@ai-sdk/openai 1.3.10 → 1.3.12

package/dist/index.mjs

@@ -1629,7 +1629,7 @@ import {
   postFormDataToApi
 } from "@ai-sdk/provider-utils";
 import { z as z6 } from "zod";
-var OpenAIProviderOptionsSchema = z6.object({
+var openAIProviderOptionsSchema = z6.object({
   include: z6.array(z6.string()).nullish(),
   language: z6.string().nullish(),
   prompt: z6.string().nullish(),
@@ -1714,7 +1714,7 @@ var OpenAITranscriptionModel = class {
     const openAIOptions = parseProviderOptions({
       provider: "openai",
       providerOptions,
-      schema: OpenAIProviderOptionsSchema
+      schema: openAIProviderOptionsSchema
     });
     const formData = new FormData();
     const blob = audio instanceof Uint8Array ? new Blob([audio]) : new Blob([convertBase64ToUint8Array(audio)]);
@@ -2621,6 +2621,110 @@ var openaiTools = {
   webSearchPreview: webSearchPreviewTool
 };
 
+// src/openai-speech-model.ts
+import {
+  combineHeaders as combineHeaders7,
+  createBinaryResponseHandler,
+  parseProviderOptions as parseProviderOptions3,
+  postJsonToApi as postJsonToApi6
+} from "@ai-sdk/provider-utils";
+import { z as z9 } from "zod";
+var OpenAIProviderOptionsSchema = z9.object({
+  instructions: z9.string().nullish(),
+  speed: z9.number().min(0.25).max(4).default(1).nullish()
+});
+var OpenAISpeechModel = class {
+  constructor(modelId, config) {
+    this.modelId = modelId;
+    this.config = config;
+    this.specificationVersion = "v1";
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  getArgs({
+    text,
+    voice = "alloy",
+    outputFormat = "mp3",
+    speed,
+    instructions,
+    providerOptions
+  }) {
+    const warnings = [];
+    const openAIOptions = parseProviderOptions3({
+      provider: "openai",
+      providerOptions,
+      schema: OpenAIProviderOptionsSchema
+    });
+    const requestBody = {
+      model: this.modelId,
+      input: text,
+      voice,
+      response_format: "mp3",
+      speed,
+      instructions
+    };
+    if (outputFormat) {
+      if (["mp3", "opus", "aac", "flac", "wav", "pcm"].includes(outputFormat)) {
+        requestBody.response_format = outputFormat;
+      } else {
+        warnings.push({
+          type: "unsupported-setting",
+          setting: "outputFormat",
+          details: `Unsupported output format: ${outputFormat}. Using mp3 instead.`
+        });
+      }
+    }
+    if (openAIOptions) {
+      const speechModelOptions = {};
+      for (const key in speechModelOptions) {
+        const value = speechModelOptions[key];
+        if (value !== void 0) {
+          requestBody[key] = value;
+        }
+      }
+    }
+    return {
+      requestBody,
+      warnings
+    };
+  }
+  async doGenerate(options) {
+    var _a, _b, _c;
+    const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
+    const { requestBody, warnings } = this.getArgs(options);
+    const {
+      value: audio,
+      responseHeaders,
+      rawValue: rawResponse
+    } = await postJsonToApi6({
+      url: this.config.url({
+        path: "/audio/speech",
+        modelId: this.modelId
+      }),
+      headers: combineHeaders7(this.config.headers(), options.headers),
+      body: requestBody,
+      failedResponseHandler: openaiFailedResponseHandler,
+      successfulResponseHandler: createBinaryResponseHandler(),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    return {
+      audio,
+      warnings,
+      request: {
+        body: JSON.stringify(requestBody)
+      },
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: rawResponse
+      }
+    };
+  }
+};
+
 // src/openai-provider.ts
 function createOpenAI(options = {}) {
   var _a, _b, _c;
@@ -2669,6 +2773,12 @@ function createOpenAI(options = {}) {
     headers: getHeaders,
     fetch: options.fetch
   });
+  const createSpeechModel = (modelId) => new OpenAISpeechModel(modelId, {
+    provider: `${providerName}.speech`,
+    url: ({ path }) => `${baseURL}${path}`,
+    headers: getHeaders,
+    fetch: options.fetch
+  });
   const createLanguageModel = (modelId, settings) => {
     if (new.target) {
       throw new Error(
@@ -2705,6 +2815,8 @@ function createOpenAI(options = {}) {
   provider.imageModel = createImageModel;
   provider.transcription = createTranscriptionModel;
   provider.transcriptionModel = createTranscriptionModel;
+  provider.speech = createSpeechModel;
+  provider.speechModel = createSpeechModel;
   provider.tools = openaiTools;
   return provider;
 }