@vectorx/ai-sdk 1.0.0 → 1.1.0

package/lib/models/index.d.ts

@@ -1,12 +1,15 @@
 import { DefaultSimpleModel } from "./Default";
 import { Qwen3VlPlus } from "./Qwen3VlPlus";
 import { Qwen25T2iPreviewModel } from "./Qwen25T2iPreview";
+import { AliyunCosyVoiceTTSModel } from "./QwenCosyVoiceTTS";
 import { QwenDocTurbo } from "./QwenDocTurbo";
 import { QwenImageModel } from "./QwenImage";
+import { QwenImage20FastModel, QwenImage20ProModel } from "./QwenImage20";
 import { QwenImageEditModel } from "./QwenImageEdit";
 import { WanxSketchToImageLiteModel } from "./QwenSketchToImage";
 import { QwenStyleRepaintV1Model } from "./QwenStyleRepaintV1";
 import { QwenVlMax } from "./QwenVlMax";
+import { SimpleTTSModel } from "./TTSModel";
 import { Wan26ImageModel } from "./Wan26Image";
 import { ReActModel } from "./react";
 export declare enum ModelName {
@@ -19,6 +22,14 @@ export declare enum ModelName {
     QwenVlOcr = "qwen-vl-ocr",
     Wanx21T2iTurbo = "wanx2.1-t2i-turbo"
 }
+export declare enum TTSModelName {
+    CosyVoiceV35Flash = "cosyvoice-v3.5-flash",
+    CosyVoiceV35Plus = "cosyvoice-v3.5-plus",
+    CosyVoiceV3Flash = "cosyvoice-v3-flash",
+    CosyVoiceV3Plus = "cosyvoice-v3-plus",
+    CosyVoiceV2 = "cosyvoice-v2",
+    CosyVoiceV1 = "cosyvoice-v1"
+}
 export declare enum MultiModalModelName {
     QvqMaxLatest = "qvq-max-latest",
     QwenVlMax = "qwen-vl-max",
@@ -33,7 +44,9 @@ export declare enum MultiModalModelName {
     Qwen25T2iPreview = "wan2.5-i2i-preview",
     Wan26Image = "wan2.6-image",
     WanxSketchToImageLite = "wanx-sketch-to-image-lite",
-    QwenStyleRepaintV1 = "wanx-style-repaint-v1"
+    QwenStyleRepaintV1 = "wanx-style-repaint-v1",
+    QwenImage20Pro = "qwen-image-2.0-pro",
+    QwenImage20 = "qwen-image-2.0"
 }
 export declare const modelName: {
     "deepseek-r1": string;
@@ -43,6 +56,12 @@ export declare const modelName: {
     "qwen-plus": string;
     "qwen-turbo": string;
     "qwen-vl-ocr": string;
+    "cosyvoice-v3.5-flash": string;
+    "cosyvoice-v3.5-plus": string;
+    "cosyvoice-v3-flash": string;
+    "cosyvoice-v3-plus": string;
+    "cosyvoice-v2": string;
+    "cosyvoice-v1": string;
     "qvq-max-latest": string;
     "qwen-vl-max": string;
     "qwen3-vl-plus": string;
@@ -57,13 +76,16 @@ export declare const modelName: {
     "wan2.6-image": string;
     "wanx-sketch-to-image-lite": string;
     "wanx-style-repaint-v1": string;
+    "qwen-image-2.0-pro": string;
+    "qwen-image-2.0": string;
 };
-export declare const isValidModel: (model: ModelName | MultiModalModelName) => model is ModelName | MultiModalModelName;
+export declare const isValidModel: (model: ModelName | MultiModalModelName | TTSModelName) => model is ModelName | MultiModalModelName | TTSModelName;
 export declare const isMultiModalModel: (model: ModelName | MultiModalModelName) => model is MultiModalModelName;
+export declare const isTTSModel: (model: ModelName | MultiModalModelName | TTSModelName) => model is TTSModelName;
 declare const toolMap: Map<string, Function>;
 export declare const MODEL_REQUEST_PATTERNS: Map<string, {
     domain: string;
     paths: string[];
 }>;
 export declare function isModelRequestUrl(url: string): boolean;
-export { DefaultSimpleModel, QwenDocTurbo, ReActModel, toolMap, WanxSketchToImageLiteModel, QwenImageModel, QwenImageEditModel, Qwen25T2iPreviewModel, Wan26ImageModel, Qwen3VlPlus, QwenVlMax, QwenStyleRepaintV1Model, };
+export { SimpleTTSModel, AliyunCosyVoiceTTSModel, DefaultSimpleModel, QwenDocTurbo, ReActModel, toolMap, WanxSketchToImageLiteModel, QwenImageModel, QwenImageEditModel, QwenImage20ProModel, QwenImage20FastModel, 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.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.QwenStyleRepaintV1Model = exports.QwenVlMax = exports.Qwen3VlPlus = exports.Wan26ImageModel = exports.Qwen25T2iPreviewModel = exports.QwenImage20FastModel = exports.QwenImage20ProModel = exports.QwenImageEditModel = exports.QwenImageModel = exports.WanxSketchToImageLiteModel = exports.toolMap = exports.ReActModel = exports.QwenDocTurbo = exports.DefaultSimpleModel = exports.AliyunCosyVoiceTTSModel = exports.SimpleTTSModel = exports.MODEL_REQUEST_PATTERNS = exports.isTTSModel = exports.isMultiModalModel = exports.isValidModel = exports.modelName = exports.MultiModalModelName = exports.TTSModelName = exports.ModelName = void 0;
 exports.isModelRequestUrl = isModelRequestUrl;
 const Default_1 = require("./Default");
 Object.defineProperty(exports, "DefaultSimpleModel", { enumerable: true, get: function () { return Default_1.DefaultSimpleModel; } });
@@ -8,10 +8,15 @@ const Qwen3VlPlus_1 = require("./Qwen3VlPlus");
 Object.defineProperty(exports, "Qwen3VlPlus", { enumerable: true, get: function () { return Qwen3VlPlus_1.Qwen3VlPlus; } });
 const Qwen25T2iPreview_1 = require("./Qwen25T2iPreview");
 Object.defineProperty(exports, "Qwen25T2iPreviewModel", { enumerable: true, get: function () { return Qwen25T2iPreview_1.Qwen25T2iPreviewModel; } });
+const QwenCosyVoiceTTS_1 = require("./QwenCosyVoiceTTS");
+Object.defineProperty(exports, "AliyunCosyVoiceTTSModel", { enumerable: true, get: function () { return QwenCosyVoiceTTS_1.AliyunCosyVoiceTTSModel; } });
 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 QwenImage20_1 = require("./QwenImage20");
+Object.defineProperty(exports, "QwenImage20FastModel", { enumerable: true, get: function () { return QwenImage20_1.QwenImage20FastModel; } });
+Object.defineProperty(exports, "QwenImage20ProModel", { enumerable: true, get: function () { return QwenImage20_1.QwenImage20ProModel; } });
 const QwenImageEdit_1 = require("./QwenImageEdit");
 Object.defineProperty(exports, "QwenImageEditModel", { enumerable: true, get: function () { return QwenImageEdit_1.QwenImageEditModel; } });
 const QwenSketchToImage_1 = require("./QwenSketchToImage");
@@ -20,6 +25,8 @@ 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 TTSModel_1 = require("./TTSModel");
+Object.defineProperty(exports, "SimpleTTSModel", { enumerable: true, get: function () { return TTSModel_1.SimpleTTSModel; } });
 const Wan26Image_1 = require("./Wan26Image");
 Object.defineProperty(exports, "Wan26ImageModel", { enumerable: true, get: function () { return Wan26Image_1.Wan26ImageModel; } });
 const react_1 = require("./react");
@@ -35,6 +42,15 @@ var ModelName;
     ModelName["QwenVlOcr"] = "qwen-vl-ocr";
     ModelName["Wanx21T2iTurbo"] = "wanx2.1-t2i-turbo";
 })(ModelName || (exports.ModelName = ModelName = {}));
+var TTSModelName;
+(function (TTSModelName) {
+    TTSModelName["CosyVoiceV35Flash"] = "cosyvoice-v3.5-flash";
+    TTSModelName["CosyVoiceV35Plus"] = "cosyvoice-v3.5-plus";
+    TTSModelName["CosyVoiceV3Flash"] = "cosyvoice-v3-flash";
+    TTSModelName["CosyVoiceV3Plus"] = "cosyvoice-v3-plus";
+    TTSModelName["CosyVoiceV2"] = "cosyvoice-v2";
+    TTSModelName["CosyVoiceV1"] = "cosyvoice-v1";
+})(TTSModelName || (exports.TTSModelName = TTSModelName = {}));
 var MultiModalModelName;
 (function (MultiModalModelName) {
     MultiModalModelName["QvqMaxLatest"] = "qvq-max-latest";
@@ -51,6 +67,8 @@ var MultiModalModelName;
     MultiModalModelName["Wan26Image"] = "wan2.6-image";
     MultiModalModelName["WanxSketchToImageLite"] = "wanx-sketch-to-image-lite";
     MultiModalModelName["QwenStyleRepaintV1"] = "wanx-style-repaint-v1";
+    MultiModalModelName["QwenImage20Pro"] = "qwen-image-2.0-pro";
+    MultiModalModelName["QwenImage20"] = "qwen-image-2.0";
 })(MultiModalModelName || (exports.MultiModalModelName = MultiModalModelName = {}));
 exports.modelName = {
     [ModelName.DeepSeekR1]: "deepseek-r1",
@@ -60,6 +78,12 @@ exports.modelName = {
     [ModelName.QwenPlus]: "qwen-plus",
     [ModelName.QwenTurbo]: "qwen-turbo",
     [ModelName.QwenVlOcr]: "qwen-vl-ocr",
+    [TTSModelName.CosyVoiceV35Flash]: "cosyvoice-v3.5-flash",
+    [TTSModelName.CosyVoiceV35Plus]: "cosyvoice-v3.5-plus",
+    [TTSModelName.CosyVoiceV3Flash]: "cosyvoice-v3-flash",
+    [TTSModelName.CosyVoiceV3Plus]: "cosyvoice-v3-plus",
+    [TTSModelName.CosyVoiceV2]: "cosyvoice-v2",
+    [TTSModelName.CosyVoiceV1]: "cosyvoice-v1",
     [MultiModalModelName.QvqMaxLatest]: "qvq-max-latest",
     [MultiModalModelName.QwenVlMax]: "qwen-vl-max",
     [MultiModalModelName.Qwen3VlPlus]: "qwen3-vl-plus",
@@ -74,6 +98,8 @@ exports.modelName = {
     [MultiModalModelName.Wan26Image]: "wan2.6-image",
     [MultiModalModelName.WanxSketchToImageLite]: "wanx-sketch-to-image-lite",
     [MultiModalModelName.QwenStyleRepaintV1]: "wanx-style-repaint-v1",
+    [MultiModalModelName.QwenImage20Pro]: "qwen-image-2.0-pro",
+    [MultiModalModelName.QwenImage20]: "qwen-image-2.0",
 };
 const isValidModel = (model) => {
     const modelNames = Object.values(exports.modelName);
@@ -84,6 +110,10 @@ const isMultiModalModel = (model) => {
     return Object.values(MultiModalModelName).includes(model);
 };
 exports.isMultiModalModel = isMultiModalModel;
+const isTTSModel = (model) => {
+    return Object.values(TTSModelName).includes(model);
+};
+exports.isTTSModel = isTTSModel;
 const toolMap = new Map();
 exports.toolMap = toolMap;
 exports.MODEL_REQUEST_PATTERNS = new Map([
@@ -136,6 +166,20 @@ exports.MODEL_REQUEST_PATTERNS = new Map([
             ],
         },
     ],
+    [
+        "qwen-image-2.0-pro",
+        {
+            domain: "https://dashscope.aliyuncs.com",
+            paths: ["api/v1/services/aigc/multimodal-generation/generation"],
+        },
+    ],
+    [
+        "qwen-image-2.0",
+        {
+            domain: "https://dashscope.aliyuncs.com",
+            paths: ["api/v1/services/aigc/multimodal-generation/generation"],
+        },
+    ],
     [
         "qwen-vl-max",
         {
@@ -154,7 +198,10 @@ exports.MODEL_REQUEST_PATTERNS = new Map([
         "qwen-doc-turbo",
         {
             domain: "https://dashscope.aliyuncs.com",
-            paths: ["api/v1/services/aigc/text-generation/generation", "compatible-mode/v1/chat/completions"],
+            paths: [
+                "api/v1/services/aigc/text-generation/generation",
+                "compatible-mode/v1/chat/completions",
+            ],
         },
     ],
 ]);

package/lib/models/react.d.ts

@@ -1,8 +1,12 @@
-import type { DoGenerateOutput, DoStreamOutput, ModelRequestOptions, ReqOptions } from "../model-type";
+import type { DoGenerateOutput, DoGenerateSpeechOutput, DoStreamOutput, DoStreamSpeechOutput, ModelRequestOptions, ReqOptions, TTSRequestOptions } from "../model-type";
 import type { SimpleChatModel } from "./Chat";
+import type { SimpleTTSModel } from "./TTSModel";
 export declare class ReActModel {
     private model;
-    constructor(model: SimpleChatModel);
+    private ttsModel;
+    constructor(model: SimpleChatModel | SimpleTTSModel);
     generateText(_input: ModelRequestOptions, options?: ReqOptions): Promise<DoGenerateOutput>;
     streamText(_input: ModelRequestOptions, options?: ReqOptions): Promise<DoStreamOutput>;
+    streamSpeech(options: TTSRequestOptions): Promise<DoStreamSpeechOutput>;
+    generateSpeech(options: TTSRequestOptions): Promise<DoGenerateSpeechOutput>;
 }

package/lib/models/react.js

@@ -13,6 +13,10 @@ exports.ReActModel = void 0;
 class ReActModel {
     constructor(model) {
         this.model = model;
+        this.ttsModel = null;
+        if ("doStreamSpeech" in model) {
+            this.ttsModel = model;
+        }
     }
     generateText(_input, options) {
         return __awaiter(this, void 0, void 0, function* () {
@@ -24,5 +28,23 @@ class ReActModel {
             return this.model.doStream(_input, options);
         });
     }
+    streamSpeech(options) {
+        return __awaiter(this, void 0, void 0, function* () {
+            console.log("[ReActModel:streamSpeech] called, hasTtsModel:", !!this.ttsModel);
+            if (!this.ttsModel) {
+                throw new Error("streamSpeech is only available for TTS models (e.g. cosyvoice-v3-flash)");
+            }
+            return this.ttsModel.doStreamSpeech(options);
+        });
+    }
+    generateSpeech(options) {
+        return __awaiter(this, void 0, void 0, function* () {
+            console.log("[ReActModel:generateSpeech] called, hasTtsModel:", !!this.ttsModel);
+            if (!this.ttsModel) {
+                throw new Error("generateSpeech is only available for TTS models (e.g. cosyvoice-v3-flash)");
+            }
+            return this.ttsModel.doGenerateSpeech(options);
+        });
+    }
 }
 exports.ReActModel = ReActModel;

package/lib/stream.js

@@ -42,7 +42,9 @@ class TextDecoderStream {
     constructor(encoding = "utf-8", options = {}) {
         this.transform = new exports.TransformStream({
             transform: (chunk, controller) => {
-                const value = this.handle.decode(new Uint8Array(chunk), { stream: true });
+                const value = this.handle.decode(new Uint8Array(chunk), {
+                    stream: true,
+                });
                 if (value) {
                     controller.enqueue(value);
                 }
@@ -98,17 +100,18 @@ function intoStandardStream(stream) {
         .pipeThrough((0, exports.createEventSourceParserTransformStream)())
         .pipeThrough(new exports.TransformStream({
         transform(chunk, controller) {
+            var _a;
             try {
+                if (chunk.data === "[DONE]") {
+                    console.log("[intoStandardStream] received [DONE], terminating");
+                    controller.terminate();
+                    return;
+                }
                 const data = JSON.parse(chunk.data);
                 controller.enqueue(data);
             }
             catch (e) {
-                if (chunk.data !== "[DONE]") {
-                    console.warn("Error when transforming event source data to json", e, chunk);
-                }
-                else {
-                    controller.terminate();
-                }
+                console.warn("[intoStandardStream] Error parsing SSE data:", e, "raw:", (_a = chunk.data) === null || _a === void 0 ? void 0 : _a.slice(0, 200));
             }
         },
     })));

package/lib/tokenManager.js

@@ -25,15 +25,25 @@ class TokenManager {
     }
     getValidToken() {
         return __awaiter(this, void 0, void 0, function* () {
+            const _debug = process.env.VECTORX_DEBUG === "1";
+            if (_debug)
+                console.log("[TokenManager:getValidToken] fullUrl:", this.fullUrl);
             if (process.env && process.env.RED_LOOK_API_KEY_TEMP) {
-                console.log("-- redlook temp key fixed --");
+                if (_debug)
+                    console.log("[TokenManager] using RED_LOOK_API_KEY_TEMP env var");
                 return process.env.RED_LOOK_API_KEY_TEMP;
             }
             const now = Math.floor(Date.now() / 1000);
             if (this.currentToken && this.tokenExpiresAt > now + this.refreshBufferSeconds) {
+                if (_debug)
+                    console.log("[TokenManager] reusing cached token, expires in:", this.tokenExpiresAt - now, "s");
                 return this.currentToken;
             }
+            if (_debug)
+                console.log("[TokenManager] refreshing token...");
             yield this.refreshToken();
+            if (_debug)
+                console.log("[TokenManager] token refreshed successfully");
             return this.currentToken;
         });
     }

package/package.json

@@ -1,12 +1,12 @@
 {
   "name": "@vectorx/ai-sdk",
-  "version": "1.0.0",
+  "version": "1.1.0",
   "description": "Cloud AI SDK",
   "main": "lib/index.js",
+  "types": "lib/index.d.ts",
   "sideEffects": false,
   "files": [
     "lib",
-    "types",
     "README.md"
   ],
   "keywords": [
@@ -18,20 +18,22 @@
   "author": "",
   "license": "ISC",
   "engines": {
-    "node": ">=18.0.0"
+    "node": ">=20.0.0"
   },
   "dependencies": {
     "@mattiasbuelens/web-streams-adapter": "^0.1.0",
-    "@vectorx/ai-types": "1.0.0",
     "langfuse": "^3.38.4",
     "openai": "^4.103.0",
     "text-encoding-shim": "^1.0.5",
     "web-streams-polyfill": "^4.1.0",
-    "zod": "^3.24.2"
+    "ws": "^8.18.0",
+    "zod": "^3.24.2",
+    "@vectorx/ai-types": "1.1.0"
   },
   "devDependencies": {
     "@types/jest": "^29.5.12",
     "@types/node": "^20.11.24",
+    "@types/ws": "^8.5.13",
     "@typescript-eslint/eslint-plugin": "^7.1.0",
     "@typescript-eslint/parser": "^7.1.0",
     "eslint": "^8.57.0",
@@ -42,8 +44,8 @@
   "scripts": {
     "dev": "tsc -w",
     "build": "rimraf lib types && tsc",
+    "scan:models": "python3 scripts/scan-models.py",
     "lint": "eslint 'src/**/*.ts'",
     "fix": "eslint 'src/**/*.ts' --fix"
-  },
-  "readme": "### 小程序/智能体 ai sdk "
+  }
 }