@ai-sdk/google 3.0.113 → 3.0.116

package/CHANGELOG.md

@@ -1,5 +1,25 @@
 # @ai-sdk/google
 
+## 3.0.116
+
+### Patch Changes
+
+- 9a1656e: fix(google): convert enum values to the Gemini schema format
+- cfdc8df: feat (provider/google, provider/google-vertex): Gemini 3.5 Transcribe support — unary transcription (`gemini-3.5-transcribe`) via generateContent with language detection, speaker diarization, word timestamps, custom vocabulary, and `mode: 'VERBATIM' | 'SMART'` transcription formatting
+
+## 3.0.115
+
+### Patch Changes
+
+- Updated dependencies [9a521b9]
+  - @ai-sdk/provider-utils@4.0.49
+
+## 3.0.114
+
+### Patch Changes
+
+- 1df2ced: Preserve recursive tool input schemas without aborting Google model calls.
+
 ## 3.0.113
 
 ### Patch Changes

package/dist/index.d.mts

@@ -1,6 +1,7 @@
 import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
-import { InferSchema, FetchFunction } from '@ai-sdk/provider-utils';
-import { ProviderV3, LanguageModelV3, ImageModelV3, EmbeddingModelV3, Experimental_VideoModelV3 } from '@ai-sdk/provider';
+import { InferSchema, Resolvable, FetchFunction } from '@ai-sdk/provider-utils';
+import { TranscriptionModelV3, ProviderV3, LanguageModelV3, ImageModelV3, EmbeddingModelV3, Experimental_VideoModelV3 } from '@ai-sdk/provider';
+import { z } from 'zod/v4';
 
 declare const googleErrorDataSchema: _ai_sdk_provider_utils.LazySchema<{
     error: {
@@ -406,6 +407,50 @@ type GoogleInteractionsProviderMetadata = {
  */
 type GoogleInteractionsAgentName = 'deep-research-pro-preview-12-2025' | 'deep-research-preview-04-2026' | 'deep-research-max-preview-04-2026' | 'antigravity-preview-05-2026';
 
+type GoogleTranscriptionModelId = 'gemini-3.5-transcribe' | 'gemini-3.5-transcribe-live' | (string & {});
+/**
+ * Speech recognition options for Gemini transcription models
+ * (`gemini-3.5-transcribe`). Maps onto Google's `AudioTranscriptionConfig`.
+ * The live variant (`gemini-3.5-transcribe-live`) requires streaming
+ * transcription, which is only available in AI SDK v7.
+ */
+declare const googleTranscriptionModelOptions: z.ZodObject<{
+    languageCodes: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    customVocabulary: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    wordTimestamp: z.ZodOptional<z.ZodBoolean>;
+    diarization: z.ZodOptional<z.ZodBoolean>;
+    mode: z.ZodOptional<z.ZodEnum<{
+        SMART: "SMART";
+        VERBATIM: "VERBATIM";
+    }>>;
+}, z.core.$strip>;
+type GoogleTranscriptionModelOptions = z.infer<typeof googleTranscriptionModelOptions>;
+
+interface GoogleTranscriptionModelConfig {
+    provider: string;
+    baseURL: string;
+    headers?: Resolvable<Record<string, string | undefined>>;
+    fetch?: FetchFunction;
+    _internal?: {
+        currentDate?: () => Date;
+    };
+}
+/**
+ * Gemini transcription (speech-to-text) via the Interactions API
+ * (e.g. `gemini-3.5-transcribe`).
+ *
+ * @see https://ai.google.dev/gemini-api/docs/transcribe
+ */
+declare class GoogleTranscriptionModel implements TranscriptionModelV3 {
+    readonly modelId: GoogleTranscriptionModelId;
+    private readonly config;
+    readonly specificationVersion = "v3";
+    get provider(): string;
+    constructor(modelId: GoogleTranscriptionModelId, config: GoogleTranscriptionModelConfig);
+    private parseOptions;
+    doGenerate(options: Parameters<TranscriptionModelV3['doGenerate']>[0]): Promise<Awaited<ReturnType<TranscriptionModelV3['doGenerate']>>>;
+}
+
 declare const googleTools: {
     /**
      * Creates a Google search tool that gives Google direct access to real-time web content.
@@ -517,6 +562,15 @@ interface GoogleGenerativeAIProvider extends ProviderV3 {
      * @deprecated Use `embeddingModel` instead.
      */
     textEmbeddingModel(modelId: GoogleGenerativeAIEmbeddingModelId): EmbeddingModelV3;
+    /**
+     * Creates a model for transcription (speech-to-text), e.g.
+     * `gemini-3.5-transcribe`.
+     */
+    transcription(modelId: GoogleTranscriptionModelId): TranscriptionModelV3;
+    /**
+     * Creates a model for transcription (speech-to-text).
+     */
+    transcriptionModel(modelId: GoogleTranscriptionModelId): TranscriptionModelV3;
     /**
      * Creates a model for video generation.
      */
@@ -581,4 +635,4 @@ declare const google: GoogleGenerativeAIProvider;
 
 declare const VERSION: string;
 
-export { type GoogleEmbeddingModelOptions, type GoogleErrorData, type GoogleEmbeddingModelOptions as GoogleGenerativeAIEmbeddingProviderOptions, type GoogleImageModelOptions as GoogleGenerativeAIImageProviderOptions, type GoogleGenerativeAIProvider, type GoogleGenerativeAIProviderMetadata, type GoogleLanguageModelOptions as GoogleGenerativeAIProviderOptions, type GoogleGenerativeAIProviderSettings, type GoogleGenerativeAIVideoModelId, type GoogleVideoModelOptions as GoogleGenerativeAIVideoProviderOptions, type GoogleImageModelOptions, type GoogleInteractionsAgentName, type GoogleInteractionsModelId, type GoogleInteractionsProviderMetadata, type GoogleLanguageModelInteractionsOptions, type GoogleLanguageModelOptions, type GoogleVideoModelOptions, VERSION, createGoogleGenerativeAI, google };
+export { type GoogleEmbeddingModelOptions, type GoogleErrorData, type GoogleEmbeddingModelOptions as GoogleGenerativeAIEmbeddingProviderOptions, type GoogleImageModelOptions as GoogleGenerativeAIImageProviderOptions, type GoogleGenerativeAIProvider, type GoogleGenerativeAIProviderMetadata, type GoogleLanguageModelOptions as GoogleGenerativeAIProviderOptions, type GoogleGenerativeAIProviderSettings, type GoogleGenerativeAIVideoModelId, type GoogleVideoModelOptions as GoogleGenerativeAIVideoProviderOptions, type GoogleImageModelOptions, type GoogleInteractionsAgentName, type GoogleInteractionsModelId, type GoogleInteractionsProviderMetadata, type GoogleLanguageModelInteractionsOptions, type GoogleLanguageModelOptions, GoogleTranscriptionModel, type GoogleTranscriptionModelId, type GoogleTranscriptionModelOptions, type GoogleVideoModelOptions, VERSION, createGoogleGenerativeAI, google };

package/dist/index.d.ts

@@ -1,6 +1,7 @@
 import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
-import { InferSchema, FetchFunction } from '@ai-sdk/provider-utils';
-import { ProviderV3, LanguageModelV3, ImageModelV3, EmbeddingModelV3, Experimental_VideoModelV3 } from '@ai-sdk/provider';
+import { InferSchema, Resolvable, FetchFunction } from '@ai-sdk/provider-utils';
+import { TranscriptionModelV3, ProviderV3, LanguageModelV3, ImageModelV3, EmbeddingModelV3, Experimental_VideoModelV3 } from '@ai-sdk/provider';
+import { z } from 'zod/v4';
 
 declare const googleErrorDataSchema: _ai_sdk_provider_utils.LazySchema<{
     error: {
@@ -406,6 +407,50 @@ type GoogleInteractionsProviderMetadata = {
  */
 type GoogleInteractionsAgentName = 'deep-research-pro-preview-12-2025' | 'deep-research-preview-04-2026' | 'deep-research-max-preview-04-2026' | 'antigravity-preview-05-2026';
 
+type GoogleTranscriptionModelId = 'gemini-3.5-transcribe' | 'gemini-3.5-transcribe-live' | (string & {});
+/**
+ * Speech recognition options for Gemini transcription models
+ * (`gemini-3.5-transcribe`). Maps onto Google's `AudioTranscriptionConfig`.
+ * The live variant (`gemini-3.5-transcribe-live`) requires streaming
+ * transcription, which is only available in AI SDK v7.
+ */
+declare const googleTranscriptionModelOptions: z.ZodObject<{
+    languageCodes: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    customVocabulary: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    wordTimestamp: z.ZodOptional<z.ZodBoolean>;
+    diarization: z.ZodOptional<z.ZodBoolean>;
+    mode: z.ZodOptional<z.ZodEnum<{
+        SMART: "SMART";
+        VERBATIM: "VERBATIM";
+    }>>;
+}, z.core.$strip>;
+type GoogleTranscriptionModelOptions = z.infer<typeof googleTranscriptionModelOptions>;
+
+interface GoogleTranscriptionModelConfig {
+    provider: string;
+    baseURL: string;
+    headers?: Resolvable<Record<string, string | undefined>>;
+    fetch?: FetchFunction;
+    _internal?: {
+        currentDate?: () => Date;
+    };
+}
+/**
+ * Gemini transcription (speech-to-text) via the Interactions API
+ * (e.g. `gemini-3.5-transcribe`).
+ *
+ * @see https://ai.google.dev/gemini-api/docs/transcribe
+ */
+declare class GoogleTranscriptionModel implements TranscriptionModelV3 {
+    readonly modelId: GoogleTranscriptionModelId;
+    private readonly config;
+    readonly specificationVersion = "v3";
+    get provider(): string;
+    constructor(modelId: GoogleTranscriptionModelId, config: GoogleTranscriptionModelConfig);
+    private parseOptions;
+    doGenerate(options: Parameters<TranscriptionModelV3['doGenerate']>[0]): Promise<Awaited<ReturnType<TranscriptionModelV3['doGenerate']>>>;
+}
+
 declare const googleTools: {
     /**
      * Creates a Google search tool that gives Google direct access to real-time web content.
@@ -517,6 +562,15 @@ interface GoogleGenerativeAIProvider extends ProviderV3 {
      * @deprecated Use `embeddingModel` instead.
      */
     textEmbeddingModel(modelId: GoogleGenerativeAIEmbeddingModelId): EmbeddingModelV3;
+    /**
+     * Creates a model for transcription (speech-to-text), e.g.
+     * `gemini-3.5-transcribe`.
+     */
+    transcription(modelId: GoogleTranscriptionModelId): TranscriptionModelV3;
+    /**
+     * Creates a model for transcription (speech-to-text).
+     */
+    transcriptionModel(modelId: GoogleTranscriptionModelId): TranscriptionModelV3;
     /**
      * Creates a model for video generation.
      */
@@ -581,4 +635,4 @@ declare const google: GoogleGenerativeAIProvider;
 
 declare const VERSION: string;
 
-export { type GoogleEmbeddingModelOptions, type GoogleErrorData, type GoogleEmbeddingModelOptions as GoogleGenerativeAIEmbeddingProviderOptions, type GoogleImageModelOptions as GoogleGenerativeAIImageProviderOptions, type GoogleGenerativeAIProvider, type GoogleGenerativeAIProviderMetadata, type GoogleLanguageModelOptions as GoogleGenerativeAIProviderOptions, type GoogleGenerativeAIProviderSettings, type GoogleGenerativeAIVideoModelId, type GoogleVideoModelOptions as GoogleGenerativeAIVideoProviderOptions, type GoogleImageModelOptions, type GoogleInteractionsAgentName, type GoogleInteractionsModelId, type GoogleInteractionsProviderMetadata, type GoogleLanguageModelInteractionsOptions, type GoogleLanguageModelOptions, type GoogleVideoModelOptions, VERSION, createGoogleGenerativeAI, google };
+export { type GoogleEmbeddingModelOptions, type GoogleErrorData, type GoogleEmbeddingModelOptions as GoogleGenerativeAIEmbeddingProviderOptions, type GoogleImageModelOptions as GoogleGenerativeAIImageProviderOptions, type GoogleGenerativeAIProvider, type GoogleGenerativeAIProviderMetadata, type GoogleLanguageModelOptions as GoogleGenerativeAIProviderOptions, type GoogleGenerativeAIProviderSettings, type GoogleGenerativeAIVideoModelId, type GoogleVideoModelOptions as GoogleGenerativeAIVideoProviderOptions, type GoogleImageModelOptions, type GoogleInteractionsAgentName, type GoogleInteractionsModelId, type GoogleInteractionsProviderMetadata, type GoogleLanguageModelInteractionsOptions, type GoogleLanguageModelOptions, GoogleTranscriptionModel, type GoogleTranscriptionModelId, type GoogleTranscriptionModelOptions, type GoogleVideoModelOptions, VERSION, createGoogleGenerativeAI, google };