@vectorx/ai-sdk 0.4.0 → 0.5.0

package/lib/ai.js

@@ -57,6 +57,8 @@ class AI {
         switch (model) {
             case models.MultiModalModelName.QwenImage:
                 return new models.ReActModel(new models.QwenImageModel(this.request, models.QwenImageModel.BASE_URL, this.tokenManager));
+            case models.MultiModalModelName.QwenImageEdit:
+                return new models.ReActModel(new models.QwenImageEditModel(this.request, models.QwenImageEditModel.BASE_URL, this.tokenManager));
             case models.MultiModalModelName.WanxSketchToImageLite:
                 return new models.ReActModel(new models.WanxSketchToImageLiteModel(this.request, models.WanxSketchToImageLiteModel.BASE_URL, this.tokenManager));
             case models.MultiModalModelName.QwenStyleRepaintV1:

package/lib/models/QwenDocTurbo/index.js

@@ -27,13 +27,6 @@ class QwenDocTurbo extends Chat_1.SimpleChatModel {
             if (isStreaming) {
                 fetchHeaders["X-DashScope-SSE"] = "enable";
             }
-            console.log("=== QwenDocTurbo Request ===", {
-                method: "post",
-                headers: Object.assign({}, fetchHeaders),
-                body: JSON.stringify(data),
-                url: `${this.baseUrl}/${this.subUrl}`,
-                stream: isStreaming,
-            });
             const { data: responseData, header } = (yield this.req.fetch({
                 method: "post",
                 headers: Object.assign({}, fetchHeaders),

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

@@ -0,0 +1,77 @@
+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 QwenImageEditParameters {
+    negative_prompt?: string;
+    watermark?: boolean;
+}
+export interface QwenImageEditAPIInput {
+    model: string;
+    input: {
+        messages: Array<{
+            role: "user";
+            content: Array<{
+                image?: string;
+                text?: string;
+            }>;
+        }>;
+    };
+    parameters?: QwenImageEditParameters;
+}
+export type QwenImageEditRequestOptions = QwenImageEditAPIInput & {
+    parameters?: QwenImageEditParameters;
+};
+export type QwenImageEditContentItem = {
+    image?: string;
+    url?: string;
+    image_url?: string | {
+        url: string;
+    };
+    b64_json?: string;
+    [key: string]: any;
+};
+export interface QwenImageEditAPIResponse {
+    async?: boolean;
+    output: {
+        choices?: Array<{
+            finish_reason: string;
+            message: {
+                role: "assistant" | "user";
+                content: QwenImageEditContentItem[];
+            };
+        }>;
+        task_status?: string;
+        task_id?: string;
+        task_metric?: {
+            TOTAL: number;
+            FAILED: number;
+            SUCCEEDED: number;
+        };
+    };
+    usage?: {
+        width?: number;
+        height?: number;
+        image_count?: number;
+    };
+    request_id?: string;
+    id?: string;
+    model?: string;
+    created?: number;
+    object?: string;
+    code?: number;
+    error?: string;
+}
+export declare class QwenImageEditModel extends SimpleChatModel {
+    static BASE_URL: string;
+    static SUB_GENERATION_URL: string;
+    modelName: string;
+    constructor(req: IAbstractRequest, baseUrl: string, tokenManager: TokenManager);
+    protected normalizeStandardImageEditCompletion(res: QwenImageEditAPIResponse, fallbackModel: string): DoGenerateOutput;
+    protected coverModelRequestToQwenInput(data: ModelRequestOptions & {
+        parameters?: QwenImageEditParameters;
+    }): QwenImageEditRequestOptions;
+    protected modelRequest(data: QwenImageEditRequestOptions, options?: ReqOptions): Promise<ReadableStream<Uint8Array> | Promise<unknown>>;
+    doGenerate(data: ModelRequestOptions, options?: ReqOptions): Promise<DoGenerateOutput>;
+    doStream(data: ModelRequestOptions, options?: ReqOptions): Promise<DoStreamOutput>;
+}

package/lib/models/QwenImageEdit/index.js

@@ -0,0 +1,205 @@
+"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.QwenImageEditModel = void 0;
+const stream_1 = require("../../stream");
+const utils_1 = require("../../utils");
+const Chat_1 = require("../Chat");
+class QwenImageEditModel extends Chat_1.SimpleChatModel {
+    constructor(req, baseUrl, tokenManager) {
+        super(req, baseUrl, QwenImageEditModel.SUB_GENERATION_URL, tokenManager);
+        this.modelName = "qwen-image-edit";
+    }
+    normalizeStandardImageEditCompletion(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: (_d = res === null || res === void 0 ? void 0 : res.error) !== null && _d !== void 0 ? _d : "",
+                code: (_e = res === null || res === void 0 ? void 0 : res.code) !== null && _e !== void 0 ? _e : 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 first = ((_g = (_f = qOutput === null || qOutput === void 0 ? void 0 : qOutput.choices) === null || _f === void 0 ? void 0 : _f[0]) !== null && _g !== void 0 ? _g : null);
+        const message = (_h = first === null || first === void 0 ? void 0 : first.message) !== null && _h !== void 0 ? _h : {};
+        const contentUrl = Array.isArray(message === null || message === void 0 ? void 0 : message.content) && ((_j = message.content[0]) === null || _j === void 0 ? void 0 : _j.image) ? String(message.content[0].image) : "";
+        const created = (_k = res === null || res === void 0 ? void 0 : res.created) !== null && _k !== void 0 ? _k : 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: (_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: (_o = res === null || res === void 0 ? void 0 : res.error) !== null && _o !== void 0 ? _o : "",
+            code: (_p = res === null || res === void 0 ? void 0 : res.code) !== null && _p !== void 0 ? _p : 0,
+            choices: [
+                {
+                    index: 0,
+                    message: {
+                        id,
+                        role: "assistant",
+                        type: "image",
+                        content: contentUrl || "",
+                        reasoning_content: "",
+                    },
+                    finish_reason: (_q = first === null || first === void 0 ? void 0 : first.finish_reason) !== null && _q !== void 0 ? _q : "stop",
+                },
+            ],
+            usage: mapUsageToStandard(res === null || res === void 0 ? void 0 : res.usage),
+        };
+        return normalized;
+    }
+    coverModelRequestToQwenInput(data) {
+        var _a, _b;
+        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 (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(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 (texts.length === 0 && data.msg) {
+            texts.push(String(data.msg));
+        }
+        if (texts.length === 0 && data.prompt) {
+            texts.push(String(data.prompt));
+        }
+        if (imageUrls.length === 0)
+            throw new Error("QwenImageEdit 需要提供至少一个图片 URL");
+        if (texts.length === 0)
+            throw new Error("QwenImageEdit 需要提供至少一个文本提示词");
+        const content = [];
+        imageUrls.forEach((url) => {
+            content.push({ image: url });
+        });
+        texts.forEach((text) => {
+            content.push({ text });
+        });
+        return {
+            parameters: data.parameters,
+            model: this.modelName,
+            input: {
+                messages: [
+                    {
+                        role: "user",
+                        content,
+                    },
+                ],
+            },
+        };
+    }
+    modelRequest(data_1) {
+        return __awaiter(this, arguments, void 0, function* (data, options = { timeout: 30 * 1000 }) {
+            const fetchHeaders = yield this.createAuthHeaders(options === null || options === void 0 ? void 0 : options.headers);
+            const joinedUrl = `${String(this.baseUrl).replace(/\/+$/, "")}/${String(this.subUrl).replace(/^\/+/, "")}`;
+            const { data: responseData, header } = (yield this.req.fetch({
+                url: joinedUrl,
+                headers: Object.assign({}, fetchHeaders),
+                body: JSON.stringify(data),
+                method: "post",
+                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 payload = this.coverModelRequestToQwenInput(data);
+            const res = (yield this.modelRequest(payload, options));
+            return this.normalizeStandardImageEditCompletion(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 || "image",
+                            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.QwenImageEditModel = QwenImageEditModel;
+QwenImageEditModel.BASE_URL = "https://dashscope.aliyuncs.com";
+QwenImageEditModel.SUB_GENERATION_URL = "api/v1/services/aigc/multimodal-generation/generation";
+function mapUsageToStandard(usage) {
+    return {
+        prompt_tokens: 0,
+        completion_tokens: 0,
+        knowledge_tokens: 0,
+        reasoning_tokens: 0,
+        total_tokens: 0,
+    };
+}

package/lib/models/index.d.ts

@@ -1,6 +1,7 @@
 import { DefaultSimpleModel } from "./Default";
 import { QwenDocTurbo } from "./QwenDocTurbo";
 import { QwenImageModel } from "./QwenImage";
+import { QwenImageEditModel } from "./QwenImageEdit";
 import { WanxSketchToImageLiteModel } from "./QwenSketchToImage";
 import { QwenStyleRepaintV1Model } from "./QwenStyleRepaintV1";
 import { QwenVlMax } from "./QwenVlMax";
@@ -23,6 +24,7 @@ export declare enum MultiModalModelName {
     Wanx21T2iPlus = "wanx2.1-t2i-plus",
     Wanx21T2iTurbo = "wanx2.1-t2i-turbo",
     QwenImage = "qwen-image",
+    QwenImageEdit = "qwen-image-edit",
     WanxSketchToImageLite = "wanx-sketch-to-image-lite",
     QwenStyleRepaintV1 = "wanx-style-repaint-v1"
 }
@@ -41,10 +43,11 @@ export declare const modelName: {
     "wanx2.1-t2i-plus": string;
     "wanx2.1-t2i-turbo": string;
     "qwen-image": string;
+    "qwen-image-edit": string;
     "wanx-sketch-to-image-lite": string;
     "wanx-style-repaint-v1": string;
 };
 export declare const isValidModel: (model: ModelName | MultiModalModelName) => model is ModelName | MultiModalModelName;
 export declare const isMultiModalModel: (model: ModelName | MultiModalModelName) => model is MultiModalModelName;
 declare const toolMap: Map<string, Function>;
-export { DefaultSimpleModel, QwenDocTurbo, ReActModel, toolMap, WanxSketchToImageLiteModel, QwenImageModel, QwenVlMax, QwenStyleRepaintV1Model, };
+export { DefaultSimpleModel, QwenDocTurbo, ReActModel, toolMap, WanxSketchToImageLiteModel, QwenImageModel, QwenImageEditModel, QwenVlMax, QwenStyleRepaintV1Model, };

package/lib/models/index.js

@@ -1,12 +1,14 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.QwenStyleRepaintV1Model = exports.QwenVlMax = exports.QwenImageModel = exports.WanxSketchToImageLiteModel = exports.toolMap = exports.ReActModel = exports.QwenDocTurbo = exports.DefaultSimpleModel = exports.isMultiModalModel = exports.isValidModel = exports.modelName = exports.MultiModalModelName = exports.ModelName = void 0;
+exports.QwenStyleRepaintV1Model = exports.QwenVlMax = exports.QwenImageEditModel = exports.QwenImageModel = exports.WanxSketchToImageLiteModel = exports.toolMap = exports.ReActModel = exports.QwenDocTurbo = exports.DefaultSimpleModel = exports.isMultiModalModel = exports.isValidModel = exports.modelName = exports.MultiModalModelName = exports.ModelName = void 0;
 const Default_1 = require("./Default");
 Object.defineProperty(exports, "DefaultSimpleModel", { enumerable: true, get: function () { return Default_1.DefaultSimpleModel; } });
 const QwenDocTurbo_1 = require("./QwenDocTurbo");
 Object.defineProperty(exports, "QwenDocTurbo", { enumerable: true, get: function () { return QwenDocTurbo_1.QwenDocTurbo; } });
 const QwenImage_1 = require("./QwenImage");
 Object.defineProperty(exports, "QwenImageModel", { enumerable: true, get: function () { return QwenImage_1.QwenImageModel; } });
+const QwenImageEdit_1 = require("./QwenImageEdit");
+Object.defineProperty(exports, "QwenImageEditModel", { enumerable: true, get: function () { return QwenImageEdit_1.QwenImageEditModel; } });
 const QwenSketchToImage_1 = require("./QwenSketchToImage");
 Object.defineProperty(exports, "WanxSketchToImageLiteModel", { enumerable: true, get: function () { return QwenSketchToImage_1.WanxSketchToImageLiteModel; } });
 const QwenStyleRepaintV1_1 = require("./QwenStyleRepaintV1");
@@ -35,6 +37,7 @@ var MultiModalModelName;
     MultiModalModelName["Wanx21T2iPlus"] = "wanx2.1-t2i-plus";
     MultiModalModelName["Wanx21T2iTurbo"] = "wanx2.1-t2i-turbo";
     MultiModalModelName["QwenImage"] = "qwen-image";
+    MultiModalModelName["QwenImageEdit"] = "qwen-image-edit";
     MultiModalModelName["WanxSketchToImageLite"] = "wanx-sketch-to-image-lite";
     MultiModalModelName["QwenStyleRepaintV1"] = "wanx-style-repaint-v1";
 })(MultiModalModelName || (exports.MultiModalModelName = MultiModalModelName = {}));
@@ -53,6 +56,7 @@ exports.modelName = {
     [MultiModalModelName.Wanx21T2iPlus]: "wanx2.1-t2i-plus",
     [MultiModalModelName.Wanx21T2iTurbo]: "wanx2.1-t2i-turbo",
     [MultiModalModelName.QwenImage]: "qwen-image",
+    [MultiModalModelName.QwenImageEdit]: "qwen-image-edit",
     [MultiModalModelName.WanxSketchToImageLite]: "wanx-sketch-to-image-lite",
     [MultiModalModelName.QwenStyleRepaintV1]: "wanx-style-repaint-v1",
 };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vectorx/ai-sdk",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "description": "Cloud AI SDK",
   "main": "lib/index.js",
   "sideEffects": false,