@ai-sdk/openai 1.3.7 → 1.3.8

package/dist/index.mjs

@@ -1611,16 +1611,201 @@ var openaiImageResponseSchema = z5.object({
   data: z5.array(z5.object({ b64_json: z5.string() }))
 });
 
-// src/responses/openai-responses-language-model.ts
+// src/openai-transcription-model.ts
 import {
   combineHeaders as combineHeaders5,
-  createEventSourceResponseHandler as createEventSourceResponseHandler3,
+  convertBase64ToUint8Array,
   createJsonResponseHandler as createJsonResponseHandler5,
-  generateId as generateId2,
   parseProviderOptions,
-  postJsonToApi as postJsonToApi5
+  postFormDataToApi
 } from "@ai-sdk/provider-utils";
 import { z as z6 } from "zod";
+var OpenAIProviderOptionsSchema = z6.object({
+  include: z6.array(z6.string()).optional().describe(
+    "Additional information to include in the transcription response."
+  ),
+  language: z6.string().optional().describe("The language of the input audio in ISO-639-1 format."),
+  prompt: z6.string().optional().describe(
+    "An optional text to guide the model's style or continue a previous audio segment."
+  ),
+  temperature: z6.number().min(0).max(1).optional().default(0).describe("The sampling temperature, between 0 and 1."),
+  timestampGranularities: z6.array(z6.enum(["word", "segment"])).optional().default(["segment"]).describe(
+    "The timestamp granularities to populate for this transcription."
+  )
+});
+var languageMap = {
+  afrikaans: "af",
+  arabic: "ar",
+  armenian: "hy",
+  azerbaijani: "az",
+  belarusian: "be",
+  bosnian: "bs",
+  bulgarian: "bg",
+  catalan: "ca",
+  chinese: "zh",
+  croatian: "hr",
+  czech: "cs",
+  danish: "da",
+  dutch: "nl",
+  english: "en",
+  estonian: "et",
+  finnish: "fi",
+  french: "fr",
+  galician: "gl",
+  german: "de",
+  greek: "el",
+  hebrew: "he",
+  hindi: "hi",
+  hungarian: "hu",
+  icelandic: "is",
+  indonesian: "id",
+  italian: "it",
+  japanese: "ja",
+  kannada: "kn",
+  kazakh: "kk",
+  korean: "ko",
+  latvian: "lv",
+  lithuanian: "lt",
+  macedonian: "mk",
+  malay: "ms",
+  marathi: "mr",
+  maori: "mi",
+  nepali: "ne",
+  norwegian: "no",
+  persian: "fa",
+  polish: "pl",
+  portuguese: "pt",
+  romanian: "ro",
+  russian: "ru",
+  serbian: "sr",
+  slovak: "sk",
+  slovenian: "sl",
+  spanish: "es",
+  swahili: "sw",
+  swedish: "sv",
+  tagalog: "tl",
+  tamil: "ta",
+  thai: "th",
+  turkish: "tr",
+  ukrainian: "uk",
+  urdu: "ur",
+  vietnamese: "vi",
+  welsh: "cy"
+};
+var OpenAITranscriptionModel = class {
+  constructor(modelId, config) {
+    this.modelId = modelId;
+    this.config = config;
+    this.specificationVersion = "v1";
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  getArgs({
+    audio,
+    mimeType,
+    providerOptions
+  }) {
+    const warnings = [];
+    const openAIOptions = parseProviderOptions({
+      provider: "openai",
+      providerOptions,
+      schema: OpenAIProviderOptionsSchema
+    });
+    const formData = new FormData();
+    const blob = audio instanceof Uint8Array ? new Blob([audio]) : new Blob([convertBase64ToUint8Array(audio)]);
+    formData.append("model", this.modelId);
+    formData.append("file", new File([blob], "audio", { type: mimeType }));
+    if (openAIOptions) {
+      const transcriptionModelOptions = {
+        include: openAIOptions.include,
+        language: openAIOptions.language,
+        prompt: openAIOptions.prompt,
+        temperature: openAIOptions.temperature,
+        timestamp_granularities: openAIOptions.timestampGranularities
+      };
+      for (const key in transcriptionModelOptions) {
+        const value = transcriptionModelOptions[key];
+        if (value !== void 0) {
+          formData.append(key, value);
+        }
+      }
+    }
+    return {
+      formData,
+      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 { formData, warnings } = this.getArgs(options);
+    const { value: response, responseHeaders } = await postFormDataToApi({
+      url: this.config.url({
+        path: "/audio/transcriptions",
+        modelId: this.modelId
+      }),
+      headers: combineHeaders5(this.config.headers(), options.headers),
+      formData,
+      failedResponseHandler: openaiFailedResponseHandler,
+      successfulResponseHandler: createJsonResponseHandler5(
+        openaiTranscriptionResponseSchema
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    let language;
+    if (response.language && response.language in languageMap) {
+      language = languageMap[response.language];
+    }
+    return {
+      text: response.text,
+      segments: response.words.map((word) => ({
+        text: word.word,
+        startSecond: word.start,
+        endSecond: word.end
+      })),
+      language,
+      durationInSeconds: response.duration,
+      warnings,
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: response
+      },
+      // When using format `verbose_json` on `whisper-1`, OpenAI includes the things like `task` and enhanced `segments` information.
+      providerMetadata: {
+        openai: {
+          transcript: response
+        }
+      }
+    };
+  }
+};
+var openaiTranscriptionResponseSchema = z6.object({
+  text: z6.string(),
+  language: z6.string().optional(),
+  duration: z6.number().optional(),
+  words: z6.array(
+    z6.object({
+      word: z6.string(),
+      start: z6.number(),
+      end: z6.number()
+    })
+  )
+});
+
+// src/responses/openai-responses-language-model.ts
+import {
+  combineHeaders as combineHeaders6,
+  createEventSourceResponseHandler as createEventSourceResponseHandler3,
+  createJsonResponseHandler as createJsonResponseHandler6,
+  generateId as generateId2,
+  parseProviderOptions as parseProviderOptions2,
+  postJsonToApi as postJsonToApi5
+} from "@ai-sdk/provider-utils";
+import { z as z7 } from "zod";
 
 // src/responses/convert-to-openai-responses-messages.ts
 import {
@@ -1911,7 +2096,7 @@ var OpenAIResponsesLanguageModel = class {
       systemMessageMode: modelConfig.systemMessageMode
     });
     warnings.push(...messageWarnings);
-    const openaiOptions = parseProviderOptions({
+    const openaiOptions = parseProviderOptions2({
       provider: "openai",
       providerOptions: providerMetadata,
       schema: openaiResponsesProviderOptionsSchema
@@ -2036,53 +2221,53 @@ var OpenAIResponsesLanguageModel = class {
         path: "/responses",
         modelId: this.modelId
       }),
-      headers: combineHeaders5(this.config.headers(), options.headers),
+      headers: combineHeaders6(this.config.headers(), options.headers),
       body,
       failedResponseHandler: openaiFailedResponseHandler,
-      successfulResponseHandler: createJsonResponseHandler5(
-        z6.object({
-          id: z6.string(),
-          created_at: z6.number(),
-          model: z6.string(),
-          output: z6.array(
-            z6.discriminatedUnion("type", [
-              z6.object({
-                type: z6.literal("message"),
-                role: z6.literal("assistant"),
-                content: z6.array(
-                  z6.object({
-                    type: z6.literal("output_text"),
-                    text: z6.string(),
-                    annotations: z6.array(
-                      z6.object({
-                        type: z6.literal("url_citation"),
-                        start_index: z6.number(),
-                        end_index: z6.number(),
-                        url: z6.string(),
-                        title: z6.string()
+      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()
                       })
                     )
                   })
                 )
               }),
-              z6.object({
-                type: z6.literal("function_call"),
-                call_id: z6.string(),
-                name: z6.string(),
-                arguments: z6.string()
+              z7.object({
+                type: z7.literal("function_call"),
+                call_id: z7.string(),
+                name: z7.string(),
+                arguments: z7.string()
               }),
-              z6.object({
-                type: z6.literal("web_search_call")
+              z7.object({
+                type: z7.literal("web_search_call")
               }),
-              z6.object({
-                type: z6.literal("computer_call")
+              z7.object({
+                type: z7.literal("computer_call")
               }),
-              z6.object({
-                type: z6.literal("reasoning")
+              z7.object({
+                type: z7.literal("reasoning")
               })
             ])
           ),
-          incomplete_details: z6.object({ reason: z6.string() }).nullable(),
+          incomplete_details: z7.object({ reason: z7.string() }).nullable(),
           usage: usageSchema
         })
       ),
@@ -2151,7 +2336,7 @@ var OpenAIResponsesLanguageModel = class {
         path: "/responses",
         modelId: this.modelId
       }),
-      headers: combineHeaders5(this.config.headers(), options.headers),
+      headers: combineHeaders6(this.config.headers(), options.headers),
       body: {
         ...body,
         stream: true
@@ -2280,79 +2465,79 @@ var OpenAIResponsesLanguageModel = class {
     };
   }
 };
-var usageSchema = z6.object({
-  input_tokens: z6.number(),
-  input_tokens_details: z6.object({ cached_tokens: z6.number().nullish() }).nullish(),
-  output_tokens: z6.number(),
-  output_tokens_details: z6.object({ reasoning_tokens: z6.number().nullish() }).nullish()
+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 textDeltaChunkSchema = z6.object({
-  type: z6.literal("response.output_text.delta"),
-  delta: z6.string()
+var textDeltaChunkSchema = z7.object({
+  type: z7.literal("response.output_text.delta"),
+  delta: z7.string()
 });
-var responseFinishedChunkSchema = z6.object({
-  type: z6.enum(["response.completed", "response.incomplete"]),
-  response: z6.object({
-    incomplete_details: z6.object({ reason: z6.string() }).nullish(),
+var responseFinishedChunkSchema = z7.object({
+  type: z7.enum(["response.completed", "response.incomplete"]),
+  response: z7.object({
+    incomplete_details: z7.object({ reason: z7.string() }).nullish(),
     usage: usageSchema
   })
 });
-var responseCreatedChunkSchema = z6.object({
-  type: z6.literal("response.created"),
-  response: z6.object({
-    id: z6.string(),
-    created_at: z6.number(),
-    model: z6.string()
+var responseCreatedChunkSchema = z7.object({
+  type: z7.literal("response.created"),
+  response: z7.object({
+    id: z7.string(),
+    created_at: z7.number(),
+    model: z7.string()
   })
 });
-var responseOutputItemDoneSchema = z6.object({
-  type: z6.literal("response.output_item.done"),
-  output_index: z6.number(),
-  item: z6.discriminatedUnion("type", [
-    z6.object({
-      type: z6.literal("message")
+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")
     }),
-    z6.object({
-      type: z6.literal("function_call"),
-      id: z6.string(),
-      call_id: z6.string(),
-      name: z6.string(),
-      arguments: z6.string(),
-      status: z6.literal("completed")
+    z7.object({
+      type: z7.literal("function_call"),
+      id: z7.string(),
+      call_id: z7.string(),
+      name: z7.string(),
+      arguments: z7.string(),
+      status: z7.literal("completed")
     })
   ])
 });
-var responseFunctionCallArgumentsDeltaSchema = z6.object({
-  type: z6.literal("response.function_call_arguments.delta"),
-  item_id: z6.string(),
-  output_index: z6.number(),
-  delta: z6.string()
+var responseFunctionCallArgumentsDeltaSchema = z7.object({
+  type: z7.literal("response.function_call_arguments.delta"),
+  item_id: z7.string(),
+  output_index: z7.number(),
+  delta: z7.string()
 });
-var responseOutputItemAddedSchema = z6.object({
-  type: z6.literal("response.output_item.added"),
-  output_index: z6.number(),
-  item: z6.discriminatedUnion("type", [
-    z6.object({
-      type: z6.literal("message")
+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")
     }),
-    z6.object({
-      type: z6.literal("function_call"),
-      id: z6.string(),
-      call_id: z6.string(),
-      name: z6.string(),
-      arguments: z6.string()
+    z7.object({
+      type: z7.literal("function_call"),
+      id: z7.string(),
+      call_id: z7.string(),
+      name: z7.string(),
+      arguments: z7.string()
     })
   ])
 });
-var responseAnnotationAddedSchema = z6.object({
-  type: z6.literal("response.output_text.annotation.added"),
-  annotation: z6.object({
-    type: z6.literal("url_citation"),
-    url: z6.string(),
-    title: z6.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 openaiResponsesChunkSchema = z6.union([
+var openaiResponsesChunkSchema = z7.union([
   textDeltaChunkSchema,
   responseFinishedChunkSchema,
   responseCreatedChunkSchema,
@@ -2360,7 +2545,7 @@ var openaiResponsesChunkSchema = z6.union([
   responseFunctionCallArgumentsDeltaSchema,
   responseOutputItemAddedSchema,
   responseAnnotationAddedSchema,
-  z6.object({ type: z6.string() }).passthrough()
+  z7.object({ type: z7.string() }).passthrough()
   // fallback for unknown chunks
 ]);
 function isTextDeltaChunk(chunk) {
@@ -2405,20 +2590,20 @@ function getResponsesModelConfig(modelId) {
     requiredAutoTruncation: false
   };
 }
-var openaiResponsesProviderOptionsSchema = z6.object({
-  metadata: z6.any().nullish(),
-  parallelToolCalls: z6.boolean().nullish(),
-  previousResponseId: z6.string().nullish(),
-  store: z6.boolean().nullish(),
-  user: z6.string().nullish(),
-  reasoningEffort: z6.string().nullish(),
-  strictSchemas: z6.boolean().nullish(),
-  instructions: z6.string().nullish()
+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()
 });
 
 // src/openai-tools.ts
-import { z as z7 } from "zod";
-var WebSearchPreviewParameters = z7.object({});
+import { z as z8 } from "zod";
+var WebSearchPreviewParameters = z8.object({});
 function webSearchPreviewTool({
   searchContextSize,
   userLocation
@@ -2479,6 +2664,12 @@ function createOpenAI(options = {}) {
     headers: getHeaders,
     fetch: options.fetch
   });
+  const createTranscriptionModel = (modelId) => new OpenAITranscriptionModel(modelId, {
+    provider: `${providerName}.transcription`,
+    url: ({ path }) => `${baseURL}${path}`,
+    headers: getHeaders,
+    fetch: options.fetch
+  });
   const createLanguageModel = (modelId, settings) => {
     if (new.target) {
       throw new Error(
@@ -2513,6 +2704,8 @@ function createOpenAI(options = {}) {
   provider.textEmbeddingModel = createEmbeddingModel;
   provider.image = createImageModel;
   provider.imageModel = createImageModel;
+  provider.transcription = createTranscriptionModel;
+  provider.transcriptionModel = createTranscriptionModel;
   provider.tools = openaiTools;
   return provider;
 }