@ai-sdk/openai 2.0.0-canary.2 → 2.0.0-canary.3

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (14) hide show
  1. package/CHANGELOG.md +15 -0
  2. package/dist/index.d.mts +7 -1
  3. package/dist/index.d.ts +7 -1
  4. package/dist/index.js +397 -215
  5. package/dist/index.js.map +1 -1
  6. package/dist/index.mjs +366 -178
  7. package/dist/index.mjs.map +1 -1
  8. package/internal/dist/index.d.mts +67 -2
  9. package/internal/dist/index.d.ts +67 -2
  10. package/internal/dist/index.js +388 -212
  11. package/internal/dist/index.js.map +1 -1
  12. package/internal/dist/index.mjs +359 -178
  13. package/internal/dist/index.mjs.map +1 -1
  14. package/package.json +3 -3
package/internal/dist/index.js CHANGED
@@ -25,18 +25,18 @@ __export(internal_exports, {
25
25
  OpenAIEmbeddingModel: () => OpenAIEmbeddingModel,
26
26
  OpenAIImageModel: () => OpenAIImageModel,
27
27
  OpenAIResponsesLanguageModel: () => OpenAIResponsesLanguageModel,
28
+ OpenAITranscriptionModel: () => OpenAITranscriptionModel,
28
29
  modelMaxImagesPerCall: () => modelMaxImagesPerCall
29
30
  });
30
31
  module.exports = __toCommonJS(internal_exports);
31
32
 
32
33
  // src/openai-chat-language-model.ts
33
34
  var import_provider3 = require("@ai-sdk/provider");
34
- var import_provider_utils3 = require("@ai-sdk/provider-utils");
35
+ var import_provider_utils2 = require("@ai-sdk/provider-utils");
35
36
  var import_zod2 = require("zod");
36
37
 
37
38
  // src/convert-to-openai-chat-messages.ts
38
39
  var import_provider = require("@ai-sdk/provider");
39
- var import_provider_utils = require("@ai-sdk/provider-utils");
40
40
  function convertToOpenAIChatMessages({
41
41
  prompt,
42
42
  useLegacyFunctionCalling = false,
@@ -80,55 +80,65 @@ function convertToOpenAIChatMessages({
80
80
  messages.push({
81
81
  role: "user",
82
82
  content: content.map((part, index) => {
83
- var _a, _b, _c, _d;
83
+ var _a, _b, _c;
84
84
  switch (part.type) {
85
85
  case "text": {
86
86
  return { type: "text", text: part.text };
87
87
  }
88
- case "image": {
89
- return {
90
- type: "image_url",
91
- image_url: {
92
- url: part.image instanceof URL ? part.image.toString() : `data:${(_a = part.mimeType) != null ? _a : "image/jpeg"};base64,${(0, import_provider_utils.convertUint8ArrayToBase64)(part.image)}`,
93
- // OpenAI specific extension: image detail
94
- detail: (_c = (_b = part.providerOptions) == null ? void 0 : _b.openai) == null ? void 0 : _c.imageDetail
95
- }
96
- };
97
- }
98
88
  case "file": {
99
- if (part.data instanceof URL) {
100
- throw new import_provider.UnsupportedFunctionalityError({
101
- functionality: "'File content parts with URL data' functionality not supported."
102
- });
103
- }
104
- switch (part.mimeType) {
105
- case "audio/wav": {
106
- return {
107
- type: "input_audio",
108
- input_audio: { data: part.data, format: "wav" }
109
- };
110
- }
111
- case "audio/mp3":
112
- case "audio/mpeg": {
113
- return {
114
- type: "input_audio",
115
- input_audio: { data: part.data, format: "mp3" }
116
- };
89
+ if (part.mediaType.startsWith("image/")) {
90
+ const mediaType = part.mediaType === "image/*" ? "image/jpeg" : part.mediaType;
91
+ return {
92
+ type: "image_url",
93
+ image_url: {
94
+ url: part.data instanceof URL ? part.data.toString() : `data:${mediaType};base64,${part.data}`,
95
+ // OpenAI specific extension: image detail
96
+ detail: (_b = (_a = part.providerOptions) == null ? void 0 : _a.openai) == null ? void 0 : _b.imageDetail
97
+ }
98
+ };
99
+ } else if (part.mediaType.startsWith("audio/")) {
100
+ if (part.data instanceof URL) {
101
+ throw new import_provider.UnsupportedFunctionalityError({
102
+ functionality: "audio file parts with URLs"
103
+ });
117
104
  }
118
- case "application/pdf": {
119
- return {
120
- type: "file",
121
- file: {
122
- filename: (_d = part.filename) != null ? _d : `part-${index}.pdf`,
123
- file_data: `data:application/pdf;base64,${part.data}`
124
- }
125
- };
105
+ switch (part.mediaType) {
106
+ case "audio/wav": {
107
+ return {
108
+ type: "input_audio",
109
+ input_audio: { data: part.data, format: "wav" }
110
+ };
111
+ }
112
+ case "audio/mp3":
113
+ case "audio/mpeg": {
114
+ return {
115
+ type: "input_audio",
116
+ input_audio: { data: part.data, format: "mp3" }
117
+ };
118
+ }
119
+ default: {
120
+ throw new import_provider.UnsupportedFunctionalityError({
121
+ functionality: `audio content parts with media type ${part.mediaType}`
122
+ });
123
+ }
126
124
  }
127
- default: {
125
+ } else if (part.mediaType === "application/pdf") {
126
+ if (part.data instanceof URL) {
128
127
  throw new import_provider.UnsupportedFunctionalityError({
129
- functionality: `File content part type ${part.mimeType} in user messages`
128
+ functionality: "PDF file parts with URLs"
130
129
  });
131
130
  }
131
+ return {
132
+ type: "file",
133
+ file: {
134
+ filename: (_c = part.filename) != null ? _c : `part-${index}.pdf`,
135
+ file_data: `data:application/pdf;base64,${part.data}`
136
+ }
137
+ };
138
+ } else {
139
+ throw new import_provider.UnsupportedFunctionalityError({
140
+ functionality: `file part media type ${part.mediaType}`
141
+ });
132
142
  }
133
143
  }
134
144
  }
@@ -237,7 +247,7 @@ function mapOpenAIFinishReason(finishReason) {
237
247
 
238
248
  // src/openai-error.ts
239
249
  var import_zod = require("zod");
240
- var import_provider_utils2 = require("@ai-sdk/provider-utils");
250
+ var import_provider_utils = require("@ai-sdk/provider-utils");
241
251
  var openaiErrorDataSchema = import_zod.z.object({
242
252
  error: import_zod.z.object({
243
253
  message: import_zod.z.string(),
@@ -249,7 +259,7 @@ var openaiErrorDataSchema = import_zod.z.object({
249
259
  code: import_zod.z.union([import_zod.z.string(), import_zod.z.number()]).nullish()
250
260
  })
251
261
  });
252
- var openaiFailedResponseHandler = (0, import_provider_utils2.createJsonErrorResponseHandler)({
262
+ var openaiFailedResponseHandler = (0, import_provider_utils.createJsonErrorResponseHandler)({
253
263
  errorSchema: openaiErrorDataSchema,
254
264
  errorToMessage: (data) => data.error.message
255
265
  });
@@ -361,7 +371,7 @@ function prepareTools({
361
371
  default: {
362
372
  const _exhaustiveCheck = type;
363
373
  throw new import_provider2.UnsupportedFunctionalityError({
364
- functionality: `Unsupported tool choice type: ${_exhaustiveCheck}`
374
+ functionality: `tool choice type: ${_exhaustiveCheck}`
365
375
  });
366
376
  }
367
377
  }
@@ -568,15 +578,15 @@ var OpenAIChatLanguageModel = class {
568
578
  responseHeaders,
569
579
  value: response,
570
580
  rawValue: rawResponse
571
- } = await (0, import_provider_utils3.postJsonToApi)({
581
+ } = await (0, import_provider_utils2.postJsonToApi)({
572
582
  url: this.config.url({
573
583
  path: "/chat/completions",
574
584
  modelId: this.modelId
575
585
  }),
576
- headers: (0, import_provider_utils3.combineHeaders)(this.config.headers(), options.headers),
586
+ headers: (0, import_provider_utils2.combineHeaders)(this.config.headers(), options.headers),
577
587
  body,
578
588
  failedResponseHandler: openaiFailedResponseHandler,
579
- successfulResponseHandler: (0, import_provider_utils3.createJsonResponseHandler)(
589
+ successfulResponseHandler: (0, import_provider_utils2.createJsonResponseHandler)(
580
590
  openaiChatResponseSchema
581
591
  ),
582
592
  abortSignal: options.abortSignal,
@@ -604,7 +614,7 @@ var OpenAIChatLanguageModel = class {
604
614
  toolCalls: this.settings.useLegacyFunctionCalling && choice.message.function_call ? [
605
615
  {
606
616
  toolCallType: "function",
607
- toolCallId: (0, import_provider_utils3.generateId)(),
617
+ toolCallId: (0, import_provider_utils2.generateId)(),
608
618
  toolName: choice.message.function_call.name,
609
619
  args: choice.message.function_call.arguments
610
620
  }
@@ -612,7 +622,7 @@ var OpenAIChatLanguageModel = class {
612
622
  var _a2;
613
623
  return {
614
624
  toolCallType: "function",
615
- toolCallId: (_a2 = toolCall.id) != null ? _a2 : (0, import_provider_utils3.generateId)(),
625
+ toolCallId: (_a2 = toolCall.id) != null ? _a2 : (0, import_provider_utils2.generateId)(),
616
626
  toolName: toolCall.function.name,
617
627
  args: toolCall.function.arguments
618
628
  };
@@ -682,15 +692,15 @@ var OpenAIChatLanguageModel = class {
682
692
  // only include stream_options when in strict compatibility mode:
683
693
  stream_options: this.config.compatibility === "strict" ? { include_usage: true } : void 0
684
694
  };
685
- const { responseHeaders, value: response } = await (0, import_provider_utils3.postJsonToApi)({
695
+ const { responseHeaders, value: response } = await (0, import_provider_utils2.postJsonToApi)({
686
696
  url: this.config.url({
687
697
  path: "/chat/completions",
688
698
  modelId: this.modelId
689
699
  }),
690
- headers: (0, import_provider_utils3.combineHeaders)(this.config.headers(), options.headers),
700
+ headers: (0, import_provider_utils2.combineHeaders)(this.config.headers(), options.headers),
691
701
  body,
692
702
  failedResponseHandler: openaiFailedResponseHandler,
693
- successfulResponseHandler: (0, import_provider_utils3.createEventSourceResponseHandler)(
703
+ successfulResponseHandler: (0, import_provider_utils2.createEventSourceResponseHandler)(
694
704
  openaiChatChunkSchema
695
705
  ),
696
706
  abortSignal: options.abortSignal,
@@ -778,7 +788,7 @@ var OpenAIChatLanguageModel = class {
778
788
  const mappedToolCalls = useLegacyFunctionCalling && delta.function_call != null ? [
779
789
  {
780
790
  type: "function",
781
- id: (0, import_provider_utils3.generateId)(),
791
+ id: (0, import_provider_utils2.generateId)(),
782
792
  function: delta.function_call,
783
793
  index: 0
784
794
  }
@@ -825,11 +835,11 @@ var OpenAIChatLanguageModel = class {
825
835
  argsTextDelta: toolCall2.function.arguments
826
836
  });
827
837
  }
828
- if ((0, import_provider_utils3.isParsableJson)(toolCall2.function.arguments)) {
838
+ if ((0, import_provider_utils2.isParsableJson)(toolCall2.function.arguments)) {
829
839
  controller.enqueue({
830
840
  type: "tool-call",
831
841
  toolCallType: "function",
832
- toolCallId: (_e = toolCall2.id) != null ? _e : (0, import_provider_utils3.generateId)(),
842
+ toolCallId: (_e = toolCall2.id) != null ? _e : (0, import_provider_utils2.generateId)(),
833
843
  toolName: toolCall2.function.name,
834
844
  args: toolCall2.function.arguments
835
845
  });
@@ -852,11 +862,11 @@ var OpenAIChatLanguageModel = class {
852
862
  toolName: toolCall.function.name,
853
863
  argsTextDelta: (_i = toolCallDelta.function.arguments) != null ? _i : ""
854
864
  });
855
- if (((_j = toolCall.function) == null ? void 0 : _j.name) != null && ((_k = toolCall.function) == null ? void 0 : _k.arguments) != null && (0, import_provider_utils3.isParsableJson)(toolCall.function.arguments)) {
865
+ if (((_j = toolCall.function) == null ? void 0 : _j.name) != null && ((_k = toolCall.function) == null ? void 0 : _k.arguments) != null && (0, import_provider_utils2.isParsableJson)(toolCall.function.arguments)) {
856
866
  controller.enqueue({
857
867
  type: "tool-call",
858
868
  toolCallType: "function",
859
- toolCallId: (_l = toolCall.id) != null ? _l : (0, import_provider_utils3.generateId)(),
869
+ toolCallId: (_l = toolCall.id) != null ? _l : (0, import_provider_utils2.generateId)(),
860
870
  toolName: toolCall.function.name,
861
871
  args: toolCall.function.arguments
862
872
  });
@@ -1026,7 +1036,7 @@ var reasoningModels = {
1026
1036
  };
1027
1037
 
1028
1038
  // src/openai-completion-language-model.ts
1029
- var import_provider_utils4 = require("@ai-sdk/provider-utils");
1039
+ var import_provider_utils3 = require("@ai-sdk/provider-utils");
1030
1040
  var import_zod3 = require("zod");
1031
1041
 
1032
1042
  // src/convert-to-openai-completion-prompt.ts
@@ -1061,13 +1071,8 @@ function convertToOpenAICompletionPrompt({
1061
1071
  case "text": {
1062
1072
  return part.text;
1063
1073
  }
1064
- case "image": {
1065
- throw new import_provider4.UnsupportedFunctionalityError({
1066
- functionality: "images"
1067
- });
1068
- }
1069
1074
  }
1070
- }).join("");
1075
+ }).filter(Boolean).join("");
1071
1076
  text += `${user}:
1072
1077
  ${userMessage}
1073
1078
 
@@ -1204,15 +1209,15 @@ var OpenAICompletionLanguageModel = class {
1204
1209
  responseHeaders,
1205
1210
  value: response,
1206
1211
  rawValue: rawResponse
1207
- } = await (0, import_provider_utils4.postJsonToApi)({
1212
+ } = await (0, import_provider_utils3.postJsonToApi)({
1208
1213
  url: this.config.url({
1209
1214
  path: "/completions",
1210
1215
  modelId: this.modelId
1211
1216
  }),
1212
- headers: (0, import_provider_utils4.combineHeaders)(this.config.headers(), options.headers),
1217
+ headers: (0, import_provider_utils3.combineHeaders)(this.config.headers(), options.headers),
1213
1218
  body: args,
1214
1219
  failedResponseHandler: openaiFailedResponseHandler,
1215
- successfulResponseHandler: (0, import_provider_utils4.createJsonResponseHandler)(
1220
+ successfulResponseHandler: (0, import_provider_utils3.createJsonResponseHandler)(
1216
1221
  openaiCompletionResponseSchema
1217
1222
  ),
1218
1223
  abortSignal: options.abortSignal,
@@ -1243,15 +1248,15 @@ var OpenAICompletionLanguageModel = class {
1243
1248
  // only include stream_options when in strict compatibility mode:
1244
1249
  stream_options: this.config.compatibility === "strict" ? { include_usage: true } : void 0
1245
1250
  };
1246
- const { responseHeaders, value: response } = await (0, import_provider_utils4.postJsonToApi)({
1251
+ const { responseHeaders, value: response } = await (0, import_provider_utils3.postJsonToApi)({
1247
1252
  url: this.config.url({
1248
1253
  path: "/completions",
1249
1254
  modelId: this.modelId
1250
1255
  }),
1251
- headers: (0, import_provider_utils4.combineHeaders)(this.config.headers(), options.headers),
1256
+ headers: (0, import_provider_utils3.combineHeaders)(this.config.headers(), options.headers),
1252
1257
  body,
1253
1258
  failedResponseHandler: openaiFailedResponseHandler,
1254
- successfulResponseHandler: (0, import_provider_utils4.createEventSourceResponseHandler)(
1259
+ successfulResponseHandler: (0, import_provider_utils3.createEventSourceResponseHandler)(
1255
1260
  openaiCompletionChunkSchema
1256
1261
  ),
1257
1262
  abortSignal: options.abortSignal,
@@ -1375,7 +1380,7 @@ var openaiCompletionChunkSchema = import_zod3.z.union([
1375
1380
 
1376
1381
  // src/openai-embedding-model.ts
1377
1382
  var import_provider5 = require("@ai-sdk/provider");
1378
- var import_provider_utils5 = require("@ai-sdk/provider-utils");
1383
+ var import_provider_utils4 = require("@ai-sdk/provider-utils");
1379
1384
  var import_zod4 = require("zod");
1380
1385
  var OpenAIEmbeddingModel = class {
1381
1386
  constructor(modelId, settings, config) {
@@ -1408,12 +1413,12 @@ var OpenAIEmbeddingModel = class {
1408
1413
  values
1409
1414
  });
1410
1415
  }
1411
- const { responseHeaders, value: response } = await (0, import_provider_utils5.postJsonToApi)({
1416
+ const { responseHeaders, value: response } = await (0, import_provider_utils4.postJsonToApi)({
1412
1417
  url: this.config.url({
1413
1418
  path: "/embeddings",
1414
1419
  modelId: this.modelId
1415
1420
  }),
1416
- headers: (0, import_provider_utils5.combineHeaders)(this.config.headers(), headers),
1421
+ headers: (0, import_provider_utils4.combineHeaders)(this.config.headers(), headers),
1417
1422
  body: {
1418
1423
  model: this.modelId,
1419
1424
  input: values,
@@ -1422,7 +1427,7 @@ var OpenAIEmbeddingModel = class {
1422
1427
  user: this.settings.user
1423
1428
  },
1424
1429
  failedResponseHandler: openaiFailedResponseHandler,
1425
- successfulResponseHandler: (0, import_provider_utils5.createJsonResponseHandler)(
1430
+ successfulResponseHandler: (0, import_provider_utils4.createJsonResponseHandler)(
1426
1431
  openaiTextEmbeddingResponseSchema
1427
1432
  ),
1428
1433
  abortSignal,
@@ -1441,7 +1446,7 @@ var openaiTextEmbeddingResponseSchema = import_zod4.z.object({
1441
1446
  });
1442
1447
 
1443
1448
  // src/openai-image-model.ts
1444
- var import_provider_utils6 = require("@ai-sdk/provider-utils");
1449
+ var import_provider_utils5 = require("@ai-sdk/provider-utils");
1445
1450
  var import_zod5 = require("zod");
1446
1451
 
1447
1452
  // src/openai-image-settings.ts
@@ -1488,12 +1493,12 @@ var OpenAIImageModel = class {
1488
1493
  warnings.push({ type: "unsupported-setting", setting: "seed" });
1489
1494
  }
1490
1495
  const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
1491
- const { value: response, responseHeaders } = await (0, import_provider_utils6.postJsonToApi)({
1496
+ const { value: response, responseHeaders } = await (0, import_provider_utils5.postJsonToApi)({
1492
1497
  url: this.config.url({
1493
1498
  path: "/images/generations",
1494
1499
  modelId: this.modelId
1495
1500
  }),
1496
- headers: (0, import_provider_utils6.combineHeaders)(this.config.headers(), headers),
1501
+ headers: (0, import_provider_utils5.combineHeaders)(this.config.headers(), headers),
1497
1502
  body: {
1498
1503
  model: this.modelId,
1499
1504
  prompt,
@@ -1503,7 +1508,7 @@ var OpenAIImageModel = class {
1503
1508
  response_format: "b64_json"
1504
1509
  },
1505
1510
  failedResponseHandler: openaiFailedResponseHandler,
1506
- successfulResponseHandler: (0, import_provider_utils6.createJsonResponseHandler)(
1511
+ successfulResponseHandler: (0, import_provider_utils5.createJsonResponseHandler)(
1507
1512
  openaiImageResponseSchema
1508
1513
  ),
1509
1514
  abortSignal,
@@ -1524,13 +1529,186 @@ var openaiImageResponseSchema = import_zod5.z.object({
1524
1529
  data: import_zod5.z.array(import_zod5.z.object({ b64_json: import_zod5.z.string() }))
1525
1530
  });
1526
1531
 
1527
- // src/responses/openai-responses-language-model.ts
1528
- var import_provider_utils8 = require("@ai-sdk/provider-utils");
1532
+ // src/openai-transcription-model.ts
1533
+ var import_provider_utils6 = require("@ai-sdk/provider-utils");
1529
1534
  var import_zod6 = require("zod");
1535
+ var OpenAIProviderOptionsSchema = import_zod6.z.object({
1536
+ include: import_zod6.z.array(import_zod6.z.string()).optional().describe(
1537
+ "Additional information to include in the transcription response."
1538
+ ),
1539
+ language: import_zod6.z.string().optional().describe("The language of the input audio in ISO-639-1 format."),
1540
+ prompt: import_zod6.z.string().optional().describe(
1541
+ "An optional text to guide the model's style or continue a previous audio segment."
1542
+ ),
1543
+ temperature: import_zod6.z.number().min(0).max(1).optional().default(0).describe("The sampling temperature, between 0 and 1."),
1544
+ timestampGranularities: import_zod6.z.array(import_zod6.z.enum(["word", "segment"])).optional().default(["segment"]).describe(
1545
+ "The timestamp granularities to populate for this transcription."
1546
+ )
1547
+ });
1548
+ var languageMap = {
1549
+ afrikaans: "af",
1550
+ arabic: "ar",
1551
+ armenian: "hy",
1552
+ azerbaijani: "az",
1553
+ belarusian: "be",
1554
+ bosnian: "bs",
1555
+ bulgarian: "bg",
1556
+ catalan: "ca",
1557
+ chinese: "zh",
1558
+ croatian: "hr",
1559
+ czech: "cs",
1560
+ danish: "da",
1561
+ dutch: "nl",
1562
+ english: "en",
1563
+ estonian: "et",
1564
+ finnish: "fi",
1565
+ french: "fr",
1566
+ galician: "gl",
1567
+ german: "de",
1568
+ greek: "el",
1569
+ hebrew: "he",
1570
+ hindi: "hi",
1571
+ hungarian: "hu",
1572
+ icelandic: "is",
1573
+ indonesian: "id",
1574
+ italian: "it",
1575
+ japanese: "ja",
1576
+ kannada: "kn",
1577
+ kazakh: "kk",
1578
+ korean: "ko",
1579
+ latvian: "lv",
1580
+ lithuanian: "lt",
1581
+ macedonian: "mk",
1582
+ malay: "ms",
1583
+ marathi: "mr",
1584
+ maori: "mi",
1585
+ nepali: "ne",
1586
+ norwegian: "no",
1587
+ persian: "fa",
1588
+ polish: "pl",
1589
+ portuguese: "pt",
1590
+ romanian: "ro",
1591
+ russian: "ru",
1592
+ serbian: "sr",
1593
+ slovak: "sk",
1594
+ slovenian: "sl",
1595
+ spanish: "es",
1596
+ swahili: "sw",
1597
+ swedish: "sv",
1598
+ tagalog: "tl",
1599
+ tamil: "ta",
1600
+ thai: "th",
1601
+ turkish: "tr",
1602
+ ukrainian: "uk",
1603
+ urdu: "ur",
1604
+ vietnamese: "vi",
1605
+ welsh: "cy"
1606
+ };
1607
+ var OpenAITranscriptionModel = class {
1608
+ constructor(modelId, config) {
1609
+ this.modelId = modelId;
1610
+ this.config = config;
1611
+ this.specificationVersion = "v1";
1612
+ }
1613
+ get provider() {
1614
+ return this.config.provider;
1615
+ }
1616
+ getArgs({
1617
+ audio,
1618
+ mediaType,
1619
+ providerOptions
1620
+ }) {
1621
+ const warnings = [];
1622
+ const openAIOptions = (0, import_provider_utils6.parseProviderOptions)({
1623
+ provider: "openai",
1624
+ providerOptions,
1625
+ schema: OpenAIProviderOptionsSchema
1626
+ });
1627
+ const formData = new FormData();
1628
+ const blob = audio instanceof Uint8Array ? new Blob([audio]) : new Blob([(0, import_provider_utils6.convertBase64ToUint8Array)(audio)]);
1629
+ formData.append("model", this.modelId);
1630
+ formData.append("file", new File([blob], "audio", { type: mediaType }));
1631
+ if (openAIOptions) {
1632
+ const transcriptionModelOptions = {
1633
+ include: openAIOptions.include,
1634
+ language: openAIOptions.language,
1635
+ prompt: openAIOptions.prompt,
1636
+ temperature: openAIOptions.temperature,
1637
+ timestamp_granularities: openAIOptions.timestampGranularities
1638
+ };
1639
+ for (const key in transcriptionModelOptions) {
1640
+ const value = transcriptionModelOptions[key];
1641
+ if (value !== void 0) {
1642
+ formData.append(key, value);
1643
+ }
1644
+ }
1645
+ }
1646
+ return {
1647
+ formData,
1648
+ warnings
1649
+ };
1650
+ }
1651
+ async doGenerate(options) {
1652
+ var _a, _b, _c, _d, _e, _f;
1653
+ const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
1654
+ const { formData, warnings } = this.getArgs(options);
1655
+ const {
1656
+ value: response,
1657
+ responseHeaders,
1658
+ rawValue: rawResponse
1659
+ } = await (0, import_provider_utils6.postFormDataToApi)({
1660
+ url: this.config.url({
1661
+ path: "/audio/transcriptions",
1662
+ modelId: this.modelId
1663
+ }),
1664
+ headers: (0, import_provider_utils6.combineHeaders)(this.config.headers(), options.headers),
1665
+ formData,
1666
+ failedResponseHandler: openaiFailedResponseHandler,
1667
+ successfulResponseHandler: (0, import_provider_utils6.createJsonResponseHandler)(
1668
+ openaiTranscriptionResponseSchema
1669
+ ),
1670
+ abortSignal: options.abortSignal,
1671
+ fetch: this.config.fetch
1672
+ });
1673
+ const language = response.language != null && response.language in languageMap ? languageMap[response.language] : void 0;
1674
+ return {
1675
+ text: response.text,
1676
+ segments: (_e = (_d = response.words) == null ? void 0 : _d.map((word) => ({
1677
+ text: word.word,
1678
+ startSecond: word.start,
1679
+ endSecond: word.end
1680
+ }))) != null ? _e : [],
1681
+ language,
1682
+ durationInSeconds: (_f = response.duration) != null ? _f : void 0,
1683
+ warnings,
1684
+ response: {
1685
+ timestamp: currentDate,
1686
+ modelId: this.modelId,
1687
+ headers: responseHeaders,
1688
+ body: rawResponse
1689
+ }
1690
+ };
1691
+ }
1692
+ };
1693
+ var openaiTranscriptionResponseSchema = import_zod6.z.object({
1694
+ text: import_zod6.z.string(),
1695
+ language: import_zod6.z.string().nullish(),
1696
+ duration: import_zod6.z.number().nullish(),
1697
+ words: import_zod6.z.array(
1698
+ import_zod6.z.object({
1699
+ word: import_zod6.z.string(),
1700
+ start: import_zod6.z.number(),
1701
+ end: import_zod6.z.number()
1702
+ })
1703
+ ).nullish()
1704
+ });
1705
+
1706
+ // src/responses/openai-responses-language-model.ts
1707
+ var import_provider_utils7 = require("@ai-sdk/provider-utils");
1708
+ var import_zod7 = require("zod");
1530
1709
 
1531
1710
  // src/responses/convert-to-openai-responses-messages.ts
1532
1711
  var import_provider6 = require("@ai-sdk/provider");
1533
- var import_provider_utils7 = require("@ai-sdk/provider-utils");
1534
1712
  function convertToOpenAIResponsesMessages({
1535
1713
  prompt,
1536
1714
  systemMessageMode
@@ -1569,38 +1747,35 @@ function convertToOpenAIResponsesMessages({
1569
1747
  messages.push({
1570
1748
  role: "user",
1571
1749
  content: content.map((part, index) => {
1572
- var _a, _b, _c, _d;
1750
+ var _a, _b, _c;
1573
1751
  switch (part.type) {
1574
1752
  case "text": {
1575
1753
  return { type: "input_text", text: part.text };
1576
1754
  }
1577
- case "image": {
1578
- return {
1579
- type: "input_image",
1580
- image_url: part.image instanceof URL ? part.image.toString() : `data:${(_a = part.mimeType) != null ? _a : "image/jpeg"};base64,${(0, import_provider_utils7.convertUint8ArrayToBase64)(part.image)}`,
1581
- // OpenAI specific extension: image detail
1582
- detail: (_c = (_b = part.providerOptions) == null ? void 0 : _b.openai) == null ? void 0 : _c.imageDetail
1583
- };
1584
- }
1585
1755
  case "file": {
1586
- if (part.data instanceof URL) {
1587
- throw new import_provider6.UnsupportedFunctionalityError({
1588
- functionality: "File URLs in user messages"
1589
- });
1590
- }
1591
- switch (part.mimeType) {
1592
- case "application/pdf": {
1593
- return {
1594
- type: "input_file",
1595
- filename: (_d = part.filename) != null ? _d : `part-${index}.pdf`,
1596
- file_data: `data:application/pdf;base64,${part.data}`
1597
- };
1598
- }
1599
- default: {
1756
+ if (part.mediaType.startsWith("image/")) {
1757
+ const mediaType = part.mediaType === "image/*" ? "image/jpeg" : part.mediaType;
1758
+ return {
1759
+ type: "input_image",
1760
+ image_url: part.data instanceof URL ? part.data.toString() : `data:${mediaType};base64,${part.data}`,
1761
+ // OpenAI specific extension: image detail
1762
+ detail: (_b = (_a = part.providerOptions) == null ? void 0 : _a.openai) == null ? void 0 : _b.imageDetail
1763
+ };
1764
+ } else if (part.mediaType === "application/pdf") {
1765
+ if (part.data instanceof URL) {
1600
1766
  throw new import_provider6.UnsupportedFunctionalityError({
1601
- functionality: "Only PDF files are supported in user messages"
1767
+ functionality: "PDF file parts with URLs"
1602
1768
  });
1603
1769
  }
1770
+ return {
1771
+ type: "input_file",
1772
+ filename: (_c = part.filename) != null ? _c : `part-${index}.pdf`,
1773
+ file_data: `data:application/pdf;base64,${part.data}`
1774
+ };
1775
+ } else {
1776
+ throw new import_provider6.UnsupportedFunctionalityError({
1777
+ functionality: `file part media type ${part.mediaType}`
1778
+ });
1604
1779
  }
1605
1780
  }
1606
1781
  }
@@ -1729,7 +1904,7 @@ function prepareResponsesTools({
1729
1904
  default: {
1730
1905
  const _exhaustiveCheck = type;
1731
1906
  throw new import_provider7.UnsupportedFunctionalityError({
1732
- functionality: `Unsupported tool choice type: ${_exhaustiveCheck}`
1907
+ functionality: `tool choice type: ${_exhaustiveCheck}`
1733
1908
  });
1734
1909
  }
1735
1910
  }
@@ -1790,7 +1965,7 @@ var OpenAIResponsesLanguageModel = class {
1790
1965
  systemMessageMode: modelConfig.systemMessageMode
1791
1966
  });
1792
1967
  warnings.push(...messageWarnings);
1793
- const openaiOptions = (0, import_provider_utils8.parseProviderOptions)({
1968
+ const openaiOptions = (0, import_provider_utils7.parseProviderOptions)({
1794
1969
  provider: "openai",
1795
1970
  providerOptions,
1796
1971
  schema: openaiResponsesProviderOptionsSchema
@@ -1871,58 +2046,58 @@ var OpenAIResponsesLanguageModel = class {
1871
2046
  responseHeaders,
1872
2047
  value: response,
1873
2048
  rawValue: rawResponse
1874
- } = await (0, import_provider_utils8.postJsonToApi)({
2049
+ } = await (0, import_provider_utils7.postJsonToApi)({
1875
2050
  url: this.config.url({
1876
2051
  path: "/responses",
1877
2052
  modelId: this.modelId
1878
2053
  }),
1879
- headers: (0, import_provider_utils8.combineHeaders)(this.config.headers(), options.headers),
2054
+ headers: (0, import_provider_utils7.combineHeaders)(this.config.headers(), options.headers),
1880
2055
  body,
1881
2056
  failedResponseHandler: openaiFailedResponseHandler,
1882
- successfulResponseHandler: (0, import_provider_utils8.createJsonResponseHandler)(
1883
- import_zod6.z.object({
1884
- id: import_zod6.z.string(),
1885
- created_at: import_zod6.z.number(),
1886
- model: import_zod6.z.string(),
1887
- output: import_zod6.z.array(
1888
- import_zod6.z.discriminatedUnion("type", [
1889
- import_zod6.z.object({
1890
- type: import_zod6.z.literal("message"),
1891
- role: import_zod6.z.literal("assistant"),
1892
- content: import_zod6.z.array(
1893
- import_zod6.z.object({
1894
- type: import_zod6.z.literal("output_text"),
1895
- text: import_zod6.z.string(),
1896
- annotations: import_zod6.z.array(
1897
- import_zod6.z.object({
1898
- type: import_zod6.z.literal("url_citation"),
1899
- start_index: import_zod6.z.number(),
1900
- end_index: import_zod6.z.number(),
1901
- url: import_zod6.z.string(),
1902
- title: import_zod6.z.string()
2057
+ successfulResponseHandler: (0, import_provider_utils7.createJsonResponseHandler)(
2058
+ import_zod7.z.object({
2059
+ id: import_zod7.z.string(),
2060
+ created_at: import_zod7.z.number(),
2061
+ model: import_zod7.z.string(),
2062
+ output: import_zod7.z.array(
2063
+ import_zod7.z.discriminatedUnion("type", [
2064
+ import_zod7.z.object({
2065
+ type: import_zod7.z.literal("message"),
2066
+ role: import_zod7.z.literal("assistant"),
2067
+ content: import_zod7.z.array(
2068
+ import_zod7.z.object({
2069
+ type: import_zod7.z.literal("output_text"),
2070
+ text: import_zod7.z.string(),
2071
+ annotations: import_zod7.z.array(
2072
+ import_zod7.z.object({
2073
+ type: import_zod7.z.literal("url_citation"),
2074
+ start_index: import_zod7.z.number(),
2075
+ end_index: import_zod7.z.number(),
2076
+ url: import_zod7.z.string(),
2077
+ title: import_zod7.z.string()
1903
2078
  })
1904
2079
  )
1905
2080
  })
1906
2081
  )
1907
2082
  }),
1908
- import_zod6.z.object({
1909
- type: import_zod6.z.literal("function_call"),
1910
- call_id: import_zod6.z.string(),
1911
- name: import_zod6.z.string(),
1912
- arguments: import_zod6.z.string()
2083
+ import_zod7.z.object({
2084
+ type: import_zod7.z.literal("function_call"),
2085
+ call_id: import_zod7.z.string(),
2086
+ name: import_zod7.z.string(),
2087
+ arguments: import_zod7.z.string()
1913
2088
  }),
1914
- import_zod6.z.object({
1915
- type: import_zod6.z.literal("web_search_call")
2089
+ import_zod7.z.object({
2090
+ type: import_zod7.z.literal("web_search_call")
1916
2091
  }),
1917
- import_zod6.z.object({
1918
- type: import_zod6.z.literal("computer_call")
2092
+ import_zod7.z.object({
2093
+ type: import_zod7.z.literal("computer_call")
1919
2094
  }),
1920
- import_zod6.z.object({
1921
- type: import_zod6.z.literal("reasoning")
2095
+ import_zod7.z.object({
2096
+ type: import_zod7.z.literal("reasoning")
1922
2097
  })
1923
2098
  ])
1924
2099
  ),
1925
- incomplete_details: import_zod6.z.object({ reason: import_zod6.z.string() }).nullable(),
2100
+ incomplete_details: import_zod7.z.object({ reason: import_zod7.z.string() }).nullable(),
1926
2101
  usage: usageSchema
1927
2102
  })
1928
2103
  ),
@@ -1943,7 +2118,7 @@ var OpenAIResponsesLanguageModel = class {
1943
2118
  var _a2, _b2, _c2;
1944
2119
  return {
1945
2120
  sourceType: "url",
1946
- id: (_c2 = (_b2 = (_a2 = this.config).generateId) == null ? void 0 : _b2.call(_a2)) != null ? _c2 : (0, import_provider_utils8.generateId)(),
2121
+ id: (_c2 = (_b2 = (_a2 = this.config).generateId) == null ? void 0 : _b2.call(_a2)) != null ? _c2 : (0, import_provider_utils7.generateId)(),
1947
2122
  url: annotation.url,
1948
2123
  title: annotation.title
1949
2124
  };
@@ -1986,18 +2161,18 @@ var OpenAIResponsesLanguageModel = class {
1986
2161
  }
1987
2162
  async doStream(options) {
1988
2163
  const { args: body, warnings } = this.getArgs(options);
1989
- const { responseHeaders, value: response } = await (0, import_provider_utils8.postJsonToApi)({
2164
+ const { responseHeaders, value: response } = await (0, import_provider_utils7.postJsonToApi)({
1990
2165
  url: this.config.url({
1991
2166
  path: "/responses",
1992
2167
  modelId: this.modelId
1993
2168
  }),
1994
- headers: (0, import_provider_utils8.combineHeaders)(this.config.headers(), options.headers),
2169
+ headers: (0, import_provider_utils7.combineHeaders)(this.config.headers(), options.headers),
1995
2170
  body: {
1996
2171
  ...body,
1997
2172
  stream: true
1998
2173
  },
1999
2174
  failedResponseHandler: openaiFailedResponseHandler,
2000
- successfulResponseHandler: (0, import_provider_utils8.createEventSourceResponseHandler)(
2175
+ successfulResponseHandler: (0, import_provider_utils7.createEventSourceResponseHandler)(
2001
2176
  openaiResponsesChunkSchema
2002
2177
  ),
2003
2178
  abortSignal: options.abortSignal,
@@ -2085,7 +2260,7 @@ var OpenAIResponsesLanguageModel = class {
2085
2260
  type: "source",
2086
2261
  source: {
2087
2262
  sourceType: "url",
2088
- id: (_h = (_g = (_f = self.config).generateId) == null ? void 0 : _g.call(_f)) != null ? _h : (0, import_provider_utils8.generateId)(),
2263
+ id: (_h = (_g = (_f = self.config).generateId) == null ? void 0 : _g.call(_f)) != null ? _h : (0, import_provider_utils7.generateId)(),
2089
2264
  url: value.annotation.url,
2090
2265
  title: value.annotation.title
2091
2266
  }
@@ -2120,79 +2295,79 @@ var OpenAIResponsesLanguageModel = class {
2120
2295
  };
2121
2296
  }
2122
2297
  };
2123
- var usageSchema = import_zod6.z.object({
2124
- input_tokens: import_zod6.z.number(),
2125
- input_tokens_details: import_zod6.z.object({ cached_tokens: import_zod6.z.number().nullish() }).nullish(),
2126
- output_tokens: import_zod6.z.number(),
2127
- output_tokens_details: import_zod6.z.object({ reasoning_tokens: import_zod6.z.number().nullish() }).nullish()
2298
+ var usageSchema = import_zod7.z.object({
2299
+ input_tokens: import_zod7.z.number(),
2300
+ input_tokens_details: import_zod7.z.object({ cached_tokens: import_zod7.z.number().nullish() }).nullish(),
2301
+ output_tokens: import_zod7.z.number(),
2302
+ output_tokens_details: import_zod7.z.object({ reasoning_tokens: import_zod7.z.number().nullish() }).nullish()
2128
2303
  });
2129
- var textDeltaChunkSchema = import_zod6.z.object({
2130
- type: import_zod6.z.literal("response.output_text.delta"),
2131
- delta: import_zod6.z.string()
2304
+ var textDeltaChunkSchema = import_zod7.z.object({
2305
+ type: import_zod7.z.literal("response.output_text.delta"),
2306
+ delta: import_zod7.z.string()
2132
2307
  });
2133
- var responseFinishedChunkSchema = import_zod6.z.object({
2134
- type: import_zod6.z.enum(["response.completed", "response.incomplete"]),
2135
- response: import_zod6.z.object({
2136
- incomplete_details: import_zod6.z.object({ reason: import_zod6.z.string() }).nullish(),
2308
+ var responseFinishedChunkSchema = import_zod7.z.object({
2309
+ type: import_zod7.z.enum(["response.completed", "response.incomplete"]),
2310
+ response: import_zod7.z.object({
2311
+ incomplete_details: import_zod7.z.object({ reason: import_zod7.z.string() }).nullish(),
2137
2312
  usage: usageSchema
2138
2313
  })
2139
2314
  });
2140
- var responseCreatedChunkSchema = import_zod6.z.object({
2141
- type: import_zod6.z.literal("response.created"),
2142
- response: import_zod6.z.object({
2143
- id: import_zod6.z.string(),
2144
- created_at: import_zod6.z.number(),
2145
- model: import_zod6.z.string()
2315
+ var responseCreatedChunkSchema = import_zod7.z.object({
2316
+ type: import_zod7.z.literal("response.created"),
2317
+ response: import_zod7.z.object({
2318
+ id: import_zod7.z.string(),
2319
+ created_at: import_zod7.z.number(),
2320
+ model: import_zod7.z.string()
2146
2321
  })
2147
2322
  });
2148
- var responseOutputItemDoneSchema = import_zod6.z.object({
2149
- type: import_zod6.z.literal("response.output_item.done"),
2150
- output_index: import_zod6.z.number(),
2151
- item: import_zod6.z.discriminatedUnion("type", [
2152
- import_zod6.z.object({
2153
- type: import_zod6.z.literal("message")
2323
+ var responseOutputItemDoneSchema = import_zod7.z.object({
2324
+ type: import_zod7.z.literal("response.output_item.done"),
2325
+ output_index: import_zod7.z.number(),
2326
+ item: import_zod7.z.discriminatedUnion("type", [
2327
+ import_zod7.z.object({
2328
+ type: import_zod7.z.literal("message")
2154
2329
  }),
2155
- import_zod6.z.object({
2156
- type: import_zod6.z.literal("function_call"),
2157
- id: import_zod6.z.string(),
2158
- call_id: import_zod6.z.string(),
2159
- name: import_zod6.z.string(),
2160
- arguments: import_zod6.z.string(),
2161
- status: import_zod6.z.literal("completed")
2330
+ import_zod7.z.object({
2331
+ type: import_zod7.z.literal("function_call"),
2332
+ id: import_zod7.z.string(),
2333
+ call_id: import_zod7.z.string(),
2334
+ name: import_zod7.z.string(),
2335
+ arguments: import_zod7.z.string(),
2336
+ status: import_zod7.z.literal("completed")
2162
2337
  })
2163
2338
  ])
2164
2339
  });
2165
- var responseFunctionCallArgumentsDeltaSchema = import_zod6.z.object({
2166
- type: import_zod6.z.literal("response.function_call_arguments.delta"),
2167
- item_id: import_zod6.z.string(),
2168
- output_index: import_zod6.z.number(),
2169
- delta: import_zod6.z.string()
2340
+ var responseFunctionCallArgumentsDeltaSchema = import_zod7.z.object({
2341
+ type: import_zod7.z.literal("response.function_call_arguments.delta"),
2342
+ item_id: import_zod7.z.string(),
2343
+ output_index: import_zod7.z.number(),
2344
+ delta: import_zod7.z.string()
2170
2345
  });
2171
- var responseOutputItemAddedSchema = import_zod6.z.object({
2172
- type: import_zod6.z.literal("response.output_item.added"),
2173
- output_index: import_zod6.z.number(),
2174
- item: import_zod6.z.discriminatedUnion("type", [
2175
- import_zod6.z.object({
2176
- type: import_zod6.z.literal("message")
2346
+ var responseOutputItemAddedSchema = import_zod7.z.object({
2347
+ type: import_zod7.z.literal("response.output_item.added"),
2348
+ output_index: import_zod7.z.number(),
2349
+ item: import_zod7.z.discriminatedUnion("type", [
2350
+ import_zod7.z.object({
2351
+ type: import_zod7.z.literal("message")
2177
2352
  }),
2178
- import_zod6.z.object({
2179
- type: import_zod6.z.literal("function_call"),
2180
- id: import_zod6.z.string(),
2181
- call_id: import_zod6.z.string(),
2182
- name: import_zod6.z.string(),
2183
- arguments: import_zod6.z.string()
2353
+ import_zod7.z.object({
2354
+ type: import_zod7.z.literal("function_call"),
2355
+ id: import_zod7.z.string(),
2356
+ call_id: import_zod7.z.string(),
2357
+ name: import_zod7.z.string(),
2358
+ arguments: import_zod7.z.string()
2184
2359
  })
2185
2360
  ])
2186
2361
  });
2187
- var responseAnnotationAddedSchema = import_zod6.z.object({
2188
- type: import_zod6.z.literal("response.output_text.annotation.added"),
2189
- annotation: import_zod6.z.object({
2190
- type: import_zod6.z.literal("url_citation"),
2191
- url: import_zod6.z.string(),
2192
- title: import_zod6.z.string()
2362
+ var responseAnnotationAddedSchema = import_zod7.z.object({
2363
+ type: import_zod7.z.literal("response.output_text.annotation.added"),
2364
+ annotation: import_zod7.z.object({
2365
+ type: import_zod7.z.literal("url_citation"),
2366
+ url: import_zod7.z.string(),
2367
+ title: import_zod7.z.string()
2193
2368
  })
2194
2369
  });
2195
- var openaiResponsesChunkSchema = import_zod6.z.union([
2370
+ var openaiResponsesChunkSchema = import_zod7.z.union([
2196
2371
  textDeltaChunkSchema,
2197
2372
  responseFinishedChunkSchema,
2198
2373
  responseCreatedChunkSchema,
@@ -2200,7 +2375,7 @@ var openaiResponsesChunkSchema = import_zod6.z.union([
2200
2375
  responseFunctionCallArgumentsDeltaSchema,
2201
2376
  responseOutputItemAddedSchema,
2202
2377
  responseAnnotationAddedSchema,
2203
- import_zod6.z.object({ type: import_zod6.z.string() }).passthrough()
2378
+ import_zod7.z.object({ type: import_zod7.z.string() }).passthrough()
2204
2379
  // fallback for unknown chunks
2205
2380
  ]);
2206
2381
  function isTextDeltaChunk(chunk) {
@@ -2245,15 +2420,15 @@ function getResponsesModelConfig(modelId) {
2245
2420
  requiredAutoTruncation: false
2246
2421
  };
2247
2422
  }
2248
- var openaiResponsesProviderOptionsSchema = import_zod6.z.object({
2249
- metadata: import_zod6.z.any().nullish(),
2250
- parallelToolCalls: import_zod6.z.boolean().nullish(),
2251
- previousResponseId: import_zod6.z.string().nullish(),
2252
- store: import_zod6.z.boolean().nullish(),
2253
- user: import_zod6.z.string().nullish(),
2254
- reasoningEffort: import_zod6.z.string().nullish(),
2255
- strictSchemas: import_zod6.z.boolean().nullish(),
2256
- instructions: import_zod6.z.string().nullish()
2423
+ var openaiResponsesProviderOptionsSchema = import_zod7.z.object({
2424
+ metadata: import_zod7.z.any().nullish(),
2425
+ parallelToolCalls: import_zod7.z.boolean().nullish(),
2426
+ previousResponseId: import_zod7.z.string().nullish(),
2427
+ store: import_zod7.z.boolean().nullish(),
2428
+ user: import_zod7.z.string().nullish(),
2429
+ reasoningEffort: import_zod7.z.string().nullish(),
2430
+ strictSchemas: import_zod7.z.boolean().nullish(),
2431
+ instructions: import_zod7.z.string().nullish()
2257
2432
  });
2258
2433
  // Annotate the CommonJS export names for ESM import in node:
2259
2434
  0 && (module.exports = {
@@ -2262,6 +2437,7 @@ var openaiResponsesProviderOptionsSchema = import_zod6.z.object({
2262
2437
  OpenAIEmbeddingModel,
2263
2438
  OpenAIImageModel,
2264
2439
  OpenAIResponsesLanguageModel,
2440
+ OpenAITranscriptionModel,
2265
2441
  modelMaxImagesPerCall
2266
2442
  });
2267
2443
  //# sourceMappingURL=index.js.map