@vectorx/ai-sdk 0.7.0 → 0.8.1

package/lib/ai.js

@@ -66,6 +66,8 @@ class AI {
                 return new models.ReActModel(new models.QwenStyleRepaintV1Model(this.request, models.QwenStyleRepaintV1Model.BASE_URL, this.tokenManager));
             case models.MultiModalModelName.Qwen25T2iPreview:
                 return new models.ReActModel(new models.Qwen25T2iPreviewModel(this.request, models.Qwen25T2iPreviewModel.BASE_URL, this.tokenManager));
+            case models.MultiModalModelName.Wan26Image:
+                return new models.ReActModel(new models.Wan26ImageModel(this.request, models.Wan26ImageModel.BASE_URL, this.tokenManager));
             case models.MultiModalModelName.QwenVlMax:
                 return new models.ReActModel(new models.QwenVlMax(this.request, models.QwenVlMax.BASE_URL, model, this.tokenManager));
             case models.MultiModalModelName.Qwen3VlPlus:

package/lib/models/Wan26Image/index.d.ts

@@ -0,0 +1,98 @@
+import type { IAbstractRequest } from "@vectorx/ai-types";
+import type { DoGenerateOutput, DoStreamOutput, ModelRequestOptions, ReqOptions } from "../../model-type";
+import { TokenManager } from "../../tokenManager";
+import { SimpleChatModel } from "../Chat";
+export interface Wan26ImageParameters {
+    negative_prompt?: string;
+    size?: string;
+    enable_interleave?: boolean;
+    n?: number;
+    max_images?: number;
+    prompt_extend?: boolean;
+    watermark?: boolean;
+    stream?: boolean;
+    seed?: number;
+}
+export interface Wan26ImageAPIInput {
+    model: string;
+    input: {
+        messages: Array<{
+            role: "user";
+            content: Array<{
+                text?: string;
+                image?: string;
+            }>;
+        }>;
+    };
+    parameters?: Wan26ImageParameters;
+}
+export type Wan26ImageRequestOptions = Wan26ImageAPIInput & {
+    parameters?: Wan26ImageParameters;
+};
+export type Wan26ImageContentItem = {
+    type?: "text" | "image" | string;
+    text?: string;
+    image?: string;
+    url?: string;
+    image_url?: string | {
+        url: string;
+    };
+    b64_json?: string;
+    [key: string]: any;
+};
+export interface Wan26ImageAPIResponse {
+    async?: boolean;
+    output?: {
+        choices?: Array<{
+            finish_reason: string;
+            message: {
+                role: "assistant" | "user";
+                content: Wan26ImageContentItem[];
+            };
+        }>;
+        finished?: boolean;
+        task_status?: string;
+        task_id?: string;
+        task_metric?: {
+            TOTAL: number;
+            FAILED: number;
+            SUCCEEDED: number;
+        };
+        results?: Array<{
+            url?: string;
+            [key: string]: any;
+        }>;
+    };
+    usage?: {
+        image_count?: number;
+        input_tokens?: number;
+        output_tokens?: number;
+        total_tokens?: number;
+        size?: string;
+        width?: number;
+        height?: number;
+    };
+    request_id?: string;
+    id?: string;
+    model?: string;
+    created?: number;
+    object?: string;
+    code?: number;
+    error?: string;
+    message?: string;
+}
+export declare class Wan26ImageModel extends SimpleChatModel {
+    static BASE_URL: string;
+    static SUB_URL: string;
+    modelName: string;
+    constructor(req: IAbstractRequest, baseUrl: string, tokenManager: TokenManager);
+    protected normalizeStandardImageCompletion(res: Wan26ImageAPIResponse, fallbackModel: string): DoGenerateOutput;
+    protected convertToWan26Input(data: ModelRequestOptions & {
+        parameters?: Wan26ImageParameters;
+    } & Record<string, any>): Wan26ImageRequestOptions;
+    protected modelRequest(data: Wan26ImageRequestOptions, options?: ReqOptions): Promise<ReadableStream<Uint8Array> | Promise<unknown>>;
+    doGenerate(data: ModelRequestOptions & {
+        parameters?: Wan26ImageParameters;
+    }, options?: ReqOptions): Promise<DoGenerateOutput>;
+    doStream(data: ModelRequestOptions, options?: ReqOptions): Promise<DoStreamOutput>;
+}

package/lib/models/Wan26Image/index.js

@@ -0,0 +1,243 @@
+"use strict";
+var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, generator) {
+    function adopt(value) { return value instanceof P ? value : new P(function (resolve) { resolve(value); }); }
+    return new (P || (P = Promise))(function (resolve, reject) {
+        function fulfilled(value) { try { step(generator.next(value)); } catch (e) { reject(e); } }
+        function rejected(value) { try { step(generator["throw"](value)); } catch (e) { reject(e); } }
+        function step(result) { result.done ? resolve(result.value) : adopt(result.value).then(fulfilled, rejected); }
+        step((generator = generator.apply(thisArg, _arguments || [])).next());
+    });
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.Wan26ImageModel = void 0;
+const stream_1 = require("../../stream");
+const utils_1 = require("../../utils");
+const Chat_1 = require("../Chat");
+const index_1 = require("../index");
+class Wan26ImageModel extends Chat_1.SimpleChatModel {
+    constructor(req, baseUrl, tokenManager) {
+        super(req, baseUrl, Wan26ImageModel.SUB_URL, tokenManager);
+        this.modelName = index_1.modelName[index_1.MultiModalModelName.Wan26Image];
+    }
+    normalizeStandardImageCompletion(res, fallbackModel) {
+        var _a, _b, _c, _d, _e, _f, _g, _h, _j, _k, _l, _m, _o, _p, _q;
+        const qOutput = ((res === null || res === void 0 ? void 0 : res.output) || {});
+        if ((qOutput === null || qOutput === void 0 ? void 0 : qOutput.task_status) && (qOutput === null || qOutput === void 0 ? void 0 : qOutput.task_id)) {
+            const created = (_a = res === null || res === void 0 ? void 0 : res.created) !== null && _a !== void 0 ? _a : Math.floor(Date.now() / 1000);
+            const id = (res === null || res === void 0 ? void 0 : res.id) || (res === null || res === void 0 ? void 0 : res.request_id) || "";
+            const normalized = {
+                id,
+                object: (_b = res === null || res === void 0 ? void 0 : res.object) !== null && _b !== void 0 ? _b : "chat.completion",
+                created,
+                model: (_c = res === null || res === void 0 ? void 0 : res.model) !== null && _c !== void 0 ? _c : fallbackModel,
+                log_id: id,
+                error: (_e = (_d = res === null || res === void 0 ? void 0 : res.error) !== null && _d !== void 0 ? _d : res === null || res === void 0 ? void 0 : res.message) !== null && _e !== void 0 ? _e : "",
+                code: (_f = res === null || res === void 0 ? void 0 : res.code) !== null && _f !== void 0 ? _f : 0,
+                choices: [
+                    {
+                        index: 0,
+                        message: {
+                            id,
+                            role: "assistant",
+                            type: "async_task",
+                            content: JSON.stringify(Object.assign(Object.assign({}, ((res === null || res === void 0 ? void 0 : res.output) || {})), { request_id: (res === null || res === void 0 ? void 0 : res.request_id) || id })),
+                            reasoning_content: "",
+                        },
+                        finish_reason: "stop",
+                    },
+                ],
+                usage: mapUsageToStandard(res === null || res === void 0 ? void 0 : res.usage),
+            };
+            return normalized;
+        }
+        const contentUrl = extractFirstImageUrl(res);
+        const created = (_g = res === null || res === void 0 ? void 0 : res.created) !== null && _g !== void 0 ? _g : Math.floor(Date.now() / 1000);
+        const id = (res === null || res === void 0 ? void 0 : res.id) || (res === null || res === void 0 ? void 0 : res.request_id) || "";
+        const finishReason = (_k = (_j = (_h = qOutput === null || qOutput === void 0 ? void 0 : qOutput.choices) === null || _h === void 0 ? void 0 : _h[0]) === null || _j === void 0 ? void 0 : _j.finish_reason) !== null && _k !== void 0 ? _k : "stop";
+        const normalized = {
+            id,
+            object: (_l = res === null || res === void 0 ? void 0 : res.object) !== null && _l !== void 0 ? _l : "chat.completion",
+            created,
+            model: (_m = res === null || res === void 0 ? void 0 : res.model) !== null && _m !== void 0 ? _m : fallbackModel,
+            log_id: id,
+            error: (_p = (_o = res === null || res === void 0 ? void 0 : res.error) !== null && _o !== void 0 ? _o : res === null || res === void 0 ? void 0 : res.message) !== null && _p !== void 0 ? _p : "",
+            code: (_q = res === null || res === void 0 ? void 0 : res.code) !== null && _q !== void 0 ? _q : 0,
+            choices: [
+                {
+                    index: 0,
+                    message: {
+                        id,
+                        role: "assistant",
+                        type: "image",
+                        content: contentUrl || "",
+                        reasoning_content: "",
+                    },
+                    finish_reason: finishReason,
+                },
+            ],
+            usage: mapUsageToStandard(res === null || res === void 0 ? void 0 : res.usage),
+        };
+        return normalized;
+    }
+    convertToWan26Input(data) {
+        var _a, _b, _c;
+        const imageUrls = [];
+        const texts = [];
+        const messages = data.messages || data.history || [];
+        if (Array.isArray(messages) && messages.length > 0) {
+            const firstUser = (_a = messages.find((m) => (m === null || m === void 0 ? void 0 : m.role) === "user")) !== null && _a !== void 0 ? _a : messages[0];
+            const c = firstUser === null || firstUser === void 0 ? void 0 : firstUser.content;
+            if (typeof c === "string" && c.trim()) {
+                texts.push(c.trim());
+            }
+            else if (Array.isArray(c)) {
+                for (const p of c) {
+                    if ((p === null || p === void 0 ? void 0 : p.type) === "image_url" && ((_b = p.image_url) === null || _b === void 0 ? void 0 : _b.url)) {
+                        imageUrls.push(String(p.image_url.url));
+                    }
+                    else if ((p === null || p === void 0 ? void 0 : p.type) === "text" && typeof p.text === "string" && p.text.trim()) {
+                        texts.push(p.text.trim());
+                    }
+                }
+            }
+        }
+        if (imageUrls.length === 0 && data.image)
+            imageUrls.push(String(data.image));
+        if (imageUrls.length === 0 && Array.isArray(data.images))
+            imageUrls.push(...data.images.map(String));
+        if (texts.length === 0 && data.msg)
+            texts.push(String(data.msg));
+        if (texts.length === 0 && data.prompt)
+            texts.push(String(data.prompt));
+        const prompt = texts.join(" ").trim();
+        if (!prompt)
+            throw new Error("wan2.6-image 需要提供一个 text 提示词");
+        const userParams = Object.assign({}, (data.parameters || {}));
+        if (!userParams.negative_prompt && data.negative_prompt) {
+            userParams.negative_prompt = String(data.negative_prompt);
+        }
+        const enableInterleave = (_c = userParams.enable_interleave) !== null && _c !== void 0 ? _c : false;
+        const parameters = Object.assign(Object.assign({ prompt_extend: true, watermark: false, n: 1 }, userParams), { enable_interleave: enableInterleave });
+        const imgCount = imageUrls.length;
+        if (enableInterleave) {
+            if (imgCount > 1)
+                throw new Error("wan2.6-image: enable_interleave=true 时最多只能输入 1 张图");
+            parameters.n = 1;
+            if (typeof parameters.max_images !== "number")
+                parameters.max_images = 5;
+        }
+        else {
+            if (imgCount < 1)
+                throw new Error("wan2.6-image: enable_interleave=false 时必须输入 1~4 张参考图");
+            if (imgCount > 4)
+                throw new Error("wan2.6-image: enable_interleave=false 时最多只能输入 4 张参考图");
+            if (typeof parameters.n !== "number")
+                parameters.n = 1;
+            if (parameters.n < 1 || parameters.n > 4)
+                throw new Error("wan2.6-image: enable_interleave=false 时 n 取值范围为 1~4");
+        }
+        const content = [{ text: prompt }];
+        imageUrls.forEach((url) => content.push({ image: url }));
+        return {
+            model: this.modelName,
+            input: { messages: [{ role: "user", content }] },
+            parameters,
+        };
+    }
+    modelRequest(data_1) {
+        return __awaiter(this, arguments, void 0, function* (data, options = {}) {
+            const fetchHeaders = yield this.createAuthHeaders(options === null || options === void 0 ? void 0 : options.headers);
+            const { data: responseData, header } = (yield this.req.fetch({
+                method: "post",
+                headers: Object.assign(Object.assign({}, fetchHeaders), ((options === null || options === void 0 ? void 0 : options.headers) || {})),
+                body: JSON.stringify(data),
+                url: `${String(this.baseUrl).replace(/\/+$/, "")}/${String(this.subUrl).replace(/^\/+/, "")}`,
+                stream: false,
+            }));
+            return (0, utils_1.handleResponseData)(responseData, header);
+        });
+    }
+    doGenerate(data, options) {
+        return __awaiter(this, void 0, void 0, function* () {
+            data.model = this.modelName;
+            const body = this.convertToWan26Input(data);
+            const headers = Object.assign(Object.assign({}, ((options === null || options === void 0 ? void 0 : options.headers) || {})), { "X-DashScope-Async": "enable" });
+            const res = (yield this.modelRequest(body, Object.assign(Object.assign({}, options), { headers })));
+            return this.normalizeStandardImageCompletion(res, this.modelName);
+        });
+    }
+    doStream(data, options) {
+        return __awaiter(this, void 0, void 0, function* () {
+            var _a, _b;
+            const nonStream = yield this.doGenerate(Object.assign({}, data), options);
+            const msg = ((_b = (_a = nonStream.choices) === null || _a === void 0 ? void 0 : _a[0]) === null || _b === void 0 ? void 0 : _b.message) || {};
+            const singleChunk = {
+                id: nonStream.id,
+                object: "chat.completion.chunk",
+                created: nonStream.created,
+                model: nonStream.model,
+                log_id: nonStream.log_id,
+                error: nonStream.error || "",
+                code: nonStream.code || 0,
+                choices: [
+                    {
+                        index: 0,
+                        message: {
+                            id: nonStream.id,
+                            role: "assistant",
+                            type: msg.type || "async_task",
+                            content: msg.content || "",
+                            reasoning_content: "",
+                        },
+                        finish_reason: "stop",
+                    },
+                ],
+                usage: nonStream.usage,
+            };
+            const stream = new stream_1.ReadableStream({
+                start(controller) {
+                    controller.enqueue(singleChunk);
+                    controller.close();
+                },
+            });
+            return (0, stream_1.createAsyncIterable)(stream);
+        });
+    }
+}
+exports.Wan26ImageModel = Wan26ImageModel;
+Wan26ImageModel.BASE_URL = "https://dashscope.aliyuncs.com";
+Wan26ImageModel.SUB_URL = "api/v1/services/aigc/image-generation/generation";
+function mapUsageToStandard(_usage) {
+    return {
+        prompt_tokens: 0,
+        completion_tokens: 0,
+        knowledge_tokens: 0,
+        reasoning_tokens: 0,
+        total_tokens: 0,
+    };
+}
+function extractFirstImageUrl(res) {
+    var _a, _b, _c;
+    const qOutput = res === null || res === void 0 ? void 0 : res.output;
+    const msg = (_b = (_a = qOutput === null || qOutput === void 0 ? void 0 : qOutput.choices) === null || _a === void 0 ? void 0 : _a[0]) === null || _b === void 0 ? void 0 : _b.message;
+    const content = msg === null || msg === void 0 ? void 0 : msg.content;
+    if (Array.isArray(content)) {
+        for (const item of content) {
+            if (!item)
+                continue;
+            if (typeof item.image === "string" && item.image)
+                return item.image;
+            if (typeof item.url === "string" && item.url)
+                return item.url;
+            if (typeof item.image_url === "string" && item.image_url)
+                return item.image_url;
+            if (item.image_url && typeof item.image_url === "object" && typeof item.image_url.url === "string") {
+                return String(item.image_url.url);
+            }
+        }
+    }
+    const firstResult = (_c = qOutput === null || qOutput === void 0 ? void 0 : qOutput.results) === null || _c === void 0 ? void 0 : _c[0];
+    if (firstResult === null || firstResult === void 0 ? void 0 : firstResult.url)
+        return String(firstResult.url);
+    return "";
+}

package/lib/models/index.d.ts

@@ -7,6 +7,7 @@ import { QwenImageEditModel } from "./QwenImageEdit";
 import { WanxSketchToImageLiteModel } from "./QwenSketchToImage";
 import { QwenStyleRepaintV1Model } from "./QwenStyleRepaintV1";
 import { QwenVlMax } from "./QwenVlMax";
+import { Wan26ImageModel } from "./Wan26Image";
 import { ReActModel } from "./react";
 export declare enum ModelName {
     DeepSeekR1 = "deepseek-r1",
@@ -30,6 +31,7 @@ export declare enum MultiModalModelName {
     QwenImageEdit = "qwen-image-edit",
     QwenImageEditPlus = "qwen-image-edit-plus",
     Qwen25T2iPreview = "wan2.5-i2i-preview",
+    Wan26Image = "wan2.6-image",
     WanxSketchToImageLite = "wanx-sketch-to-image-lite",
     QwenStyleRepaintV1 = "wanx-style-repaint-v1"
 }
@@ -52,6 +54,7 @@ export declare const modelName: {
     "qwen-image-edit": string;
     "qwen-image-edit-plus": string;
     "wan2.5-i2i-preview": string;
+    "wan2.6-image": string;
     "wanx-sketch-to-image-lite": string;
     "wanx-style-repaint-v1": string;
 };
@@ -63,4 +66,4 @@ export declare const MODEL_REQUEST_PATTERNS: Map<string, {
     paths: string[];
 }>;
 export declare function isModelRequestUrl(url: string): boolean;
-export { DefaultSimpleModel, QwenDocTurbo, ReActModel, toolMap, WanxSketchToImageLiteModel, QwenImageModel, QwenImageEditModel, Qwen25T2iPreviewModel, Qwen3VlPlus, QwenVlMax, QwenStyleRepaintV1Model, };
+export { DefaultSimpleModel, QwenDocTurbo, ReActModel, toolMap, WanxSketchToImageLiteModel, QwenImageModel, QwenImageEditModel, Qwen25T2iPreviewModel, Wan26ImageModel, Qwen3VlPlus, QwenVlMax, QwenStyleRepaintV1Model, };

package/lib/models/index.js

@@ -1,6 +1,6 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.QwenStyleRepaintV1Model = exports.QwenVlMax = exports.Qwen3VlPlus = exports.Qwen25T2iPreviewModel = exports.QwenImageEditModel = exports.QwenImageModel = exports.WanxSketchToImageLiteModel = exports.toolMap = exports.ReActModel = exports.QwenDocTurbo = exports.DefaultSimpleModel = exports.MODEL_REQUEST_PATTERNS = exports.isMultiModalModel = exports.isValidModel = exports.modelName = exports.MultiModalModelName = exports.ModelName = void 0;
+exports.QwenStyleRepaintV1Model = exports.QwenVlMax = exports.Qwen3VlPlus = exports.Wan26ImageModel = exports.Qwen25T2iPreviewModel = exports.QwenImageEditModel = exports.QwenImageModel = exports.WanxSketchToImageLiteModel = exports.toolMap = exports.ReActModel = exports.QwenDocTurbo = exports.DefaultSimpleModel = exports.MODEL_REQUEST_PATTERNS = exports.isMultiModalModel = exports.isValidModel = exports.modelName = exports.MultiModalModelName = exports.ModelName = void 0;
 exports.isModelRequestUrl = isModelRequestUrl;
 const Default_1 = require("./Default");
 Object.defineProperty(exports, "DefaultSimpleModel", { enumerable: true, get: function () { return Default_1.DefaultSimpleModel; } });
@@ -20,6 +20,8 @@ const QwenStyleRepaintV1_1 = require("./QwenStyleRepaintV1");
 Object.defineProperty(exports, "QwenStyleRepaintV1Model", { enumerable: true, get: function () { return QwenStyleRepaintV1_1.QwenStyleRepaintV1Model; } });
 const QwenVlMax_1 = require("./QwenVlMax");
 Object.defineProperty(exports, "QwenVlMax", { enumerable: true, get: function () { return QwenVlMax_1.QwenVlMax; } });
+const Wan26Image_1 = require("./Wan26Image");
+Object.defineProperty(exports, "Wan26ImageModel", { enumerable: true, get: function () { return Wan26Image_1.Wan26ImageModel; } });
 const react_1 = require("./react");
 Object.defineProperty(exports, "ReActModel", { enumerable: true, get: function () { return react_1.ReActModel; } });
 var ModelName;
@@ -46,6 +48,7 @@ var MultiModalModelName;
     MultiModalModelName["QwenImageEdit"] = "qwen-image-edit";
     MultiModalModelName["QwenImageEditPlus"] = "qwen-image-edit-plus";
     MultiModalModelName["Qwen25T2iPreview"] = "wan2.5-i2i-preview";
+    MultiModalModelName["Wan26Image"] = "wan2.6-image";
     MultiModalModelName["WanxSketchToImageLite"] = "wanx-sketch-to-image-lite";
     MultiModalModelName["QwenStyleRepaintV1"] = "wanx-style-repaint-v1";
 })(MultiModalModelName || (exports.MultiModalModelName = MultiModalModelName = {}));
@@ -68,6 +71,7 @@ exports.modelName = {
     [MultiModalModelName.QwenImageEdit]: "qwen-image-edit",
     [MultiModalModelName.QwenImageEditPlus]: "qwen-image-edit-plus",
     [MultiModalModelName.Qwen25T2iPreview]: "wan2.5-i2i-preview",
+    [MultiModalModelName.Wan26Image]: "wan2.6-image",
     [MultiModalModelName.WanxSketchToImageLite]: "wanx-sketch-to-image-lite",
     [MultiModalModelName.QwenStyleRepaintV1]: "wanx-style-repaint-v1",
 };
@@ -122,6 +126,16 @@ exports.MODEL_REQUEST_PATTERNS = new Map([
             paths: ["api/v1/services/aigc/image2image/image-synthesis"],
         },
     ],
+    [
+        "wan2.6-image",
+        {
+            domain: "https://dashscope.aliyuncs.com",
+            paths: [
+                "api/v1/services/aigc/image-generation/generation",
+                "api/v1/services/aigc/multimodal-generation/generation",
+            ],
+        },
+    ],
     [
         "qwen-vl-max",
         {

package/lib/tokenManager.js

@@ -25,6 +25,10 @@ class TokenManager {
     }
     getValidToken() {
         return __awaiter(this, void 0, void 0, function* () {
+            if (process.env && process.env.RED_LOOK_API_KEY_TEMP) {
+                console.log("-- redlook temp key fixed --");
+                return process.env.RED_LOOK_API_KEY_TEMP;
+            }
             const now = Math.floor(Date.now() / 1000);
             if (this.currentToken && this.tokenExpiresAt > now + this.refreshBufferSeconds) {
                 return this.currentToken;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vectorx/ai-sdk",
-  "version": "0.7.0",
+  "version": "0.8.1",
   "description": "Cloud AI SDK",
   "main": "lib/index.js",
   "sideEffects": false,
@@ -22,7 +22,7 @@
   },
   "dependencies": {
     "@mattiasbuelens/web-streams-adapter": "^0.1.0",
-    "@vectorx/ai-types": "0.6.0",
+    "@vectorx/ai-types": "0.7.1",
     "langfuse": "^3.38.4",
     "openai": "^4.103.0",
     "text-encoding-shim": "^1.0.5",