@ai-sdk/openai 2.1.0-beta.10 → 2.1.0-beta.12

package/CHANGELOG.md

@@ -1,5 +1,28 @@
 # @ai-sdk/openai
 
+## 2.1.0-beta.12
+
+### Patch Changes
+
+- 046aa3b: feat(provider): speech model v3 spec
+- e8109d3: feat: tool execution approval
+- 21e20c0: feat(provider): transcription model v3 spec
+- Updated dependencies [046aa3b]
+- Updated dependencies [e8109d3]
+  - @ai-sdk/provider@2.1.0-beta.5
+  - @ai-sdk/provider-utils@3.1.0-beta.7
+
+## 2.1.0-beta.11
+
+### Patch Changes
+
+- 0adc679: feat(provider): shared spec v3
+- 2b0caef: feat(provider/openai): preview image generation results
+- Updated dependencies [0adc679]
+- Updated dependencies [2b0caef]
+  - @ai-sdk/provider-utils@3.1.0-beta.6
+  - @ai-sdk/provider@2.1.0-beta.4
+
 ## 2.1.0-beta.10
 
 ### Patch Changes

package/dist/index.d.mts

@@ -1,4 +1,4 @@
-import { ProviderV3, LanguageModelV3, EmbeddingModelV3, ImageModelV3, TranscriptionModelV2, SpeechModelV2 } from '@ai-sdk/provider';
+import { ProviderV3, LanguageModelV3, EmbeddingModelV3, ImageModelV3, TranscriptionModelV3, SpeechModelV3 } from '@ai-sdk/provider';
 import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
 import { FetchFunction } from '@ai-sdk/provider-utils';
 import { z } from 'zod/v4';
@@ -182,11 +182,16 @@ declare const openaiTools: {
      *
      * Must have name `image_generation`.
      *
-     * @param size - Image dimensions (e.g., 1024x1024, 1024x1536)
-     * @param quality - Rendering quality (e.g. low, medium, high)
-     * @param format - File output format
-     * @param compression - Compression level (0-100%) for JPEG and WebP formats
-     * @param background - Transparent or opaque
+     * @param background - Background type for the generated image. One of 'auto', 'opaque', or 'transparent'.
+     * @param inputFidelity - Input fidelity for the generated image. One of 'low' or 'high'.
+     * @param inputImageMask - Optional mask for inpainting. Contains fileId and/or imageUrl.
+     * @param model - The image generation model to use. Default: gpt-image-1.
+     * @param moderation - Moderation level for the generated image. Default: 'auto'.
+     * @param outputCompression - Compression level for the output image (0-100).
+     * @param outputFormat - The output format of the generated image. One of 'png', 'jpeg', or 'webp'.
+     * @param partialImages - Number of partial images to generate in streaming mode (0-3).
+     * @param quality - The quality of the generated image. One of 'auto', 'low', 'medium', or 'high'.
+     * @param size - The size of the generated image. One of 'auto', '1024x1024', '1024x1536', or '1536x1024'.
      */
     imageGeneration: (args?: {
         background?: "auto" | "opaque" | "transparent";
@@ -199,6 +204,7 @@ declare const openaiTools: {
         moderation?: "auto";
         outputCompression?: number;
         outputFormat?: "png" | "jpeg" | "webp";
+        partialImages?: number;
         quality?: "auto" | "low" | "medium" | "high";
         size?: "auto" | "1024x1024" | "1024x1536" | "1536x1024";
     }) => _ai_sdk_provider_utils.Tool<{}, {
@@ -305,11 +311,11 @@ interface OpenAIProvider extends ProviderV3 {
     /**
   Creates a model for transcription.
      */
-    transcription(modelId: OpenAITranscriptionModelId): TranscriptionModelV2;
+    transcription(modelId: OpenAITranscriptionModelId): TranscriptionModelV3;
     /**
   Creates a model for speech generation.
      */
-    speech(modelId: OpenAISpeechModelId): SpeechModelV2;
+    speech(modelId: OpenAISpeechModelId): SpeechModelV3;
     /**
   OpenAI-specific tools.
      */

package/dist/index.d.ts

@@ -1,4 +1,4 @@
-import { ProviderV3, LanguageModelV3, EmbeddingModelV3, ImageModelV3, TranscriptionModelV2, SpeechModelV2 } from '@ai-sdk/provider';
+import { ProviderV3, LanguageModelV3, EmbeddingModelV3, ImageModelV3, TranscriptionModelV3, SpeechModelV3 } from '@ai-sdk/provider';
 import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
 import { FetchFunction } from '@ai-sdk/provider-utils';
 import { z } from 'zod/v4';
@@ -182,11 +182,16 @@ declare const openaiTools: {
      *
      * Must have name `image_generation`.
      *
-     * @param size - Image dimensions (e.g., 1024x1024, 1024x1536)
-     * @param quality - Rendering quality (e.g. low, medium, high)
-     * @param format - File output format
-     * @param compression - Compression level (0-100%) for JPEG and WebP formats
-     * @param background - Transparent or opaque
+     * @param background - Background type for the generated image. One of 'auto', 'opaque', or 'transparent'.
+     * @param inputFidelity - Input fidelity for the generated image. One of 'low' or 'high'.
+     * @param inputImageMask - Optional mask for inpainting. Contains fileId and/or imageUrl.
+     * @param model - The image generation model to use. Default: gpt-image-1.
+     * @param moderation - Moderation level for the generated image. Default: 'auto'.
+     * @param outputCompression - Compression level for the output image (0-100).
+     * @param outputFormat - The output format of the generated image. One of 'png', 'jpeg', or 'webp'.
+     * @param partialImages - Number of partial images to generate in streaming mode (0-3).
+     * @param quality - The quality of the generated image. One of 'auto', 'low', 'medium', or 'high'.
+     * @param size - The size of the generated image. One of 'auto', '1024x1024', '1024x1536', or '1536x1024'.
      */
     imageGeneration: (args?: {
         background?: "auto" | "opaque" | "transparent";
@@ -199,6 +204,7 @@ declare const openaiTools: {
         moderation?: "auto";
         outputCompression?: number;
         outputFormat?: "png" | "jpeg" | "webp";
+        partialImages?: number;
         quality?: "auto" | "low" | "medium" | "high";
         size?: "auto" | "1024x1024" | "1024x1536" | "1536x1024";
     }) => _ai_sdk_provider_utils.Tool<{}, {
@@ -305,11 +311,11 @@ interface OpenAIProvider extends ProviderV3 {
     /**
   Creates a model for transcription.
      */
-    transcription(modelId: OpenAITranscriptionModelId): TranscriptionModelV2;
+    transcription(modelId: OpenAITranscriptionModelId): TranscriptionModelV3;
     /**
   Creates a model for speech generation.
      */
-    speech(modelId: OpenAISpeechModelId): SpeechModelV2;
+    speech(modelId: OpenAISpeechModelId): SpeechModelV3;
     /**
   OpenAI-specific tools.
      */

package/dist/index.js

@@ -60,6 +60,7 @@ function convertToOpenAIChatMessages({
   prompt,
   systemMessageMode = "system"
 }) {
+  var _a;
   const messages = [];
   const warnings = [];
   for (const { role, content } of prompt) {
@@ -98,7 +99,7 @@ function convertToOpenAIChatMessages({
         messages.push({
           role: "user",
           content: content.map((part, index) => {
-            var _a, _b, _c;
+            var _a2, _b, _c;
             switch (part.type) {
               case "text": {
                 return { type: "text", text: part.text };
@@ -111,7 +112,7 @@ function convertToOpenAIChatMessages({
                     image_url: {
                       url: part.data instanceof URL ? part.data.toString() : `data:${mediaType};base64,${(0, import_provider_utils2.convertToBase64)(part.data)}`,
                       // OpenAI specific extension: image detail
-                      detail: (_b = (_a = part.providerOptions) == null ? void 0 : _a.openai) == null ? void 0 : _b.imageDetail
+                      detail: (_b = (_a2 = part.providerOptions) == null ? void 0 : _a2.openai) == null ? void 0 : _b.imageDetail
                     }
                   };
                 } else if (part.mediaType.startsWith("audio/")) {
@@ -208,6 +209,9 @@ function convertToOpenAIChatMessages({
             case "error-text":
               contentValue = output.value;
               break;
+            case "execution-denied":
+              contentValue = (_a = output.reason) != null ? _a : "Tool execution denied.";
+              break;
             case "content":
             case "json":
             case "error-json":
@@ -1767,6 +1771,7 @@ var imageGenerationArgsSchema = import_v411.z.object({
   moderation: import_v411.z.enum(["auto"]).optional(),
   outputCompression: import_v411.z.number().int().min(0).max(100).optional(),
   outputFormat: import_v411.z.enum(["png", "jpeg", "webp"]).optional(),
+  partialImages: import_v411.z.number().int().min(0).max(3).optional(),
   quality: import_v411.z.enum(["auto", "low", "medium", "high"]).optional(),
   size: import_v411.z.enum(["1024x1024", "1024x1536", "1536x1024", "auto"]).optional()
 }).strict();
@@ -1938,11 +1943,16 @@ var openaiTools = {
    *
    * Must have name `image_generation`.
    *
-   * @param size - Image dimensions (e.g., 1024x1024, 1024x1536)
-   * @param quality - Rendering quality (e.g. low, medium, high)
-   * @param format - File output format
-   * @param compression - Compression level (0-100%) for JPEG and WebP formats
-   * @param background - Transparent or opaque
+   * @param background - Background type for the generated image. One of 'auto', 'opaque', or 'transparent'.
+   * @param inputFidelity - Input fidelity for the generated image. One of 'low' or 'high'.
+   * @param inputImageMask - Optional mask for inpainting. Contains fileId and/or imageUrl.
+   * @param model - The image generation model to use. Default: gpt-image-1.
+   * @param moderation - Moderation level for the generated image. Default: 'auto'.
+   * @param outputCompression - Compression level for the output image (0-100).
+   * @param outputFormat - The output format of the generated image. One of 'png', 'jpeg', or 'webp'.
+   * @param partialImages - Number of partial images to generate in streaming mode (0-3).
+   * @param quality - The quality of the generated image. One of 'auto', 'low', 'medium', or 'high'.
+   * @param size - The size of the generated image. One of 'auto', '1024x1024', '1024x1536', or '1536x1024'.
    */
   imageGeneration,
   /**
@@ -1999,7 +2009,7 @@ async function convertToOpenAIResponsesInput({
   store,
   hasLocalShellTool = false
 }) {
-  var _a, _b, _c, _d, _e, _f, _g, _h, _i;
+  var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j;
   const input = [];
   const warnings = [];
   for (const { role, content } of prompt) {
@@ -2202,6 +2212,9 @@ async function convertToOpenAIResponsesInput({
             case "error-text":
               contentValue = output.value;
               break;
+            case "execution-denied":
+              contentValue = (_j = output.reason) != null ? _j : "Tool execution denied.";
+              break;
             case "content":
             case "json":
             case "error-json":
@@ -2331,11 +2344,12 @@ function prepareResponsesTools({
                 image_url: args.inputImageMask.imageUrl
               } : void 0,
               model: args.model,
-              size: args.size,
-              quality: args.quality,
               moderation: args.moderation,
+              partial_images: args.partialImages,
+              quality: args.quality,
+              output_compression: args.outputCompression,
               output_format: args.outputFormat,
-              output_compression: args.outputCompression
+              size: args.size
             });
             break;
           }
@@ -3294,6 +3308,17 @@ var OpenAIResponsesLanguageModel = class {
                   delta: value.delta
                 });
               }
+            } else if (isResponseImageGenerationCallPartialImageChunk(value)) {
+              controller.enqueue({
+                type: "tool-result",
+                toolCallId: value.item_id,
+                toolName: "image_generation",
+                result: {
+                  result: value.partial_image_b64
+                },
+                providerExecuted: true,
+                preliminary: true
+              });
             } else if (isResponseCodeInterpreterCallCodeDeltaChunk(value)) {
               const toolCall = ongoingToolCalls[value.output_index];
               if (toolCall != null) {
@@ -3566,6 +3591,12 @@ var responseFunctionCallArgumentsDeltaSchema = import_v416.z.object({
   output_index: import_v416.z.number(),
   delta: import_v416.z.string()
 });
+var responseImageGenerationCallPartialImageSchema = import_v416.z.object({
+  type: import_v416.z.literal("response.image_generation_call.partial_image"),
+  item_id: import_v416.z.string(),
+  output_index: import_v416.z.number(),
+  partial_image_b64: import_v416.z.string()
+});
 var responseCodeInterpreterCallCodeDeltaSchema = import_v416.z.object({
   type: import_v416.z.literal("response.code_interpreter_call_code.delta"),
   item_id: import_v416.z.string(),
@@ -3615,6 +3646,7 @@ var openaiResponsesChunkSchema = import_v416.z.union([
   responseOutputItemAddedSchema,
   responseOutputItemDoneSchema,
   responseFunctionCallArgumentsDeltaSchema,
+  responseImageGenerationCallPartialImageSchema,
   responseCodeInterpreterCallCodeDeltaSchema,
   responseCodeInterpreterCallCodeDoneSchema,
   responseAnnotationAddedSchema,
@@ -3642,6 +3674,9 @@ function isResponseCreatedChunk(chunk) {
 function isResponseFunctionCallArgumentsDeltaChunk(chunk) {
   return chunk.type === "response.function_call_arguments.delta";
 }
+function isResponseImageGenerationCallPartialImageChunk(chunk) {
+  return chunk.type === "response.image_generation_call.partial_image";
+}
 function isResponseCodeInterpreterCallCodeDeltaChunk(chunk) {
   return chunk.type === "response.code_interpreter_call_code.delta";
 }
@@ -3753,7 +3788,7 @@ var OpenAISpeechModel = class {
   constructor(modelId, config) {
     this.modelId = modelId;
     this.config = config;
-    this.specificationVersion = "v2";
+    this.specificationVersion = "v3";
   }
   get provider() {
     return this.config.provider;
@@ -3944,7 +3979,7 @@ var OpenAITranscriptionModel = class {
   constructor(modelId, config) {
     this.modelId = modelId;
     this.config = config;
-    this.specificationVersion = "v2";
+    this.specificationVersion = "v3";
   }
   get provider() {
     return this.config.provider;
@@ -4074,7 +4109,7 @@ var openaiTranscriptionResponseSchema = import_v419.z.object({
 });
 
 // src/version.ts
-var VERSION = true ? "2.1.0-beta.10" : "0.0.0-test";
+var VERSION = true ? "2.1.0-beta.12" : "0.0.0-test";
 
 // src/openai-provider.ts
 function createOpenAI(options = {}) {