchema: configSchema || (TranscriptionConfigSchema as any),\n    info,\n    config,\n    namespace,\n  });\n}\n"],"mappings":"AAqBA,SAAS,8BAA8B,SAAS,UAAU,SAAS;AAEnE,SAAS,aAAa;AAStB,SAAS,sCAAsC,mBAAmB;AAW3D,MAAM,2BAA2B;AAAA,EACtC,UAAU;AAAA,IACR,OAAO;AAAA,IACP,QAAQ,CAAC,QAAQ,MAAM;AAAA,IACvB,WAAW;AAAA,IACX,YAAY;AAAA,IACZ,OAAO;AAAA,EACT;AACF;AAEO,MAAM,oBAA+B;AAAA,EAC1C,UAAU;AAAA,IACR,OAAO;AAAA,IACP,QAAQ,CAAC,OAAO;AAAA,IAChB,WAAW;AAAA,IACX,YAAY;AAAA,IACZ,OAAO;AAAA,EACT;AACF;AAEA,MAAM,yBAAyB,EAAE,OAAO;AAAA,EACtC,MAAM,EAAE,OAAO;AAAA,EACf,mBAAmB,EAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EAC7C,qBAAqB,EAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EAC/C,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,IAAI,CAAG,EAAE,SAAS;AACjD,CAAC;AACM,MAAM,4BAA4B,6BAA6B,KAAK;AAAA,EACzE,aAAa;AACf,CAAC,EAAE,OAAO;AAAA,EACR,mBAAmB,EAChB,MAAM,CAAC,EAAE,QAAQ,MAAM,GAAG,sBAAsB,CAAC,EACjD,SAAS;AAAA,EACZ,SAAS,EAAE,MAAM,EAAE,IAAI,CAAC,EAAE,SAAS;AAAA,EACnC,UAAU,EAAE,OAAO,EAAE,SAAS;AAAA,EAC9B,yBAAyB,EAAE,MAAM,EAAE,KAAK,CAAC,QAAQ,SAAS,CAAC,CAAC,EAAE,SAAS;AAAA,EACvE,iBAAiB,EACd,KAAK,CAAC,QAAQ,QAAQ,OAAO,gBAAgB,KAAK,CAAC,EACnD,SAAS;AAAA;AAEd,CAAC;AAEM,MAAM,qBAAqB,EAAE,OAAO;AAAA,EACzC,OAAO,EACJ,KAAK,CAAC,SAAS,QAAQ,SAAS,QAAQ,QAAQ,SAAS,CAAC,EAC1D,QAAQ,OAAO;AAAA,EAClB,OAAO,EAAE,OAAO,EAAE,IAAI,IAAI,EAAE,IAAI,CAAG,EAAE,SAAS;AAAA,EAC9C,iBAAiB,EACd,KAAK,CAAC,OAAO,QAAQ,OAAO,QAAQ,OAAO,KAAK,CAAC,EACjD,SAAS;AACd,CAAC;AAKM,MAAM,8BAA8B;AAAA,EACzC,KAAK;AAAA,EACL,MAAM;AAAA,EACN,KAAK;AAAA,EACL,MAAM;AAAA,EACN,KAAK;AAAA,EACL,KAAK;AACP;AAEA,SAAS,aACP,WACA,SACA,gBACoB;AACpB,QAAM;AAAA,IACJ;AAAA,IACA,SAAS;AAAA,IACT;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,GAAG;AAAA,EACL,IAAI,QAAQ,UAAU,CAAC;AAEvB,MAAI,UAA8B;AAAA,IAChC,OAAO,gBAAgB;AAAA,IACvB,OAAO,IAAI,QAAQ,QAAQ,SAAS,CAAC,CAAC,EAAE;AAAA,IACxC,OAAO,SAAS;AAAA,EAClB;AACA,MAAI,gBAAgB;AAClB,mBAAe,SAAS,OAAO;AAAA,EACjC,OAAO;AACL,cAAU;AAAA,MACR,GAAG;AAAA,MACH,GAAG;AAAA;AAAA,IACL;AAAA,EACF;AACA,aAAW,KAAK,SAAS;AACvB,QAAI,QAAQ,CAAC,MAAM,QAAW;AAC5B,aAAO,QAAQ,CAAC;AAAA,IAClB;AAAA,EACF;AACA,SAAO;AACT;AAEA,eAAe,mBACb,UACA,iBAAkE,OACnC;AAC/B,QAAM,oBAAoB,MAAM,SAAS,YAAY;AACrD,QAAM,eAAe,OAAO,KAAK,IAAI,WAAW,iBAAiB,CAAC;AAClE,QAAM,YAAY,4BAA4B,cAAc;AAC5D,SAAO;AAAA,IACL,SAAS;AAAA,MACP,MAAM;AAAA,MACN,SAAS;AAAA,QACP;AAAA,UACE,OAAO;AAAA,YACL,aAAa;AAAA,YACb,KAAK,QAAQ,SAAS,WAAW,aAAa,SAAS,QAAQ,CAAC;AAAA,UAClE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,IACA,cAAc;AAAA,IACd,KAAK;AAAA,EACP;AACF;AAiBO,SAAS,8BAEd,QAMc;AACd,QAAM;AAAA,IACJ;AAAA,IACA,QAAQ;AAAA,IACR;AAAA,IACA,UAAAA;AAAA,IACA;AAAA,EACF,IAAI;AACJ,QAAM,YAAY,YAAY,MAAM,eAAe,IAAI;AACvD,QAAM,aAAa,GAAG,eAAe,QAAQ,YAAY,IAAI,SAAS;AAEtE,SAAO;AAAA,IACL;AAAA,MACE,MAAM;AAAA,MACN,GAAGA,WAAU;AAAA,MACb,cAAcA,WAAU;AAAA,IAC1B;AAAA,IACA,OAAO,SAAS,EAAE,YAAY,MAAM;AAClC,YAAM,aAAa,aAAa,WAAW,SAAS,cAAc;AAClE,YAAM,SAAS;AAAA,QACb;AAAA,QACA;AAAA,QACA;AAAA,MACF;AACA,YAAM,SAAS,MAAM,OAAO,MAAM,OAAO,OAAO,YAAY;AAAA,QAC1D,QAAQ;AAAA,MACV,CAAC;AACD,aAAO,MAAM,mBAAmB,QAAQ,WAAW,eAAe;AAAA,IACpE;AAAA,EACF;AACF;AAIO,SAAS,wBAEd,QAMC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,OAAO;AAAA,IACP;AAAA,IACA,SAAS;AAAA,IACT;AAAA,EACF,IAAI;AACJ,SAAO,SAAS;AAAA,IACd;AAAA,IACA,cAAc,gBAAiB;AAAA,IAC/B;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AAEA,SAAS,aACP,WACA,SACA,gBAC2B;AAC3B,QAAM,UAAU,IAAI,QAAQ,QAAQ,SAAS,CAAC,CAAC;AAC/C,QAAM,QAAQ,QAAQ;AACtB,MAAI,CAAC,OAAO,KAAK;AACf,UAAM,IAAI,MAAM,+BAA+B;AAAA,EACjD;AACA,QAAM,cAAc,OAAO;AAAA,IACzB,MAAM,IAAI,MAAM,MAAM,IAAI,QAAQ,GAAG,IAAI,CAAC;AAAA,IAC1C;AAAA,EACF;AACA,QAAM,YAAY,IAAI,KAAK,CAAC,WAAW,GAAG,SAAS;AAAA,IACjD,MACE,MAAM,eACN,MAAM,IAAI,MAAM,QAAQ,QAAQ,MAAM,IAAI,QAAQ,GAAG,CAAC;AAAA,EAC1D,CAAC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,SAAS;AAAA,IACT;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,GAAG;AAAA,EACL,IAAI,QAAQ,UAAU,CAAC;AAEvB,MAAI,UAAqC;AAAA,IACvC,OAAO,gBAAgB;AAAA,IACvB,MAAM;AAAA,IACN,QAAQ,QAAQ;AAAA,IAChB;AAAA,EACF;AACA,MAAI,gBAAgB;AAClB,mBAAe,SAAS,OAAO;AAAA,EACjC,OAAO;AACL,cAAU;AAAA,MACR,GAAG;AAAA,MACH,GAAG;AAAA;AAAA,IACL;AAAA,EACF;AACA,QAAM,eAAe,QAAQ,QAAQ;AACrC,QAAM,eAAe,QAAQ,QAAQ;AACrC,MAAI,gBAAgB,cAAc;AAChC,QACE,iBAAiB,UACjB,iBAAiB,UACjB,iBAAiB,gBACjB;AACA,YAAM,IAAI;AAAA,QACR,0BAA0B,YAAY,yCAAyC,YAAY;AAAA,MAC7F;AAAA,IACF;AAAA,EACF;AACA,MAAI,iBAAiB,SAAS;AAC5B,UAAM,IAAI,MAAM,iBAAiB,YAAY,oBAAoB;AAAA,EACnE;AACA,UAAQ,kBAAkB,gBAAgB,gBAAgB;AAC1D,aAAW,KAAK,SAAS;AACvB,QAAI,QAAQ,CAAC,MAAM,QAAW;AAC5B,aAAO,QAAQ,CAAC;AAAA,IAClB;AAAA,EACF;AACA,SAAO;AACT;AAEA,SAAS,gCACP,QACsB;AACtB,SAAO;AAAA,IACL,SAAS;AAAA,MACP,MAAM;AAAA,MACN,SAAS;AAAA,QACP;AAAA,UACE,MAAM,OAAO,WAAW,WAAW,SAAS,OAAO;AAAA,QACrD;AAAA,MACF;AAAA,IACF;AAAA,IACA,cAAc;AAAA,IACd,KAAK;AAAA,EACP;AACF;AAkBO,SAAS,qCAEd,QAMc;AACd,QAAM;AAAA,IACJ,MAAM;AAAA,IACN;AAAA,IACA,QAAQ;AAAA,IACR,UAAAA;AAAA,IACA;AAAA,EACF,IAAI;AAEJ,QAAM,aACJA,WAAU,QAAQ,GAAG,eAAe,QAAQ,YAAY,IAAI,SAAS;AACvE,SAAO;AAAA,IACL;AAAA,MACE,MAAM;AAAA,MACN,GAAGA,WAAU;AAAA,MACb,cAAcA,WAAU;AAAA,IAC1B;AAAA,IACA,OAAO,SAAS,EAAE,YAAY,MAAM;AAClC,YAAMC,UAAS,aAAa,WAAW,SAAS,cAAc;AAC9D,YAAM,SAAS;AAAA,QACb;AAAA,QACA;AAAA,QACA;AAAA,MACF;AAEA,YAAM,SAAS,MAAM,OAAO,MAAM,eAAe;AAAA,QAC/C;AAAA,UACE,GAAGA;AAAA,UACH,QAAQ;AAAA,QACV;AAAA,QACA,EAAE,QAAQ,YAAY;AAAA,MACxB;AACA,aAAO,gCAAgC,MAAM;AAAA,IAC/C;AAAA,EACF;AACF;AAIO,SAAS,+BAEd,QAMC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,OAAO;AAAA,IACP;AAAA,IACA,SAAS;AAAA,IACT;AAAA,EACF,IAAI;AACJ,SAAO,SAAS;AAAA,IACd;AAAA,IACA,cAAc,gBAAiB;AAAA,IAC/B;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;","names":["modelRef","params"]}
+{"version":3,"sources":["../src/audio.ts"],"sourcesContent":["/**\n * Copyright 2024 The Fire Company\n * Copyright 2024 Google LLC\n *\n * Licensed under the Apache License, Version 2.0 (the \"License\");\n * you may not use this file except in compliance with the License.\n * You may obtain a copy of the License at\n *\n *     http://www.apache.org/licenses/LICENSE-2.0\n *\n * Unless required by applicable law or agreed to in writing, software\n * distributed under the License is distributed on an \"AS IS\" BASIS,\n * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n * See the License for the specific language governing permissions and\n * limitations under the License.\n */\nimport type {\n  GenerateRequest,\n  GenerateResponseData,\n  ModelReference,\n} from 'genkit';\nimport { GenerationCommonConfigSchema, Message, modelRef, z } from 'genkit';\nimport type { ModelAction, ModelInfo } from 'genkit/model';\nimport { model } from 'genkit/plugin';\nimport OpenAI from 'openai';\nimport { Response } from 'openai/core.mjs';\nimport type {\n  SpeechCreateParams,\n  Transcription,\n  TranscriptionCreateParams,\n} from 'openai/resources/audio/index.mjs';\nimport { PluginOptions } from './index.js';\nimport { maybeCreateRequestScopedOpenAIClient, toModelName } from './utils.js';\n\nexport type SpeechRequestBuilder = (\n  req: GenerateRequest,\n  params: SpeechCreateParams\n) => void;\nexport type TranscriptionRequestBuilder = (\n  req: GenerateRequest,\n  params: TranscriptionCreateParams\n) => void;\n\nexport const TRANSCRIPTION_MODEL_INFO = {\n  supports: {\n    media: true,\n    output: ['text', 'json'],\n    multiturn: false,\n    systemRole: false,\n    tools: false,\n  },\n};\n\nexport const SPEECH_MODEL_INFO: ModelInfo = {\n  supports: {\n    media: false,\n    output: ['media'],\n    multiturn: false,\n    systemRole: false,\n    tools: false,\n  },\n};\n\nconst ChunkingStrategySchema = z.object({\n  type: z.string(),\n  prefix_padding_ms: z.number().int().optional(),\n  silence_duration_ms: z.number().int().optional(),\n  threshold: z.number().min(0).max(1.0).optional(),\n});\nexport const TranscriptionConfigSchema = GenerationCommonConfigSchema.pick({\n  temperature: true,\n}).extend({\n  chunking_strategy: z\n    .union([z.literal('auto'), ChunkingStrategySchema])\n    .optional(),\n  include: z.array(z.any()).optional(),\n  language: z.string().optional(),\n  timestamp_granularities: z.array(z.enum(['word', 'segment'])).optional(),\n  response_format: z\n    .enum(['json', 'text', 'srt', 'verbose_json', 'vtt'])\n    .optional(),\n  // TODO stream support\n});\n\nexport const SpeechConfigSchema = z.object({\n  voice: z\n    .enum(['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'])\n    .default('alloy'),\n  speed: z.number().min(0.25).max(4.0).optional(),\n  response_format: z\n    .enum(['mp3', 'opus', 'aac', 'flac', 'wav', 'pcm'])\n    .optional(),\n});\n\n/**\n * Supported media formats for Audio generation\n */\nexport const RESPONSE_FORMAT_MEDIA_TYPES = {\n  mp3: 'audio/mpeg',\n  opus: 'audio/opus',\n  aac: 'audio/aac',\n  flac: 'audio/flac',\n  wav: 'audio/wav',\n  pcm: 'audio/L16',\n};\n\nexport function toTTSRequest(\n  modelName: string,\n  request: GenerateRequest,\n  requestBuilder?: SpeechRequestBuilder\n): SpeechCreateParams {\n  const {\n    voice,\n    version: modelVersion,\n    temperature,\n    maxOutputTokens,\n    stopSequences,\n    topK,\n    topP,\n    ...restOfConfig\n  } = request.config ?? {};\n\n  let options: SpeechCreateParams = {\n    model: modelVersion ?? modelName,\n    input: new Message(request.messages[0]).text,\n    voice: voice ?? 'alloy',\n  };\n  if (requestBuilder) {\n    requestBuilder(request, options);\n  } else {\n    options = {\n      ...options,\n      ...restOfConfig, // passthorugh rest of the config\n    };\n  }\n  for (const k in options) {\n    if (options[k] === undefined) {\n      delete options[k];\n    }\n  }\n  return options;\n}\n\nexport async function speechToGenerateResponse(\n  response: Response,\n  responseFormat: 'mp3' | 'opus' | 'aac' | 'flac' | 'wav' | 'pcm' = 'mp3'\n): Promise<GenerateResponseData> {\n  const resultArrayBuffer = await response.arrayBuffer();\n  const resultBuffer = Buffer.from(new Uint8Array(resultArrayBuffer));\n  const mediaType = RESPONSE_FORMAT_MEDIA_TYPES[responseFormat];\n  return {\n    message: {\n      role: 'model',\n      content: [\n        {\n          media: {\n            contentType: mediaType,\n            url: `data:${mediaType};base64,${resultBuffer.toString('base64')}`,\n          },\n        },\n      ],\n    },\n    finishReason: 'stop',\n    raw: response,\n  };\n}\n\n/**\n * Method to define a new Genkit Model that is compatible with the Open AI Audio\n * API. \n *\n * These models are to be used to create audio speech from a given request.\n * @param params An object containing parameters for defining the OpenAI speech\n * model.\n * @param params.ai The Genkit AI instance.\n * @param params.name The name of the model.\n * @param params.client The OpenAI client instance.\n * @param params.modelRef Optional reference to the model's configuration and\n * custom options.\n\n * @returns the created {@link ModelAction}\n */\nexport function defineCompatOpenAISpeechModel<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  client: OpenAI;\n  modelRef?: ModelReference<CustomOptions>;\n  requestBuilder?: SpeechRequestBuilder;\n  pluginOptions: PluginOptions;\n}): ModelAction {\n  const {\n    name,\n    client: defaultClient,\n    pluginOptions,\n    modelRef,\n    requestBuilder,\n  } = params;\n  const modelName = toModelName(name, pluginOptions?.name);\n  const actionName = `${pluginOptions?.name ?? 'compat-oai'}/${modelName}`;\n\n  return model(\n    {\n      name: actionName,\n      ...modelRef?.info,\n      configSchema: modelRef?.configSchema,\n    },\n    async (request, { abortSignal }) => {\n      const ttsRequest = toTTSRequest(modelName, request, requestBuilder);\n      const client = maybeCreateRequestScopedOpenAIClient(\n        pluginOptions,\n        request,\n        defaultClient\n      );\n      const result = await client.audio.speech.create(ttsRequest, {\n        signal: abortSignal,\n      });\n      return await speechToGenerateResponse(result, ttsRequest.response_format);\n    }\n  );\n}\n\n/** Speech generation ModelRef helper, with reasonable defaults for\n * OpenAI-compatible providers */\nexport function compatOaiSpeechModelRef<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  info?: ModelInfo;\n  configSchema?: CustomOptions;\n  config?: any;\n  namespace?: string;\n}) {\n  const {\n    name,\n    info = SPEECH_MODEL_INFO,\n    configSchema,\n    config = undefined,\n    namespace,\n  } = params;\n  return modelRef({\n    name,\n    configSchema: configSchema || (SpeechConfigSchema as any),\n    info,\n    config,\n    namespace,\n  });\n}\n\nexport function toSttRequest(\n  modelName: string,\n  request: GenerateRequest,\n  requestBuilder?: TranscriptionRequestBuilder\n): TranscriptionCreateParams {\n  const message = new Message(request.messages[0]);\n  const media = message.media;\n  if (!media?.url) {\n    throw new Error('No media found in the request');\n  }\n  const mediaBuffer = Buffer.from(\n    media.url.slice(media.url.indexOf(',') + 1),\n    'base64'\n  );\n  const mediaFile = new File([mediaBuffer], 'input', {\n    type:\n      media.contentType ??\n      media.url.slice('data:'.length, media.url.indexOf(';')),\n  });\n  const {\n    temperature,\n    version: modelVersion,\n    maxOutputTokens,\n    stopSequences,\n    topK,\n    topP,\n    ...restOfConfig\n  } = request.config ?? {};\n\n  let options: TranscriptionCreateParams = {\n    model: modelVersion ?? modelName,\n    file: mediaFile,\n    prompt: message.text,\n    temperature,\n  };\n  if (requestBuilder) {\n    requestBuilder(request, options);\n  } else {\n    options = {\n      ...options,\n      ...restOfConfig, // passthrough rest of the config\n    };\n  }\n  const outputFormat = request.output?.format as 'json' | 'text' | 'media';\n  const customFormat = request.config?.response_format;\n  if (outputFormat && customFormat) {\n    if (\n      outputFormat === 'json' &&\n      customFormat !== 'json' &&\n      customFormat !== 'verbose_json'\n    ) {\n      throw new Error(\n        `Custom response format ${customFormat} is not compatible with output format ${outputFormat}`\n      );\n    }\n  }\n  if (outputFormat === 'media') {\n    throw new Error(`Output format ${outputFormat} is not supported.`);\n  }\n  options.response_format = customFormat || outputFormat || 'text';\n  for (const k in options) {\n    if (options[k] === undefined) {\n      delete options[k];\n    }\n  }\n  return options;\n}\n\nexport function transcriptionToGenerateResponse(\n  result: Transcription | string\n): GenerateResponseData {\n  return {\n    message: {\n      role: 'model',\n      content: [\n        {\n          text: typeof result === 'string' ? result : result.text,\n        },\n      ],\n    },\n    finishReason: 'stop',\n    raw: result,\n  };\n}\n\n/**\n * Method to define a new Genkit Model that is compatible with Open AI\n * Transcriptions API. \n *\n * These models are to be used to transcribe audio to text.\n *\n * @param params An object containing parameters for defining the OpenAI\n * transcription model.\n * @param params.ai The Genkit AI instance.\n * @param params.name The name of the model.\n * @param params.client The OpenAI client instance.\n * @param params.modelRef Optional reference to the model's configuration and\n * custom options.\n\n * @returns the created {@link ModelAction}\n */\nexport function defineCompatOpenAITranscriptionModel<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  client: OpenAI;\n  pluginOptions?: PluginOptions;\n  modelRef?: ModelReference<CustomOptions>;\n  requestBuilder?: TranscriptionRequestBuilder;\n}): ModelAction {\n  const {\n    name,\n    pluginOptions,\n    client: defaultClient,\n    modelRef,\n    requestBuilder,\n  } = params;\n  const modelName = toModelName(name, pluginOptions?.name);\n  const actionName =\n    modelRef?.name ?? `${pluginOptions?.name ?? 'compat-oai'}/${modelName}`;\n\n  return model(\n    {\n      name: actionName,\n      ...modelRef?.info,\n      configSchema: modelRef?.configSchema,\n    },\n    async (request, { abortSignal }) => {\n      const params = toSttRequest(modelName, request, requestBuilder);\n      const client = maybeCreateRequestScopedOpenAIClient(\n        pluginOptions,\n        request,\n        defaultClient\n      );\n      // Explicitly setting stream to false ensures we use the non-streaming overload\n      const result = await client.audio.transcriptions.create(\n        {\n          ...params,\n          stream: false,\n        },\n        { signal: abortSignal }\n      );\n      return transcriptionToGenerateResponse(result);\n    }\n  );\n}\n\n/** Transcription ModelRef helper, with reasonable defaults for\n * OpenAI-compatible providers */\nexport function compatOaiTranscriptionModelRef<\n  CustomOptions extends z.ZodTypeAny = z.ZodTypeAny,\n>(params: {\n  name: string;\n  info?: ModelInfo;\n  configSchema?: CustomOptions;\n  config?: any;\n  namespace?: string;\n}) {\n  const {\n    name,\n    info = TRANSCRIPTION_MODEL_INFO,\n    configSchema,\n    config = undefined,\n    namespace,\n  } = params;\n  return modelRef({\n    name,\n    configSchema: configSchema || (TranscriptionConfigSchema as any),\n    info,\n    config,\n    namespace,\n  });\n}\n"],"mappings":"AAqBA,SAAS,8BAA8B,SAAS,UAAU,SAAS;AAEnE,SAAS,aAAa;AAStB,SAAS,sCAAsC,mBAAmB;AAW3D,MAAM,2BAA2B;AAAA,EACtC,UAAU;AAAA,IACR,OAAO;AAAA,IACP,QAAQ,CAAC,QAAQ,MAAM;AAAA,IACvB,WAAW;AAAA,IACX,YAAY;AAAA,IACZ,OAAO;AAAA,EACT;AACF;AAEO,MAAM,oBAA+B;AAAA,EAC1C,UAAU;AAAA,IACR,OAAO;AAAA,IACP,QAAQ,CAAC,OAAO;AAAA,IAChB,WAAW;AAAA,IACX,YAAY;AAAA,IACZ,OAAO;AAAA,EACT;AACF;AAEA,MAAM,yBAAyB,EAAE,OAAO;AAAA,EACtC,MAAM,EAAE,OAAO;AAAA,EACf,mBAAmB,EAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EAC7C,qBAAqB,EAAE,OAAO,EAAE,IAAI,EAAE,SAAS;AAAA,EAC/C,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,IAAI,CAAG,EAAE,SAAS;AACjD,CAAC;AACM,MAAM,4BAA4B,6BAA6B,KAAK;AAAA,EACzE,aAAa;AACf,CAAC,EAAE,OAAO;AAAA,EACR,mBAAmB,EAChB,MAAM,CAAC,EAAE,QAAQ,MAAM,GAAG,sBAAsB,CAAC,EACjD,SAAS;AAAA,EACZ,SAAS,EAAE,MAAM,EAAE,IAAI,CAAC,EAAE,SAAS;AAAA,EACnC,UAAU,EAAE,OAAO,EAAE,SAAS;AAAA,EAC9B,yBAAyB,EAAE,MAAM,EAAE,KAAK,CAAC,QAAQ,SAAS,CAAC,CAAC,EAAE,SAAS;AAAA,EACvE,iBAAiB,EACd,KAAK,CAAC,QAAQ,QAAQ,OAAO,gBAAgB,KAAK,CAAC,EACnD,SAAS;AAAA;AAEd,CAAC;AAEM,MAAM,qBAAqB,EAAE,OAAO;AAAA,EACzC,OAAO,EACJ,KAAK,CAAC,SAAS,QAAQ,SAAS,QAAQ,QAAQ,SAAS,CAAC,EAC1D,QAAQ,OAAO;AAAA,EAClB,OAAO,EAAE,OAAO,EAAE,IAAI,IAAI,EAAE,IAAI,CAAG,EAAE,SAAS;AAAA,EAC9C,iBAAiB,EACd,KAAK,CAAC,OAAO,QAAQ,OAAO,QAAQ,OAAO,KAAK,CAAC,EACjD,SAAS;AACd,CAAC;AAKM,MAAM,8BAA8B;AAAA,EACzC,KAAK;AAAA,EACL,MAAM;AAAA,EACN,KAAK;AAAA,EACL,MAAM;AAAA,EACN,KAAK;AAAA,EACL,KAAK;AACP;AAEO,SAAS,aACd,WACA,SACA,gBACoB;AACpB,QAAM;AAAA,IACJ;AAAA,IACA,SAAS;AAAA,IACT;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,GAAG;AAAA,EACL,IAAI,QAAQ,UAAU,CAAC;AAEvB,MAAI,UAA8B;AAAA,IAChC,OAAO,gBAAgB;AAAA,IACvB,OAAO,IAAI,QAAQ,QAAQ,SAAS,CAAC,CAAC,EAAE;AAAA,IACxC,OAAO,SAAS;AAAA,EAClB;AACA,MAAI,gBAAgB;AAClB,mBAAe,SAAS,OAAO;AAAA,EACjC,OAAO;AACL,cAAU;AAAA,MACR,GAAG;AAAA,MACH,GAAG;AAAA;AAAA,IACL;AAAA,EACF;AACA,aAAW,KAAK,SAAS;AACvB,QAAI,QAAQ,CAAC,MAAM,QAAW;AAC5B,aAAO,QAAQ,CAAC;AAAA,IAClB;AAAA,EACF;AACA,SAAO;AACT;AAEA,eAAsB,yBACpB,UACA,iBAAkE,OACnC;AAC/B,QAAM,oBAAoB,MAAM,SAAS,YAAY;AACrD,QAAM,eAAe,OAAO,KAAK,IAAI,WAAW,iBAAiB,CAAC;AAClE,QAAM,YAAY,4BAA4B,cAAc;AAC5D,SAAO;AAAA,IACL,SAAS;AAAA,MACP,MAAM;AAAA,MACN,SAAS;AAAA,QACP;AAAA,UACE,OAAO;AAAA,YACL,aAAa;AAAA,YACb,KAAK,QAAQ,SAAS,WAAW,aAAa,SAAS,QAAQ,CAAC;AAAA,UAClE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,IACA,cAAc;AAAA,IACd,KAAK;AAAA,EACP;AACF;AAiBO,SAAS,8BAEd,QAMc;AACd,QAAM;AAAA,IACJ;AAAA,IACA,QAAQ;AAAA,IACR;AAAA,IACA,UAAAA;AAAA,IACA;AAAA,EACF,IAAI;AACJ,QAAM,YAAY,YAAY,MAAM,eAAe,IAAI;AACvD,QAAM,aAAa,GAAG,eAAe,QAAQ,YAAY,IAAI,SAAS;AAEtE,SAAO;AAAA,IACL;AAAA,MACE,MAAM;AAAA,MACN,GAAGA,WAAU;AAAA,MACb,cAAcA,WAAU;AAAA,IAC1B;AAAA,IACA,OAAO,SAAS,EAAE,YAAY,MAAM;AAClC,YAAM,aAAa,aAAa,WAAW,SAAS,cAAc;AAClE,YAAM,SAAS;AAAA,QACb;AAAA,QACA;AAAA,QACA;AAAA,MACF;AACA,YAAM,SAAS,MAAM,OAAO,MAAM,OAAO,OAAO,YAAY;AAAA,QAC1D,QAAQ;AAAA,MACV,CAAC;AACD,aAAO,MAAM,yBAAyB,QAAQ,WAAW,eAAe;AAAA,IAC1E;AAAA,EACF;AACF;AAIO,SAAS,wBAEd,QAMC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,OAAO;AAAA,IACP;AAAA,IACA,SAAS;AAAA,IACT;AAAA,EACF,IAAI;AACJ,SAAO,SAAS;AAAA,IACd;AAAA,IACA,cAAc,gBAAiB;AAAA,IAC/B;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;AAEO,SAAS,aACd,WACA,SACA,gBAC2B;AAC3B,QAAM,UAAU,IAAI,QAAQ,QAAQ,SAAS,CAAC,CAAC;AAC/C,QAAM,QAAQ,QAAQ;AACtB,MAAI,CAAC,OAAO,KAAK;AACf,UAAM,IAAI,MAAM,+BAA+B;AAAA,EACjD;AACA,QAAM,cAAc,OAAO;AAAA,IACzB,MAAM,IAAI,MAAM,MAAM,IAAI,QAAQ,GAAG,IAAI,CAAC;AAAA,IAC1C;AAAA,EACF;AACA,QAAM,YAAY,IAAI,KAAK,CAAC,WAAW,GAAG,SAAS;AAAA,IACjD,MACE,MAAM,eACN,MAAM,IAAI,MAAM,QAAQ,QAAQ,MAAM,IAAI,QAAQ,GAAG,CAAC;AAAA,EAC1D,CAAC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,SAAS;AAAA,IACT;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,GAAG;AAAA,EACL,IAAI,QAAQ,UAAU,CAAC;AAEvB,MAAI,UAAqC;AAAA,IACvC,OAAO,gBAAgB;AAAA,IACvB,MAAM;AAAA,IACN,QAAQ,QAAQ;AAAA,IAChB;AAAA,EACF;AACA,MAAI,gBAAgB;AAClB,mBAAe,SAAS,OAAO;AAAA,EACjC,OAAO;AACL,cAAU;AAAA,MACR,GAAG;AAAA,MACH,GAAG;AAAA;AAAA,IACL;AAAA,EACF;AACA,QAAM,eAAe,QAAQ,QAAQ;AACrC,QAAM,eAAe,QAAQ,QAAQ;AACrC,MAAI,gBAAgB,cAAc;AAChC,QACE,iBAAiB,UACjB,iBAAiB,UACjB,iBAAiB,gBACjB;AACA,YAAM,IAAI;AAAA,QACR,0BAA0B,YAAY,yCAAyC,YAAY;AAAA,MAC7F;AAAA,IACF;AAAA,EACF;AACA,MAAI,iBAAiB,SAAS;AAC5B,UAAM,IAAI,MAAM,iBAAiB,YAAY,oBAAoB;AAAA,EACnE;AACA,UAAQ,kBAAkB,gBAAgB,gBAAgB;AAC1D,aAAW,KAAK,SAAS;AACvB,QAAI,QAAQ,CAAC,MAAM,QAAW;AAC5B,aAAO,QAAQ,CAAC;AAAA,IAClB;AAAA,EACF;AACA,SAAO;AACT;AAEO,SAAS,gCACd,QACsB;AACtB,SAAO;AAAA,IACL,SAAS;AAAA,MACP,MAAM;AAAA,MACN,SAAS;AAAA,QACP;AAAA,UACE,MAAM,OAAO,WAAW,WAAW,SAAS,OAAO;AAAA,QACrD;AAAA,MACF;AAAA,IACF;AAAA,IACA,cAAc;AAAA,IACd,KAAK;AAAA,EACP;AACF;AAkBO,SAAS,qCAEd,QAMc;AACd,QAAM;AAAA,IACJ;AAAA,IACA;AAAA,IACA,QAAQ;AAAA,IACR,UAAAA;AAAA,IACA;AAAA,EACF,IAAI;AACJ,QAAM,YAAY,YAAY,MAAM,eAAe,IAAI;AACvD,QAAM,aACJA,WAAU,QAAQ,GAAG,eAAe,QAAQ,YAAY,IAAI,SAAS;AAEvE,SAAO;AAAA,IACL;AAAA,MACE,MAAM;AAAA,MACN,GAAGA,WAAU;AAAA,MACb,cAAcA,WAAU;AAAA,IAC1B;AAAA,IACA,OAAO,SAAS,EAAE,YAAY,MAAM;AAClC,YAAMC,UAAS,aAAa,WAAW,SAAS,cAAc;AAC9D,YAAM,SAAS;AAAA,QACb;AAAA,QACA;AAAA,QACA;AAAA,MACF;AAEA,YAAM,SAAS,MAAM,OAAO,MAAM,eAAe;AAAA,QAC/C;AAAA,UACE,GAAGA;AAAA,UACH,QAAQ;AAAA,QACV;AAAA,QACA,EAAE,QAAQ,YAAY;AAAA,MACxB;AACA,aAAO,gCAAgC,MAAM;AAAA,IAC/C;AAAA,EACF;AACF;AAIO,SAAS,+BAEd,QAMC;AACD,QAAM;AAAA,IACJ;AAAA,IACA,OAAO;AAAA,IACP;AAAA,IACA,SAAS;AAAA,IACT;AAAA,EACF,IAAI;AACJ,SAAO,SAAS;AAAA,IACd;AAAA,IACA,cAAc,gBAAiB;AAAA,IAC/B;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AACH;","names":["modelRef","params"]}

package/lib/deepseek/deepseek.d.mts

@@ -1,7 +1,8 @@
 import { z } from 'genkit';
 import { ModelInfo, ModelReference } from 'genkit/model';
-import { M as ModelRequestBuilder } from '../audio-CUuCwm-y.mjs';
+import { M as ModelRequestBuilder } from '../audio-CJ8rzf35.mjs';
 import 'openai';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'genkit/plugin';
 import 'genkit/registry';

package/lib/deepseek/deepseek.d.ts

@@ -1,7 +1,8 @@
 import { z } from 'genkit';
 import { ModelInfo, ModelReference } from 'genkit/model';
-import { M as ModelRequestBuilder } from '../audio-CUuCwm-y.js';
+import { M as ModelRequestBuilder } from '../audio-CJ8rzf35.js';
 import 'openai';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'genkit/plugin';
 import 'genkit/registry';

package/lib/deepseek/index.d.mts

@@ -1,9 +1,10 @@
 import { z, ModelReference } from 'genkit';
 import { GenkitPluginV2 } from 'genkit/plugin';
-import { P as PluginOptions } from '../audio-CUuCwm-y.mjs';
+import { P as PluginOptions } from '../audio-CJ8rzf35.mjs';
 import { SUPPORTED_DEEPSEEK_MODELS, DeepSeekChatCompletionConfigSchema } from './deepseek.mjs';
 import 'genkit/model';
 import 'openai';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'genkit/registry';
 import 'openai/resources/images.mjs';

package/lib/deepseek/index.d.ts

@@ -1,9 +1,10 @@
 import { z, ModelReference } from 'genkit';
 import { GenkitPluginV2 } from 'genkit/plugin';
-import { P as PluginOptions } from '../audio-CUuCwm-y.js';
+import { P as PluginOptions } from '../audio-CJ8rzf35.js';
 import { SUPPORTED_DEEPSEEK_MODELS, DeepSeekChatCompletionConfigSchema } from './deepseek.js';
 import 'genkit/model';
 import 'openai';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'genkit/registry';
 import 'openai/resources/images.mjs';

package/lib/embedder.d.mts

@@ -1,7 +1,8 @@
 import 'genkit';
 import 'openai';
-export { d as defineCompatOpenAIEmbedder } from './audio-CUuCwm-y.mjs';
+export { d as defineCompatOpenAIEmbedder } from './audio-CJ8rzf35.mjs';
 import 'genkit/model';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'genkit/plugin';
 import 'genkit/registry';

package/lib/embedder.d.ts

@@ -1,7 +1,8 @@
 import 'genkit';
 import 'openai';
-export { d as defineCompatOpenAIEmbedder } from './audio-CUuCwm-y.js';
+export { d as defineCompatOpenAIEmbedder } from './audio-CJ8rzf35.js';
 import 'genkit/model';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'genkit/plugin';
 import 'genkit/registry';

package/lib/image.d.mts

@@ -2,7 +2,8 @@ import 'genkit';
 import 'genkit/model';
 import 'openai';
 import 'openai/resources/images.mjs';
-export { b as IMAGE_GENERATION_MODEL_INFO, I as ImageGenerationCommonConfigSchema, a as ImageRequestBuilder, e as compatOaiImageModelRef, c as defineCompatOpenAIImageModel } from './audio-CUuCwm-y.mjs';
+export { b as IMAGE_GENERATION_MODEL_INFO, I as ImageGenerationCommonConfigSchema, a as ImageRequestBuilder, e as compatOaiImageModelRef, c as defineCompatOpenAIImageModel } from './audio-CJ8rzf35.mjs';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'genkit/plugin';
 import 'genkit/registry';

package/lib/image.d.ts

@@ -2,7 +2,8 @@ import 'genkit';
 import 'genkit/model';
 import 'openai';
 import 'openai/resources/images.mjs';
-export { b as IMAGE_GENERATION_MODEL_INFO, I as ImageGenerationCommonConfigSchema, a as ImageRequestBuilder, e as compatOaiImageModelRef, c as defineCompatOpenAIImageModel } from './audio-CUuCwm-y.js';
+export { b as IMAGE_GENERATION_MODEL_INFO, I as ImageGenerationCommonConfigSchema, a as ImageRequestBuilder, e as compatOaiImageModelRef, c as defineCompatOpenAIImageModel } from './audio-CJ8rzf35.js';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'genkit/plugin';
 import 'genkit/registry';

package/lib/index.d.mts

@@ -2,8 +2,9 @@ import 'genkit/plugin';
 import 'genkit';
 import 'genkit/registry';
 import 'openai';
-export { C as ChatCompletionCommonConfigSchema, I as ImageGenerationCommonConfigSchema, a as ImageRequestBuilder, M as ModelRequestBuilder, P as PluginOptions, S as SpeechConfigSchema, v as SpeechRequestBuilder, T as TranscriptionConfigSchema, w as TranscriptionRequestBuilder, e as compatOaiImageModelRef, n as compatOaiModelRef, q as compatOaiSpeechModelRef, r as compatOaiTranscriptionModelRef, p as default, d as defineCompatOpenAIEmbedder, c as defineCompatOpenAIImageModel, m as defineCompatOpenAIModel, s as defineCompatOpenAISpeechModel, u as defineCompatOpenAITranscriptionModel, p as openAICompatible, o as openAIModelRunner } from './audio-CUuCwm-y.mjs';
+export { C as ChatCompletionCommonConfigSchema, I as ImageGenerationCommonConfigSchema, a as ImageRequestBuilder, M as ModelRequestBuilder, P as PluginOptions, S as SpeechConfigSchema, v as SpeechRequestBuilder, T as TranscriptionConfigSchema, w as TranscriptionRequestBuilder, e as compatOaiImageModelRef, n as compatOaiModelRef, q as compatOaiSpeechModelRef, r as compatOaiTranscriptionModelRef, p as default, d as defineCompatOpenAIEmbedder, c as defineCompatOpenAIImageModel, m as defineCompatOpenAIModel, s as defineCompatOpenAISpeechModel, u as defineCompatOpenAITranscriptionModel, p as openAICompatible, o as openAIModelRunner } from './audio-CJ8rzf35.mjs';
 import 'genkit/model';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'openai/resources/images.mjs';
 import 'openai/resources/index.mjs';

package/lib/index.d.ts

@@ -2,8 +2,9 @@ import 'genkit/plugin';
 import 'genkit';
 import 'genkit/registry';
 import 'openai';
-export { C as ChatCompletionCommonConfigSchema, I as ImageGenerationCommonConfigSchema, a as ImageRequestBuilder, M as ModelRequestBuilder, P as PluginOptions, S as SpeechConfigSchema, v as SpeechRequestBuilder, T as TranscriptionConfigSchema, w as TranscriptionRequestBuilder, e as compatOaiImageModelRef, n as compatOaiModelRef, q as compatOaiSpeechModelRef, r as compatOaiTranscriptionModelRef, p as default, d as defineCompatOpenAIEmbedder, c as defineCompatOpenAIImageModel, m as defineCompatOpenAIModel, s as defineCompatOpenAISpeechModel, u as defineCompatOpenAITranscriptionModel, p as openAICompatible, o as openAIModelRunner } from './audio-CUuCwm-y.js';
+export { C as ChatCompletionCommonConfigSchema, I as ImageGenerationCommonConfigSchema, a as ImageRequestBuilder, M as ModelRequestBuilder, P as PluginOptions, S as SpeechConfigSchema, v as SpeechRequestBuilder, T as TranscriptionConfigSchema, w as TranscriptionRequestBuilder, e as compatOaiImageModelRef, n as compatOaiModelRef, q as compatOaiSpeechModelRef, r as compatOaiTranscriptionModelRef, p as default, d as defineCompatOpenAIEmbedder, c as defineCompatOpenAIImageModel, m as defineCompatOpenAIModel, s as defineCompatOpenAISpeechModel, u as defineCompatOpenAITranscriptionModel, p as openAICompatible, o as openAIModelRunner } from './audio-CJ8rzf35.js';
 import 'genkit/model';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'openai/resources/images.mjs';
 import 'openai/resources/index.mjs';

package/lib/model.d.mts

@@ -2,7 +2,8 @@ import 'genkit';
 import 'genkit/model';
 import 'openai';
 import 'openai/resources/index.mjs';
-export { C as ChatCompletionCommonConfigSchema, M as ModelRequestBuilder, n as compatOaiModelRef, m as defineCompatOpenAIModel, j as fromOpenAIChoice, k as fromOpenAIChunkChoice, i as fromOpenAIToolCall, o as openAIModelRunner, h as toOpenAIMessages, l as toOpenAIRequestBody, t as toOpenAIRole, g as toOpenAITextAndMedia, f as toOpenAITool } from './audio-CUuCwm-y.mjs';
+export { C as ChatCompletionCommonConfigSchema, M as ModelRequestBuilder, n as compatOaiModelRef, m as defineCompatOpenAIModel, j as fromOpenAIChoice, k as fromOpenAIChunkChoice, i as fromOpenAIToolCall, o as openAIModelRunner, h as toOpenAIMessages, l as toOpenAIRequestBody, t as toOpenAIRole, g as toOpenAITextAndMedia, f as toOpenAITool } from './audio-CJ8rzf35.mjs';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'genkit/plugin';
 import 'genkit/registry';

package/lib/model.d.ts

@@ -2,7 +2,8 @@ import 'genkit';
 import 'genkit/model';
 import 'openai';
 import 'openai/resources/index.mjs';
-export { C as ChatCompletionCommonConfigSchema, M as ModelRequestBuilder, n as compatOaiModelRef, m as defineCompatOpenAIModel, j as fromOpenAIChoice, k as fromOpenAIChunkChoice, i as fromOpenAIToolCall, o as openAIModelRunner, h as toOpenAIMessages, l as toOpenAIRequestBody, t as toOpenAIRole, g as toOpenAITextAndMedia, f as toOpenAITool } from './audio-CUuCwm-y.js';
+export { C as ChatCompletionCommonConfigSchema, M as ModelRequestBuilder, n as compatOaiModelRef, m as defineCompatOpenAIModel, j as fromOpenAIChoice, k as fromOpenAIChunkChoice, i as fromOpenAIToolCall, o as openAIModelRunner, h as toOpenAIMessages, l as toOpenAIRequestBody, t as toOpenAIRole, g as toOpenAITextAndMedia, f as toOpenAITool } from './audio-CJ8rzf35.js';
+import 'openai/core.mjs';
 import 'openai/resources/audio/index.mjs';
 import 'genkit/plugin';
 import 'genkit/registry';

package/lib/model.js

@@ -103,7 +103,7 @@ function toOpenAITextAndMedia(part, visualDetailLevel) {
         contentType = extracted.contentType;
       }
     }
-    if (isImageContentType(contentType)) {
+    if (!contentType || isImageContentType(contentType)) {
       return {
         type: "image_url",
         image_url: {