@ai-sdk/google 3.0.136 → 3.0.138

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-sdk/google",
-  "version": "3.0.136",
+  "version": "3.0.138",
   "license": "Apache-2.0",
   "sideEffects": false,
   "main": "./dist/index.js",
@@ -37,7 +37,7 @@
   },
   "dependencies": {
     "@ai-sdk/provider": "3.0.19",
-    "@ai-sdk/provider-utils": "4.0.61"
+    "@ai-sdk/provider-utils": "4.0.62"
   },
   "devDependencies": {
     "@types/node": "20.17.24",

package/src/convert-json-schema-to-openapi-schema.ts

@@ -109,8 +109,18 @@ function convertJSONSchemaDefinition(
       if (nonNullTypes.length === 0) {
         // Only null type
         result.type = 'null';
+      } else if (nonNullTypes.length === 1) {
+        // A single type stays a plain node so sibling keywords (items,
+        // format, ...) remain valid: Gemini rejects fields beside anyOf and
+        // items on a non-array node.
+        result.type = nonNullTypes[0];
+        if (hasNull) {
+          result.nullable = true;
+        }
       } else {
-        // One or more non-null types: always use anyOf
+        // Multiple non-null types: use anyOf. Type-specific keywords
+        // (items, format, ...) still land beside it, which Gemini rejects;
+        // splitting them per branch is not handled yet.
         result.anyOf = nonNullTypes.map(t => ({ type: t }));
         if (hasNull) {
           result.nullable = true;

package/src/google-generative-ai-embedding-options.ts

@@ -10,21 +10,22 @@ export type GoogleGenerativeAIEmbeddingModelId =
   | 'gemini-embedding-2-preview'
   | (string & {});
 
-const googleEmbeddingContentPartSchema = z.union([
-  z.object({ text: z.string() }),
-  z.object({
-    inlineData: z.object({
-      mimeType: z.string(),
-      data: z.string(),
+const googleEmbeddingContentPartSchema = () =>
+  z.union([
+    z.object({ text: z.string() }),
+    z.object({
+      inlineData: z.object({
+        mimeType: z.string(),
+        data: z.string(),
+      }),
     }),
-  }),
-  z.object({
-    fileData: z.object({
-      fileUri: z.string(),
-      mimeType: z.string(),
+    z.object({
+      fileData: z.object({
+        fileUri: z.string(),
+        mimeType: z.string(),
+      }),
     }),
-  }),
-]);
+  ]);
 
 export const googleEmbeddingModelOptions = lazySchema(() =>
   zodSchema(
@@ -70,7 +71,7 @@ export const googleEmbeddingModelOptions = lazySchema(() =>
        * the case of a single embedding, the array length must be 1.
        */
       content: z
-        .array(z.array(googleEmbeddingContentPartSchema).min(1).nullable())
+        .array(z.array(googleEmbeddingContentPartSchema()).min(1).nullable())
         .optional(),
     }),
   ),

package/src/google-generative-ai-image-model.ts

@@ -418,7 +418,7 @@ const googleImageModelOptionsSchema = lazySchema(() =>
        * dedicated escape hatch for grounding image generation the same way
        * `generateText` does.
        */
-      googleSearch: googleSearchToolArgsBaseSchema.optional(),
+      googleSearch: googleSearchToolArgsBaseSchema().optional(),
     }),
   ),
 );

package/src/google-generative-ai-language-model.ts

@@ -1573,14 +1573,15 @@ export const getGroundingMetadataSchema = () =>
       .nullish(),
   });
 
-const partialArgSchema = z.object({
-  jsonPath: z.string(),
-  stringValue: z.string().nullish(),
-  numberValue: z.number().nullish(),
-  boolValue: z.boolean().nullish(),
-  nullValue: z.unknown().nullish(),
-  willContinue: z.boolean().nullish(),
-});
+const partialArgSchema = () =>
+  z.object({
+    jsonPath: z.string(),
+    stringValue: z.string().nullish(),
+    numberValue: z.number().nullish(),
+    boolValue: z.boolean().nullish(),
+    nullValue: z.unknown().nullish(),
+    willContinue: z.boolean().nullish(),
+  });
 
 const getContentSchema = () =>
   z.object({
@@ -1593,7 +1594,7 @@ const getContentSchema = () =>
               id: z.string().nullish(),
               name: z.string().nullish(),
               args: z.unknown().nullish(),
-              partialArgs: z.array(partialArgSchema).nullish(),
+              partialArgs: z.array(partialArgSchema()).nullish(),
               willContinue: z.boolean().nullish(),
             }),
             thoughtSignature: z.string().nullish(),
@@ -1655,35 +1656,37 @@ const getSafetyRatingSchema = () =>
     blocked: z.boolean().nullish(),
   });
 
-const tokenDetailsSchema = z
-  .array(
-    z
-      .object({
-        modality: z.string(),
-        tokenCount: z.number(),
-      })
-      .catchall(z.json()),
-  )
-  .nullish();
-
-const usageSchema = z
-  .object({
-    cachedContentTokenCount: z.number().nullish(),
-    thoughtsTokenCount: z.number().nullish(),
-    promptTokenCount: z.number().nullish(),
-    candidatesTokenCount: z.number().nullish(),
-    toolUsePromptTokenCount: z.number().nullish(),
-    totalTokenCount: z.number().nullish(),
-    // https://cloud.google.com/vertex-ai/generative-ai/docs/reference/rest/v1/GenerateContentResponse#TrafficType
-    trafficType: z.string().nullish(),
-    serviceTier: z.string().nullish(),
-    // https://ai.google.dev/api/generate-content#Modality
-    promptTokensDetails: tokenDetailsSchema,
-    cacheTokensDetails: tokenDetailsSchema,
-    candidatesTokensDetails: tokenDetailsSchema,
-    toolUsePromptTokensDetails: tokenDetailsSchema,
-  })
-  .catchall(z.json());
+const tokenDetailsSchema = () =>
+  z
+    .array(
+      z
+        .object({
+          modality: z.string(),
+          tokenCount: z.number(),
+        })
+        .catchall(z.json()),
+    )
+    .nullish();
+
+const usageSchema = () =>
+  z
+    .object({
+      cachedContentTokenCount: z.number().nullish(),
+      thoughtsTokenCount: z.number().nullish(),
+      promptTokenCount: z.number().nullish(),
+      candidatesTokenCount: z.number().nullish(),
+      toolUsePromptTokenCount: z.number().nullish(),
+      totalTokenCount: z.number().nullish(),
+      // https://cloud.google.com/vertex-ai/generative-ai/docs/reference/rest/v1/GenerateContentResponse#TrafficType
+      trafficType: z.string().nullish(),
+      serviceTier: z.string().nullish(),
+      // https://ai.google.dev/api/generate-content#Modality
+      promptTokensDetails: tokenDetailsSchema(),
+      cacheTokensDetails: tokenDetailsSchema(),
+      candidatesTokensDetails: tokenDetailsSchema(),
+      toolUsePromptTokensDetails: tokenDetailsSchema(),
+    })
+    .catchall(z.json());
 
 // https://ai.google.dev/api/generate-content#UrlRetrievalMetadata
 export const getUrlContextMetadataSchema = () =>
@@ -1715,7 +1718,7 @@ const responseSchema = lazySchema(() =>
           }),
         )
         .nullish(),
-      usageMetadata: usageSchema.nullish(),
+      usageMetadata: usageSchema().nullish(),
       promptFeedback: z
         .object({
           blockReason: z.string().nullish(),
@@ -1779,7 +1782,7 @@ const chunkSchema = lazySchema(() =>
           }),
         )
         .nullish(),
-      usageMetadata: usageSchema.nullish(),
+      usageMetadata: usageSchema().nullish(),
       promptFeedback: z
         .object({
           blockReason: z.string().nullish(),

package/src/google-generative-ai-video-model.ts

@@ -408,36 +408,40 @@ export class GoogleGenerativeAIVideoModel implements Experimental_VideoModelV3 {
   }
 }
 
-const googleOperationSchema = z.object({
-  name: z.string().nullish(),
-  done: z.boolean().nullish(),
-  error: z
-    .object({
-      code: z.number().nullish(),
-      message: z.string(),
-      status: z.string().nullish(),
-    })
-    .nullish(),
-  response: z
-    .object({
-      generateVideoResponse: z
+const googleOperationSchema = lazySchema(() =>
+  zodSchema(
+    z.object({
+      name: z.string().nullish(),
+      done: z.boolean().nullish(),
+      error: z
+        .object({
+          code: z.number().nullish(),
+          message: z.string(),
+          status: z.string().nullish(),
+        })
+        .nullish(),
+      response: z
         .object({
-          generatedSamples: z
-            .array(
-              z.object({
-                video: z
-                  .object({
-                    uri: z.string().nullish(),
-                  })
-                  .nullish(),
-              }),
-            )
+          generateVideoResponse: z
+            .object({
+              generatedSamples: z
+                .array(
+                  z.object({
+                    video: z
+                      .object({
+                        uri: z.string().nullish(),
+                      })
+                      .nullish(),
+                  }),
+                )
+                .nullish(),
+            })
             .nullish(),
         })
         .nullish(),
-    })
-    .nullish(),
-});
+    }),
+  ),
+);
 
 const googleVideoModelOptionsSchema = lazySchema(() =>
   zodSchema(

package/src/tool/file-search.ts

@@ -6,40 +6,43 @@ import {
 import { z } from 'zod/v4';
 
 /** Tool to retrieve knowledge from the File Search Stores. */
-const fileSearchArgsBaseSchema = z
-  .object({
-    /** The names of the file_search_stores to retrieve from.
-     *  Example: `fileSearchStores/my-file-search-store-123`
-     */
-    fileSearchStoreNames: z
-      .array(z.string())
-      .describe(
-        'The names of the file_search_stores to retrieve from. Example: `fileSearchStores/my-file-search-store-123`',
-      ),
-    /** The number of file search retrieval chunks to retrieve. */
-    topK: z
-      .number()
-      .int()
-      .positive()
-      .describe('The number of file search retrieval chunks to retrieve.')
-      .optional(),
+const fileSearchArgsBaseSchema = () =>
+  z
+    .object({
+      /** The names of the file_search_stores to retrieve from.
+       *  Example: `fileSearchStores/my-file-search-store-123`
+       */
+      fileSearchStoreNames: z
+        .array(z.string())
+        .describe(
+          'The names of the file_search_stores to retrieve from. Example: `fileSearchStores/my-file-search-store-123`',
+        ),
+      /** The number of file search retrieval chunks to retrieve. */
+      topK: z
+        .number()
+        .int()
+        .positive()
+        .describe('The number of file search retrieval chunks to retrieve.')
+        .optional(),
 
-    /** Metadata filter to apply to the file search retrieval documents.
-     *  See https://google.aip.dev/160 for the syntax of the filter expression.
-     */
-    metadataFilter: z
-      .string()
-      .describe(
-        'Metadata filter to apply to the file search retrieval documents. See https://google.aip.dev/160 for the syntax of the filter expression.',
-      )
-      .optional(),
-  })
-  .passthrough();
+      /** Metadata filter to apply to the file search retrieval documents.
+       *  See https://google.aip.dev/160 for the syntax of the filter expression.
+       */
+      metadataFilter: z
+        .string()
+        .describe(
+          'Metadata filter to apply to the file search retrieval documents. See https://google.aip.dev/160 for the syntax of the filter expression.',
+        )
+        .optional(),
+    })
+    .passthrough();
 
-export type GoogleFileSearchToolArgs = z.infer<typeof fileSearchArgsBaseSchema>;
+export type GoogleFileSearchToolArgs = z.infer<
+  ReturnType<typeof fileSearchArgsBaseSchema>
+>;
 
 const fileSearchArgsSchema = lazySchema(() =>
-  zodSchema(fileSearchArgsBaseSchema),
+  zodSchema(fileSearchArgsBaseSchema()),
 );
 
 export const fileSearch = createProviderToolFactory<

package/src/tool/google-search.ts

@@ -9,30 +9,31 @@ import { z } from 'zod/v4';
 // https://ai.google.dev/api/generate-content#GroundingSupport
 // https://cloud.google.com/vertex-ai/generative-ai/docs/grounding/grounding-with-google-search
 
-export const googleSearchToolArgsBaseSchema = z
-  .object({
-    searchTypes: z
-      .object({
-        webSearch: z.object({}).optional(),
-        imageSearch: z.object({}).optional(),
-      })
-      .optional(),
+export const googleSearchToolArgsBaseSchema = () =>
+  z
+    .object({
+      searchTypes: z
+        .object({
+          webSearch: z.object({}).optional(),
+          imageSearch: z.object({}).optional(),
+        })
+        .optional(),
 
-    timeRangeFilter: z
-      .object({
-        startTime: z.string(),
-        endTime: z.string(),
-      })
-      .optional(),
-  })
-  .passthrough();
+      timeRangeFilter: z
+        .object({
+          startTime: z.string(),
+          endTime: z.string(),
+        })
+        .optional(),
+    })
+    .passthrough();
 
 export type GoogleSearchToolArgs = z.infer<
-  typeof googleSearchToolArgsBaseSchema
+  ReturnType<typeof googleSearchToolArgsBaseSchema>
 >;
 
 const googleSearchToolArgsSchema = lazySchema(() =>
-  zodSchema(googleSearchToolArgsBaseSchema),
+  zodSchema(googleSearchToolArgsBaseSchema()),
 );
 
 export const googleSearch = createProviderToolFactory<{}, GoogleSearchToolArgs>(

package/src/transcription/google-transcription-model-options.ts

@@ -1,3 +1,8 @@
+import {
+  lazySchema,
+  zodSchema,
+  type InferSchema,
+} from '@ai-sdk/provider-utils';
 import { z } from 'zod/v4';
 
 export type GoogleTranscriptionModelId =
@@ -11,41 +16,45 @@ export type GoogleTranscriptionModelId =
  * The live variant (`gemini-3.5-transcribe-live`) requires streaming
  * transcription, which is only available in AI SDK v7.
  */
-export const googleTranscriptionModelOptions = 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: z.array(z.string()).optional(),
+export const googleTranscriptionModelOptions = lazySchema(() =>
+  zodSchema(
+    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: z.array(z.string()).optional(),
 
-  /**
-   * Custom vocabulary phrases, which bias the speech recognition model
-   * toward recognizing specific terms.
-   */
-  customVocabulary: z.array(z.string()).optional(),
+      /**
+       * Custom vocabulary phrases, which bias the speech recognition model
+       * toward recognizing specific terms.
+       */
+      customVocabulary: z.array(z.string()).optional(),
 
-  /**
-   * Enables word-level timestamp generation.
-   */
-  wordTimestamp: z.boolean().optional(),
+      /**
+       * Enables word-level timestamp generation.
+       */
+      wordTimestamp: z.boolean().optional(),
 
-  /**
-   * Enables speaker diarization.
-   */
-  diarization: z.boolean().optional(),
+      /**
+       * Enables speaker diarization.
+       */
+      diarization: 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: z.enum(['SMART', 'VERBATIM']).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: z.enum(['SMART', 'VERBATIM']).optional(),
+    }),
+  ),
+);
 
-export type GoogleTranscriptionModelOptions = z.infer<
+export type GoogleTranscriptionModelOptions = InferSchema<
   typeof googleTranscriptionModelOptions
 >;

package/src/transcription/google-transcription-model.ts

@@ -5,6 +5,8 @@ import {
   type TranscriptionModelV3,
 } from '@ai-sdk/provider';
 import {
+  lazySchema,
+  zodSchema,
   combineHeaders,
   convertToBase64,
   createJsonResponseHandler,
@@ -211,33 +213,38 @@ function parseOffsetSeconds(
   return Number.isFinite(parsed) ? parsed : undefined;
 }
 
-const googleInteractionsWordAnnotationSchema = z.object({
-  type: z.string().nullish(),
-  text: z.string().nullish(),
-  speaker: z.string().nullish(),
-  start_offset: z.string().nullish(),
-  end_offset: z.string().nullish(),
-});
-
-const googleInteractionsTranscriptionResponseSchema = z.object({
-  status: z.string().nullish(),
-  steps: z
-    .array(
-      z.object({
-        type: z.string().nullish(),
-        content: z
-          .array(
-            z.object({
-              type: z.string().nullish(),
-              text: z.string().nullish(),
-              annotations: z
-                .array(googleInteractionsWordAnnotationSchema)
-                .nullish(),
-            }),
-          )
-          .nullish(),
-      }),
-    )
-    .nullish(),
-  usage: z.record(z.string(), z.unknown()).nullish(),
-});
+const googleInteractionsWordAnnotationSchema = () =>
+  z.object({
+    type: z.string().nullish(),
+    text: z.string().nullish(),
+    speaker: z.string().nullish(),
+    start_offset: z.string().nullish(),
+    end_offset: z.string().nullish(),
+  });
+
+const googleInteractionsTranscriptionResponseSchema = lazySchema(() =>
+  zodSchema(
+    z.object({
+      status: z.string().nullish(),
+      steps: z
+        .array(
+          z.object({
+            type: z.string().nullish(),
+            content: z
+              .array(
+                z.object({
+                  type: z.string().nullish(),
+                  text: z.string().nullish(),
+                  annotations: z
+                    .array(googleInteractionsWordAnnotationSchema())
+                    .nullish(),
+                }),
+              )
+              .nullish(),
+          }),
+        )
+        .nullish(),
+      usage: z.record(z.string(), z.unknown()).nullish(),
+    }),
+  ),
+);