fix(xiaoyunque): support reference video assets

packages/tools/xiaoyunque-mcp/scripts/live-image-smoke.ts

@@ -55,7 +55,7 @@ function notify(method: string, params: unknown) {
 try {
   const initialized = await request("initialize", {
     capabilities: {},
-    clientInfo: { name: "convax-xiaoyunque-live-smoke", version: "0.3.2" },
+    clientInfo: { name: "convax-xiaoyunque-live-smoke", version: "0.3.3" },
     protocolVersion: "2025-03-26",
   })
   if ((initialized.result as Record<string, unknown> | undefined)?.protocolVersion !== "2025-03-26") {

packages/tools/xiaoyunque-mcp/scripts/live-video-smoke.ts

@@ -117,7 +117,7 @@ try {
   progress("mcp.initialize")
   const initialized = await request("initialize", {
     capabilities: {},
-    clientInfo: { name: "convax-xiaoyunque-live-video-smoke", version: "0.3.2" },
+    clientInfo: { name: "convax-xiaoyunque-live-video-smoke", version: "0.3.3" },
     protocolVersion: "2025-03-26",
   })
   if ((initialized.result as Record<string, unknown> | undefined)?.protocolVersion !== "2025-03-26") {

packages/tools/xiaoyunque-mcp/src/mcp-server.ts

@@ -153,7 +153,9 @@ export function publicGenerationErrorMessage(error: unknown) {
     return "The selected XiaoYunque image model is no longer available. Choose another image model and try again."
   }
   if (error instanceof XiaoYunqueReferenceAssetRegistrationError) {
-    return "XiaoYunque could not prepare the reference image for generation. No generation was submitted; try again."
+    return error.referenceType === "video"
+      ? "XiaoYunque could not prepare the reference video for generation. No generation was submitted; try again."
+      : "XiaoYunque could not prepare the reference image for generation. No generation was submitted; try again."
   }
   if (error instanceof XiaoYunqueRequestRejectedError) {
     return "XiaoYunque did not accept this generation request. Refresh Services and try a model listed for this capability."
@@ -169,6 +171,7 @@ export const safeGenerationDiagnosticCodes = [
   "status-check-timeout",
   "unsupported-image-model",
   "reference-image-registration-failed",
+  "reference-video-registration-failed",
   "upstream-envelope-rejected",
   "upstream-http-rejected",
   "upstream-request-rejected",
@@ -185,7 +188,9 @@ export function safeGenerationDiagnosticCode(error: unknown): SafeGenerationDiag
   if (error instanceof XiaoYunqueQueryTimeoutError) return "status-check-timeout"
   if (error instanceof XiaoYunqueUnsupportedImageModelError) return "unsupported-image-model"
   if (error instanceof XiaoYunqueReferenceAssetRegistrationError) {
-    return "reference-image-registration-failed"
+    return error.referenceType === "video"
+      ? "reference-video-registration-failed"
+      : "reference-image-registration-failed"
   }
   if (error instanceof XiaoYunqueRequestRejectedError) return error.diagnosticCode
   return "unclassified-failure"

packages/tools/xiaoyunque-mcp/src/xiaoyunque-api.ts

@@ -37,9 +37,11 @@ export class XiaoYunqueQueryTimeoutError extends Error {
  */
 export class XiaoYunqueReferenceAssetRegistrationError extends Error {
   override name = "XiaoYunqueReferenceAssetRegistrationError"
+  readonly referenceType: "image" | "video"
 
-  constructor() {
-    super("XiaoYunque reference image asset registration failed")
+  constructor(referenceType: "image" | "video" = "image") {
+    super(`XiaoYunque reference ${referenceType} asset registration failed`)
+    this.referenceType = referenceType
   }
 }
 
@@ -136,6 +138,8 @@ interface XiaoYunqueUserInfo {
 }
 
 type FetchLike = typeof fetch
+type UploadAssetType = 1 | 2 | 4
+type RegisteredAssetType = 1 | 2
 
 function record(value: unknown) {
   return value && typeof value === "object" && !Array.isArray(value)
@@ -316,7 +320,7 @@ async function boundedJson(response: Response, label: string) {
   }
 }
 
-function uploadAssetType(reference: FileGenerationReference) {
+function uploadAssetType(reference: FileGenerationReference): UploadAssetType {
   if (reference.role === "reference_video" || reference.mime_type.startsWith("video/")) return 1
   if (reference.role === "audio" || reference.mime_type.startsWith("audio/")) return 4
   if (
@@ -328,10 +332,21 @@ function uploadAssetType(reference: FileGenerationReference) {
   throw new Error("XiaoYunque does not support this reference type")
 }
 
+function registeredAssetTypeForUpload(assetType: UploadAssetType): RegisteredAssetType | undefined {
+  switch (assetType) {
+    case 1: // video
+      return 2
+    case 2: // image
+      return 1
+    case 4: // audio does not require AssetCreateV2 registration
+      return undefined
+  }
+}
+
 function uploadedAsset(
   value: unknown,
   reference: FileGenerationReference,
-  assetType: 1 | 2 | 4,
+  assetType: UploadAssetType,
   allowLoopbackHttp: boolean,
 ): UploadedAsset {
   const data = record(value)
@@ -631,27 +646,36 @@ export class XiaoYunqueApi {
     const response = await this.#request(uploadPath, session, signal, { body, method: "POST" })
     const payload = requireSuccess(response.value, response.status, "XiaoYunque reference upload")
     const asset = uploadedAsset(payload.data, reference, assetType, this.#allowLoopbackTest)
-    if (assetType !== 2 || asset.pippitAssetId !== undefined) return asset
+    const registeredAssetType = registeredAssetTypeForUpload(assetType)
+    if (registeredAssetType === undefined || asset.pippitAssetId !== undefined) return asset
 
     return {
       ...asset,
-      pippitAssetId: await this.#registerUploadedImage(asset.assetId, session, signal),
+      pippitAssetId: await this.#registerUploadedAsset(
+        asset.assetId,
+        registeredAssetType,
+        session,
+        signal,
+      ),
     }
   }
 
-  async #registerUploadedImage(
+  async #registerUploadedAsset(
     assetId: string,
+    assetType: RegisteredAssetType,
     session: StoredWebSession,
     signal: AbortSignal,
   ) {
     try {
       // `upload_file.asset_id` is an EverPhoto source id. The first-party Web
       // product converts it into the identity consumed by image and video generation via
-      // AssetCreateV2; the two ids are deliberately never treated as aliases.
+      // AssetCreateV2; the two ids are deliberately never treated as aliases. The
+      // upload_file and AssetCreateV2 asset-type enums differ: upload 2/1 maps to
+      // registered image/video 1/2 respectively.
       const response = await this.#jsonRequest(assetCreateV2Path, {
         asset_source_type: 3,
         asset_source_id: assetId,
-        asset_type: 1,
+        asset_type: assetType,
         Base: { Client: "web" },
       }, session, signal)
       const payload = requireAssetRegistrationSuccess(response.value, response.status)
@@ -664,7 +688,7 @@ export class XiaoYunqueApi {
     } catch (error) {
       if (signal.aborted) throw cancellationError(signal)
       if (error instanceof XiaoYunqueAuthenticationError) throw error
-      throw new XiaoYunqueReferenceAssetRegistrationError()
+      throw new XiaoYunqueReferenceAssetRegistrationError(assetType === 1 ? "image" : "video")
     }
   }
 

packages/tools/xiaoyunque-mcp/test/generator.test.ts

@@ -482,6 +482,110 @@ describe("XiaoYunque generation engine", () => {
     }
   })
 
+  test("registers and submits one Canvas reference video for video generation", async () => {
+    const directory = await mkdtemp(path.join(os.tmpdir(), "xiaoyunque-reference-video-"))
+    directories.push(directory)
+    const output = path.join(directory, "output")
+    const referenceVideo = path.join(directory, "reference.mp4")
+    await Promise.all([mkdir(output), writeFile(referenceVideo, mp4)])
+    let uploadCount = 0
+    let registrationCount = 0
+    let submittedBody: Record<string, unknown> | undefined
+    let submittedTask: ReturnType<typeof taskFromSubmitBody> | undefined
+    const server = Bun.serve({
+      hostname: "127.0.0.1",
+      port: 0,
+      fetch: async (request): Promise<Response> => {
+        const url = new URL(request.url)
+        if (url.pathname === "/api/web/v1/common/upload_file") {
+          uploadCount += 1
+          const form = await request.formData()
+          expect(path.basename((form.get("file") as File | null)?.name ?? "")).toBe("reference.mp4")
+          expect(form.get("asset_type")).toBe("1")
+          return Response.json({
+            ret: 0,
+            data: {
+              asset_id: "everphoto-reference-video",
+              download_url: `${url.origin}/uploaded-reference.mp4`,
+              duration_ms: 5_000,
+            },
+          })
+        }
+        if (url.pathname === "/api/biz/v1/asset/create_v2") {
+          registrationCount += 1
+          expect(await request.json()).toEqual({
+            asset_source_type: 3,
+            asset_source_id: "everphoto-reference-video",
+            asset_type: 2,
+            Base: { Client: "web" },
+          })
+          return Response.json({ ret: 0, data: { PippitAssetID: "pippit-reference-video" } })
+        }
+        const accountPreflight = await accountPreflightResponse(request)
+        if (accountPreflight) return accountPreflight
+        if (url.pathname === "/api/biz/v1/agent/submit_run") {
+          submittedBody = await request.json() as Record<string, unknown>
+          submittedTask = taskFromSubmitBody(submittedBody)
+          return Response.json({ ret: 0, data: { accepted: true } })
+        }
+        if (url.pathname === "/api/biz/v1/agent/get_thread") {
+          return Response.json({
+            ret: 0,
+            data: {
+              thread: {
+                thread_id: submittedTask?.threadId,
+                run_list: [{
+                  entry_list: [artifactEntry("biz/x_data_video", `${url.origin}/result.mp4`)],
+                  run_id: submittedTask?.runId,
+                  state: 3,
+                  thread_id: submittedTask?.threadId,
+                }],
+              },
+            },
+          })
+        }
+        if (url.pathname === "/result.mp4") return new Response(mp4, { headers: { "Content-Type": "video/mp4" } })
+        return new Response("not found", { status: 404 })
+      },
+    })
+    const store = new OperationStore(path.join(directory, "state", "operations.json"))
+    const engine = new GenerationEngine({
+      api: new XiaoYunqueApi(`http://127.0.0.1:${server.port}`),
+      authorizer: { session: async () => testSession },
+      operationStore: store,
+      pollIntervalMs: 1,
+    })
+    try {
+      const artifacts = await engine.generate(call({
+        operation_id: "reference-video-generation",
+        output: "video",
+        output_directory: output,
+        references: [{
+          kind: "file",
+          mime_type: "video/mp4",
+          name: "reference.mp4",
+          node_id: "reference-video-node",
+          path: referenceVideo,
+          role: "reference_video",
+        }],
+      }), videoModel, new AbortController().signal)
+
+      expect(artifacts[0]?.mimeType).toBe("video/mp4")
+      expect(uploadCount).toBe(1)
+      expect(registrationCount).toBe(1)
+      expect(directVideoParameters(submittedBody!)).toMatchObject({
+        model: "Seedance_2.0_mini_lite",
+        videos: [{
+          asset_id: "everphoto-reference-video",
+          pippit_asset_id: "pippit-reference-video",
+        }],
+      })
+      expect(await store.find("reference-video-generation")).toMatchObject({ status: "submitted" })
+    } finally {
+      server.stop(true)
+    }
+  })
+
   test("rejects a last frame without a first frame before any API request", async () => {
     const directory = await mkdtemp(path.join(os.tmpdir(), "xiaoyunque-last-frame-only-"))
     directories.push(directory)

packages/tools/xiaoyunque-mcp/test/mcp-server-errors.test.ts

@@ -38,6 +38,10 @@ describe("safe public generation errors", () => {
       .toBe(
         "XiaoYunque could not prepare the reference image for generation. No generation was submitted; try again.",
       )
+    expect(publicGenerationErrorMessage(new XiaoYunqueReferenceAssetRegistrationError("video")))
+      .toBe(
+        "XiaoYunque could not prepare the reference video for generation. No generation was submitted; try again.",
+      )
     expect(publicGenerationErrorMessage(new XiaoYunqueRequestRejectedError("private rejection detail")))
       .toBe(
         "XiaoYunque did not accept this generation request. Refresh Services and try a model listed for this capability.",
@@ -51,6 +55,7 @@ describe("safe public generation errors", () => {
       publicGenerationErrorMessage(new XiaoYunqueObservationRejectedError("upstream-envelope-rejected")),
       publicGenerationErrorMessage(new XiaoYunqueQueryTimeoutError()),
       publicGenerationErrorMessage(new XiaoYunqueReferenceAssetRegistrationError()),
+      publicGenerationErrorMessage(new XiaoYunqueReferenceAssetRegistrationError("video")),
       publicGenerationErrorMessage(new XiaoYunqueRequestRejectedError("private rejection detail")),
       publicGenerationErrorMessage(new XiaoYunqueUnsupportedImageModelError()),
     ]
@@ -74,6 +79,7 @@ describe("safe public generation errors", () => {
       safeGenerationDiagnosticCode(new XiaoYunqueObservationRejectedError("upstream-http-rejected")),
       safeGenerationDiagnosticCode(new XiaoYunqueQueryTimeoutError(privateDetail)),
       safeGenerationDiagnosticCode(new XiaoYunqueReferenceAssetRegistrationError()),
+      safeGenerationDiagnosticCode(new XiaoYunqueReferenceAssetRegistrationError("video")),
       safeGenerationDiagnosticCode(new XiaoYunqueUnsupportedImageModelError()),
       safeGenerationDiagnosticCode(new XiaoYunqueRequestRejectedError(
         privateDetail,
@@ -87,6 +93,7 @@ describe("safe public generation errors", () => {
       "status-check-rejected",
       "status-check-timeout",
       "reference-image-registration-failed",
+      "reference-video-registration-failed",
       "unsupported-image-model",
       "upstream-envelope-rejected",
       "unclassified-failure",