@ai-sdk/xai 3.0.75 → 3.0.77

package/CHANGELOG.md

@@ -1,5 +1,19 @@
 # @ai-sdk/xai
 
+## 3.0.77
+
+### Patch Changes
+
+- bacdae4: feat(provider/xai): add video extension and reference-to-video (R2V) support
+
+## 3.0.76
+
+### Patch Changes
+
+- Updated dependencies [0469aed]
+  - @ai-sdk/provider-utils@4.0.22
+  - @ai-sdk/openai-compatible@2.0.38
+
 ## 3.0.75
 
 ### Patch Changes

package/dist/index.d.mts

@@ -97,13 +97,76 @@ type XaiImageModelOptions = z.infer<typeof xaiImageModelOptions>;
 
 type XaiVideoModelId = 'grok-imagine-video' | (string & {});
 
-type XaiVideoModelOptions = {
+declare const resolutionSchema: z.ZodEnum<{
+    "480p": "480p";
+    "720p": "720p";
+}>;
+type XaiVideoResolution = z.infer<typeof resolutionSchema>;
+interface XaiVideoSharedOptions {
     pollIntervalMs?: number | null;
     pollTimeoutMs?: number | null;
-    resolution?: '480p' | '720p' | null;
-    videoUrl?: string | null;
-    [key: string]: unknown;
-};
+    resolution?: XaiVideoResolution | null;
+}
+interface XaiVideoEditModeOptions extends XaiVideoSharedOptions {
+    /**
+     * Select edit-video mode explicitly for best autocomplete and narrowing.
+     */
+    mode: 'edit-video';
+    /** Source video URL to edit. */
+    videoUrl: string;
+}
+interface XaiVideoExtendModeOptions extends XaiVideoSharedOptions {
+    /**
+     * Select extend-video mode explicitly for best autocomplete and narrowing.
+     */
+    mode: 'extend-video';
+    /** Source video URL to extend from its last frame. */
+    videoUrl: string;
+}
+interface XaiVideoReferenceToVideoOptions extends XaiVideoSharedOptions {
+    /**
+     * Select reference-to-video mode explicitly for best autocomplete and narrowing.
+     */
+    mode: 'reference-to-video';
+    /** Reference image URLs (1-7) for R2V generation. */
+    referenceImageUrls: string[];
+}
+interface XaiVideoGenerationOptions extends XaiVideoSharedOptions {
+    mode?: undefined;
+    videoUrl?: undefined;
+    referenceImageUrls?: undefined;
+}
+interface XaiLegacyEditVideoOptions extends XaiVideoSharedOptions {
+    /**
+     * Legacy backward-compatible shape: omitting `mode` while providing
+     * `videoUrl` behaves like edit-video.
+     */
+    mode?: undefined;
+    videoUrl: string;
+}
+interface XaiLegacyReferenceToVideoOptions extends XaiVideoSharedOptions {
+    /**
+     * Legacy backward-compatible shape: omitting `mode` while providing
+     * `referenceImageUrls` behaves like reference-to-video.
+     */
+    mode?: undefined;
+    referenceImageUrls: string[];
+}
+/**
+ * Provider options for xAI video generation.
+ *
+ * Use the `mode` option to select the operation:
+ *
+ * - `'edit-video'`         + `videoUrl`           -- video editing   (`POST /v1/videos/edits`)
+ * - `'extend-video'`       + `videoUrl`           -- video extension (`POST /v1/videos/extensions`)
+ * - `'reference-to-video'` + `referenceImageUrls` -- R2V generation  (`POST /v1/videos/generations`)
+ * - no `mode`                                     -- standard generation from text prompts or image input
+ *
+ * Runtime remains backward compatible with legacy auto-detected provider
+ * options, but the public TypeScript type is intentionally explicit so editors
+ * can suggest valid modes and flag invalid field combinations.
+ */
+type XaiVideoModelOptions = XaiVideoGenerationOptions | XaiVideoEditModeOptions | XaiVideoExtendModeOptions | XaiVideoReferenceToVideoOptions | XaiLegacyEditVideoOptions | XaiLegacyReferenceToVideoOptions;
 
 type XaiImageModelId = 'grok-imagine-image' | 'grok-imagine-image-pro' | (string & {});
 

package/dist/index.d.ts

@@ -97,13 +97,76 @@ type XaiImageModelOptions = z.infer<typeof xaiImageModelOptions>;
 
 type XaiVideoModelId = 'grok-imagine-video' | (string & {});
 
-type XaiVideoModelOptions = {
+declare const resolutionSchema: z.ZodEnum<{
+    "480p": "480p";
+    "720p": "720p";
+}>;
+type XaiVideoResolution = z.infer<typeof resolutionSchema>;
+interface XaiVideoSharedOptions {
     pollIntervalMs?: number | null;
     pollTimeoutMs?: number | null;
-    resolution?: '480p' | '720p' | null;
-    videoUrl?: string | null;
-    [key: string]: unknown;
-};
+    resolution?: XaiVideoResolution | null;
+}
+interface XaiVideoEditModeOptions extends XaiVideoSharedOptions {
+    /**
+     * Select edit-video mode explicitly for best autocomplete and narrowing.
+     */
+    mode: 'edit-video';
+    /** Source video URL to edit. */
+    videoUrl: string;
+}
+interface XaiVideoExtendModeOptions extends XaiVideoSharedOptions {
+    /**
+     * Select extend-video mode explicitly for best autocomplete and narrowing.
+     */
+    mode: 'extend-video';
+    /** Source video URL to extend from its last frame. */
+    videoUrl: string;
+}
+interface XaiVideoReferenceToVideoOptions extends XaiVideoSharedOptions {
+    /**
+     * Select reference-to-video mode explicitly for best autocomplete and narrowing.
+     */
+    mode: 'reference-to-video';
+    /** Reference image URLs (1-7) for R2V generation. */
+    referenceImageUrls: string[];
+}
+interface XaiVideoGenerationOptions extends XaiVideoSharedOptions {
+    mode?: undefined;
+    videoUrl?: undefined;
+    referenceImageUrls?: undefined;
+}
+interface XaiLegacyEditVideoOptions extends XaiVideoSharedOptions {
+    /**
+     * Legacy backward-compatible shape: omitting `mode` while providing
+     * `videoUrl` behaves like edit-video.
+     */
+    mode?: undefined;
+    videoUrl: string;
+}
+interface XaiLegacyReferenceToVideoOptions extends XaiVideoSharedOptions {
+    /**
+     * Legacy backward-compatible shape: omitting `mode` while providing
+     * `referenceImageUrls` behaves like reference-to-video.
+     */
+    mode?: undefined;
+    referenceImageUrls: string[];
+}
+/**
+ * Provider options for xAI video generation.
+ *
+ * Use the `mode` option to select the operation:
+ *
+ * - `'edit-video'`         + `videoUrl`           -- video editing   (`POST /v1/videos/edits`)
+ * - `'extend-video'`       + `videoUrl`           -- video extension (`POST /v1/videos/extensions`)
+ * - `'reference-to-video'` + `referenceImageUrls` -- R2V generation  (`POST /v1/videos/generations`)
+ * - no `mode`                                     -- standard generation from text prompts or image input
+ *
+ * Runtime remains backward compatible with legacy auto-detected provider
+ * options, but the public TypeScript type is intentionally explicit so editors
+ * can suggest valid modes and flag invalid field combinations.
+ */
+type XaiVideoModelOptions = XaiVideoGenerationOptions | XaiVideoEditModeOptions | XaiVideoExtendModeOptions | XaiVideoReferenceToVideoOptions | XaiLegacyEditVideoOptions | XaiLegacyReferenceToVideoOptions;
 
 type XaiImageModelId = 'grok-imagine-image' | 'grok-imagine-image-pro' | (string & {});
 

package/dist/index.js

@@ -2719,7 +2719,7 @@ var xaiTools = {
 };
 
 // src/version.ts
-var VERSION = true ? "3.0.75" : "0.0.0-test";
+var VERSION = true ? "3.0.77" : "0.0.0-test";
 
 // src/xai-video-model.ts
 var import_provider6 = require("@ai-sdk/provider");
@@ -2729,15 +2729,52 @@ var import_v416 = require("zod/v4");
 // src/xai-video-options.ts
 var import_provider_utils15 = require("@ai-sdk/provider-utils");
 var import_v415 = require("zod/v4");
+var nonEmptyStringSchema = import_v415.z.string().min(1);
+var resolutionSchema = import_v415.z.enum(["480p", "720p"]);
+var modeSchema = import_v415.z.enum(["edit-video", "extend-video", "reference-to-video"]);
+var baseFields = {
+  pollIntervalMs: import_v415.z.number().positive().nullish(),
+  pollTimeoutMs: import_v415.z.number().positive().nullish(),
+  resolution: resolutionSchema.nullish()
+};
+var editVideoSchema = import_v415.z.object({
+  ...baseFields,
+  mode: import_v415.z.literal("edit-video"),
+  videoUrl: nonEmptyStringSchema,
+  referenceImageUrls: import_v415.z.undefined().optional()
+});
+var extendVideoSchema = import_v415.z.object({
+  ...baseFields,
+  mode: import_v415.z.literal("extend-video"),
+  videoUrl: nonEmptyStringSchema,
+  referenceImageUrls: import_v415.z.undefined().optional()
+});
+var referenceToVideoSchema = import_v415.z.object({
+  ...baseFields,
+  mode: import_v415.z.literal("reference-to-video"),
+  referenceImageUrls: import_v415.z.array(nonEmptyStringSchema).min(1).max(7),
+  videoUrl: import_v415.z.undefined().optional()
+});
+var autoDetectSchema = import_v415.z.object({
+  ...baseFields,
+  mode: import_v415.z.undefined().optional(),
+  videoUrl: nonEmptyStringSchema.optional(),
+  referenceImageUrls: import_v415.z.array(nonEmptyStringSchema).min(1).max(7).optional()
+});
+var xaiVideoModelOptions = import_v415.z.union([
+  editVideoSchema,
+  extendVideoSchema,
+  referenceToVideoSchema,
+  autoDetectSchema
+]);
+var runtimeSchema = import_v415.z.object({
+  mode: modeSchema.optional(),
+  videoUrl: nonEmptyStringSchema.optional(),
+  referenceImageUrls: import_v415.z.array(nonEmptyStringSchema).min(1).max(7).optional(),
+  ...baseFields
+}).passthrough();
 var xaiVideoModelOptionsSchema = (0, import_provider_utils15.lazySchema)(
-  () => (0, import_provider_utils15.zodSchema)(
-    import_v415.z.object({
-      pollIntervalMs: import_v415.z.number().positive().nullish(),
-      pollTimeoutMs: import_v415.z.number().positive().nullish(),
-      resolution: import_v415.z.enum(["480p", "720p"]).nullish(),
-      videoUrl: import_v415.z.string().nullish()
-    }).passthrough()
-  )
+  () => (0, import_provider_utils15.zodSchema)(runtimeSchema)
 );
 
 // src/xai-video-model.ts
@@ -2746,6 +2783,18 @@ var RESOLUTION_MAP = {
   "854x480": "480p",
   "640x480": "480p"
 };
+function resolveVideoMode(options) {
+  if ((options == null ? void 0 : options.mode) != null) {
+    return options.mode;
+  }
+  if ((options == null ? void 0 : options.videoUrl) != null) {
+    return "edit-video";
+  }
+  if ((options == null ? void 0 : options.referenceImageUrls) != null && options.referenceImageUrls.length > 0) {
+    return "reference-to-video";
+  }
+  return void 0;
+}
 var XaiVideoModel = class {
   constructor(modelId, config) {
     this.modelId = modelId;
@@ -2765,7 +2814,10 @@ var XaiVideoModel = class {
       providerOptions: options.providerOptions,
       schema: xaiVideoModelOptionsSchema
     });
-    const isEdit = (xaiOptions == null ? void 0 : xaiOptions.videoUrl) != null;
+    const effectiveMode = resolveVideoMode(xaiOptions);
+    const isEdit = effectiveMode === "edit-video";
+    const isExtension = effectiveMode === "extend-video";
+    const hasReferenceImages = effectiveMode === "reference-to-video";
     if (options.fps != null) {
       warnings.push({
         type: "unsupported",
@@ -2808,19 +2860,36 @@ var XaiVideoModel = class {
         details: "xAI video editing does not support custom resolution."
       });
     }
+    if (isExtension && options.aspectRatio != null) {
+      warnings.push({
+        type: "unsupported",
+        feature: "aspectRatio",
+        details: "xAI video extension does not support custom aspect ratio."
+      });
+    }
+    if (isExtension && ((xaiOptions == null ? void 0 : xaiOptions.resolution) != null || options.resolution != null)) {
+      warnings.push({
+        type: "unsupported",
+        feature: "resolution",
+        details: "xAI video extension does not support custom resolution."
+      });
+    }
     const body = {
       model: this.modelId,
       prompt: options.prompt
     };
-    if (!isEdit && options.duration != null) {
+    const allowDuration = !isEdit;
+    const allowAspectRatio = !isEdit && !isExtension;
+    const allowResolution = !isEdit && !isExtension;
+    if (allowDuration && options.duration != null) {
       body.duration = options.duration;
     }
-    if (!isEdit && options.aspectRatio != null) {
+    if (allowAspectRatio && options.aspectRatio != null) {
       body.aspect_ratio = options.aspectRatio;
     }
-    if (!isEdit && (xaiOptions == null ? void 0 : xaiOptions.resolution) != null) {
+    if (allowResolution && (xaiOptions == null ? void 0 : xaiOptions.resolution) != null) {
       body.resolution = xaiOptions.resolution;
-    } else if (!isEdit && options.resolution != null) {
+    } else if (allowResolution && options.resolution != null) {
       const mapped = RESOLUTION_MAP[options.resolution];
       if (mapped != null) {
         body.resolution = mapped;
@@ -2832,7 +2901,10 @@ var XaiVideoModel = class {
         });
       }
     }
-    if ((xaiOptions == null ? void 0 : xaiOptions.videoUrl) != null) {
+    if (isEdit) {
+      body.video = { url: xaiOptions.videoUrl };
+    }
+    if (isExtension) {
       body.video = { url: xaiOptions.videoUrl };
     }
     if (options.image != null) {
@@ -2845,21 +2917,36 @@ var XaiVideoModel = class {
         };
       }
     }
+    if (hasReferenceImages) {
+      body.reference_images = xaiOptions.referenceImageUrls.map((url) => ({
+        url
+      }));
+    }
     if (xaiOptions != null) {
       for (const [key, value] of Object.entries(xaiOptions)) {
         if (![
+          "mode",
           "pollIntervalMs",
           "pollTimeoutMs",
           "resolution",
-          "videoUrl"
+          "videoUrl",
+          "referenceImageUrls"
         ].includes(key)) {
           body[key] = value;
         }
       }
     }
     const baseURL = (_d = this.config.baseURL) != null ? _d : "https://api.x.ai/v1";
+    let endpoint;
+    if (isEdit) {
+      endpoint = `${baseURL}/videos/edits`;
+    } else if (isExtension) {
+      endpoint = `${baseURL}/videos/extensions`;
+    } else {
+      endpoint = `${baseURL}/videos/generations`;
+    }
     const { value: createResponse } = await (0, import_provider_utils16.postJsonToApi)({
-      url: `${baseURL}/videos/${isEdit ? "edits" : "generations"}`,
+      url: endpoint,
       headers: (0, import_provider_utils16.combineHeaders)(this.config.headers(), options.headers),
       body,
       failedResponseHandler: xaiFailedResponseHandler,
@@ -2931,7 +3018,8 @@ var XaiVideoModel = class {
               requestId,
               videoUrl: statusResponse.video.url,
               ...statusResponse.video.duration != null ? { duration: statusResponse.video.duration } : {},
-              ...((_j = statusResponse.usage) == null ? void 0 : _j.cost_in_usd_ticks) != null ? { costInUsdTicks: statusResponse.usage.cost_in_usd_ticks } : {}
+              ...((_j = statusResponse.usage) == null ? void 0 : _j.cost_in_usd_ticks) != null ? { costInUsdTicks: statusResponse.usage.cost_in_usd_ticks } : {},
+              ...statusResponse.progress != null ? { progress: statusResponse.progress } : {}
             }
           }
         };
@@ -2942,6 +3030,12 @@ var XaiVideoModel = class {
           message: "Video generation request expired."
         });
       }
+      if (statusResponse.status === "failed") {
+        throw new import_provider6.AISDKError({
+          name: "XAI_VIDEO_GENERATION_FAILED",
+          message: "Video generation failed."
+        });
+      }
     }
   }
 };
@@ -2958,6 +3052,11 @@ var xaiVideoStatusResponseSchema = import_v416.z.object({
   model: import_v416.z.string().nullish(),
   usage: import_v416.z.object({
     cost_in_usd_ticks: import_v416.z.number().nullish()
+  }).nullish(),
+  progress: import_v416.z.number().nullish(),
+  error: import_v416.z.object({
+    code: import_v416.z.string().nullish(),
+    message: import_v416.z.string().nullish()
   }).nullish()
 });