@ai-sdk/google-vertex 5.0.49 → 5.0.51

package/dist/index.js

@@ -1,5 +1,5 @@
 // src/google-vertex-provider.ts
-import { loadOptionalSetting as loadOptionalSetting2, resolve as resolve6 } from "@ai-sdk/provider-utils";
+import { loadOptionalSetting as loadOptionalSetting2, resolve as resolve7 } from "@ai-sdk/provider-utils";
 
 // src/google-vertex-auth-google-auth-library.ts
 import { GoogleAuth } from "google-auth-library";
@@ -27,13 +27,13 @@ import {
   loadOptionalSetting,
   loadSetting,
   normalizeHeaders,
-  resolve as resolve5,
+  resolve as resolve6,
   withoutTrailingSlash,
   withUserAgentSuffix
 } from "@ai-sdk/provider-utils";
 
 // src/version.ts
-var VERSION = true ? "5.0.49" : "0.0.0-test";
+var VERSION = true ? "5.0.51" : "0.0.0-test";
 
 // src/google-vertex-embedding-model.ts
 import {
@@ -644,6 +644,127 @@ function getBase64Data(file) {
   return convertUint8ArrayToBase64(file.data);
 }
 
+// src/google-vertex-cloud-tts-speech-model.ts
+import {
+  combineHeaders as combineHeaders3,
+  convertBase64ToUint8Array,
+  createJsonResponseHandler as createJsonResponseHandler3,
+  postJsonToApi as postJsonToApi3,
+  resolve as resolve3,
+  serializeModelOptions as serializeModelOptions3,
+  WORKFLOW_DESERIALIZE as WORKFLOW_DESERIALIZE3,
+  WORKFLOW_SERIALIZE as WORKFLOW_SERIALIZE3
+} from "@ai-sdk/provider-utils";
+import { z as z6 } from "zod/v4";
+var DEFAULT_VOICE = "Kore";
+var DEFAULT_LANGUAGE = "en-US";
+var CHIRP3_HD_VOICE_INFIX = "Chirp3-HD";
+var CLOUD_TTS_SYNTHESIZE_URL = "https://texttospeech.googleapis.com/v1/text:synthesize";
+var GoogleVertexCloudTTSSpeechModel = class _GoogleVertexCloudTTSSpeechModel {
+  constructor(modelId, config) {
+    this.modelId = modelId;
+    this.config = config;
+    this.specificationVersion = "v4";
+  }
+  static [WORKFLOW_SERIALIZE3](model) {
+    return serializeModelOptions3({
+      modelId: model.modelId,
+      config: model.config
+    });
+  }
+  static [WORKFLOW_DESERIALIZE3](options) {
+    return new _GoogleVertexCloudTTSSpeechModel(options.modelId, options.config);
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  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 warnings = [];
+    const {
+      text,
+      voice = DEFAULT_VOICE,
+      outputFormat,
+      instructions,
+      speed,
+      language
+    } = options;
+    let voiceName;
+    let languageCode;
+    if (voice.includes(CHIRP3_HD_VOICE_INFIX)) {
+      voiceName = voice;
+      const localePrefix = voice.split(CHIRP3_HD_VOICE_INFIX)[0].replace(/-$/, "");
+      languageCode = language != null ? language : localePrefix || DEFAULT_LANGUAGE;
+    } else {
+      languageCode = language != null ? language : DEFAULT_LANGUAGE;
+      voiceName = `${languageCode}-${CHIRP3_HD_VOICE_INFIX}-${voice}`;
+    }
+    if (instructions != null) {
+      warnings.push({
+        type: "unsupported",
+        feature: "instructions",
+        details: "Google Cloud Text-to-Speech Chirp 3: HD voices do not support the `instructions` option. It was ignored."
+      });
+    }
+    if (outputFormat != null && outputFormat !== "wav") {
+      warnings.push({
+        type: "unsupported",
+        feature: "outputFormat",
+        details: `Unsupported output format: ${outputFormat}. Using wav instead.`
+      });
+    }
+    const requestBody = {
+      input: { text },
+      voice: { languageCode, name: voiceName },
+      audioConfig: {
+        audioEncoding: "LINEAR16",
+        ...speed != null ? { speakingRate: speed } : {}
+      }
+    };
+    const {
+      value: response,
+      responseHeaders,
+      rawValue: rawResponse
+    } = await postJsonToApi3({
+      url: CLOUD_TTS_SYNTHESIZE_URL,
+      headers: combineHeaders3(
+        this.config.headers ? await resolve3(this.config.headers) : void 0,
+        options.headers
+      ),
+      body: requestBody,
+      failedResponseHandler: googleVertexFailedResponseHandler,
+      successfulResponseHandler: createJsonResponseHandler3(
+        googleVertexCloudTTSResponseSchema
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    const audio = response.audioContent != null ? convertBase64ToUint8Array(response.audioContent) : new Uint8Array(0);
+    return {
+      audio,
+      warnings,
+      request: {
+        body: JSON.stringify(requestBody)
+      },
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: rawResponse
+      },
+      providerMetadata: {
+        google: {
+          mimeType: "audio/wav"
+        }
+      }
+    };
+  }
+};
+var googleVertexCloudTTSResponseSchema = z6.object({
+  audioContent: z6.string().nullish()
+});
+
 // src/google-vertex-tools.ts
 import { googleTools } from "@ai-sdk/google/internal";
 var googleVertexTools = {
@@ -658,31 +779,31 @@ var googleVertexTools = {
 
 // src/google-vertex-transcription-model.ts
 import {
-  combineHeaders as combineHeaders3,
+  combineHeaders as combineHeaders4,
   convertUint8ArrayToBase64 as convertUint8ArrayToBase642,
-  createJsonResponseHandler as createJsonResponseHandler3,
+  createJsonResponseHandler as createJsonResponseHandler4,
   parseProviderOptions as parseProviderOptions3,
-  postJsonToApi as postJsonToApi3,
-  resolve as resolve3,
-  serializeModelOptions as serializeModelOptions3,
-  WORKFLOW_DESERIALIZE as WORKFLOW_DESERIALIZE3,
-  WORKFLOW_SERIALIZE as WORKFLOW_SERIALIZE3
+  postJsonToApi as postJsonToApi4,
+  resolve as resolve4,
+  serializeModelOptions as serializeModelOptions4,
+  WORKFLOW_DESERIALIZE as WORKFLOW_DESERIALIZE4,
+  WORKFLOW_SERIALIZE as WORKFLOW_SERIALIZE4
 } from "@ai-sdk/provider-utils";
-import { z as z7 } from "zod/v4";
+import { z as z8 } from "zod/v4";
 
 // src/google-vertex-transcription-model-options.ts
-import { z as z6 } from "zod/v4";
-var googleVertexTranscriptionProviderOptionsSchema = z6.object({
+import { z as z7 } from "zod/v4";
+var googleVertexTranscriptionProviderOptionsSchema = z7.object({
   /**
    * BCP-47 language codes to recognize (e.g. `['en-US']`), or `['auto']` to let
    * Chirp auto-detect the spoken language. Defaults to `['auto']`. For
    * `telephony`, pass a supported explicit language code.
    */
-  languageCodes: z6.array(z6.string()).optional(),
+  languageCodes: z7.array(z7.string()).optional(),
   /**
    * Whether to add punctuation to the transcript. Defaults to `true`.
    */
-  enableAutomaticPunctuation: z6.boolean().optional(),
+  enableAutomaticPunctuation: z7.boolean().optional(),
   /**
    * Whether to include word-level timestamps. Defaults to `true` so the
    * transcription result can include segments.
@@ -690,7 +811,7 @@ var googleVertexTranscriptionProviderOptionsSchema = z6.object({
    * Enabling word-level timestamps can reduce transcription quality and speed
    * for Chirp models.
    */
-  enableWordTimeOffsets: z6.boolean().optional(),
+  enableWordTimeOffsets: z7.boolean().optional(),
   /**
    * The Cloud Speech-to-Text region for the request (e.g. `'us'`, `'eu'`,
    * `'us-central1'`). Defaults to the provider `location`.
@@ -699,7 +820,7 @@ var googleVertexTranscriptionProviderOptionsSchema = z6.object({
    * available in specific Speech-to-Text regions and is not available in the
    * `global` location.
    */
-  region: z6.string().optional()
+  region: z7.string().optional()
 });
 
 // src/google-vertex-transcription-model.ts
@@ -727,13 +848,13 @@ var GoogleVertexTranscriptionModel = class _GoogleVertexTranscriptionModel {
     this.config = config;
     this.specificationVersion = "v4";
   }
-  static [WORKFLOW_SERIALIZE3](model) {
-    return serializeModelOptions3({
+  static [WORKFLOW_SERIALIZE4](model) {
+    return serializeModelOptions4({
       modelId: model.modelId,
       config: model.config
     });
   }
-  static [WORKFLOW_DESERIALIZE3](options) {
+  static [WORKFLOW_DESERIALIZE4](options) {
     return new _GoogleVertexTranscriptionModel(options.modelId, options.config);
   }
   get provider() {
@@ -777,15 +898,15 @@ var GoogleVertexTranscriptionModel = class _GoogleVertexTranscriptionModel {
       value: response,
       responseHeaders,
       rawValue: rawResponse
-    } = await postJsonToApi3({
+    } = await postJsonToApi4({
       url,
-      headers: combineHeaders3(
-        this.config.headers ? await resolve3(this.config.headers) : void 0,
+      headers: combineHeaders4(
+        this.config.headers ? await resolve4(this.config.headers) : void 0,
         options.headers
       ),
       body: requestBody,
       failedResponseHandler: googleVertexFailedResponseHandler,
-      successfulResponseHandler: createJsonResponseHandler3(
+      successfulResponseHandler: createJsonResponseHandler4(
         googleVertexTranscriptionResponseSchema
       ),
       abortSignal: options.abortSignal,
@@ -824,26 +945,26 @@ var GoogleVertexTranscriptionModel = class _GoogleVertexTranscriptionModel {
     };
   }
 };
-var googleVertexTranscriptionResponseSchema = z7.object({
-  results: z7.array(
-    z7.object({
-      alternatives: z7.array(
-        z7.object({
-          transcript: z7.string().nullish(),
-          words: z7.array(
-            z7.object({
-              word: z7.string().nullish(),
-              startOffset: z7.string().nullish(),
-              endOffset: z7.string().nullish()
+var googleVertexTranscriptionResponseSchema = z8.object({
+  results: z8.array(
+    z8.object({
+      alternatives: z8.array(
+        z8.object({
+          transcript: z8.string().nullish(),
+          words: z8.array(
+            z8.object({
+              word: z8.string().nullish(),
+              startOffset: z8.string().nullish(),
+              endOffset: z8.string().nullish()
             })
           ).nullish()
         })
       ).nullish(),
-      languageCode: z7.string().nullish()
+      languageCode: z8.string().nullish()
     })
   ).nullish(),
-  metadata: z7.object({
-    totalBilledDuration: z7.string().nullish()
+  metadata: z8.object({
+    totalBilledDuration: z8.string().nullish()
   }).nullish()
 });
 
@@ -852,31 +973,31 @@ import {
   AISDKError
 } from "@ai-sdk/provider";
 import {
-  combineHeaders as combineHeaders4,
+  combineHeaders as combineHeaders5,
   convertUint8ArrayToBase64 as convertUint8ArrayToBase643,
-  createJsonResponseHandler as createJsonResponseHandler4,
+  createJsonResponseHandler as createJsonResponseHandler5,
   parseProviderOptions as parseProviderOptions4,
-  postJsonToApi as postJsonToApi4,
-  resolve as resolve4
+  postJsonToApi as postJsonToApi5,
+  resolve as resolve5
 } from "@ai-sdk/provider-utils";
-import { z as z9 } from "zod/v4";
+import { z as z10 } from "zod/v4";
 
 // src/google-vertex-video-model-options.ts
 import { lazySchema, zodSchema } from "@ai-sdk/provider-utils";
-import { z as z8 } from "zod/v4";
+import { z as z9 } from "zod/v4";
 var googleVertexVideoModelOptionsSchema = lazySchema(
   () => zodSchema(
-    z8.looseObject({
-      pollIntervalMs: z8.number().positive().nullish(),
-      pollTimeoutMs: z8.number().positive().nullish(),
-      personGeneration: z8.enum(["dont_allow", "allow_adult", "allow_all"]).nullish(),
-      negativePrompt: z8.string().nullish(),
-      generateAudio: z8.boolean().nullish(),
-      gcsOutputDirectory: z8.string().nullish(),
-      referenceImages: z8.array(
-        z8.object({
-          bytesBase64Encoded: z8.string().nullish(),
-          gcsUri: z8.string().nullish()
+    z9.looseObject({
+      pollIntervalMs: z9.number().positive().nullish(),
+      pollTimeoutMs: z9.number().positive().nullish(),
+      personGeneration: z9.enum(["dont_allow", "allow_adult", "allow_all"]).nullish(),
+      negativePrompt: z9.string().nullish(),
+      generateAudio: z9.boolean().nullish(),
+      gcsOutputDirectory: z9.string().nullish(),
+      referenceImages: z9.array(
+        z9.object({
+          bytesBase64Encoded: z9.string().nullish(),
+          gcsUri: z9.string().nullish()
         })
       ).nullish()
     })
@@ -1097,17 +1218,17 @@ var GoogleVertexVideoModel = class {
     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 { instances, parameters, warnings } = await this.buildRequest(options);
-    const { value: operation, responseHeaders } = await postJsonToApi4({
+    const { value: operation, responseHeaders } = await postJsonToApi5({
       url: `${this.config.baseURL}/models/${this.modelId}:predictLongRunning`,
-      headers: combineHeaders4(
-        await resolve4(this.config.headers),
+      headers: combineHeaders5(
+        await resolve5(this.config.headers),
         options.headers
       ),
       body: {
         instances,
         parameters
       },
-      successfulResponseHandler: createJsonResponseHandler4(
+      successfulResponseHandler: createJsonResponseHandler5(
         googleVertexOperationSchema
       ),
       failedResponseHandler: googleVertexFailedResponseHandler,
@@ -1135,16 +1256,16 @@ var GoogleVertexVideoModel = class {
     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 { operationName } = options.operation;
-    const { value: statusOperation, responseHeaders } = await postJsonToApi4({
+    const { value: statusOperation, responseHeaders } = await postJsonToApi5({
       url: `${this.config.baseURL}/models/${this.modelId}:fetchPredictOperation`,
-      headers: combineHeaders4(
-        await resolve4(this.config.headers),
+      headers: combineHeaders5(
+        await resolve5(this.config.headers),
         options.headers
       ),
       body: {
         operationName
       },
-      successfulResponseHandler: createJsonResponseHandler4(
+      successfulResponseHandler: createJsonResponseHandler5(
         googleVertexOperationSchema
       ),
       failedResponseHandler: googleVertexFailedResponseHandler,
@@ -1180,23 +1301,23 @@ var GoogleVertexVideoModel = class {
     });
   }
 };
-var googleVertexOperationSchema = z9.object({
-  name: z9.string().nullish(),
-  done: z9.boolean().nullish(),
-  error: z9.object({
-    code: z9.number().nullish(),
-    message: z9.string(),
-    status: z9.string().nullish()
+var googleVertexOperationSchema = z10.object({
+  name: z10.string().nullish(),
+  done: z10.boolean().nullish(),
+  error: z10.object({
+    code: z10.number().nullish(),
+    message: z10.string(),
+    status: z10.string().nullish()
   }).nullish(),
-  response: z9.object({
-    videos: z9.array(
-      z9.object({
-        bytesBase64Encoded: z9.string().nullish(),
-        gcsUri: z9.string().nullish(),
-        mimeType: z9.string().nullish()
+  response: z10.object({
+    videos: z10.array(
+      z10.object({
+        bytesBase64Encoded: z10.string().nullish(),
+        gcsUri: z10.string().nullish(),
+        mimeType: z10.string().nullish()
       })
     ).nullish(),
-    raiMediaFilteredCount: z9.number().nullish()
+    raiMediaFilteredCount: z10.number().nullish()
   }).nullish()
 });
 
@@ -1256,7 +1377,7 @@ function createGoogleVertex(options = {}) {
   const createConfig = (name, { endpoint = false } = {}) => {
     const getHeaders = async () => {
       var _a;
-      const originalHeaders = await resolve5((_a = options.headers) != null ? _a : {});
+      const originalHeaders = await resolve6((_a = options.headers) != null ? _a : {});
       return withUserAgentSuffix(
         originalHeaders,
         `ai-sdk/google-vertex/${VERSION}`
@@ -1324,7 +1445,22 @@ function createGoogleVertex(options = {}) {
       generateId: (_a = options.generateId) != null ? _a : generateId
     });
   };
-  const createSpeechModel = (modelId) => new GoogleSpeechModel(modelId, createConfig("speech"));
+  const createSpeechModel = (modelId) => {
+    if (modelId.startsWith("chirp")) {
+      if (apiKey) {
+        throw new Error(
+          "Google Vertex Chirp speech models do not support Express Mode API keys. Use standard Google Cloud credentials instead."
+        );
+      }
+      const config = createConfig("speech");
+      return new GoogleVertexCloudTTSSpeechModel(modelId, {
+        provider: config.provider,
+        headers: config.headers,
+        fetch: config.fetch
+      });
+    }
+    return new GoogleSpeechModel(modelId, createConfig("speech"));
+  };
   const createTranscriptionModel = (modelId) => {
     if (apiKey) {
       throw new Error(
@@ -1383,7 +1519,7 @@ function createGoogleVertex2(options = {}) {
     ...options,
     headers: async () => ({
       Authorization: `Bearer ${await generateAuthToken()}`,
-      ...await resolve6(options.headers)
+      ...await resolve7(options.headers)
     })
   });
 }