Merge branch 'feature/add_provider_popi' of https://git.yuanzoo.cn/weitao/popiart-node-canvas into feature/add_provider_popi

src/app/api/generate/providers/__tests__/popiserver.test.ts

@@ -425,4 +425,35 @@ describe("popiserver generation provider", () => {
       ]);
     }
   });
+
+  it("does not include reference images from completed image tasks", async () => {
+    vi.mocked(global.fetch).mockResolvedValue({
+      ok: true,
+      json: () => Promise.resolve({
+        status: "0000",
+        message: "ok",
+        data: {
+          list: [
+            {
+              id: 1192,
+              status: 2,
+              images: ["https://example.com/reference.jpg"],
+              resultList: [
+                { images: ["https://example.com/generated.png"] },
+              ],
+            },
+          ],
+        },
+      }),
+    } as Response);
+
+    const result = await checkPopiTaskOnce(makeRequest(), "login-token", "1192", "image");
+
+    expect(result.status).toBe("completed");
+    if (result.status === "completed") {
+      expect(result.output.outputs).toEqual([
+        { type: "image", data: "", url: "https://example.com/generated.png" },
+      ]);
+    }
+  });
 });

src/app/api/generate/providers/popiserver.ts

@@ -557,11 +557,14 @@ function extractOutputFromTask(task: PopiTask, mediaType: string): GenerationOut
       ? { success: true, outputs: [{ type: "audio", data: "", url }] }
       : { success: false, error: "PopiServer audio task completed without output" };
   }
+  const resultImages = collectStrings(
+    Array.isArray(task.resultList) ? task.resultList.map((result) => result.images) : []
+  );
+  // task.images is the original input/reference image list in Popiserver task
+  // payloads. Only use it as a defensive fallback for older responses that do
+  // not put generated images under resultList.
   const images = [
-    ...new Set(collectStrings([
+    ...new Set(resultImages.length > 0 ? resultImages : collectStrings(task.images)),
-      ...(Array.isArray(task.resultList) ? task.resultList.map((result) => result.images) : []),
-      task.images,
-    ])),
   ];
   return images.length > 0
     ? { success: true, outputs: images.map((url) => ({ type: "image", data: "", url })) }

src/app/api/llm/__tests__/route.test.ts

@@ -478,7 +478,7 @@ describe("/api/llm route", () => {
         prompt: "请描述这张图",
         provider: "popiserver",
         model: "kimi-k2.6",
-        images: ["data:image/png;base64,abc"],
+        images: ["https://static.popi.art/uploads/input.png"],
       });
 
       const response = await POST(request);
@@ -498,7 +498,7 @@ describe("/api/llm route", () => {
                 role: "user",
                 content: [
                   { type: "text", text: "请描述这张图" },
-                  { type: "image_url", image_url: { url: "data:image/png;base64,abc" } },
+                  { type: "image_url", image_url: { url: "https://static.popi.art/uploads/input.png" } },
                 ],
               },
             ],

src/store/execution/__tests__/llmGenerateExecutor.test.ts

@@ -188,6 +188,67 @@ describe("executeLlmGenerate", () => {
     expect((loadingCall![1] as Record<string, unknown>).inputVideos).toEqual(["data:video/mp4;base64,vid1"]);
   });
 
+  it("uploads Popiserver LLM media before calling /api/llm", async () => {
+    const node = makeNode({
+      selectedModel: {
+        provider: "popiserver",
+        modelId: "31",
+        displayName: "Kimi K2.6",
+        metadata: { aiModelId: 31 },
+      },
+    });
+    const ctx = makeCtx(node, {
+      providerSettings: {
+        providers: {
+          ...defaultProviderSettings.providers,
+          popiserver: { enabled: true },
+        },
+      } as any,
+      getConnectedInputs: vi.fn().mockReturnValue({
+        images: ["data:image/png;base64,img1"],
+        videos: [],
+        audio: [],
+        text: "describe this",
+        dynamicInputs: {},
+        easeCurve: null,
+      }),
+    });
+    mockFetch.mockResolvedValueOnce({
+      ok: true,
+      json: () => Promise.resolve({ success: true, url: "https://static.popi.art/uploads/input.png" }),
+    });
+    mockFetch.mockResolvedValueOnce({
+      ok: true,
+      json: () => Promise.resolve({ success: true, text: "description" }),
+    });
+
+    const originalMode = process.env.NEXT_PUBLIC_PROVIDER_MODE;
+    process.env.NEXT_PUBLIC_PROVIDER_MODE = "popi";
+    try {
+      await executeLlmGenerate(ctx);
+    } finally {
+      if (originalMode === undefined) {
+        delete process.env.NEXT_PUBLIC_PROVIDER_MODE;
+      } else {
+        process.env.NEXT_PUBLIC_PROVIDER_MODE = originalMode;
+      }
+    }
+
+    expect(mockFetch).toHaveBeenNthCalledWith(
+      1,
+      "/api/popi/media/upload",
+      expect.objectContaining({ method: "POST" })
+    );
+    expect(mockFetch).toHaveBeenNthCalledWith(
+      2,
+      "/api/llm",
+      expect.objectContaining({ method: "POST" })
+    );
+    const body = JSON.parse(mockFetch.mock.calls[1][1].body);
+    expect(body.provider).toBe("popiserver");
+    expect(body.images).toEqual(["https://static.popi.art/uploads/input.png"]);
+  });
+
   it("should not include images field when none connected", async () => {
     const node = makeNode();
     mockFetch.mockResolvedValueOnce({

src/store/execution/__tests__/nanoBananaExecutor.test.ts

@@ -232,6 +232,11 @@ describe("executeNanoBanana", () => {
     expect(body.prompt).toBe("config prompt");
     expect(body.aspectRatio).toBe("9:16");
     expect(body.resolution).toBe("4K");
+    expect(body.model).toBe("32");
+    expect(body.selectedModel).toMatchObject({
+      provider: "popiserver",
+      modelId: "32",
+    });
     expect(body.parameters).toMatchObject({
       aspectRatio: "9:16",
       aspect_ratio: "9:16",

src/store/execution/generateVideoExecutor.ts

@@ -28,8 +28,7 @@ import {
 import { readVideoGenerationConfig, removeFirstClassGenerationParameters } from "@/utils/generationConfig";
 import {
   cleanupTemporaryImages,
-  uploadPopiserverDynamicInputsForGeneration,
+  uploadPopiserverGenerationInputsForGeneration,
-  uploadPopiserverImagesForGeneration,
   uploadTemporaryDynamicInputsForGeneration,
   uploadTemporaryImagesForGeneration,
 } from "@/utils/temporaryImageUpload";
@@ -258,13 +257,16 @@ export async function executeGenerateVideo(
     let temporaryMediaIds: string[] = [];
 
     try {
-      const preparedImages = provider === "popiserver"
+      const preparedPopiserverInputs = provider === "popiserver"
-        ? await uploadPopiserverImagesForGeneration(images)
+        ? await uploadPopiserverGenerationInputsForGeneration(images, requestDynamicInputs)
+        : null;
+      const preparedImages = preparedPopiserverInputs
+        ? { images: preparedPopiserverInputs.images, cleanupIds: preparedPopiserverInputs.cleanupIds }
         : provider === "newapiwg"
           ? await uploadTemporaryImagesForGeneration(images)
           : { images, cleanupIds: [] };
-      const preparedDynamicInputs = provider === "popiserver"
+      const preparedDynamicInputs = preparedPopiserverInputs
-        ? await uploadPopiserverDynamicInputsForGeneration(requestDynamicInputs)
+        ? { dynamicInputs: preparedPopiserverInputs.dynamicInputs, cleanupIds: preparedPopiserverInputs.cleanupIds }
         : provider === "newapiwg"
           ? await uploadTemporaryDynamicInputsForGeneration(requestDynamicInputs)
           : { dynamicInputs: requestDynamicInputs, cleanupIds: [] };

src/store/execution/llmGenerateExecutor.ts

@@ -17,6 +17,7 @@ import { isPopiProviderMode } from "@/lib/providerMode";
 import { runWithFallback } from "./runWithFallback";
 import type { NodeExecutionContext } from "./types";
 import { mergeConfiguredParameters } from "./generationParameters";
+import { uploadPopiserverMediaForGeneration } from "@/utils/temporaryImageUpload";
 
 export interface LlmGenerateOptions {
   /** When true, falls back to stored inputImages/inputPrompt if no connections provide them. */
@@ -113,13 +114,22 @@ export async function executeLlmGenerate(
     });
 
     try {
+      const preparedMedia = llmProvider === "popiserver"
+        ? await uploadPopiserverMediaForGeneration([...images, ...videos])
+        : { media: [...images, ...videos], cleanupIds: [] };
+      const preparedImages = images.length > 0
+        ? preparedMedia.media.slice(0, images.length)
+        : images;
+      const preparedVideos = videos.length > 0
+        ? preparedMedia.media.slice(images.length, images.length + videos.length)
+        : videos;
       const response = await fetch("/api/llm", {
         method: "POST",
         headers,
         body: JSON.stringify({
           prompt: finalText,
-          ...(images.length > 0 && { images }),
+          ...(preparedImages.length > 0 && { images: preparedImages }),
-          ...(videos.length > 0 && { videos }),
+          ...(preparedVideos.length > 0 && { videos: preparedVideos }),
           provider: llmProvider,
           model: llmModel,
           ...(aiModelId !== undefined && { aiModelId }),

src/store/execution/nanoBananaExecutor.ts

@@ -29,8 +29,7 @@ import {
 import { isHttpMediaUrl } from "@/utils/mediaResolver";
 import {
   cleanupTemporaryImages,
-  uploadPopiserverDynamicInputsForGeneration,
+  uploadPopiserverGenerationInputsForGeneration,
-  uploadPopiserverImagesForGeneration,
   uploadTemporaryDynamicInputsForGeneration,
   uploadTemporaryImagesForGeneration,
 } from "@/utils/temporaryImageUpload";
@@ -159,13 +158,16 @@ export async function executeNanoBanana(
     let temporaryImageIds: string[] = [];
 
     try {
-      const preparedImages = provider === "popiserver"
+      const preparedPopiserverInputs = provider === "popiserver"
-        ? await uploadPopiserverImagesForGeneration(images)
+        ? await uploadPopiserverGenerationInputsForGeneration(images, sanitizedDynamicInputs)
+        : null;
+      const preparedImages = preparedPopiserverInputs
+        ? { images: preparedPopiserverInputs.images, cleanupIds: preparedPopiserverInputs.cleanupIds }
         : provider === "newapiwg"
           ? await uploadTemporaryImagesForGeneration(images)
           : { images, cleanupIds: [] };
-      const preparedDynamicInputs = provider === "popiserver"
+      const preparedDynamicInputs = preparedPopiserverInputs
-        ? await uploadPopiserverDynamicInputsForGeneration(sanitizedDynamicInputs)
+        ? { dynamicInputs: preparedPopiserverInputs.dynamicInputs, cleanupIds: preparedPopiserverInputs.cleanupIds }
         : provider === "newapiwg"
           ? await uploadTemporaryDynamicInputsForGeneration(sanitizedDynamicInputs)
           : { dynamicInputs: sanitizedDynamicInputs, cleanupIds: [] };
@@ -196,12 +198,13 @@ export async function executeNanoBanana(
             batchSize: requestedImageCount,
           }
         : mergeConfiguredParametersOverride(normalizedBaseParameters, configuredParameters);
+      const usedModelId = modelToUse.modelId || nodeData.model;
       const requestPayload = {
         images: preparedImages.images,
         prompt: finalPrompt,
         aspectRatio: (parametersOverride?.aspectRatio as string) ?? nodeConfig.aspectRatio,
         resolution: (parametersOverride?.resolution as string) ?? nodeConfig.resolution,
-        model: nodeData.model,
+        model: usedModelId,
         useGoogleSearch: (parametersOverride?.useGoogleSearch as boolean) ?? nodeData.useGoogleSearch,
         useImageSearch: (parametersOverride?.useImageSearch as boolean) ?? nodeData.useImageSearch,
         selectedModel: modelToUse,
@@ -219,7 +222,6 @@ export async function executeNanoBanana(
 
       const usePopiserverBatch = provider === "popiserver" && requestedImageCount > 1;
       const requestCount = usePopiserverBatch ? 1 : requestedImageCount;
-      const usedModelId = modelToUse.modelId || nodeData.model;
       const usedModelDisplayName = modelToUse.displayName || usedModelId;
       const batchTimestamp = Date.now();
       const generatedItems: Array<{

src/utils/temporaryImageUpload.ts

@@ -15,6 +15,12 @@ type UploadedTemporaryDynamicInputs = {
   cleanupIds: string[];
 };
 
+type UploadedPopiserverGenerationInputs = {
+  images: string[];
+  dynamicInputs: Record<string, string | string[]>;
+  cleanupIds: string[];
+};
+
 function dataUrlMimeType(dataUrl: string): string | null {
   return dataUrl.match(/^data:([^;]+);base64,/)?.[1] ?? null;
 }
@@ -242,6 +248,39 @@ export async function uploadPopiserverDynamicInputsForGeneration(
   return { dynamicInputs: Object.fromEntries(entries), cleanupIds: [] };
 }
 
+export async function uploadPopiserverGenerationInputsForGeneration(
+  images: string[],
+  dynamicInputs: Record<string, string | string[]>
+): Promise<UploadedPopiserverGenerationInputs> {
+  const uploadCache = new Map<string, Promise<string>>();
+  const prepareMedia = async (media: string): Promise<string> => {
+    if (!shouldUploadPopiserverMedia(media)) return media;
+    let upload = uploadCache.get(media);
+    if (!upload) {
+      upload = uploadPopiserverMedia(media);
+      uploadCache.set(media, upload);
+    }
+    return upload;
+  };
+
+  const preparedImages = await Promise.all(images.map(prepareMedia));
+  const entries = await Promise.all(
+    Object.entries(dynamicInputs).map(async ([key, value]) => {
+      if (typeof value === "string") {
+        return [key, await prepareMedia(value)] as const;
+      }
+
+      return [key, await Promise.all(value.map(prepareMedia))] as const;
+    })
+  );
+
+  return {
+    images: preparedImages,
+    dynamicInputs: Object.fromEntries(entries),
+    cleanupIds: [],
+  };
+}
+
 export async function cleanupTemporaryImages(ids: string[]): Promise<void> {
   if (ids.length === 0) return;