@ai-sdk/google 4.0.0-canary.63 → 4.0.0-canary.65

package/docs/15-google.mdx

@@ -1154,21 +1154,28 @@ The following optional provider options are available:
   Whether the model returns synthesized thought summaries on reasoning
   parts. Defaults to the API default.
 
-- **imageConfig** _\{ aspectRatio?: string; imageSize?: '1K' | '2K' | '4K' | '512' \}_
+- **responseFormat** _Array\<\{ type: 'text' | 'image' | 'audio'; mimeType?: string; schema?: unknown; aspectRatio?: string; imageSize?: '1K' \| '2K' \| '4K' \| '512' \}\>_
 
-  Image generation configuration when `responseModalities` includes
-  `'image'`. `aspectRatio` accepts `1:1`, `2:3`, `3:2`, `3:4`, `4:3`,
-  `4:5`, `5:4`, `9:16`, `16:9`, `21:9`, `1:8`, `8:1`, `1:4`, `4:1`.
+  Output-format entries that map directly to the API's `response_format`
+  array. Use this for fine-grained control over image, audio, or non-JSON
+  text outputs (e.g. `aspectRatio` and `imageSize` for image generation).
+  The AI SDK call-level `responseFormat: { type: 'json', schema }` still
+  drives JSON-mode automatically and prepends a matching text entry;
+  entries listed here are appended.
 
-- **mediaResolution** _'low' | 'medium' | 'high' | 'ultra_high'_
+  `aspectRatio` accepts `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`,
+  `9:16`, `16:9`, `21:9`, `1:8`, `8:1`, `1:4`, `4:1`.
 
-  Media resolution applied to image inputs / outputs.
+- **imageConfig** _\{ aspectRatio?: string; imageSize?: '1K' | '2K' | '4K' | '512' \}_ (deprecated)
 
-- **responseModalities** _Array\<'text' | 'image' | 'audio' | 'video' | 'document'\>_
+  Use **responseFormat** with a `{ type: 'image', ... }` entry instead.
+  Retained for backwards compatibility; the SDK translates `imageConfig`
+  into a matching `response_format` image entry and emits a warning when
+  set. Ignored when `responseFormat` already supplies an image entry.
 
-  The modalities the model may emit. Defaults to text-only. Pass
-  `['image']` (or `['text', 'image']`) to enable native image output. See
-  [Image output](#image-output-via-interactions).
+- **mediaResolution** _'low' | 'medium' | 'high' | 'ultra_high'_
+
+  Media resolution applied to image inputs / outputs.
 
 - **serviceTier** _'flex' | 'standard' | 'priority'_
 
@@ -1321,9 +1328,10 @@ const { text, toolCalls } = await generateText({
 
 ### Image output via Interactions
 
-Set `responseModalities: ['image']` on a Gemini image-capable model to get
-images as `LanguageModelV4FilePart` files in the response. No tool wrapping
-is required.
+Add a `{ type: 'image' }` entry to `responseFormat` on a Gemini
+image-capable model to get images as `LanguageModelV4FilePart` files in
+the response. No tool wrapping is required, and the entry doubles as the
+place to set `aspectRatio`, `imageSize`, and `mimeType`.
 
 ```ts
 import { google } from '@ai-sdk/google';
@@ -1334,7 +1342,7 @@ const result = await generateText({
   prompt: 'Generate an image of a comic cat in a spaceship.',
   providerOptions: {
     google: {
-      responseModalities: ['image'],
+      responseFormat: [{ type: 'image' }],
     },
   },
 });
@@ -1346,6 +1354,54 @@ for (const file of result.files) {
 }
 ```
 
+To control aspect ratio, image size, or output mime type, add those
+fields to the same image entry:
+
+```ts
+const result = await generateText({
+  model: google.interactions('gemini-3-pro-image-preview'),
+  prompt: 'Generate a high-quality landscape photo of mountains at sunset.',
+  providerOptions: {
+    google: {
+      responseFormat: [
+        {
+          type: 'image',
+          aspectRatio: '16:9',
+          imageSize: '4K',
+        },
+      ],
+    },
+  },
+});
+```
+
+For multimodal output, list one entry per modality. The model returns
+text in `result.text` and the accompanying image(s) in `result.files`:
+
+```ts
+import { google } from '@ai-sdk/google';
+import { generateText } from 'ai';
+
+const result = await generateText({
+  model: google.interactions('gemini-2.5-flash-image'),
+  prompt:
+    'Tell me a three sentence bedtime story about a unicorn, accompanied by a suitable illustration.',
+  providerOptions: {
+    google: {
+      responseFormat: [
+        { type: 'text' },
+        { type: 'image', aspectRatio: '16:9' },
+      ],
+    },
+  },
+});
+
+console.log(result.text);
+
+const images = result.files.filter(file => file.mediaType.startsWith('image/'));
+// images[0].uint8Array | images[0].base64 | images[0].mediaType
+```
+
 Iterative image editing pairs naturally with stateful chaining — keep
 `previousInteractionId` set across turns and the model edits its prior
 output:
@@ -1359,7 +1415,7 @@ const model = google.interactions('gemini-3-pro-image-preview');
 const turn1 = await generateText({
   model,
   prompt: 'Generate an image of a comic cat in a spaceship.',
-  providerOptions: { google: { responseModalities: ['image'] } },
+  providerOptions: { google: { responseFormat: [{ type: 'image' }] } },
 });
 
 const interactionId = turn1.providerMetadata?.google?.interactionId as
@@ -1371,7 +1427,7 @@ const turn2 = await generateText({
   prompt: 'now make the cat red',
   providerOptions: {
     google: {
-      responseModalities: ['image'],
+      responseFormat: [{ type: 'image' }],
       previousInteractionId: interactionId,
     },
   },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-sdk/google",
-  "version": "4.0.0-canary.63",
+  "version": "4.0.0-canary.65",
   "type": "module",
   "license": "Apache-2.0",
   "sideEffects": false,

package/src/convert-to-google-messages.ts

@@ -64,6 +64,7 @@ function appendToolResultParts(
     LanguageModelV4ToolResultOutput,
     { type: 'content' }
   >['value'],
+  toolCallId?: string,
 ): void {
   const functionResponseParts: GoogleFunctionResponsePart[] = [];
   const responseTextParts: string[] = [];
@@ -106,6 +107,7 @@ function appendToolResultParts(
 
   parts.push({
     functionResponse: {
+      ...(toolCallId != null ? { id: toolCallId } : {}),
       name: toolName,
       response: {
         name: toolName,
@@ -133,12 +135,14 @@ function appendLegacyToolResultParts(
     LanguageModelV4ToolResultOutput,
     { type: 'content' }
   >['value'],
+  toolCallId?: string,
 ): void {
   for (const contentPart of outputValue) {
     switch (contentPart.type) {
       case 'text':
         parts.push({
           functionResponse: {
+            ...(toolCallId != null ? { id: toolCallId } : {}),
             name: toolName,
             response: {
               name: toolName,
@@ -447,6 +451,9 @@ export function convertToGoogleMessages(
 
                   return {
                     functionCall: {
+                      ...(part.toolCallId != null
+                        ? { id: part.toolCallId }
+                        : {}),
                       name: part.toolName,
                       args: part.input,
                     },
@@ -533,13 +540,24 @@ export function convertToGoogleMessages(
 
           if (output.type === 'content') {
             if (supportsFunctionResponseParts) {
-              appendToolResultParts(parts, part.toolName, output.value);
+              appendToolResultParts(
+                parts,
+                part.toolName,
+                output.value,
+                part.toolCallId,
+              );
             } else {
-              appendLegacyToolResultParts(parts, part.toolName, output.value);
+              appendLegacyToolResultParts(
+                parts,
+                part.toolName,
+                output.value,
+                part.toolCallId,
+              );
             }
           } else {
             parts.push({
               functionResponse: {
+                ...(part.toolCallId != null ? { id: part.toolCallId } : {}),
                 name: part.toolName,
                 response: {
                   name: part.toolName,

package/src/google-language-model.ts

@@ -387,7 +387,7 @@ export class GoogleLanguageModel implements LanguageModelV4 {
       } else if ('functionCall' in part && part.functionCall.name != null) {
         content.push({
           type: 'tool-call' as const,
-          toolCallId: this.config.generateId(),
+          toolCallId: part.functionCall.id ?? this.config.generateId(),
           toolName: part.functionCall.name,
           input: JSON.stringify(part.functionCall.args ?? {}),
           providerMetadata: part.thoughtSignature
@@ -838,7 +838,7 @@ export class GoogleLanguageModel implements LanguageModelV4 {
                     part.functionCall.name != null &&
                     part.functionCall.willContinue === true
                   ) {
-                    const toolCallId = generateId();
+                    const toolCallId = part.functionCall.id ?? generateId();
                     const accumulator = new GoogleJSONAccumulator();
                     activeStreamingToolCalls.push({
                       toolCallId,
@@ -920,7 +920,7 @@ export class GoogleLanguageModel implements LanguageModelV4 {
 
                   hasToolCalls = true;
                 } else if (isCompleteCall) {
-                  const toolCallId = generateId();
+                  const toolCallId = part.functionCall.id ?? generateId();
                   const toolName = part.functionCall.name!;
                   const args =
                     typeof part.functionCall.args === 'string'
@@ -957,7 +957,7 @@ export class GoogleLanguageModel implements LanguageModelV4 {
 
                   hasToolCalls = true;
                 } else if (isNoArgsCompleteCall) {
-                  const toolCallId = generateId();
+                  const toolCallId = part.functionCall.id ?? generateId();
                   const toolName = part.functionCall.name!;
 
                   controller.enqueue({
@@ -1333,6 +1333,7 @@ const getContentSchema = () =>
           // note: order matters since text can be fully empty
           z.object({
             functionCall: z.object({
+              id: z.string().nullish(),
               name: z.string().nullish(),
               args: z.unknown().nullish(),
               partialArgs: z.array(partialArgSchema).nullish(),

package/src/google-prompt.ts

@@ -27,9 +27,13 @@ export type GoogleContentPart =
       thought?: boolean;
       thoughtSignature?: string;
     }
-  | { functionCall: { name: string; args: unknown }; thoughtSignature?: string }
+  | {
+      functionCall: { id?: string; name: string; args: unknown };
+      thoughtSignature?: string;
+    }
   | {
       functionResponse: {
+        id?: string;
         name: string;
         response: unknown;
         parts?: Array<GoogleFunctionResponsePart>;