@ai-sdk/google-vertex 4.0.189 → 4.0.191

package/dist/index.js

@@ -20,14 +20,236 @@ var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: tru
 // src/index.ts
 var index_exports = {};
 __export(index_exports, {
+  GoogleVertexGeminiTranscriptionModel: () => GoogleVertexGeminiTranscriptionModel,
   VERSION: () => VERSION,
   createVertex: () => createVertex2,
   vertex: () => vertex
 });
 module.exports = __toCommonJS(index_exports);
 
+// src/gemini-transcription/google-vertex-gemini-transcription-model.ts
+var import_provider = require("@ai-sdk/provider");
+var import_provider_utils2 = require("@ai-sdk/provider-utils");
+var import_v43 = require("zod/v4");
+
+// src/google-vertex-error.ts
+var import_provider_utils = require("@ai-sdk/provider-utils");
+var import_v4 = require("zod/v4");
+var googleVertexErrorDataSchema = import_v4.z.object({
+  error: import_v4.z.object({
+    code: import_v4.z.number().nullable(),
+    message: import_v4.z.string(),
+    status: import_v4.z.string()
+  })
+});
+var googleVertexFailedResponseHandler = (0, import_provider_utils.createJsonErrorResponseHandler)(
+  {
+    errorSchema: googleVertexErrorDataSchema,
+    errorToMessage: (data) => data.error.message
+  }
+);
+
+// src/gemini-transcription/google-vertex-gemini-transcription-model-options.ts
+var import_v42 = require("zod/v4");
+var googleVertexGeminiTranscriptionModelOptions = import_v42.z.object({
+  /**
+   * BCP-47 language codes providing hints about the languages present in the
+   * audio. If omitted or empty, defaults to automatic language detection.
+   */
+  languageCodes: import_v42.z.array(import_v42.z.string()).optional(),
+  /**
+   * Custom vocabulary phrases, which bias the speech recognition model
+   * toward recognizing specific terms.
+   */
+  customVocabulary: import_v42.z.array(import_v42.z.string()).optional(),
+  /**
+   * Enables word-level timestamp generation.
+   */
+  wordTimestamp: import_v42.z.boolean().optional(),
+  /**
+   * Enables speaker diarization.
+   */
+  diarization: import_v42.z.boolean().optional(),
+  /**
+   * Transcription output formatting mode.
+   *
+   * - `VERBATIM` (default): exact literal transcript preserving filler
+   *   words, repetitions, and false starts.
+   * - `SMART`: cleans up and structures the transcript in real time —
+   *   disfluency removal, inline self-corrections, structured formatting
+   *   (lists, numbers, dates, paragraph breaks), and grammar/casing polish.
+   */
+  mode: import_v42.z.enum(["SMART", "VERBATIM"]).optional()
+});
+
+// src/gemini-transcription/google-vertex-gemini-transcription-model.ts
+function isLiveTranscriptionModelId(modelId) {
+  return modelId.includes("-live");
+}
+var GoogleVertexGeminiTranscriptionModel = class {
+  constructor(modelId, config) {
+    this.modelId = modelId;
+    this.config = config;
+    this.specificationVersion = "v3";
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  async parseOptions(providerOptions) {
+    for (const provider of ["googleVertex", "vertex", "google"]) {
+      const parsed = await (0, import_provider_utils2.parseProviderOptions)({
+        provider,
+        providerOptions,
+        schema: googleVertexGeminiTranscriptionModelOptions
+      });
+      if (parsed != null) return parsed;
+    }
+    return void 0;
+  }
+  async doGenerate(options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i;
+    if (isLiveTranscriptionModelId(this.modelId)) {
+      throw new import_provider.InvalidArgumentError({
+        argument: "modelId",
+        message: `Model '${this.modelId}' only supports streaming transcription, which requires AI SDK v7. Use a unary model such as 'gemini-3.5-transcribe'.`
+      });
+    }
+    const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
+    const warnings = [];
+    const googleOptions = await this.parseOptions(options.providerOptions);
+    const audioTranscriptionConfig = buildAudioTranscriptionConfig(googleOptions);
+    const requestBody = {
+      contents: [
+        {
+          role: "user",
+          parts: [
+            {
+              inlineData: {
+                mimeType: options.mediaType,
+                data: (0, import_provider_utils2.convertToBase64)(options.audio)
+              }
+            }
+          ]
+        }
+      ],
+      ...audioTranscriptionConfig != null ? { generationConfig: { audioTranscriptionConfig } } : {}
+    };
+    const {
+      value: response,
+      responseHeaders,
+      rawValue: rawResponse
+    } = await (0, import_provider_utils2.postJsonToApi)({
+      url: `${this.config.baseURL}/models/${this.modelId}:generateContent`,
+      headers: (0, import_provider_utils2.combineHeaders)(
+        this.config.headers ? await (0, import_provider_utils2.resolve)(this.config.headers) : void 0,
+        options.headers
+      ),
+      body: requestBody,
+      failedResponseHandler: googleVertexFailedResponseHandler,
+      successfulResponseHandler: (0, import_provider_utils2.createJsonResponseHandler)(
+        googleVertexGeminiTranscriptionResponseSchema
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    const parts = (_g = (_f = (_e = (_d = response.candidates) == null ? void 0 : _d[0]) == null ? void 0 : _e.content) == null ? void 0 : _f.parts) != null ? _g : [];
+    const plainText = parts.map((part) => {
+      var _a2;
+      return (_a2 = part.text) != null ? _a2 : "";
+    }).join("");
+    const transcriptionText = parts.map((part) => {
+      var _a2, _b2;
+      return (_b2 = (_a2 = part.audioTranscription) == null ? void 0 : _a2.text) != null ? _b2 : "";
+    }).join("");
+    const text = plainText !== "" ? plainText : transcriptionText;
+    let language;
+    const segments = [];
+    for (const part of parts) {
+      const transcription = part.audioTranscription;
+      if (transcription == null) continue;
+      language != null ? language : language = (_h = transcription.languageCode) != null ? _h : void 0;
+      for (const word of (_i = transcription.words) != null ? _i : []) {
+        const startSecond = parseOffsetSeconds(word.startOffset);
+        const endSecond = parseOffsetSeconds(word.endOffset);
+        if (word.word == null || startSecond == null || endSecond == null) {
+          continue;
+        }
+        segments.push({ text: word.word, startSecond, endSecond });
+      }
+    }
+    return {
+      text,
+      segments,
+      language,
+      durationInSeconds: void 0,
+      warnings,
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: rawResponse
+      },
+      ...response.usageMetadata != null ? {
+        providerMetadata: {
+          google: { usageMetadata: response.usageMetadata }
+        }
+      } : {}
+    };
+  }
+};
+function buildAudioTranscriptionConfig(options) {
+  if (options == null) return void 0;
+  const config = {};
+  if (options.languageCodes != null) {
+    config.languageCodes = options.languageCodes;
+  }
+  if (options.customVocabulary != null) {
+    config.customVocabulary = options.customVocabulary;
+  }
+  if (options.wordTimestamp != null) {
+    config.wordTimestamp = options.wordTimestamp;
+  }
+  if (options.diarization != null) {
+    config.diarization = options.diarization;
+  }
+  if (options.mode != null) {
+    config.mode = options.mode;
+  }
+  return Object.keys(config).length > 0 ? config : void 0;
+}
+function parseOffsetSeconds(offset) {
+  if (offset == null) return void 0;
+  const parsed = Number.parseFloat(offset);
+  return Number.isFinite(parsed) ? parsed : void 0;
+}
+var googleVertexGeminiTranscriptionWordSchema = import_v43.z.object({
+  word: import_v43.z.string().nullish(),
+  startOffset: import_v43.z.string().nullish(),
+  endOffset: import_v43.z.string().nullish()
+});
+var googleVertexGeminiTranscriptionResponseSchema = import_v43.z.object({
+  candidates: import_v43.z.array(
+    import_v43.z.object({
+      content: import_v43.z.object({
+        parts: import_v43.z.array(
+          import_v43.z.object({
+            text: import_v43.z.string().nullish(),
+            audioTranscription: import_v43.z.object({
+              text: import_v43.z.string().nullish(),
+              languageCode: import_v43.z.string().nullish(),
+              speakerLabel: import_v43.z.string().nullish(),
+              words: import_v43.z.array(googleVertexGeminiTranscriptionWordSchema).nullish()
+            }).nullish()
+          })
+        ).nullish()
+      }).nullish()
+    })
+  ).nullish(),
+  usageMetadata: import_v43.z.record(import_v43.z.string(), import_v43.z.unknown()).nullish()
+});
+
 // src/google-vertex-provider-node.ts
-var import_provider_utils7 = require("@ai-sdk/provider-utils");
+var import_provider_utils8 = require("@ai-sdk/provider-utils");
 
 // src/google-vertex-auth-google-auth-library.ts
 var import_google_auth_library = require("google-auth-library");
@@ -46,41 +268,24 @@ function createAuthTokenGenerator(options) {
 
 // src/google-vertex-provider.ts
 var import_internal3 = require("@ai-sdk/google/internal");
-var import_provider_utils6 = require("@ai-sdk/provider-utils");
+var import_provider_utils7 = require("@ai-sdk/provider-utils");
 
 // src/version.ts
-var VERSION = true ? "4.0.189" : "0.0.0-test";
+var VERSION = true ? "4.0.191" : "0.0.0-test";
 
 // src/google-vertex-embedding-model.ts
-var import_provider = require("@ai-sdk/provider");
-var import_provider_utils2 = require("@ai-sdk/provider-utils");
-var import_v43 = require("zod/v4");
-
-// src/google-vertex-error.ts
-var import_provider_utils = require("@ai-sdk/provider-utils");
-var import_v4 = require("zod/v4");
-var googleVertexErrorDataSchema = import_v4.z.object({
-  error: import_v4.z.object({
-    code: import_v4.z.number().nullable(),
-    message: import_v4.z.string(),
-    status: import_v4.z.string()
-  })
-});
-var googleVertexFailedResponseHandler = (0, import_provider_utils.createJsonErrorResponseHandler)(
-  {
-    errorSchema: googleVertexErrorDataSchema,
-    errorToMessage: (data) => data.error.message
-  }
-);
+var import_provider2 = require("@ai-sdk/provider");
+var import_provider_utils3 = require("@ai-sdk/provider-utils");
+var import_v45 = require("zod/v4");
 
 // src/google-vertex-embedding-options.ts
-var import_v42 = require("zod/v4");
-var googleVertexEmbeddingModelOptions = import_v42.z.object({
+var import_v44 = require("zod/v4");
+var googleVertexEmbeddingModelOptions = import_v44.z.object({
   /**
    * Optional. Optional reduced dimension for the output embedding.
    * If set, excessive values in the output embedding are truncated from the end.
    */
-  outputDimensionality: import_v42.z.number().optional(),
+  outputDimensionality: import_v44.z.number().optional(),
   /**
    * Optional. Specifies the task type for generating embeddings.
    * Supported task types:
@@ -93,7 +298,7 @@ var googleVertexEmbeddingModelOptions = import_v42.z.object({
    * - FACT_VERIFICATION: Optimized for verifying factual information.
    * - CODE_RETRIEVAL_QUERY: Optimized for retrieving code blocks based on natural language queries.
    */
-  taskType: import_v42.z.enum([
+  taskType: import_v44.z.enum([
     "SEMANTIC_SIMILARITY",
     "CLASSIFICATION",
     "CLUSTERING",
@@ -108,12 +313,12 @@ var googleVertexEmbeddingModelOptions = import_v42.z.object({
    * Only valid when task_type is set to 'RETRIEVAL_DOCUMENT'.
    * Helps the model produce better embeddings by providing additional context.
    */
-  title: import_v42.z.string().optional(),
+  title: import_v44.z.string().optional(),
   /**
    * Optional. When set to true, input text will be truncated. When set to false,
    * an error is returned if the input text is longer than the maximum length supported by the model. Defaults to true.
    */
-  autoTruncate: import_v42.z.boolean().optional()
+  autoTruncate: import_v44.z.boolean().optional()
 });
 
 // src/google-vertex-embedding-model.ts
@@ -139,13 +344,13 @@ var GoogleVertexEmbeddingModel = class {
     providerOptions
   }) {
     var _a;
-    let googleOptions = await (0, import_provider_utils2.parseProviderOptions)({
+    let googleOptions = await (0, import_provider_utils3.parseProviderOptions)({
       provider: "vertex",
       providerOptions,
       schema: googleVertexEmbeddingModelOptions
     });
     if (googleOptions == null) {
-      googleOptions = await (0, import_provider_utils2.parseProviderOptions)({
+      googleOptions = await (0, import_provider_utils3.parseProviderOptions)({
         provider: "google",
         providerOptions,
         schema: googleVertexEmbeddingModelOptions
@@ -153,15 +358,15 @@ var GoogleVertexEmbeddingModel = class {
     }
     googleOptions = googleOptions != null ? googleOptions : {};
     if (values.length > this.maxEmbeddingsPerCall) {
-      throw new import_provider.TooManyEmbeddingValuesForCallError({
+      throw new import_provider2.TooManyEmbeddingValuesForCallError({
         provider: this.provider,
         modelId: this.modelId,
         maxEmbeddingsPerCall: this.maxEmbeddingsPerCall,
         values
       });
     }
-    const mergedHeaders = (0, import_provider_utils2.combineHeaders)(
-      await (0, import_provider_utils2.resolve)(this.config.headers),
+    const mergedHeaders = (0, import_provider_utils3.combineHeaders)(
+      await (0, import_provider_utils3.resolve)(this.config.headers),
       headers
     );
     if (usesEmbedContentEndpoint(this.modelId)) {
@@ -169,7 +374,7 @@ var GoogleVertexEmbeddingModel = class {
         responseHeaders: responseHeaders2,
         value: response2,
         rawValue: rawValue2
-      } = await (0, import_provider_utils2.postJsonToApi)({
+      } = await (0, import_provider_utils3.postJsonToApi)({
         url: `${this.config.baseURL}/models/${this.modelId}:embedContent`,
         headers: mergedHeaders,
         body: {
@@ -182,7 +387,7 @@ var GoogleVertexEmbeddingModel = class {
           }
         },
         failedResponseHandler: googleVertexFailedResponseHandler,
-        successfulResponseHandler: (0, import_provider_utils2.createJsonResponseHandler)(
+        successfulResponseHandler: (0, import_provider_utils3.createJsonResponseHandler)(
           googleVertexEmbedContentResponseSchema
         ),
         abortSignal,
@@ -200,7 +405,7 @@ var GoogleVertexEmbeddingModel = class {
       responseHeaders,
       value: response,
       rawValue
-    } = await (0, import_provider_utils2.postJsonToApi)({
+    } = await (0, import_provider_utils3.postJsonToApi)({
       url,
       headers: mergedHeaders,
       body: {
@@ -215,7 +420,7 @@ var GoogleVertexEmbeddingModel = class {
         }
       },
       failedResponseHandler: googleVertexFailedResponseHandler,
-      successfulResponseHandler: (0, import_provider_utils2.createJsonResponseHandler)(
+      successfulResponseHandler: (0, import_provider_utils3.createJsonResponseHandler)(
         googleVertexTextEmbeddingResponseSchema
       ),
       abortSignal,
@@ -236,24 +441,24 @@ var GoogleVertexEmbeddingModel = class {
     };
   }
 };
-var googleVertexTextEmbeddingResponseSchema = import_v43.z.object({
-  predictions: import_v43.z.array(
-    import_v43.z.object({
-      embeddings: import_v43.z.object({
-        values: import_v43.z.array(import_v43.z.number()),
-        statistics: import_v43.z.object({
-          token_count: import_v43.z.number()
+var googleVertexTextEmbeddingResponseSchema = import_v45.z.object({
+  predictions: import_v45.z.array(
+    import_v45.z.object({
+      embeddings: import_v45.z.object({
+        values: import_v45.z.array(import_v45.z.number()),
+        statistics: import_v45.z.object({
+          token_count: import_v45.z.number()
         })
       })
     })
   )
 });
-var googleVertexEmbedContentResponseSchema = import_v43.z.object({
-  embedding: import_v43.z.object({
-    values: import_v43.z.array(import_v43.z.number())
+var googleVertexEmbedContentResponseSchema = import_v45.z.object({
+  embedding: import_v45.z.object({
+    values: import_v45.z.array(import_v45.z.number())
   }),
-  usageMetadata: import_v43.z.object({
-    promptTokenCount: import_v43.z.number().nullish()
+  usageMetadata: import_v45.z.object({
+    promptTokenCount: import_v45.z.number().nullish()
   }).nullish()
 });
 function usesEmbedContentEndpoint(modelId) {
@@ -262,8 +467,8 @@ function usesEmbedContentEndpoint(modelId) {
 
 // src/google-vertex-image-model.ts
 var import_internal = require("@ai-sdk/google/internal");
-var import_provider_utils3 = require("@ai-sdk/provider-utils");
-var import_v44 = require("zod/v4");
+var import_provider_utils4 = require("@ai-sdk/provider-utils");
+var import_v46 = require("zod/v4");
 var GoogleVertexImageModel = class {
   constructor(modelId, config) {
     this.modelId = modelId;
@@ -306,7 +511,7 @@ var GoogleVertexImageModel = class {
         details: "This model does not support the `size` option. Use `aspectRatio` instead."
       });
     }
-    const vertexImageOptions = await (0, import_provider_utils3.parseProviderOptions)({
+    const vertexImageOptions = await (0, import_provider_utils4.parseProviderOptions)({
       provider: "vertex",
       providerOptions,
       schema: googleVertexImageModelOptionsSchema
@@ -368,12 +573,12 @@ var GoogleVertexImageModel = class {
       };
     }
     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 { value: response, responseHeaders } = await (0, import_provider_utils3.postJsonToApi)({
+    const { value: response, responseHeaders } = await (0, import_provider_utils4.postJsonToApi)({
       url: `${this.config.baseURL}/models/${this.modelId}:predict`,
-      headers: (0, import_provider_utils3.combineHeaders)(await (0, import_provider_utils3.resolve)(this.config.headers), headers),
+      headers: (0, import_provider_utils4.combineHeaders)(await (0, import_provider_utils4.resolve)(this.config.headers), headers),
       body,
       failedResponseHandler: googleVertexFailedResponseHandler,
-      successfulResponseHandler: (0, import_provider_utils3.createJsonResponseHandler)(
+      successfulResponseHandler: (0, import_provider_utils4.createJsonResponseHandler)(
         vertexImageResponseSchema
       ),
       abortSignal,
@@ -462,7 +667,7 @@ var GoogleVertexImageModel = class {
       baseURL: this.config.baseURL,
       headers: (_a = this.config.headers) != null ? _a : {},
       fetch: this.config.fetch,
-      generateId: (_b = this.config.generateId) != null ? _b : import_provider_utils3.generateId,
+      generateId: (_b = this.config.generateId) != null ? _b : import_provider_utils4.generateId,
       supportedUrls: () => ({
         "*": [/^https?:\/\/.*$/, /^gs:\/\/.*$/]
       })
@@ -486,7 +691,7 @@ var GoogleVertexImageModel = class {
     const images = [];
     for (const part of result.content) {
       if (part.type === "file" && part.mediaType.startsWith("image/")) {
-        images.push((0, import_provider_utils3.convertToBase64)(part.data));
+        images.push((0, import_provider_utils4.convertToBase64)(part.data));
       }
     }
     return {
@@ -513,41 +718,41 @@ var GoogleVertexImageModel = class {
 function isGeminiModel(modelId) {
   return modelId.startsWith("gemini-");
 }
-var vertexImageResponseSchema = import_v44.z.object({
-  predictions: import_v44.z.array(
-    import_v44.z.object({
-      bytesBase64Encoded: import_v44.z.string(),
-      mimeType: import_v44.z.string(),
-      prompt: import_v44.z.string().nullish()
+var vertexImageResponseSchema = import_v46.z.object({
+  predictions: import_v46.z.array(
+    import_v46.z.object({
+      bytesBase64Encoded: import_v46.z.string(),
+      mimeType: import_v46.z.string(),
+      prompt: import_v46.z.string().nullish()
     })
   ).nullish()
 });
-var googleVertexImageModelOptionsSchema = import_v44.z.object({
-  negativePrompt: import_v44.z.string().nullish(),
-  personGeneration: import_v44.z.enum(["dont_allow", "allow_adult", "allow_all"]).nullish(),
-  safetySetting: import_v44.z.enum([
+var googleVertexImageModelOptionsSchema = import_v46.z.object({
+  negativePrompt: import_v46.z.string().nullish(),
+  personGeneration: import_v46.z.enum(["dont_allow", "allow_adult", "allow_all"]).nullish(),
+  safetySetting: import_v46.z.enum([
     "block_low_and_above",
     "block_medium_and_above",
     "block_only_high",
     "block_none"
   ]).nullish(),
-  addWatermark: import_v44.z.boolean().nullish(),
-  storageUri: import_v44.z.string().nullish(),
-  sampleImageSize: import_v44.z.enum(["1K", "2K"]).nullish(),
+  addWatermark: import_v46.z.boolean().nullish(),
+  storageUri: import_v46.z.string().nullish(),
+  sampleImageSize: import_v46.z.enum(["1K", "2K"]).nullish(),
   /**
    * Configuration for image editing operations
    */
-  edit: import_v44.z.object({
+  edit: import_v46.z.object({
     /**
      * An integer that represents the number of sampling steps.
      * A higher value offers better image quality, a lower value offers better latency.
      * Try 35 steps to start. If the quality doesn't meet your requirements,
      * increase the value towards an upper limit of 75.
      */
-    baseSteps: import_v44.z.number().nullish(),
+    baseSteps: import_v46.z.number().nullish(),
     // Edit mode options
     // https://cloud.google.com/vertex-ai/generative-ai/docs/image/edit-insert-objects
-    mode: import_v44.z.enum([
+    mode: import_v46.z.enum([
       "EDIT_MODE_INPAINT_INSERTION",
       "EDIT_MODE_INPAINT_REMOVAL",
       "EDIT_MODE_OUTPAINT",
@@ -563,7 +768,7 @@ var googleVertexImageModelOptionsSchema = import_v44.z.object({
      * - `MASK_MODE_CLOTHING_AREA` - Masks from segmenting the clothing area with open-vocab segmentation.
      * - `MASK_MODE_PARSED_PERSON` - Masks from segmenting the person body and clothing using the person-parsing model.
      */
-    maskMode: import_v44.z.enum([
+    maskMode: import_v46.z.enum([
       "MASK_MODE_DEFAULT",
       "MASK_MODE_USER_PROVIDED",
       "MASK_MODE_DETECTION_BOX",
@@ -575,7 +780,7 @@ var googleVertexImageModelOptionsSchema = import_v44.z.object({
      * percentage of the image width to grow the mask by. Using dilation helps
      * compensate for imprecise masks. We recommend a value of 0.01.
      */
-    maskDilation: import_v44.z.number().nullish()
+    maskDilation: import_v46.z.number().nullish()
   }).nullish()
 });
 function getBase64Data(file) {
@@ -587,7 +792,7 @@ function getBase64Data(file) {
   if (typeof file.data === "string") {
     return file.data;
   }
-  return (0, import_provider_utils3.convertUint8ArrayToBase64)(file.data);
+  return (0, import_provider_utils4.convertUint8ArrayToBase64)(file.data);
 }
 
 // src/google-vertex-tools.ts
@@ -603,22 +808,22 @@ var googleVertexTools = {
 };
 
 // src/google-vertex-transcription-model.ts
-var import_provider_utils4 = require("@ai-sdk/provider-utils");
-var import_v46 = require("zod/v4");
+var import_provider_utils5 = require("@ai-sdk/provider-utils");
+var import_v48 = require("zod/v4");
 
 // src/google-vertex-transcription-model-options.ts
-var import_v45 = require("zod/v4");
-var googleVertexTranscriptionProviderOptionsSchema = import_v45.z.object({
+var import_v47 = require("zod/v4");
+var googleVertexTranscriptionProviderOptionsSchema = import_v47.z.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: import_v45.z.array(import_v45.z.string()).optional(),
+  languageCodes: import_v47.z.array(import_v47.z.string()).optional(),
   /**
    * Whether to add punctuation to the transcript. Defaults to `true`.
    */
-  enableAutomaticPunctuation: import_v45.z.boolean().optional(),
+  enableAutomaticPunctuation: import_v47.z.boolean().optional(),
   /**
    * Whether to include word-level timestamps. Defaults to `true` so the
    * transcription result can include segments.
@@ -626,7 +831,7 @@ var googleVertexTranscriptionProviderOptionsSchema = import_v45.z.object({
    * Enabling word-level timestamps can reduce transcription quality and speed
    * for Chirp models.
    */
-  enableWordTimeOffsets: import_v45.z.boolean().optional(),
+  enableWordTimeOffsets: import_v47.z.boolean().optional(),
   /**
    * The Cloud Speech-to-Text region for the request (e.g. `'us'`, `'eu'`,
    * `'us-central1'`). Defaults to the provider `location`.
@@ -635,7 +840,7 @@ var googleVertexTranscriptionProviderOptionsSchema = import_v45.z.object({
    * available in specific Speech-to-Text regions and is not available in the
    * `global` location.
    */
-  region: import_v45.z.string().optional()
+  region: import_v47.z.string().optional()
 });
 
 // src/google-vertex-transcription-model.ts
@@ -672,7 +877,7 @@ var GoogleVertexTranscriptionModel = class {
     const warnings = [];
     let googleOptions;
     for (const provider of ["googleVertex", "vertex", "google"]) {
-      googleOptions = await (0, import_provider_utils4.parseProviderOptions)({
+      googleOptions = await (0, import_provider_utils5.parseProviderOptions)({
         provider,
         providerOptions: options.providerOptions,
         schema: googleVertexTranscriptionProviderOptionsSchema
@@ -683,7 +888,7 @@ var GoogleVertexTranscriptionModel = class {
     }
     const region = (_d = googleOptions == null ? void 0 : googleOptions.region) != null ? _d : this.config.location;
     const languageCodes = (_e = googleOptions == null ? void 0 : googleOptions.languageCodes) != null ? _e : ["auto"];
-    const content = typeof options.audio === "string" ? options.audio : (0, import_provider_utils4.convertUint8ArrayToBase64)(options.audio);
+    const content = typeof options.audio === "string" ? options.audio : (0, import_provider_utils5.convertUint8ArrayToBase64)(options.audio);
     const requestBody = {
       config: {
         model: this.modelId,
@@ -704,15 +909,15 @@ var GoogleVertexTranscriptionModel = class {
       value: response,
       responseHeaders,
       rawValue: rawResponse
-    } = await (0, import_provider_utils4.postJsonToApi)({
+    } = await (0, import_provider_utils5.postJsonToApi)({
       url,
-      headers: (0, import_provider_utils4.combineHeaders)(
-        this.config.headers ? await (0, import_provider_utils4.resolve)(this.config.headers) : void 0,
+      headers: (0, import_provider_utils5.combineHeaders)(
+        this.config.headers ? await (0, import_provider_utils5.resolve)(this.config.headers) : void 0,
         options.headers
       ),
       body: requestBody,
       failedResponseHandler: googleVertexFailedResponseHandler,
-      successfulResponseHandler: (0, import_provider_utils4.createJsonResponseHandler)(
+      successfulResponseHandler: (0, import_provider_utils5.createJsonResponseHandler)(
         googleVertexTranscriptionResponseSchema
       ),
       abortSignal: options.abortSignal,
@@ -751,33 +956,33 @@ var GoogleVertexTranscriptionModel = class {
     };
   }
 };
-var googleVertexTranscriptionResponseSchema = import_v46.z.object({
-  results: import_v46.z.array(
-    import_v46.z.object({
-      alternatives: import_v46.z.array(
-        import_v46.z.object({
-          transcript: import_v46.z.string().nullish(),
-          words: import_v46.z.array(
-            import_v46.z.object({
-              word: import_v46.z.string().nullish(),
-              startOffset: import_v46.z.string().nullish(),
-              endOffset: import_v46.z.string().nullish()
+var googleVertexTranscriptionResponseSchema = import_v48.z.object({
+  results: import_v48.z.array(
+    import_v48.z.object({
+      alternatives: import_v48.z.array(
+        import_v48.z.object({
+          transcript: import_v48.z.string().nullish(),
+          words: import_v48.z.array(
+            import_v48.z.object({
+              word: import_v48.z.string().nullish(),
+              startOffset: import_v48.z.string().nullish(),
+              endOffset: import_v48.z.string().nullish()
             })
           ).nullish()
         })
       ).nullish(),
-      languageCode: import_v46.z.string().nullish()
+      languageCode: import_v48.z.string().nullish()
     })
   ).nullish(),
-  metadata: import_v46.z.object({
-    totalBilledDuration: import_v46.z.string().nullish()
+  metadata: import_v48.z.object({
+    totalBilledDuration: import_v48.z.string().nullish()
   }).nullish()
 });
 
 // src/google-vertex-video-model.ts
-var import_provider2 = require("@ai-sdk/provider");
-var import_provider_utils5 = require("@ai-sdk/provider-utils");
-var import_v47 = require("zod/v4");
+var import_provider3 = require("@ai-sdk/provider");
+var import_provider_utils6 = require("@ai-sdk/provider-utils");
+var import_v49 = require("zod/v4");
 function getFirstFrameImage(options) {
   var _a, _b;
   return (_b = (_a = options.frameImages) == null ? void 0 : _a.find((frame) => frame.frameType === "first_frame")) == null ? void 0 : _b.image;
@@ -811,7 +1016,7 @@ function convertFileToVertexImage(file, warnings) {
     });
     return void 0;
   }
-  const base64Data = typeof file.data === "string" ? file.data : (0, import_provider_utils5.convertUint8ArrayToBase64)(file.data);
+  const base64Data = typeof file.data === "string" ? file.data : (0, import_provider_utils6.convertUint8ArrayToBase64)(file.data);
   return {
     bytesBase64Encoded: base64Data,
     mimeType: file.mediaType || "image/png"
@@ -837,7 +1042,7 @@ var GoogleVertexVideoModel = class {
     var _a, _b, _c, _d, _e, _f, _g;
     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 vertexOptions = await (0, import_provider_utils5.parseProviderOptions)({
+    const vertexOptions = await (0, import_provider_utils6.parseProviderOptions)({
       provider: "vertex",
       providerOptions: options.providerOptions,
       schema: googleVertexVideoModelOptionsSchema
@@ -919,17 +1124,17 @@ var GoogleVertexVideoModel = class {
         }
       }
     }
-    const { value: operation } = await (0, import_provider_utils5.postJsonToApi)({
+    const { value: operation } = await (0, import_provider_utils6.postJsonToApi)({
       url: `${this.config.baseURL}/models/${this.modelId}:predictLongRunning`,
-      headers: (0, import_provider_utils5.combineHeaders)(
-        await (0, import_provider_utils5.resolve)(this.config.headers),
+      headers: (0, import_provider_utils6.combineHeaders)(
+        await (0, import_provider_utils6.resolve)(this.config.headers),
         options.headers
       ),
       body: {
         instances,
         parameters
       },
-      successfulResponseHandler: (0, import_provider_utils5.createJsonResponseHandler)(
+      successfulResponseHandler: (0, import_provider_utils6.createJsonResponseHandler)(
         vertexOperationSchema
       ),
       failedResponseHandler: googleVertexFailedResponseHandler,
@@ -938,7 +1143,7 @@ var GoogleVertexVideoModel = class {
     });
     const operationName = operation.name;
     if (!operationName) {
-      throw new import_provider2.AISDKError({
+      throw new import_provider3.AISDKError({
         name: "VERTEX_VIDEO_GENERATION_ERROR",
         message: "No operation name returned from API"
       });
@@ -950,28 +1155,28 @@ var GoogleVertexVideoModel = class {
     let responseHeaders;
     while (!finalOperation.done) {
       if (Date.now() - startTime > pollTimeoutMs) {
-        throw new import_provider2.AISDKError({
+        throw new import_provider3.AISDKError({
           name: "VERTEX_VIDEO_GENERATION_TIMEOUT",
           message: `Video generation timed out after ${pollTimeoutMs}ms`
         });
       }
-      await (0, import_provider_utils5.delay)(pollIntervalMs);
+      await (0, import_provider_utils6.delay)(pollIntervalMs);
       if ((_g = options.abortSignal) == null ? void 0 : _g.aborted) {
-        throw new import_provider2.AISDKError({
+        throw new import_provider3.AISDKError({
           name: "VERTEX_VIDEO_GENERATION_ABORTED",
           message: "Video generation request was aborted"
         });
       }
-      const { value: statusOperation, responseHeaders: pollHeaders } = await (0, import_provider_utils5.postJsonToApi)({
+      const { value: statusOperation, responseHeaders: pollHeaders } = await (0, import_provider_utils6.postJsonToApi)({
         url: `${this.config.baseURL}/models/${this.modelId}:fetchPredictOperation`,
-        headers: (0, import_provider_utils5.combineHeaders)(
-          await (0, import_provider_utils5.resolve)(this.config.headers),
+        headers: (0, import_provider_utils6.combineHeaders)(
+          await (0, import_provider_utils6.resolve)(this.config.headers),
           options.headers
         ),
         body: {
           operationName
         },
-        successfulResponseHandler: (0, import_provider_utils5.createJsonResponseHandler)(
+        successfulResponseHandler: (0, import_provider_utils6.createJsonResponseHandler)(
           vertexOperationSchema
         ),
         failedResponseHandler: googleVertexFailedResponseHandler,
@@ -982,14 +1187,14 @@ var GoogleVertexVideoModel = class {
       responseHeaders = pollHeaders;
     }
     if (finalOperation.error) {
-      throw new import_provider2.AISDKError({
+      throw new import_provider3.AISDKError({
         name: "VERTEX_VIDEO_GENERATION_FAILED",
         message: `Video generation failed: ${finalOperation.error.message}`
       });
     }
     const response = finalOperation.response;
     if (!(response == null ? void 0 : response.videos) || response.videos.length === 0) {
-      throw new import_provider2.AISDKError({
+      throw new import_provider3.AISDKError({
         name: "VERTEX_VIDEO_GENERATION_ERROR",
         message: `No videos in response. Response: ${JSON.stringify(finalOperation)}`
       });
@@ -1019,7 +1224,7 @@ var GoogleVertexVideoModel = class {
       }
     }
     if (videos.length === 0) {
-      throw new import_provider2.AISDKError({
+      throw new import_provider3.AISDKError({
         name: "VERTEX_VIDEO_GENERATION_ERROR",
         message: "No valid videos in response"
       });
@@ -1040,38 +1245,38 @@ var GoogleVertexVideoModel = class {
     };
   }
 };
-var vertexOperationSchema = import_v47.z.object({
-  name: import_v47.z.string().nullish(),
-  done: import_v47.z.boolean().nullish(),
-  error: import_v47.z.object({
-    code: import_v47.z.number().nullish(),
-    message: import_v47.z.string(),
-    status: import_v47.z.string().nullish()
+var vertexOperationSchema = import_v49.z.object({
+  name: import_v49.z.string().nullish(),
+  done: import_v49.z.boolean().nullish(),
+  error: import_v49.z.object({
+    code: import_v49.z.number().nullish(),
+    message: import_v49.z.string(),
+    status: import_v49.z.string().nullish()
   }).nullish(),
-  response: import_v47.z.object({
-    videos: import_v47.z.array(
-      import_v47.z.object({
-        bytesBase64Encoded: import_v47.z.string().nullish(),
-        gcsUri: import_v47.z.string().nullish(),
-        mimeType: import_v47.z.string().nullish()
+  response: import_v49.z.object({
+    videos: import_v49.z.array(
+      import_v49.z.object({
+        bytesBase64Encoded: import_v49.z.string().nullish(),
+        gcsUri: import_v49.z.string().nullish(),
+        mimeType: import_v49.z.string().nullish()
       })
     ).nullish(),
-    raiMediaFilteredCount: import_v47.z.number().nullish()
+    raiMediaFilteredCount: import_v49.z.number().nullish()
   }).nullish()
 });
-var googleVertexVideoModelOptionsSchema = (0, import_provider_utils5.lazySchema)(
-  () => (0, import_provider_utils5.zodSchema)(
-    import_v47.z.object({
-      pollIntervalMs: import_v47.z.number().positive().nullish(),
-      pollTimeoutMs: import_v47.z.number().positive().nullish(),
-      personGeneration: import_v47.z.enum(["dont_allow", "allow_adult", "allow_all"]).nullish(),
-      negativePrompt: import_v47.z.string().nullish(),
-      generateAudio: import_v47.z.boolean().nullish(),
-      gcsOutputDirectory: import_v47.z.string().nullish(),
-      referenceImages: import_v47.z.array(
-        import_v47.z.object({
-          bytesBase64Encoded: import_v47.z.string().nullish(),
-          gcsUri: import_v47.z.string().nullish()
+var googleVertexVideoModelOptionsSchema = (0, import_provider_utils6.lazySchema)(
+  () => (0, import_provider_utils6.zodSchema)(
+    import_v49.z.object({
+      pollIntervalMs: import_v49.z.number().positive().nullish(),
+      pollTimeoutMs: import_v49.z.number().positive().nullish(),
+      personGeneration: import_v49.z.enum(["dont_allow", "allow_adult", "allow_all"]).nullish(),
+      negativePrompt: import_v49.z.string().nullish(),
+      generateAudio: import_v49.z.boolean().nullish(),
+      gcsOutputDirectory: import_v49.z.string().nullish(),
+      referenceImages: import_v49.z.array(
+        import_v49.z.object({
+          bytesBase64Encoded: import_v49.z.string().nullish(),
+          gcsUri: import_v49.z.string().nullish()
         })
       ).nullish()
     }).passthrough()
@@ -1085,7 +1290,7 @@ function createExpressModeFetch(apiKey, customFetch) {
     const modifiedInit = {
       ...init,
       headers: {
-        ...(init == null ? void 0 : init.headers) ? (0, import_provider_utils6.normalizeHeaders)(init.headers) : {},
+        ...(init == null ? void 0 : init.headers) ? (0, import_provider_utils7.normalizeHeaders)(init.headers) : {},
         "x-goog-api-key": apiKey
       }
     };
@@ -1093,17 +1298,17 @@ function createExpressModeFetch(apiKey, customFetch) {
   };
 }
 function createVertex(options = {}) {
-  const apiKey = (0, import_provider_utils6.loadOptionalSetting)({
+  const apiKey = (0, import_provider_utils7.loadOptionalSetting)({
     settingValue: options.apiKey,
     environmentVariableName: "GOOGLE_VERTEX_API_KEY"
   });
-  const loadVertexProject = () => (0, import_provider_utils6.loadSetting)({
+  const loadVertexProject = () => (0, import_provider_utils7.loadSetting)({
     settingValue: options.project,
     settingName: "project",
     environmentVariableName: "GOOGLE_VERTEX_PROJECT",
     description: "Google Vertex project"
   });
-  const loadVertexLocation = () => (0, import_provider_utils6.loadSetting)({
+  const loadVertexLocation = () => (0, import_provider_utils7.loadSetting)({
     settingValue: options.location,
     settingName: "location",
     environmentVariableName: "GOOGLE_VERTEX_LOCATION",
@@ -1112,7 +1317,7 @@ function createVertex(options = {}) {
   const loadBaseURL = ({ endpoint = false } = {}) => {
     var _a, _b;
     if (apiKey) {
-      return (_a = (0, import_provider_utils6.withoutTrailingSlash)(options.baseURL)) != null ? _a : EXPRESS_MODE_BASE_URL;
+      return (_a = (0, import_provider_utils7.withoutTrailingSlash)(options.baseURL)) != null ? _a : EXPRESS_MODE_BASE_URL;
     }
     const region = loadVertexLocation();
     const project = loadVertexProject();
@@ -1125,13 +1330,13 @@ function createVertex(options = {}) {
         return `${region}-aiplatform.googleapis.com`;
       }
     };
-    return (_b = (0, import_provider_utils6.withoutTrailingSlash)(options.baseURL)) != null ? _b : `https://${getHost()}/v1beta1/projects/${project}/locations/${region}${endpoint ? "" : "/publishers/google"}`;
+    return (_b = (0, import_provider_utils7.withoutTrailingSlash)(options.baseURL)) != null ? _b : `https://${getHost()}/v1beta1/projects/${project}/locations/${region}${endpoint ? "" : "/publishers/google"}`;
   };
   const createConfig = (name, { endpoint = false } = {}) => {
     const getHeaders = async () => {
       var _a;
-      const originalHeaders = await (0, import_provider_utils6.resolve)((_a = options.headers) != null ? _a : {});
-      return (0, import_provider_utils6.withUserAgentSuffix)(
+      const originalHeaders = await (0, import_provider_utils7.resolve)((_a = options.headers) != null ? _a : {});
+      return (0, import_provider_utils7.withUserAgentSuffix)(
         originalHeaders,
         `ai-sdk/google-vertex/${VERSION}`
       );
@@ -1147,7 +1352,7 @@ function createVertex(options = {}) {
     var _a;
     return new import_internal3.GoogleGenerativeAILanguageModel(modelId, {
       ...createConfig("chat"),
-      generateId: (_a = options.generateId) != null ? _a : import_provider_utils6.generateId,
+      generateId: (_a = options.generateId) != null ? _a : import_provider_utils7.generateId,
       supportedUrls: () => ({
         "*": [
           // HTTP URLs:
@@ -1173,7 +1378,7 @@ function createVertex(options = {}) {
         // endpoint-style base URL without the `/publishers/google` suffix that
         // the base-model paths carry.
         ...createConfig("interactions", { endpoint: true }),
-        generateId: (_a = options.generateId) != null ? _a : import_provider_utils6.generateId
+        generateId: (_a = options.generateId) != null ? _a : import_provider_utils7.generateId
       }
     );
   };
@@ -1182,14 +1387,14 @@ function createVertex(options = {}) {
     var _a;
     return new GoogleVertexImageModel(modelId, {
       ...createConfig("image"),
-      generateId: (_a = options.generateId) != null ? _a : import_provider_utils6.generateId
+      generateId: (_a = options.generateId) != null ? _a : import_provider_utils7.generateId
     });
   };
   const createVideoModel = (modelId) => {
     var _a;
     return new GoogleVertexVideoModel(modelId, {
       ...createConfig("video"),
-      generateId: (_a = options.generateId) != null ? _a : import_provider_utils6.generateId
+      generateId: (_a = options.generateId) != null ? _a : import_provider_utils7.generateId
     });
   };
   const createTranscriptionModel = (modelId) => {
@@ -1199,6 +1404,14 @@ function createVertex(options = {}) {
       );
     }
     const config = createConfig("transcription");
+    if (modelId.startsWith("gemini")) {
+      return new GoogleVertexGeminiTranscriptionModel(modelId, {
+        provider: config.provider,
+        baseURL: loadBaseURL(),
+        headers: config.headers,
+        fetch: config.fetch
+      });
+    }
     return new GoogleVertexTranscriptionModel(modelId, {
       provider: config.provider,
       headers: config.headers,
@@ -1232,7 +1445,7 @@ function createVertex(options = {}) {
 
 // src/google-vertex-provider-node.ts
 function createVertex2(options = {}) {
-  const apiKey = (0, import_provider_utils7.loadOptionalSetting)({
+  const apiKey = (0, import_provider_utils8.loadOptionalSetting)({
     settingValue: options.apiKey,
     environmentVariableName: "GOOGLE_VERTEX_API_KEY"
   });
@@ -1248,13 +1461,14 @@ function createVertex2(options = {}) {
     ...options,
     headers: async () => ({
       Authorization: `Bearer ${await generateAuthToken()}`,
-      ...await (0, import_provider_utils7.resolve)(options.headers)
+      ...await (0, import_provider_utils8.resolve)(options.headers)
     })
   });
 }
 var vertex = createVertex2();
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
+  GoogleVertexGeminiTranscriptionModel,
   VERSION,
   createVertex,
   vertex