@ai-sdk/xai 5.0.6 → 5.0.8

package/CHANGELOG.md

@@ -1,5 +1,22 @@
 # @ai-sdk/xai
 
+## 5.0.8
+
+### Patch Changes
+
+- 5434a34: feat(xai): add missing video provider options
+- 0a5dd0f: feat(xai): add missing Responses API provider options
+
+## 5.0.7
+
+### Patch Changes
+
+- 771e74b: chore: enable dead code lint rules
+- Updated dependencies [fe07867]
+- Updated dependencies [a4b0940]
+- Updated dependencies [771e74b]
+  - @ai-sdk/provider-utils@5.0.47
+
 ## 5.0.6
 
 ### Patch Changes

package/dist/index.d.ts

@@ -46,14 +46,24 @@ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
     }>>;
     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
+    minP: z.ZodOptional<z.ZodNumber>;
+    maxTurns: z.ZodOptional<z.ZodNumber>;
+    parallelToolCalls: z.ZodOptional<z.ZodBoolean>;
+    promptCacheKey: z.ZodOptional<z.ZodString>;
+    safetyIdentifier: z.ZodOptional<z.ZodString>;
     serviceTier: z.ZodOptional<z.ZodEnum<{
         default: "default";
         priority: "priority";
     }>>;
     store: z.ZodOptional<z.ZodBoolean>;
     previousResponseId: z.ZodOptional<z.ZodString>;
+    user: z.ZodOptional<z.ZodString>;
     include: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodEnum<{
         "file_search_call.results": "file_search_call.results";
+        "web_search_call.action.sources": "web_search_call.action.sources";
+        "code_interpreter_call.outputs": "code_interpreter_call.outputs";
+        "reasoning.encrypted_content": "reasoning.encrypted_content";
+        no_inline_citations: "no_inline_citations";
     }>>>>;
 }, z.core.$strip>;
 type XaiLanguageModelResponsesOptions = z.infer<typeof xaiLanguageModelResponsesOptions>;
@@ -83,10 +93,25 @@ declare const resolutionSchema: z.ZodEnum<{
     "1080p": "1080p";
 }>;
 type XaiVideoResolution = z.infer<typeof resolutionSchema>;
+declare const keyframeSchema: z.ZodObject<{
+    imageUrl: z.ZodString;
+    timestampSeconds: z.ZodNumber;
+}, z.core.$strip>;
+declare const storageOptionsSchema: z.ZodObject<{
+    filename: z.ZodString;
+    expiresAfter: z.ZodOptional<z.ZodNumber>;
+    publicUrl: z.ZodOptional<z.ZodUnion<readonly [z.ZodBoolean, z.ZodObject<{
+        expiresAfter: z.ZodOptional<z.ZodNumber>;
+    }, z.core.$strip>]>>;
+}, z.core.$strip>;
+type XaiVideoKeyframe = z.infer<typeof keyframeSchema>;
+type XaiVideoStorageOptions = z.infer<typeof storageOptionsSchema>;
 interface XaiVideoSharedOptions {
     pollIntervalMs?: number | null;
     pollTimeoutMs?: number | null;
     resolution?: XaiVideoResolution | null;
+    /** Store the generated video in the xAI Files API. */
+    storageOptions?: XaiVideoStorageOptions;
 }
 interface XaiVideoUserOptions {
     /**
@@ -126,6 +151,8 @@ interface XaiVideoGenerationOptions extends XaiVideoSharedOptions, XaiVideoUserO
     mode?: undefined;
     videoUrl?: undefined;
     referenceImageUrls?: undefined;
+    /** Mid-video image anchors (up to 4). */
+    keyframes?: XaiVideoKeyframe[];
 }
 interface XaiLegacyEditVideoOptions extends XaiVideoSharedOptions, XaiVideoUserOptions {
     /**

package/dist/index.js

@@ -806,7 +806,9 @@ var xaiResponsesResponseSchema = z6.object({
   output: z6.array(outputItemSchema),
   usage: xaiResponsesUsageSchema.nullish(),
   status: z6.string(),
-  service_tier: z6.string().nullish()
+  service_tier: z6.string().nullish(),
+  prompt_cache_key: z6.string().nullish(),
+  safety_identifier: z6.string().nullish()
 });
 var xaiResponsesChunkSchema = z6.union([
   z6.object({
@@ -1090,7 +1092,9 @@ var xaiResponsesChunkSchema = z6.union([
     response: z6.object({
       incomplete_details: z6.object({ reason: z6.string() }).nullish(),
       usage: xaiResponsesUsageSchema.nullish(),
-      service_tier: z6.string().nullish()
+      service_tier: z6.string().nullish(),
+      prompt_cache_key: z6.string().nullish(),
+      safety_identifier: z6.string().nullish()
     })
   }),
   z6.object({
@@ -1133,9 +1137,23 @@ var xaiLanguageModelResponsesOptions = z7.object({
    * @see https://docs.x.ai/docs/guides/reasoning
    */
   reasoningEffort: z7.enum(["none", "low", "medium", "high", "xhigh"]).optional(),
+  /**
+   * @deprecated xAI ignores the requested summary level and always returns a
+   * detailed reasoning summary.
+   */
   reasoningSummary: z7.enum(["auto", "concise", "detailed"]).optional(),
   logprobs: z7.boolean().optional(),
   topLogprobs: z7.number().int().min(0).max(8).optional(),
+  /** Min-p sampling threshold between 0 and 1. */
+  minP: z7.number().min(0).max(1).optional(),
+  /** Maximum number of agentic tool-calling turns. */
+  maxTurns: z7.number().int().optional(),
+  /** Whether the model may call tools in parallel. @default true */
+  parallelToolCalls: z7.boolean().optional(),
+  /** Cache key used to route requests with shared prompt prefixes. */
+  promptCacheKey: z7.string().optional(),
+  /** Stable identifier used to attribute policy violations to an end user. */
+  safetyIdentifier: z7.string().optional(),
   serviceTier: z7.enum(["default", "priority"]).optional(),
   /**
    * Whether to store the input message(s) and model response for later retrieval.
@@ -1148,11 +1166,21 @@ var xaiLanguageModelResponsesOptions = z7.object({
    * The ID of the previous response from the model.
    */
   previousResponseId: z7.string().optional(),
+  /** A unique identifier for the end user, used for abuse monitoring. */
+  user: z7.string().optional(),
   /**
    * Specify additional output data to include in the model response.
    * Example values: 'file_search_call.results'.
    */
-  include: z7.array(z7.enum(["file_search_call.results"])).nullish()
+  include: z7.array(
+    z7.enum([
+      "file_search_call.results",
+      "web_search_call.action.sources",
+      "code_interpreter_call.outputs",
+      "reasoning.encrypted_content",
+      "no_inline_citations"
+    ])
+  ).nullish()
 });
 
 // src/responses/xai-responses-prepare-tools.ts
@@ -1625,9 +1653,6 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
       providerOptions,
       schema: xaiLanguageModelResponsesOptions
     })) != null ? _a : {};
-    if (topK != null) {
-      warnings.push({ type: "unsupported", feature: "topK" });
-    }
     if (frequencyPenalty != null) {
       warnings.push({ type: "unsupported", feature: "frequencyPenalty" });
     }
@@ -1709,6 +1734,8 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
       max_output_tokens: maxOutputTokens,
       temperature,
       top_p: topP,
+      top_k: topK,
+      min_p: options.minP,
       seed,
       ...(responseFormat == null ? void 0 : responseFormat.type) === "json" && {
         text: {
@@ -1721,14 +1748,9 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
           } : { type: "json_object" }
         }
       },
-      ...(resolvedReasoningEffort != null || options.reasoningSummary != null) && {
+      ...resolvedReasoningEffort != null && {
         reasoning: {
-          ...resolvedReasoningEffort != null && {
-            effort: resolvedReasoningEffort
-          },
-          ...options.reasoningSummary != null && {
-            summary: options.reasoningSummary
-          }
+          effort: resolvedReasoningEffort
         }
       },
       ...options.store === false && {
@@ -1740,8 +1762,23 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
       ...options.previousResponseId != null && {
         previous_response_id: options.previousResponseId
       },
+      ...options.maxTurns != null && {
+        max_turns: options.maxTurns
+      },
+      ...options.parallelToolCalls != null && {
+        parallel_tool_calls: options.parallelToolCalls
+      },
+      ...options.promptCacheKey != null && {
+        prompt_cache_key: options.promptCacheKey
+      },
+      ...options.safetyIdentifier != null && {
+        safety_identifier: options.safetyIdentifier
+      },
       ...options.serviceTier != null && {
         service_tier: options.serviceTier
+      },
+      ...options.user != null && {
+        user: options.user
       }
     };
     if (xaiTools2 && xaiTools2.length > 0) {
@@ -1964,7 +2001,7 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
         inputTokens: { total: 0, noCache: 0, cacheRead: 0, cacheWrite: 0 },
         outputTokens: { total: 0, text: 0, reasoning: 0 }
       },
-      ...(((_q = response.usage) == null ? void 0 : _q.cost_in_usd_ticks) != null || response.service_tier != null) && {
+      ...(((_q = response.usage) == null ? void 0 : _q.cost_in_usd_ticks) != null || response.service_tier != null || response.prompt_cache_key != null || response.safety_identifier != null) && {
         providerMetadata: {
           xai: {
             ...((_r = response.usage) == null ? void 0 : _r.cost_in_usd_ticks) != null && {
@@ -1972,6 +2009,12 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
             },
             ...response.service_tier != null && {
               serviceTier: response.service_tier
+            },
+            ...response.prompt_cache_key != null && {
+              promptCacheKey: response.prompt_cache_key
+            },
+            ...response.safety_identifier != null && {
+              safetyIdentifier: response.safety_identifier
             }
           }
         }
@@ -2020,6 +2063,8 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
     let usage = void 0;
     let costInUsdTicks = void 0;
     let serviceTier = void 0;
+    let promptCacheKey = void 0;
+    let safetyIdentifier = void 0;
     let isFirstChunk = true;
     const contentBlocks = {};
     const seenToolCalls = /* @__PURE__ */ new Set();
@@ -2033,7 +2078,7 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
             controller.enqueue({ type: "stream-start", warnings });
           },
           transform(chunk, controller) {
-            var _a2, _b2, _c2, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q;
+            var _a2, _b2, _c2, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s;
             if (options.includeRawChunks) {
               controller.enqueue({ type: "raw", rawValue: chunk.rawValue });
             }
@@ -2165,8 +2210,10 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
                 costInUsdTicks = (_c2 = response2.usage.cost_in_usd_ticks) != null ? _c2 : void 0;
               }
               serviceTier = (_d = response2.service_tier) != null ? _d : void 0;
+              promptCacheKey = (_e = response2.prompt_cache_key) != null ? _e : void 0;
+              safetyIdentifier = (_f = response2.safety_identifier) != null ? _f : void 0;
               if (event.type === "response.incomplete") {
-                const reason = "incomplete_details" in response2 ? (_e = response2.incomplete_details) == null ? void 0 : _e.reason : void 0;
+                const reason = "incomplete_details" in response2 ? (_g = response2.incomplete_details) == null ? void 0 : _g.reason : void 0;
                 finishReason = {
                   unified: reason ? mapXaiResponsesFinishReason(reason) : "other",
                   raw: reason != null ? reason : "incomplete"
@@ -2180,7 +2227,7 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
               return;
             }
             if (event.type === "response.failed") {
-              const reason = (_f = event.response.incomplete_details) == null ? void 0 : _f.reason;
+              const reason = (_h = event.response.incomplete_details) == null ? void 0 : _h.reason;
               finishReason = {
                 unified: reason ? mapXaiResponsesFinishReason(reason) : "error",
                 raw: reason != null ? reason : "error"
@@ -2323,13 +2370,13 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
                     toolCallId: part.id,
                     toolName,
                     result: {
-                      queries: (_g = part.queries) != null ? _g : [],
-                      results: (_i = (_h = part.results) == null ? void 0 : _h.map((result) => ({
+                      queries: (_i = part.queries) != null ? _i : [],
+                      results: (_k = (_j = part.results) == null ? void 0 : _j.map((result) => ({
                         fileId: result.file_id,
                         filename: result.filename,
                         score: result.score,
                         text: result.text
-                      }))) != null ? _i : null
+                      }))) != null ? _k : null
                     }
                   });
                 }
@@ -2396,17 +2443,17 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
                   "x_semantic_search",
                   "x_thread_fetch"
                 ];
-                let toolName = (_j = part.name) != null ? _j : "";
-                if (webSearchSubTools.includes((_k = part.name) != null ? _k : "") || part.type === "web_search_call") {
+                let toolName = (_l = part.name) != null ? _l : "";
+                if (webSearchSubTools.includes((_m = part.name) != null ? _m : "") || part.type === "web_search_call") {
                   toolName = webSearchToolName != null ? webSearchToolName : "web_search";
-                } else if (xSearchSubTools.includes((_l = part.name) != null ? _l : "") || part.type === "x_search_call") {
+                } else if (xSearchSubTools.includes((_n = part.name) != null ? _n : "") || part.type === "x_search_call") {
                   toolName = xSearchToolName != null ? xSearchToolName : "x_search";
                 } else if (part.name === "code_execution" || part.type === "code_interpreter_call" || part.type === "code_execution_call") {
                   toolName = codeExecutionToolName != null ? codeExecutionToolName : "code_execution";
                 } else if (part.type === "mcp_call") {
-                  toolName = (_m = mcpToolName != null ? mcpToolName : part.name) != null ? _m : "mcp";
+                  toolName = (_o = mcpToolName != null ? mcpToolName : part.name) != null ? _o : "mcp";
                 }
-                const toolInput = part.type === "custom_tool_call" ? (_n = part.input) != null ? _n : "" : part.type === "mcp_call" ? (_o = part.arguments) != null ? _o : "" : (_p = part.arguments) != null ? _p : "";
+                const toolInput = part.type === "custom_tool_call" ? (_p = part.input) != null ? _p : "" : part.type === "mcp_call" ? (_q = part.arguments) != null ? _q : "" : (_r = part.arguments) != null ? _r : "";
                 const shouldEmit = part.type === "custom_tool_call" ? event.type === "response.output_item.done" : !seenToolCalls.has(part.id);
                 if (shouldEmit && !seenToolCalls.has(part.id)) {
                   seenToolCalls.add(part.id);
@@ -2467,7 +2514,7 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
                           sourceType: "url",
                           id: self.config.generateId(),
                           url: annotation.url,
-                          title: (_q = annotation.title) != null ? _q : annotation.url
+                          title: (_s = annotation.title) != null ? _s : annotation.url
                         });
                       }
                     }
@@ -2522,11 +2569,13 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
                 },
                 outputTokens: { total: 0, text: 0, reasoning: 0 }
               },
-              ...(costInUsdTicks != null || serviceTier != null) && {
+              ...(costInUsdTicks != null || serviceTier != null || promptCacheKey != null || safetyIdentifier != null) && {
                 providerMetadata: {
                   xai: {
                     ...costInUsdTicks != null && { costInUsdTicks },
-                    ...serviceTier != null && { serviceTier }
+                    ...serviceTier != null && { serviceTier },
+                    ...promptCacheKey != null && { promptCacheKey },
+                    ...safetyIdentifier != null && { safetyIdentifier }
                   }
                 }
               }
@@ -3670,7 +3719,7 @@ var xaiTools = {
 };
 
 // src/version.ts
-var VERSION = true ? "5.0.6" : "0.0.0-test";
+var VERSION = true ? "5.0.8" : "0.0.0-test";
 
 // src/files/xai-files.ts
 import {
@@ -3947,6 +3996,20 @@ import { z as z18 } from "zod/v4";
 var nonEmptyStringSchema = z18.string().min(1);
 var resolutionSchema = z18.enum(["480p", "720p", "1080p"]);
 var modeSchema = z18.enum(["edit-video", "extend-video", "reference-to-video"]);
+var keyframeSchema = z18.object({
+  imageUrl: nonEmptyStringSchema,
+  timestampSeconds: z18.number().positive()
+});
+var storageOptionsSchema = z18.object({
+  filename: nonEmptyStringSchema,
+  expiresAfter: z18.number().int().positive().max(2592e3).optional(),
+  publicUrl: z18.union([
+    z18.boolean(),
+    z18.object({
+      expiresAfter: z18.number().int().min(3600).max(2592e3).optional()
+    })
+  ]).optional()
+});
 var baseFields = {
   pollIntervalMs: z18.number().positive().nullish(),
   pollTimeoutMs: z18.number().positive().nullish(),
@@ -3957,6 +4020,8 @@ var runtimeSchema = z18.looseObject({
   videoUrl: nonEmptyStringSchema.optional(),
   referenceImageUrls: z18.array(nonEmptyStringSchema).min(1).max(7).optional(),
   referenceVoiceIds: z18.array(nonEmptyStringSchema).max(3).optional(),
+  keyframes: z18.array(keyframeSchema).max(4).optional(),
+  storageOptions: storageOptionsSchema.optional(),
   user: z18.string().optional(),
   ...baseFields
 });
@@ -3989,6 +4054,7 @@ function resolveStartImage(options) {
 }
 var isVideoFile = (file) => file.mediaType != null && getTopLevelMediaType2(file.mediaType) === "video";
 var isImageReference = (file) => file.mediaType == null || getTopLevelMediaType2(file.mediaType) === "image";
+var isAudioReference = (file) => file.mediaType != null && getTopLevelMediaType2(file.mediaType) === "audio";
 function fileToXaiUrl(file) {
   if (file.type === "url") {
     return file.url;
@@ -4000,11 +4066,14 @@ function resolveReferences(options, xaiOptions, warnings) {
   if (options.inputReferences != null && options.inputReferences.length > 0) {
     const imageFiles = [];
     for (const reference of options.inputReferences) {
+      if (isAudioReference(reference)) {
+        continue;
+      }
       if (!isImageReference(reference)) {
         warnings.push({
           type: "unsupported",
           feature: "inputReferences",
-          details: isVideoFile(reference) ? 'xAI reference-to-video accepts image references only. The video reference was ignored. Use providerOptions.xai.mode "extend-video" to continue from a video.' : "xAI reference-to-video accepts image references only. The non-image reference was ignored."
+          details: 'xAI reference-to-video does not accept video references. The video reference was ignored. Use providerOptions.xai.mode "extend-video" to continue from a video.'
         });
         continue;
       }
@@ -4015,12 +4084,20 @@ function resolveReferences(options, xaiOptions, warnings) {
   if ((xaiOptions == null ? void 0 : xaiOptions.referenceImageUrls) != null && xaiOptions.referenceImageUrls.length > 0) {
     return xaiOptions.referenceImageUrls.map((url) => ({ url }));
   }
-  return void 0;
+}
+function resolveReferenceAudios(options) {
+  var _a;
+  const audioReferences = (_a = options.inputReferences) == null ? void 0 : _a.filter(isAudioReference);
+  return audioReferences != null && audioReferences.length > 0 ? audioReferences.map((reference) => ({ url: fileToXaiUrl(reference) })) : void 0;
 }
 function hasImageInputReference(options) {
   var _a, _b;
   return (_b = (_a = options.inputReferences) == null ? void 0 : _a.some(isImageReference)) != null ? _b : false;
 }
+function hasAudioInputReference(options) {
+  var _a, _b;
+  return (_b = (_a = options.inputReferences) == null ? void 0 : _a.some(isAudioReference)) != null ? _b : false;
+}
 function resolveVideoMode(options, xaiOptions) {
   if ((xaiOptions == null ? void 0 : xaiOptions.mode) != null) {
     return xaiOptions.mode;
@@ -4028,12 +4105,10 @@ function resolveVideoMode(options, xaiOptions) {
   if ((xaiOptions == null ? void 0 : xaiOptions.videoUrl) != null) {
     return "edit-video";
   }
-  const hasFrameImages = options.frameImages != null && options.frameImages.length > 0;
   const hasLegacyReferenceUrls = (xaiOptions == null ? void 0 : xaiOptions.referenceImageUrls) != null && xaiOptions.referenceImageUrls.length > 0;
-  if (!hasFrameImages && (hasImageInputReference(options) || hasLegacyReferenceUrls)) {
+  if (hasImageInputReference(options) || hasAudioInputReference(options) || hasLegacyReferenceUrls) {
     return "reference-to-video";
   }
-  return void 0;
 }
 var XaiVideoModel = class {
   constructor(modelId, config) {
@@ -4139,6 +4214,41 @@ var XaiVideoModel = class {
         });
       }
     }
+    if (options.generateAudio != null && !isEdit && !isExtension) {
+      body.generate_audio = options.generateAudio;
+    } else if (options.generateAudio != null) {
+      warnings.push({
+        type: "unsupported",
+        feature: "generateAudio",
+        details: `xAI ${isEdit ? "video editing" : "video extension"} does not support generateAudio.`
+      });
+    }
+    if ((xaiOptions == null ? void 0 : xaiOptions.storageOptions) != null) {
+      const { filename, expiresAfter, publicUrl } = xaiOptions.storageOptions;
+      body.storage_options = {
+        filename,
+        ...expiresAfter != null ? { expires_after: expiresAfter } : {},
+        ...publicUrl != null ? {
+          public_url: typeof publicUrl === "boolean" ? publicUrl : {
+            ...publicUrl.expiresAfter != null ? { expires_after: publicUrl.expiresAfter } : {}
+          }
+        } : {}
+      };
+    }
+    if ((xaiOptions == null ? void 0 : xaiOptions.keyframes) != null && xaiOptions.keyframes.length > 0) {
+      if (this.modelId !== "grok-imagine-video-1.5" || isEdit || isExtension) {
+        warnings.push({
+          type: "unsupported",
+          feature: "keyframes",
+          details: this.modelId !== "grok-imagine-video-1.5" ? 'xAI only supports keyframes with "grok-imagine-video-1.5".' : `xAI ${isEdit ? "video editing" : "video extension"} does not support keyframes.`
+        });
+      } else {
+        body.keyframes = xaiOptions.keyframes.map((keyframe) => ({
+          image: { url: keyframe.imageUrl },
+          timestamp_s: keyframe.timestampSeconds
+        }));
+      }
+    }
     if (isEdit) {
       body.video = { url: xaiOptions.videoUrl };
     }
@@ -4160,17 +4270,22 @@ var XaiVideoModel = class {
     }
     const lastFrameImage = getLastFrameImage(options);
     if (lastFrameImage != null) {
-      warnings.push({
-        type: "unsupported",
-        feature: "frameImages",
-        details: isVideoFile(lastFrameImage) ? 'xAI does not accept a video as a start/frame image. The video last frame was ignored. Use providerOptions.xai.mode "extend-video" to continue from a video instead.' : `xAI video models do not support last_frame. Use providerOptions.xai.mode "extend-video" to continue from a video's last frame. The last frame image was ignored.`
-      });
+      if (this.modelId !== "grok-imagine-video-1.5" || isEdit || isExtension || isVideoFile(lastFrameImage)) {
+        warnings.push({
+          type: "unsupported",
+          feature: "frameImages",
+          details: this.modelId !== "grok-imagine-video-1.5" ? 'xAI only supports last_frame with "grok-imagine-video-1.5". The last frame was ignored.' : "xAI only accepts an image last_frame for video generation. The last frame was ignored."
+        });
+      } else {
+        body.last_frame = { url: fileToXaiUrl(lastFrameImage) };
+      }
     }
     if (hasReferenceImages) {
       const referenceImages = resolveReferences(options, xaiOptions, warnings);
+      const referenceAudios = resolveReferenceAudios(options);
       if (referenceImages != null) {
         body.reference_images = referenceImages;
-      } else {
+      } else if (referenceAudios == null) {
         warnings.push({
           type: "unsupported",
           feature: "referenceImages",
@@ -4178,10 +4293,22 @@ var XaiVideoModel = class {
         });
       }
       const referenceVoiceIds = xaiOptions == null ? void 0 : xaiOptions.referenceVoiceIds;
-      if (referenceVoiceIds != null && referenceVoiceIds.length > 0) {
-        body.reference_audios = referenceVoiceIds.map((voiceId) => ({
-          voice_id: voiceId
-        }));
+      const referenceVoices = referenceVoiceIds == null ? void 0 : referenceVoiceIds.map((voiceId) => ({
+        voice_id: voiceId
+      }));
+      const referenceAudioInputs = [
+        ...referenceAudios != null ? referenceAudios : [],
+        ...referenceVoices != null ? referenceVoices : []
+      ];
+      if (referenceAudioInputs.length > 0) {
+        if (referenceAudioInputs.length > 3) {
+          warnings.push({
+            type: "unsupported",
+            feature: "inputReferences",
+            details: "xAI reference-to-video supports at most 3 audio references. Only the first 3 were used."
+          });
+        }
+        body.reference_audios = referenceAudioInputs.slice(0, 3);
       }
       if (body.resolution === "1080p") {
         warnings.push({
@@ -4203,7 +4330,7 @@ var XaiVideoModel = class {
       warnings.push({
         type: "unsupported",
         feature: "inputReferences",
-        details: hasImageInputReference(options) ? "xAI only supports inputReferences for reference-to-video generation. The reference images were ignored." : "xAI reference-to-video requires at least one image reference. The references were ignored."
+        details: hasImageInputReference(options) || hasAudioInputReference(options) ? "xAI only supports inputReferences for reference-to-video generation. The references were ignored." : "xAI reference-to-video requires at least one image or audio reference. The references were ignored."
       });
     }
     if ((xaiOptions == null ? void 0 : xaiOptions.referenceVoiceIds) != null && xaiOptions.referenceVoiceIds.length > 0 && !hasReferenceImages) {
@@ -4226,6 +4353,8 @@ var XaiVideoModel = class {
           "videoUrl",
           "referenceImageUrls",
           "referenceVoiceIds",
+          "keyframes",
+          "storageOptions",
           "user"
         ].includes(key)) {
           body[key] = value;
@@ -4284,7 +4413,7 @@ var XaiVideoModel = class {
     };
   }
   async doStatus(options) {
-    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k;
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l;
     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 { requestId } = options.operation;
     const baseURL = (_d = this.config.baseURL) != null ? _d : "https://api.x.ai/v1";
@@ -4321,7 +4450,8 @@ var XaiVideoModel = class {
       };
     }
     if (statusResponse.status === "done" || statusResponse.status == null && ((_h = statusResponse.video) == null ? void 0 : _h.url)) {
-      if (((_i = statusResponse.video) == null ? void 0 : _i.respect_moderation) === false) {
+      const video = statusResponse.video;
+      if ((video == null ? void 0 : video.respect_moderation) === false) {
         return {
           status: "error",
           error: "Video generation was blocked due to a content policy violation.",
@@ -4332,7 +4462,8 @@ var XaiVideoModel = class {
           }
         };
       }
-      if (!((_j = statusResponse.video) == null ? void 0 : _j.url)) {
+      const videoUrl = (_k = (_j = video == null ? void 0 : video.url) != null ? _j : (_i = video == null ? void 0 : video.file_output) == null ? void 0 : _i.public_url) != null ? _k : void 0;
+      if (!videoUrl) {
         return {
           status: "error",
           error: "Video generation completed but no video URL was returned.",
@@ -4345,13 +4476,7 @@ var XaiVideoModel = class {
       }
       return {
         status: "completed",
-        videos: [
-          {
-            type: "url",
-            url: statusResponse.video.url,
-            mediaType: "video/mp4"
-          }
-        ],
+        videos: [{ type: "url", url: videoUrl, mediaType: "video/mp4" }],
         warnings: [],
         response: {
           timestamp: currentDate,
@@ -4361,10 +4486,29 @@ var XaiVideoModel = class {
         providerMetadata: {
           xai: {
             requestId,
-            videoUrl: statusResponse.video.url,
-            ...statusResponse.video.duration != null ? { duration: statusResponse.video.duration } : {},
-            ...((_k = statusResponse.usage) == null ? void 0 : _k.cost_in_usd_ticks) != null ? { costInUsdTicks: statusResponse.usage.cost_in_usd_ticks } : {},
-            ...statusResponse.progress != null ? { progress: statusResponse.progress } : {}
+            videoUrl,
+            ...(video == null ? void 0 : video.duration) != null ? { duration: video.duration } : {},
+            ...((_l = statusResponse.usage) == null ? void 0 : _l.cost_in_usd_ticks) != null ? { costInUsdTicks: statusResponse.usage.cost_in_usd_ticks } : {},
+            ...statusResponse.progress != null ? { progress: statusResponse.progress } : {},
+            ...(video == null ? void 0 : video.file_output) != null ? {
+              fileOutput: {
+                fileId: video.file_output.file_id,
+                filename: video.file_output.filename,
+                ...video.file_output.expires_at != null ? {
+                  expiresAt: video.file_output.expires_at
+                } : {},
+                ...video.file_output.public_url != null ? {
+                  publicUrl: video.file_output.public_url
+                } : {},
+                ...video.file_output.public_url_error != null ? {
+                  publicUrlError: video.file_output.public_url_error
+                } : {},
+                ...video.file_output.public_url_expires_at != null ? {
+                  publicUrlExpiresAt: video.file_output.public_url_expires_at
+                } : {}
+              }
+            } : {},
+            ...(video == null ? void 0 : video.storage_error) != null ? { storageError: video.storage_error } : {}
           }
         }
       };
@@ -4385,9 +4529,18 @@ var xaiCreateVideoResponseSchema = z19.object({
 var xaiVideoStatusResponseSchema = z19.object({
   status: z19.string().nullish(),
   video: z19.object({
-    url: z19.string(),
+    url: z19.string().nullish(),
     duration: z19.number().nullish(),
-    respect_moderation: z19.boolean().nullish()
+    respect_moderation: z19.boolean().nullish(),
+    file_output: z19.object({
+      file_id: z19.string(),
+      filename: z19.string(),
+      expires_at: z19.number().nullish(),
+      public_url: z19.string().nullish(),
+      public_url_error: z19.string().nullish(),
+      public_url_expires_at: z19.number().nullish()
+    }).nullish(),
+    storage_error: z19.string().nullish()
   }).nullish(),
   model: z19.string().nullish(),
   usage: z19.object({