@ai-sdk/google 3.0.86 → 3.0.88

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-sdk/google",
-  "version": "3.0.86",
+  "version": "3.0.88",
   "license": "Apache-2.0",
   "sideEffects": false,
   "main": "./dist/index.js",
@@ -36,16 +36,16 @@
     }
   },
   "dependencies": {
-    "@ai-sdk/provider": "3.0.12",
-    "@ai-sdk/provider-utils": "4.0.33"
+    "@ai-sdk/provider": "3.0.13",
+    "@ai-sdk/provider-utils": "4.0.35"
   },
   "devDependencies": {
     "@types/node": "20.17.24",
     "tsup": "^8",
     "typescript": "5.8.3",
     "zod": "3.25.76",
-    "@ai-sdk/test-server": "1.0.6",
-    "@vercel/ai-tsconfig": "0.0.0"
+    "@vercel/ai-tsconfig": "0.0.0",
+    "@ai-sdk/test-server": "1.0.6"
   },
   "peerDependencies": {
     "zod": "^3.25.76 || ^4.1.8"

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

@@ -1,6 +1,7 @@
 import {
   AISDKError,
   type Experimental_VideoModelV3,
+  type Experimental_VideoModelV3File,
   type SharedV3Warning,
 } from '@ai-sdk/provider';
 import {
@@ -51,6 +52,103 @@ interface GoogleGenerativeAIVideoModelConfig {
   };
 }
 
+function getFirstFrameImage(
+  options: Parameters<Experimental_VideoModelV3['doGenerate']>[0],
+): Experimental_VideoModelV3File | undefined {
+  return options.frameImages?.find(frame => frame.frameType === 'first_frame')
+    ?.image;
+}
+
+function resolveStartImage(
+  options: Parameters<Experimental_VideoModelV3['doGenerate']>[0],
+): Experimental_VideoModelV3File | undefined {
+  return getFirstFrameImage(options) ?? options.image;
+}
+
+function getLastFrameImage(
+  options: Parameters<Experimental_VideoModelV3['doGenerate']>[0],
+): Experimental_VideoModelV3File | undefined {
+  return options.frameImages?.find(frame => frame.frameType === 'last_frame')
+    ?.image;
+}
+
+function getInputReferences(
+  options: Parameters<Experimental_VideoModelV3['doGenerate']>[0],
+): Array<Experimental_VideoModelV3File> | undefined {
+  if (options.frameImages != null && options.frameImages.length > 0) {
+    return undefined;
+  }
+
+  return options.inputReferences != null && options.inputReferences.length > 0
+    ? options.inputReferences
+    : undefined;
+}
+
+function convertFileToGoogleImage(
+  file: Experimental_VideoModelV3File,
+  warnings: SharedV3Warning[],
+): Record<string, unknown> | undefined {
+  if (file.type === 'url') {
+    if (file.url.startsWith('gs://')) {
+      return {
+        gcsUri: file.url,
+        mimeType: 'image/png',
+      };
+    }
+
+    warnings.push({
+      type: 'unsupported',
+      feature: 'URL-based image input',
+      details:
+        'Google Generative AI video models require base64-encoded images or GCS URIs. URL will be ignored.',
+    });
+    return undefined;
+  }
+
+  const base64Data =
+    typeof file.data === 'string'
+      ? file.data
+      : convertUint8ArrayToBase64(file.data);
+
+  // The Gemini Developer API (generativelanguage.googleapis.com) requires
+  // inline image bytes wrapped as `inlineData`.
+  return {
+    inlineData: {
+      mimeType: file.mediaType || 'image/png',
+      data: base64Data,
+    },
+  };
+}
+
+function convertProviderReferenceImage(
+  refImg: NonNullable<GoogleVideoModelOptions['referenceImages']>[number],
+): Record<string, unknown> {
+  if (refImg.bytesBase64Encoded) {
+    return {
+      inlineData: {
+        mimeType: 'image/png',
+        data: refImg.bytesBase64Encoded,
+      },
+    };
+  }
+
+  if (refImg.gcsUri) {
+    return {
+      gcsUri: refImg.gcsUri,
+    };
+  }
+
+  return refImg;
+}
+
+function convertInputReferenceImage(
+  file: Experimental_VideoModelV3File,
+  warnings: SharedV3Warning[],
+): Record<string, unknown> | undefined {
+  const image = convertFileToGoogleImage(file, warnings);
+  return image != null ? { image, referenceType: 'asset' } : undefined;
+}
+
 export class GoogleGenerativeAIVideoModel implements Experimental_VideoModelV3 {
   readonly specificationVersion = 'v3';
 
@@ -87,46 +185,32 @@ export class GoogleGenerativeAIVideoModel implements Experimental_VideoModelV3 {
       instance.prompt = options.prompt;
     }
 
-    // Handle image-to-video: convert image to base64
-    if (options.image != null) {
-      if (options.image.type === 'url') {
-        warnings.push({
-          type: 'unsupported',
-          feature: 'URL-based image input',
-          details:
-            'Google Generative AI video models require base64-encoded images. URL will be ignored.',
-        });
-      } else {
-        const base64Data =
-          typeof options.image.data === 'string'
-            ? options.image.data
-            : convertUint8ArrayToBase64(options.image.data);
-
-        instance.image = {
-          inlineData: {
-            mimeType: options.image.mediaType || 'image/png',
-            data: base64Data,
-          },
-        };
+    const startImage = resolveStartImage(options);
+    if (startImage != null) {
+      const image = convertFileToGoogleImage(startImage, warnings);
+      if (image != null) {
+        instance.image = image;
       }
     }
 
-    if (googleOptions?.referenceImages != null) {
-      instance.referenceImages = googleOptions.referenceImages.map(refImg => {
-        if (refImg.bytesBase64Encoded) {
-          return {
-            inlineData: {
-              mimeType: 'image/png',
-              data: refImg.bytesBase64Encoded,
-            },
-          };
-        } else if (refImg.gcsUri) {
-          return {
-            gcsUri: refImg.gcsUri,
-          };
-        }
-        return refImg;
+    const lastFrameImage = getLastFrameImage(options);
+    if (lastFrameImage != null) {
+      const lastFrame = convertFileToGoogleImage(lastFrameImage, warnings);
+      if (lastFrame != null) {
+        instance.lastFrame = lastFrame;
+      }
+    }
+
+    const inputReferences = getInputReferences(options);
+    if (inputReferences != null) {
+      instance.referenceImages = inputReferences.flatMap(reference => {
+        const converted = convertInputReferenceImage(reference, warnings);
+        return converted != null ? [converted] : [];
       });
+    } else if (googleOptions?.referenceImages != null) {
+      instance.referenceImages = googleOptions.referenceImages.map(refImg =>
+        convertProviderReferenceImage(refImg),
+      );
     }
 
     const parameters: Record<string, unknown> = {