@runpod/ai-sdk-provider 1.2.0 → 1.3.0

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.
package/dist/index.mjs CHANGED
@@ -5,7 +5,7 @@ import {
5
5
  } from "@ai-sdk/openai-compatible";
6
6
  import {
7
7
  loadApiKey,
8
- withoutTrailingSlash as withoutTrailingSlash3
8
+ withoutTrailingSlash as withoutTrailingSlash4
9
9
  } from "@ai-sdk/provider-utils";
10
10
 
11
11
  // src/runpod-image-model.ts
@@ -54,6 +54,10 @@ var runpodTranscriptionFailedResponseHandler = createJsonErrorResponseHandler({
54
54
  errorSchema: runpodImageErrorSchema,
55
55
  errorToMessage: extractErrorMessage
56
56
  });
57
+ var runpodVideoFailedResponseHandler = createJsonErrorResponseHandler({
58
+ errorSchema: runpodImageErrorSchema,
59
+ errorToMessage: extractErrorMessage
60
+ });
57
61
 
58
62
  // src/runpod-image-model.ts
59
63
  var SUPPORTED_ASPECT_RATIOS = {
@@ -86,6 +90,50 @@ var SUPPORTED_SIZES = /* @__PURE__ */ new Set([
86
90
  "1024*768",
87
91
  "768*1024"
88
92
  ]);
93
+ var Z_IMAGE_TURBO_SUPPORTED_SIZES = /* @__PURE__ */ new Set([
94
+ "1328*1328",
95
+ // 1:1
96
+ "1472*1140",
97
+ // 4:3
98
+ "1140*1472",
99
+ // 3:4
100
+ "512*512",
101
+ "768*768",
102
+ "1024*1024",
103
+ "1280*1280",
104
+ "1536*1536",
105
+ "512*768",
106
+ "768*512",
107
+ "1024*768",
108
+ "768*1024",
109
+ "768*432",
110
+ "1024*576",
111
+ "1280*720",
112
+ "1536*864",
113
+ "432*768",
114
+ "576*1024",
115
+ "720*1280",
116
+ "864*1536"
117
+ ]);
118
+ var Z_IMAGE_TURBO_ASPECT_RATIOS = {
119
+ "1:1": "1328*1328",
120
+ "4:3": "1472*1140",
121
+ "3:4": "1140*1472",
122
+ "3:2": "768*512",
123
+ "2:3": "512*768",
124
+ "16:9": "1280*720",
125
+ "9:16": "720*1280"
126
+ };
127
+ var MODEL_SUPPORTED_SIZES = {
128
+ "tongyi-mai/z-image-turbo": Z_IMAGE_TURBO_SUPPORTED_SIZES,
129
+ "z-image-turbo": Z_IMAGE_TURBO_SUPPORTED_SIZES
130
+ // alias, not advertised
131
+ };
132
+ var MODEL_SUPPORTED_ASPECT_RATIOS = {
133
+ "tongyi-mai/z-image-turbo": Z_IMAGE_TURBO_ASPECT_RATIOS,
134
+ "z-image-turbo": Z_IMAGE_TURBO_ASPECT_RATIOS
135
+ // alias, not advertised
136
+ };
89
137
  var WAN_ASPECT_RATIOS = {
90
138
  "1:1": "1280*1280",
91
139
  // 1,638,400 pixels
@@ -191,23 +239,27 @@ var RunpodImageModel = class {
191
239
  }
192
240
  } else if (size) {
193
241
  const runpodSizeCandidate = size.replace("x", "*");
194
- if (!SUPPORTED_SIZES.has(runpodSizeCandidate)) {
242
+ const supportedSizes = MODEL_SUPPORTED_SIZES[this.modelId] ?? SUPPORTED_SIZES;
243
+ if (!supportedSizes.has(runpodSizeCandidate)) {
195
244
  throw new InvalidArgumentError({
196
245
  argument: "size",
197
246
  message: `Size ${size} is not supported by Runpod. Supported sizes: ${Array.from(
198
- SUPPORTED_SIZES
247
+ supportedSizes
199
248
  ).map((s) => s.replace("*", "x")).join(", ")}`
200
249
  });
201
250
  }
202
251
  runpodSize = runpodSizeCandidate;
203
252
  } else if (aspectRatio) {
204
- if (!SUPPORTED_ASPECT_RATIOS[aspectRatio]) {
253
+ const supportedAspectRatios = MODEL_SUPPORTED_ASPECT_RATIOS[this.modelId] ?? SUPPORTED_ASPECT_RATIOS;
254
+ if (!supportedAspectRatios[aspectRatio]) {
205
255
  throw new InvalidArgumentError({
206
256
  argument: "aspectRatio",
207
- message: `Aspect ratio ${aspectRatio} is not supported by Runpod. Supported aspect ratios: ${Object.keys(SUPPORTED_ASPECT_RATIOS).join(", ")}`
257
+ message: `Aspect ratio ${aspectRatio} is not supported by Runpod. Supported aspect ratios: ${Object.keys(
258
+ supportedAspectRatios
259
+ ).join(", ")}`
208
260
  });
209
261
  }
210
- runpodSize = SUPPORTED_ASPECT_RATIOS[aspectRatio];
262
+ runpodSize = supportedAspectRatios[aspectRatio];
211
263
  } else {
212
264
  runpodSize = "1328*1328";
213
265
  }
@@ -510,6 +562,25 @@ var RunpodImageModel = class {
510
562
  }
511
563
  return qwenEdit2511Payload;
512
564
  }
565
+ const isZImageTurbo = this.modelId === "tongyi-mai/z-image-turbo" || this.modelId === "z-image-turbo";
566
+ if (isZImageTurbo) {
567
+ const zImageTurboPayload = {
568
+ prompt,
569
+ size: runpodSize,
570
+ seed: seed ?? -1,
571
+ output_format: runpodOptions?.output_format ?? "png",
572
+ enable_safety_checker: runpodOptions?.enable_safety_checker ?? true,
573
+ ...runpodOptions
574
+ };
575
+ if (standardizedImages && standardizedImages.length > 0) {
576
+ if (standardizedImages.length === 1) {
577
+ zImageTurboPayload.image = standardizedImages[0];
578
+ } else {
579
+ zImageTurboPayload.images = standardizedImages;
580
+ }
581
+ }
582
+ return zImageTurboPayload;
583
+ }
513
584
  const isWanModel = this.modelId.includes("wan-2");
514
585
  if (isWanModel) {
515
586
  return {
@@ -928,6 +999,215 @@ var RunpodTranscriptionModel = class {
928
999
  }
929
1000
  };
930
1001
 
1002
+ // src/runpod-video-model.ts
1003
+ import {
1004
+ withoutTrailingSlash as withoutTrailingSlash3,
1005
+ createJsonResponseHandler as createJsonResponseHandler3,
1006
+ getFromApi as getFromApi3,
1007
+ postJsonToApi as postJsonToApi3
1008
+ } from "@ai-sdk/provider-utils";
1009
+ import { z as z4 } from "zod";
1010
+ function isRecord3(value) {
1011
+ return typeof value === "object" && value !== null;
1012
+ }
1013
+ var runpodVideoJobResponseSchema = z4.object({
1014
+ id: z4.string(),
1015
+ status: z4.string().optional()
1016
+ });
1017
+ var runpodVideoStatusSchema = z4.object({
1018
+ id: z4.string().optional(),
1019
+ status: z4.string(),
1020
+ output: z4.unknown().optional(),
1021
+ error: z4.string().optional()
1022
+ });
1023
+ var RunpodVideoModel = class {
1024
+ constructor(modelId, config) {
1025
+ this.modelId = modelId;
1026
+ this.config = config;
1027
+ this.specificationVersion = "v3";
1028
+ this.maxVideosPerCall = 1;
1029
+ }
1030
+ get provider() {
1031
+ return this.config.provider;
1032
+ }
1033
+ getRunpodRunUrl() {
1034
+ const baseURL = withoutTrailingSlash3(this.config.baseURL) ?? this.config.baseURL;
1035
+ if (baseURL.endsWith("/run") || baseURL.endsWith("/runsync")) {
1036
+ return baseURL;
1037
+ }
1038
+ return `${baseURL}/run`;
1039
+ }
1040
+ async doGenerate(options) {
1041
+ const currentDate = this.config._internal?.currentDate?.() ?? /* @__PURE__ */ new Date();
1042
+ const warnings = [];
1043
+ if (options.n > 1) {
1044
+ warnings.push({
1045
+ type: "unsupported",
1046
+ feature: "n > 1",
1047
+ details: "Runpod video models only support generating 1 video per call. Only 1 video will be generated."
1048
+ });
1049
+ }
1050
+ const { providerOptions, abortSignal, headers } = options;
1051
+ const runpodOptions = this.extractRunpodOptions(providerOptions);
1052
+ const input = this.buildInputPayload(options, runpodOptions);
1053
+ const requestBody = { input };
1054
+ const url = this.getRunpodRunUrl();
1055
+ const effectiveBaseURL = withoutTrailingSlash3(this.config.baseURL) ?? this.config.baseURL;
1056
+ const { value: response, responseHeaders } = await postJsonToApi3({
1057
+ url,
1058
+ headers: {
1059
+ ...this.config.headers(),
1060
+ ...headers
1061
+ },
1062
+ body: requestBody,
1063
+ failedResponseHandler: runpodVideoFailedResponseHandler,
1064
+ successfulResponseHandler: createJsonResponseHandler3(
1065
+ runpodVideoJobResponseSchema
1066
+ ),
1067
+ abortSignal,
1068
+ fetch: this.config.fetch
1069
+ });
1070
+ const typedResponse = response;
1071
+ const jobId = typedResponse.id;
1072
+ if (!jobId) {
1073
+ throw new Error("Runpod video response did not include a job id.");
1074
+ }
1075
+ const pollOptions = {
1076
+ maxAttempts: runpodOptions.maxPollAttempts ?? 120,
1077
+ pollIntervalMillis: runpodOptions.pollIntervalMillis ?? 5e3
1078
+ };
1079
+ const result = await this.pollForCompletion(
1080
+ jobId,
1081
+ abortSignal,
1082
+ pollOptions,
1083
+ effectiveBaseURL
1084
+ );
1085
+ const videoUrl = this.extractVideoUrl(result.output);
1086
+ const providerMetadata = {
1087
+ runpod: {
1088
+ jobId
1089
+ }
1090
+ };
1091
+ return {
1092
+ videos: [{ type: "url", url: videoUrl, mediaType: "video/mp4" }],
1093
+ warnings,
1094
+ response: {
1095
+ timestamp: currentDate,
1096
+ modelId: this.modelId,
1097
+ headers: responseHeaders
1098
+ },
1099
+ providerMetadata
1100
+ };
1101
+ }
1102
+ buildInputPayload(options, runpodOptions) {
1103
+ const apiOptions = Object.fromEntries(
1104
+ Object.entries(runpodOptions).filter(
1105
+ ([key]) => key !== "maxPollAttempts" && key !== "pollIntervalMillis"
1106
+ )
1107
+ );
1108
+ const input = {
1109
+ ...apiOptions
1110
+ };
1111
+ if (options.prompt) {
1112
+ input.prompt = options.prompt;
1113
+ }
1114
+ if (options.duration !== void 0) {
1115
+ input.duration = options.duration;
1116
+ }
1117
+ if (options.fps !== void 0) {
1118
+ input.fps = options.fps;
1119
+ }
1120
+ if (options.seed !== void 0) {
1121
+ input.seed = options.seed;
1122
+ }
1123
+ if (options.resolution) {
1124
+ input.size = options.resolution.replace("x", "*");
1125
+ } else if (options.aspectRatio) {
1126
+ input.aspect_ratio = options.aspectRatio;
1127
+ }
1128
+ if (options.image) {
1129
+ input.image = this.convertFileToRunpodFormat(options.image);
1130
+ }
1131
+ return input;
1132
+ }
1133
+ convertFileToRunpodFormat(file) {
1134
+ if (file.type === "url") {
1135
+ return file.url;
1136
+ }
1137
+ const mediaType = file.mediaType;
1138
+ const data = file.data;
1139
+ if (typeof data === "string") {
1140
+ return `data:${mediaType};base64,${data}`;
1141
+ }
1142
+ const base64 = this.uint8ArrayToBase64(data);
1143
+ return `data:${mediaType};base64,${base64}`;
1144
+ }
1145
+ uint8ArrayToBase64(data) {
1146
+ if (typeof Buffer !== "undefined") {
1147
+ return Buffer.from(data).toString("base64");
1148
+ }
1149
+ let binary = "";
1150
+ for (let i = 0; i < data.length; i++) {
1151
+ binary += String.fromCharCode(data[i]);
1152
+ }
1153
+ return btoa(binary);
1154
+ }
1155
+ extractRunpodOptions(providerOptions) {
1156
+ if (!providerOptions) return {};
1157
+ const runpod2 = providerOptions.runpod;
1158
+ if (isRecord3(runpod2)) {
1159
+ return runpod2;
1160
+ }
1161
+ return {};
1162
+ }
1163
+ extractVideoUrl(output) {
1164
+ if (isRecord3(output)) {
1165
+ if (typeof output.video_url === "string") return output.video_url;
1166
+ if (typeof output.result === "string") return output.result;
1167
+ if (typeof output.url === "string") return output.url;
1168
+ }
1169
+ if (typeof output === "string") {
1170
+ return output;
1171
+ }
1172
+ throw new Error(
1173
+ `Runpod video generation completed but no video URL was found in the output: ${JSON.stringify(output)}`
1174
+ );
1175
+ }
1176
+ async pollForCompletion(jobId, abortSignal, pollOptions, effectiveBaseURL) {
1177
+ const maxAttempts = pollOptions?.maxAttempts ?? 120;
1178
+ const pollInterval = pollOptions?.pollIntervalMillis ?? 5e3;
1179
+ const baseURL = effectiveBaseURL ?? this.config.baseURL;
1180
+ for (let attempt = 0; attempt < maxAttempts; attempt++) {
1181
+ if (abortSignal?.aborted) {
1182
+ throw new Error("Video generation was aborted");
1183
+ }
1184
+ const { value: statusResponse } = await getFromApi3({
1185
+ url: `${baseURL}/status/${jobId}`,
1186
+ headers: this.config.headers(),
1187
+ successfulResponseHandler: createJsonResponseHandler3(
1188
+ runpodVideoStatusSchema
1189
+ ),
1190
+ failedResponseHandler: runpodVideoFailedResponseHandler,
1191
+ abortSignal,
1192
+ fetch: this.config.fetch
1193
+ });
1194
+ const typedStatusResponse = statusResponse;
1195
+ if (typedStatusResponse.status === "COMPLETED") {
1196
+ return typedStatusResponse;
1197
+ }
1198
+ if (typedStatusResponse.status === "FAILED") {
1199
+ throw new Error(
1200
+ `Video generation failed: ${typedStatusResponse.error || "Unknown error"}`
1201
+ );
1202
+ }
1203
+ await new Promise((resolve) => setTimeout(resolve, pollInterval));
1204
+ }
1205
+ throw new Error(
1206
+ `Video generation timed out after ${maxAttempts} attempts (${maxAttempts * pollInterval / 1e3}s)`
1207
+ );
1208
+ }
1209
+ };
1210
+
931
1211
  // src/runpod-provider.ts
932
1212
  var MODEL_ID_TO_ENDPOINT_URL = {
933
1213
  "qwen/qwen3-32b-awq": "https://api.runpod.ai/v2/qwen3-32b-awq/openai/v1",
@@ -951,6 +1231,10 @@ var IMAGE_MODEL_ID_TO_ENDPOINT_URL = {
951
1231
  "black-forest-labs/flux-1-dev": "https://api.runpod.ai/v2/black-forest-labs-flux-1-dev",
952
1232
  // Alibaba Wan 2.6 (t2i)
953
1233
  "alibaba/wan-2.6": "https://api.runpod.ai/v2/wan-2-6-t2i",
1234
+ // Tongyi Z-Image Turbo (t2i)
1235
+ "tongyi-mai/z-image-turbo": "https://api.runpod.ai/v2/z-image-turbo",
1236
+ "z-image-turbo": "https://api.runpod.ai/v2/z-image-turbo",
1237
+ // alias, not advertised
954
1238
  // Nano Banana (edit only)
955
1239
  "google/nano-banana-edit": "https://api.runpod.ai/v2/nano-banana-edit",
956
1240
  "nano-banana-edit": "https://api.runpod.ai/v2/nano-banana-edit",
@@ -967,6 +1251,23 @@ var SPEECH_MODEL_ID_TO_ENDPOINT_URL = {
967
1251
  var TRANSCRIPTION_MODEL_ID_TO_ENDPOINT_URL = {
968
1252
  "pruna/whisper-v3-large": "https://api.runpod.ai/v2/whisper-v3-large"
969
1253
  };
1254
+ var VIDEO_MODEL_ID_TO_ENDPOINT_URL = {
1255
+ "pruna/p-video": "https://api.runpod.ai/v2/p-video",
1256
+ "vidu/q3-t2v": "https://api.runpod.ai/v2/vidu-q3-t2v",
1257
+ "vidu/q3-i2v": "https://api.runpod.ai/v2/vidu-q3-i2v",
1258
+ "kwaivgi/kling-v2.6-std-motion-control": "https://api.runpod.ai/v2/kling-v2-6-std-motion-control",
1259
+ "kwaivgi/kling-video-o1-r2v": "https://api.runpod.ai/v2/kling-video-o1-r2v",
1260
+ "kwaivgi/kling-v2.1-i2v-pro": "https://api.runpod.ai/v2/kling-v2-1-i2v-pro",
1261
+ "alibaba/wan-2.6-t2v": "https://api.runpod.ai/v2/wan-2-6-t2v",
1262
+ "alibaba/wan-2.6-i2v": "https://api.runpod.ai/v2/wan-2-6-i2v",
1263
+ "alibaba/wan-2.5": "https://api.runpod.ai/v2/wan-2-5",
1264
+ "alibaba/wan-2.2-t2v-720-lora": "https://api.runpod.ai/v2/wan-2-2-t2v-720-lora",
1265
+ "alibaba/wan-2.2-i2v-720": "https://api.runpod.ai/v2/wan-2-2-i2v-720",
1266
+ "alibaba/wan-2.1-i2v-720": "https://api.runpod.ai/v2/wan-2-1-i2v-720",
1267
+ "bytedance/seedance-v1.5-pro-i2v": "https://api.runpod.ai/v2/seedance-v1-5-pro-i2v",
1268
+ "openai/sora-2-pro-i2v": "https://api.runpod.ai/v2/sora-2-pro-i2v",
1269
+ "openai/sora-2-i2v": "https://api.runpod.ai/v2/sora-2-i2v"
1270
+ };
970
1271
  var MODEL_ID_TO_OPENAI_NAME = {
971
1272
  "qwen/qwen3-32b-awq": "Qwen/Qwen3-32B-AWQ",
972
1273
  "deepcogito/cogito-671b-v2.1-fp8": "deepcogito/cogito-671b-v2.1-FP8",
@@ -1025,7 +1326,7 @@ function createRunpod(options = {}) {
1025
1326
  }
1026
1327
  return {
1027
1328
  provider: `runpod.${modelType}`,
1028
- url: ({ path }) => `${withoutTrailingSlash3(baseURL)}${path}`,
1329
+ url: ({ path }) => `${withoutTrailingSlash4(baseURL)}${path}`,
1029
1330
  headers: getHeaders,
1030
1331
  fetch: runpodFetch
1031
1332
  };
@@ -1082,6 +1383,18 @@ function createRunpod(options = {}) {
1082
1383
  fetch: runpodFetch
1083
1384
  });
1084
1385
  };
1386
+ const createVideoModel = (modelId) => {
1387
+ const endpointIdFromConsole = parseRunpodConsoleEndpointId(modelId);
1388
+ const normalizedModelId = endpointIdFromConsole ?? modelId;
1389
+ const mappedBaseURL = VIDEO_MODEL_ID_TO_ENDPOINT_URL[normalizedModelId];
1390
+ const baseURL = mappedBaseURL ?? (normalizedModelId.startsWith("http") ? normalizedModelId : `https://api.runpod.ai/v2/${normalizedModelId}`);
1391
+ return new RunpodVideoModel(normalizedModelId, {
1392
+ provider: "runpod.video",
1393
+ baseURL,
1394
+ headers: getHeaders,
1395
+ fetch: options.fetch
1396
+ });
1397
+ };
1085
1398
  const provider = (modelId) => createChatModel(modelId);
1086
1399
  provider.completionModel = createCompletionModel;
1087
1400
  provider.languageModel = createChatModel;
@@ -1092,6 +1405,8 @@ function createRunpod(options = {}) {
1092
1405
  provider.speech = createSpeechModel;
1093
1406
  provider.transcriptionModel = createTranscriptionModel;
1094
1407
  provider.transcription = createTranscriptionModel;
1408
+ provider.videoModel = createVideoModel;
1409
+ provider.video = createVideoModel;
1095
1410
  return provider;
1096
1411
  }
1097
1412
  var runpod = createRunpod();