@ai-sdk/openai 1.3.10 → 1.3.12

package/internal/dist/index.mjs

@@ -1623,7 +1623,7 @@ import {
   postFormDataToApi
 } from "@ai-sdk/provider-utils";
 import { z as z6 } from "zod";
-var OpenAIProviderOptionsSchema = z6.object({
+var openAIProviderOptionsSchema = z6.object({
   include: z6.array(z6.string()).nullish(),
   language: z6.string().nullish(),
   prompt: z6.string().nullish(),
@@ -1708,7 +1708,7 @@ var OpenAITranscriptionModel = class {
     const openAIOptions = parseProviderOptions({
       provider: "openai",
       providerOptions,
-      schema: OpenAIProviderOptionsSchema
+      schema: openAIProviderOptionsSchema
     });
     const formData = new FormData();
     const blob = audio instanceof Uint8Array ? new Blob([audio]) : new Blob([convertBase64ToUint8Array(audio)]);
@@ -1789,16 +1789,120 @@ var openaiTranscriptionResponseSchema = z6.object({
   ).nullish()
 });
 
-// src/responses/openai-responses-language-model.ts
+// src/openai-speech-model.ts
 import {
   combineHeaders as combineHeaders6,
-  createEventSourceResponseHandler as createEventSourceResponseHandler3,
-  createJsonResponseHandler as createJsonResponseHandler6,
-  generateId as generateId2,
+  createBinaryResponseHandler,
   parseProviderOptions as parseProviderOptions2,
   postJsonToApi as postJsonToApi5
 } from "@ai-sdk/provider-utils";
 import { z as z7 } from "zod";
+var OpenAIProviderOptionsSchema = z7.object({
+  instructions: z7.string().nullish(),
+  speed: z7.number().min(0.25).max(4).default(1).nullish()
+});
+var OpenAISpeechModel = class {
+  constructor(modelId, config) {
+    this.modelId = modelId;
+    this.config = config;
+    this.specificationVersion = "v1";
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  getArgs({
+    text,
+    voice = "alloy",
+    outputFormat = "mp3",
+    speed,
+    instructions,
+    providerOptions
+  }) {
+    const warnings = [];
+    const openAIOptions = parseProviderOptions2({
+      provider: "openai",
+      providerOptions,
+      schema: OpenAIProviderOptionsSchema
+    });
+    const requestBody = {
+      model: this.modelId,
+      input: text,
+      voice,
+      response_format: "mp3",
+      speed,
+      instructions
+    };
+    if (outputFormat) {
+      if (["mp3", "opus", "aac", "flac", "wav", "pcm"].includes(outputFormat)) {
+        requestBody.response_format = outputFormat;
+      } else {
+        warnings.push({
+          type: "unsupported-setting",
+          setting: "outputFormat",
+          details: `Unsupported output format: ${outputFormat}. Using mp3 instead.`
+        });
+      }
+    }
+    if (openAIOptions) {
+      const speechModelOptions = {};
+      for (const key in speechModelOptions) {
+        const value = speechModelOptions[key];
+        if (value !== void 0) {
+          requestBody[key] = value;
+        }
+      }
+    }
+    return {
+      requestBody,
+      warnings
+    };
+  }
+  async doGenerate(options) {
+    var _a, _b, _c;
+    const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
+    const { requestBody, warnings } = this.getArgs(options);
+    const {
+      value: audio,
+      responseHeaders,
+      rawValue: rawResponse
+    } = await postJsonToApi5({
+      url: this.config.url({
+        path: "/audio/speech",
+        modelId: this.modelId
+      }),
+      headers: combineHeaders6(this.config.headers(), options.headers),
+      body: requestBody,
+      failedResponseHandler: openaiFailedResponseHandler,
+      successfulResponseHandler: createBinaryResponseHandler(),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    return {
+      audio,
+      warnings,
+      request: {
+        body: JSON.stringify(requestBody)
+      },
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: rawResponse
+      }
+    };
+  }
+};
+
+// src/responses/openai-responses-language-model.ts
+import {
+  combineHeaders as combineHeaders7,
+  createEventSourceResponseHandler as createEventSourceResponseHandler3,
+  createJsonResponseHandler as createJsonResponseHandler6,
+  generateId as generateId2,
+  parseProviderOptions as parseProviderOptions3,
+  postJsonToApi as postJsonToApi6
+} from "@ai-sdk/provider-utils";
+import { z as z8 } from "zod";
 
 // src/responses/convert-to-openai-responses-messages.ts
 import {
@@ -2089,7 +2193,7 @@ var OpenAIResponsesLanguageModel = class {
       systemMessageMode: modelConfig.systemMessageMode
     });
     warnings.push(...messageWarnings);
-    const openaiOptions = parseProviderOptions2({
+    const openaiOptions = parseProviderOptions3({
       provider: "openai",
       providerOptions: providerMetadata,
       schema: openaiResponsesProviderOptionsSchema
@@ -2209,58 +2313,58 @@ var OpenAIResponsesLanguageModel = class {
       responseHeaders,
       value: response,
       rawValue: rawResponse
-    } = await postJsonToApi5({
+    } = await postJsonToApi6({
       url: this.config.url({
         path: "/responses",
         modelId: this.modelId
       }),
-      headers: combineHeaders6(this.config.headers(), options.headers),
+      headers: combineHeaders7(this.config.headers(), options.headers),
       body,
       failedResponseHandler: openaiFailedResponseHandler,
       successfulResponseHandler: createJsonResponseHandler6(
-        z7.object({
-          id: z7.string(),
-          created_at: z7.number(),
-          model: z7.string(),
-          output: z7.array(
-            z7.discriminatedUnion("type", [
-              z7.object({
-                type: z7.literal("message"),
-                role: z7.literal("assistant"),
-                content: z7.array(
-                  z7.object({
-                    type: z7.literal("output_text"),
-                    text: z7.string(),
-                    annotations: z7.array(
-                      z7.object({
-                        type: z7.literal("url_citation"),
-                        start_index: z7.number(),
-                        end_index: z7.number(),
-                        url: z7.string(),
-                        title: z7.string()
+        z8.object({
+          id: z8.string(),
+          created_at: z8.number(),
+          model: z8.string(),
+          output: z8.array(
+            z8.discriminatedUnion("type", [
+              z8.object({
+                type: z8.literal("message"),
+                role: z8.literal("assistant"),
+                content: z8.array(
+                  z8.object({
+                    type: z8.literal("output_text"),
+                    text: z8.string(),
+                    annotations: z8.array(
+                      z8.object({
+                        type: z8.literal("url_citation"),
+                        start_index: z8.number(),
+                        end_index: z8.number(),
+                        url: z8.string(),
+                        title: z8.string()
                       })
                     )
                   })
                 )
               }),
-              z7.object({
-                type: z7.literal("function_call"),
-                call_id: z7.string(),
-                name: z7.string(),
-                arguments: z7.string()
+              z8.object({
+                type: z8.literal("function_call"),
+                call_id: z8.string(),
+                name: z8.string(),
+                arguments: z8.string()
               }),
-              z7.object({
-                type: z7.literal("web_search_call")
+              z8.object({
+                type: z8.literal("web_search_call")
               }),
-              z7.object({
-                type: z7.literal("computer_call")
+              z8.object({
+                type: z8.literal("computer_call")
               }),
-              z7.object({
-                type: z7.literal("reasoning")
+              z8.object({
+                type: z8.literal("reasoning")
               })
             ])
           ),
-          incomplete_details: z7.object({ reason: z7.string() }).nullable(),
+          incomplete_details: z8.object({ reason: z8.string() }).nullable(),
           usage: usageSchema
         })
       ),
@@ -2324,12 +2428,12 @@ var OpenAIResponsesLanguageModel = class {
   }
   async doStream(options) {
     const { args: body, warnings } = this.getArgs(options);
-    const { responseHeaders, value: response } = await postJsonToApi5({
+    const { responseHeaders, value: response } = await postJsonToApi6({
       url: this.config.url({
         path: "/responses",
         modelId: this.modelId
       }),
-      headers: combineHeaders6(this.config.headers(), options.headers),
+      headers: combineHeaders7(this.config.headers(), options.headers),
       body: {
         ...body,
         stream: true
@@ -2458,79 +2562,79 @@ var OpenAIResponsesLanguageModel = class {
     };
   }
 };
-var usageSchema = z7.object({
-  input_tokens: z7.number(),
-  input_tokens_details: z7.object({ cached_tokens: z7.number().nullish() }).nullish(),
-  output_tokens: z7.number(),
-  output_tokens_details: z7.object({ reasoning_tokens: z7.number().nullish() }).nullish()
+var usageSchema = z8.object({
+  input_tokens: z8.number(),
+  input_tokens_details: z8.object({ cached_tokens: z8.number().nullish() }).nullish(),
+  output_tokens: z8.number(),
+  output_tokens_details: z8.object({ reasoning_tokens: z8.number().nullish() }).nullish()
 });
-var textDeltaChunkSchema = z7.object({
-  type: z7.literal("response.output_text.delta"),
-  delta: z7.string()
+var textDeltaChunkSchema = z8.object({
+  type: z8.literal("response.output_text.delta"),
+  delta: z8.string()
 });
-var responseFinishedChunkSchema = z7.object({
-  type: z7.enum(["response.completed", "response.incomplete"]),
-  response: z7.object({
-    incomplete_details: z7.object({ reason: z7.string() }).nullish(),
+var responseFinishedChunkSchema = z8.object({
+  type: z8.enum(["response.completed", "response.incomplete"]),
+  response: z8.object({
+    incomplete_details: z8.object({ reason: z8.string() }).nullish(),
     usage: usageSchema
   })
 });
-var responseCreatedChunkSchema = z7.object({
-  type: z7.literal("response.created"),
-  response: z7.object({
-    id: z7.string(),
-    created_at: z7.number(),
-    model: z7.string()
+var responseCreatedChunkSchema = z8.object({
+  type: z8.literal("response.created"),
+  response: z8.object({
+    id: z8.string(),
+    created_at: z8.number(),
+    model: z8.string()
   })
 });
-var responseOutputItemDoneSchema = z7.object({
-  type: z7.literal("response.output_item.done"),
-  output_index: z7.number(),
-  item: z7.discriminatedUnion("type", [
-    z7.object({
-      type: z7.literal("message")
+var responseOutputItemDoneSchema = z8.object({
+  type: z8.literal("response.output_item.done"),
+  output_index: z8.number(),
+  item: z8.discriminatedUnion("type", [
+    z8.object({
+      type: z8.literal("message")
     }),
-    z7.object({
-      type: z7.literal("function_call"),
-      id: z7.string(),
-      call_id: z7.string(),
-      name: z7.string(),
-      arguments: z7.string(),
-      status: z7.literal("completed")
+    z8.object({
+      type: z8.literal("function_call"),
+      id: z8.string(),
+      call_id: z8.string(),
+      name: z8.string(),
+      arguments: z8.string(),
+      status: z8.literal("completed")
     })
   ])
 });
-var responseFunctionCallArgumentsDeltaSchema = z7.object({
-  type: z7.literal("response.function_call_arguments.delta"),
-  item_id: z7.string(),
-  output_index: z7.number(),
-  delta: z7.string()
+var responseFunctionCallArgumentsDeltaSchema = z8.object({
+  type: z8.literal("response.function_call_arguments.delta"),
+  item_id: z8.string(),
+  output_index: z8.number(),
+  delta: z8.string()
 });
-var responseOutputItemAddedSchema = z7.object({
-  type: z7.literal("response.output_item.added"),
-  output_index: z7.number(),
-  item: z7.discriminatedUnion("type", [
-    z7.object({
-      type: z7.literal("message")
+var responseOutputItemAddedSchema = z8.object({
+  type: z8.literal("response.output_item.added"),
+  output_index: z8.number(),
+  item: z8.discriminatedUnion("type", [
+    z8.object({
+      type: z8.literal("message")
     }),
-    z7.object({
-      type: z7.literal("function_call"),
-      id: z7.string(),
-      call_id: z7.string(),
-      name: z7.string(),
-      arguments: z7.string()
+    z8.object({
+      type: z8.literal("function_call"),
+      id: z8.string(),
+      call_id: z8.string(),
+      name: z8.string(),
+      arguments: z8.string()
     })
   ])
 });
-var responseAnnotationAddedSchema = z7.object({
-  type: z7.literal("response.output_text.annotation.added"),
-  annotation: z7.object({
-    type: z7.literal("url_citation"),
-    url: z7.string(),
-    title: z7.string()
+var responseAnnotationAddedSchema = z8.object({
+  type: z8.literal("response.output_text.annotation.added"),
+  annotation: z8.object({
+    type: z8.literal("url_citation"),
+    url: z8.string(),
+    title: z8.string()
   })
 });
-var openaiResponsesChunkSchema = z7.union([
+var openaiResponsesChunkSchema = z8.union([
   textDeltaChunkSchema,
   responseFinishedChunkSchema,
   responseCreatedChunkSchema,
@@ -2538,7 +2642,7 @@ var openaiResponsesChunkSchema = z7.union([
   responseFunctionCallArgumentsDeltaSchema,
   responseOutputItemAddedSchema,
   responseAnnotationAddedSchema,
-  z7.object({ type: z7.string() }).passthrough()
+  z8.object({ type: z8.string() }).passthrough()
   // fallback for unknown chunks
 ]);
 function isTextDeltaChunk(chunk) {
@@ -2583,15 +2687,15 @@ function getResponsesModelConfig(modelId) {
     requiredAutoTruncation: false
   };
 }
-var openaiResponsesProviderOptionsSchema = z7.object({
-  metadata: z7.any().nullish(),
-  parallelToolCalls: z7.boolean().nullish(),
-  previousResponseId: z7.string().nullish(),
-  store: z7.boolean().nullish(),
-  user: z7.string().nullish(),
-  reasoningEffort: z7.string().nullish(),
-  strictSchemas: z7.boolean().nullish(),
-  instructions: z7.string().nullish()
+var openaiResponsesProviderOptionsSchema = z8.object({
+  metadata: z8.any().nullish(),
+  parallelToolCalls: z8.boolean().nullish(),
+  previousResponseId: z8.string().nullish(),
+  store: z8.boolean().nullish(),
+  user: z8.string().nullish(),
+  reasoningEffort: z8.string().nullish(),
+  strictSchemas: z8.boolean().nullish(),
+  instructions: z8.string().nullish()
 });
 export {
   OpenAIChatLanguageModel,
@@ -2599,6 +2703,7 @@ export {
   OpenAIEmbeddingModel,
   OpenAIImageModel,
   OpenAIResponsesLanguageModel,
+  OpenAISpeechModel,
   OpenAITranscriptionModel,
   modelMaxImagesPerCall
 };