@vectorx/ai-sdk 0.3.0 → 0.4.0

package/lib/ai.js

@@ -63,6 +63,8 @@ class AI {
                 return new models.ReActModel(new models.QwenStyleRepaintV1Model(this.request, models.QwenStyleRepaintV1Model.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.QwenDocTurbo:
+                return new models.ReActModel(new models.QwenDocTurbo(this.request, models.QwenDocTurbo.BASE_URL, model, this.tokenManager));
             default:
                 return new models.ReActModel(new models.DefaultSimpleModel(this.request, this.baseUrl, model));
         }

package/lib/model-type.d.ts

@@ -1,6 +1,6 @@
-import type { GetAgentInfoResponse, GetConversationsResponse, GetHistoryMessagesParams, GetHistoryMessagesResponse, QueryTasksResponse, SendMessageInput } from "@vectorx/ai-types";
+import type { GetAgentInfoResponse, GetConversationsResponse, GetHistoryMessagesParams, GetHistoryMessagesResponse, QueryTasksResponse, SendMessageInput, UploadFileInput, UploadFileResponse } from "@vectorx/ai-types";
 import type { ModelName } from "./models";
-export type { GetAgentInfoResponse, GetConversationsResponse, GetHistoryMessagesParams, GetHistoryMessagesResponse, SendMessageInput, QueryTasksResponse, };
+export type { GetAgentInfoResponse, GetConversationsResponse, GetHistoryMessagesParams, GetHistoryMessagesResponse, SendMessageInput, QueryTasksResponse, UploadFileInput, UploadFileResponse, };
 export type DoGenerateOutput = BaseDoGenerateOutput;
 export type DoStreamOutput = AsyncIterableReadableStream<BaseDoStreamOutputChunk>;
 export type AsyncIterableReadableStream<T> = ReadableStream<T> & {
@@ -8,6 +8,51 @@ export type AsyncIterableReadableStream<T> = ReadableStream<T> & {
         next(): Promise<IteratorResult<T>>;
     };
 };
+export interface TextContent {
+    type: "text";
+    text: string;
+}
+export interface ImageUrlContent {
+    type: "image_url";
+    image_url: {
+        url: string;
+        detail?: "auto" | "low" | "high";
+    };
+}
+export interface DocUrlContent {
+    type: "doc_url";
+    doc_url: string[];
+    file_parsing_strategy?: "auto" | "text_only";
+}
+export interface VideoUrlContent {
+    type: "video_url";
+    video_url: {
+        url: string;
+    };
+}
+export interface AudioUrlContent {
+    type: "audio_url";
+    audio_url: {
+        url: string;
+    };
+}
+export interface VideoContent {
+    type: "video";
+    video: string[];
+}
+export interface AudioContent {
+    type: "audio";
+    audio: string[];
+}
+export type MultiModalContent = TextContent | ImageUrlContent | DocUrlContent | VideoUrlContent | AudioUrlContent | VideoContent | AudioContent;
+export type SupportedContentType = MultiModalContent["type"];
+export declare function filterContentByTypes<T extends SupportedContentType>(content: MultiModalContent[], supportedTypes: T[]): Extract<MultiModalContent, {
+    type: T;
+}>[];
+export declare function hasContentType(content: string | MultiModalContent[], type: SupportedContentType): boolean;
+export declare function extractContentByType<T extends SupportedContentType>(content: string | MultiModalContent[], type: T): Extract<MultiModalContent, {
+    type: T;
+}>[];
 export interface ModelRequestOptions {
     async?: boolean;
     model: ModelName | string;
@@ -25,15 +70,7 @@ export interface ModelRequestOptions {
     }>;
     messages?: Array<{
         role: string;
-        content: string | Array<{
-            type: "text";
-            text: string;
-        } | {
-            type: "image_url";
-            image_url: {
-                url: string;
-            };
-        }>;
+        content: string | MultiModalContent[];
     }>;
     knowledge_base?: Array<{
         knowledge_base_id: string;
@@ -44,6 +81,7 @@ export interface ModelRequestOptions {
     db_base?: any[];
     enable_thinking?: boolean;
     enable_search?: boolean;
+    parameters?: any;
 }
 export type ModelReq = <T extends ModelRequestOptions>(props: T) => T["stream"] extends true ? Promise<ReadableStream<Uint8Array>> : Promise<Object>;
 export interface IAgentReqInput {
@@ -155,6 +193,7 @@ export interface IAgent {
     getAgentInfo?(): Promise<GetAgentInfoResponse>;
     getConversations?(): Promise<GetConversationsResponse>;
     queryTasks?(task_id?: string): Promise<QueryTasksResponse>;
+    uploadFile?(input: UploadFileInput): Promise<UploadFileResponse>;
 }
 export declare enum IAgentEnv {
     Production = "production",

package/lib/model-type.js

@@ -1,6 +1,22 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.IAgentEnv = void 0;
+exports.filterContentByTypes = filterContentByTypes;
+exports.hasContentType = hasContentType;
+exports.extractContentByType = extractContentByType;
+function filterContentByTypes(content, supportedTypes) {
+    return content.filter((item) => supportedTypes.includes(item.type));
+}
+function hasContentType(content, type) {
+    if (typeof content === "string")
+        return false;
+    return content.some((item) => item.type === type);
+}
+function extractContentByType(content, type) {
+    if (typeof content === "string")
+        return [];
+    return content.filter((item) => item.type === type);
+}
 var IAgentEnv;
 (function (IAgentEnv) {
     IAgentEnv["Production"] = "production";

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

@@ -0,0 +1,84 @@
+import type { IAbstractRequest } from "@vectorx/ai-types";
+import type { DoGenerateOutput, DoStreamOutput, DocUrlContent, ModelRequestOptions, ReqOptions, TextContent } from "../../model-type";
+import type { TokenManager } from "../../tokenManager";
+import { SimpleChatModel } from "../Chat";
+import type { MultiModalModelName } from "../index";
+export interface QwenDocTurboMessage {
+    role: "system" | "user" | "assistant";
+    content: string | Array<TextContent | DocUrlContent>;
+}
+export interface QwenDocTurboParameters {
+    temperature?: number;
+    top_p?: number;
+    max_tokens?: number;
+    n?: number;
+    stream?: boolean;
+    stop?: string | string[];
+    frequency_penalty?: number;
+    presence_penalty?: number;
+    user?: string;
+    stream_options?: {
+        include_usage?: boolean;
+    };
+    response_format?: {
+        type: "text" | "json_object";
+    };
+    tools?: Array<object>;
+    tool_choice?: string | object;
+    parallel_tool_calls?: boolean;
+}
+export interface QwenDocTurboAPIInput {
+    model: string;
+    input: {
+        messages: QwenDocTurboMessage[];
+    };
+    parameters?: QwenDocTurboParameters;
+}
+export interface QwenDocTurboResponse {
+    request_id?: string;
+    output?: {
+        text?: string;
+        choices?: Array<{
+            finish_reason: string;
+            message: {
+                role: string;
+                content: string;
+            };
+        }>;
+    };
+    usage?: {
+        input_tokens?: number;
+        output_tokens?: number;
+        total_tokens?: number;
+    };
+}
+export interface QwenDocTurboStreamChunk {
+    request_id?: string;
+    output?: {
+        text?: string;
+        choices?: Array<{
+            finish_reason: string;
+            message: {
+                role?: string;
+                content?: string;
+            };
+        }>;
+    };
+    usage?: {
+        input_tokens?: number;
+        output_tokens?: number;
+        total_tokens?: number;
+    };
+}
+export declare class QwenDocTurbo extends SimpleChatModel {
+    static BASE_URL: string;
+    static SUB_URL: string;
+    modelName: MultiModalModelName;
+    constructor(req: IAbstractRequest, baseUrl: string, modelName: MultiModalModelName, tokenManager: TokenManager);
+    protected modelRequest(data: QwenDocTurboAPIInput, options?: ReqOptions): Promise<ReadableStream<Uint8Array> | Promise<unknown>>;
+    protected normalizeResponse(response: QwenDocTurboResponse): DoGenerateOutput;
+    doGenerate(data: ModelRequestOptions, options?: ReqOptions): Promise<DoGenerateOutput>;
+    doStream(data: ModelRequestOptions, options?: ReqOptions): Promise<DoStreamOutput>;
+    private normalizeStreamChunks;
+    private convertToQwenDocTurboRequestOptions;
+}

package/lib/models/QwenDocTurbo/index.js

@@ -0,0 +1,178 @@
+"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.QwenDocTurbo = void 0;
+const model_type_1 = require("../../model-type");
+const stream_1 = require("../../stream");
+const utils_1 = require("../../utils");
+const Chat_1 = require("../Chat");
+class QwenDocTurbo extends Chat_1.SimpleChatModel {
+    constructor(req, baseUrl, modelName, tokenManager) {
+        super(req, baseUrl, QwenDocTurbo.SUB_URL, tokenManager);
+        this.modelName = modelName;
+    }
+    modelRequest(data, options) {
+        return __awaiter(this, void 0, void 0, function* () {
+            var _a, _b;
+            const fetchHeaders = yield this.createAuthHeaders(options === null || options === void 0 ? void 0 : options.headers);
+            const isStreaming = (_b = (_a = data.parameters) === null || _a === void 0 ? void 0 : _a.stream) !== null && _b !== void 0 ? _b : false;
+            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),
+                body: JSON.stringify(data),
+                url: `${this.baseUrl}/${this.subUrl}`,
+                stream: isStreaming,
+            }));
+            return (0, utils_1.handleResponseData)(responseData, header);
+        });
+    }
+    normalizeResponse(response) {
+        var _a, _b, _c, _d, _e;
+        const output = response.output || {};
+        const choice = (_a = output.choices) === null || _a === void 0 ? void 0 : _a[0];
+        const requestId = response.request_id || "";
+        const created = Math.floor(Date.now() / 1000);
+        const content = ((_b = choice === null || choice === void 0 ? void 0 : choice.message) === null || _b === void 0 ? void 0 : _b.content) || output.text || "";
+        return {
+            id: requestId,
+            object: "chat.completion",
+            created: created,
+            model: this.modelName,
+            log_id: requestId,
+            error: "",
+            code: 0,
+            choices: [
+                {
+                    index: 0,
+                    message: {
+                        id: requestId,
+                        role: "assistant",
+                        type: "answer",
+                        content: content,
+                        reasoning_content: "",
+                    },
+                    finish_reason: (choice === null || choice === void 0 ? void 0 : choice.finish_reason) || "stop",
+                },
+            ],
+            usage: {
+                prompt_tokens: ((_c = response.usage) === null || _c === void 0 ? void 0 : _c.input_tokens) || 0,
+                completion_tokens: ((_d = response.usage) === null || _d === void 0 ? void 0 : _d.output_tokens) || 0,
+                knowledge_tokens: 0,
+                reasoning_tokens: 0,
+                total_tokens: ((_e = response.usage) === null || _e === void 0 ? void 0 : _e.total_tokens) || 0,
+            },
+        };
+    }
+    doGenerate(data, options) {
+        return __awaiter(this, void 0, void 0, function* () {
+            const qwenDocTurboData = this.convertToQwenDocTurboRequestOptions(data, false);
+            const res = yield this.modelRequest(qwenDocTurboData, options);
+            return this.normalizeResponse(res);
+        });
+    }
+    doStream(data, options) {
+        return __awaiter(this, void 0, void 0, function* () {
+            const qwenDocTurboData = this.convertToQwenDocTurboRequestOptions(data, true);
+            const _stream = (yield this.modelRequest(qwenDocTurboData, options));
+            const stream = (0, stream_1.toPolyfillReadable)(_stream);
+            const rawStream = (0, stream_1.intoStandardStream)(stream);
+            const standardStream = this.normalizeStreamChunks(rawStream);
+            return (0, stream_1.createAsyncIterable)(standardStream);
+        });
+    }
+    normalizeStreamChunks(stream) {
+        const modelName = this.modelName;
+        let previousContent = "";
+        return stream.pipeThrough(new stream_1.TransformStream({
+            transform(chunk, controller) {
+                var _a, _b, _c, _d, _e;
+                const output = chunk.output || {};
+                const choice = (_a = output.choices) === null || _a === void 0 ? void 0 : _a[0];
+                const requestId = chunk.request_id || "";
+                const created = Math.floor(Date.now() / 1000);
+                const fullContent = ((_b = choice === null || choice === void 0 ? void 0 : choice.message) === null || _b === void 0 ? void 0 : _b.content) || output.text || "";
+                const deltaContent = fullContent.slice(previousContent.length);
+                previousContent = fullContent;
+                if (!deltaContent && (choice === null || choice === void 0 ? void 0 : choice.finish_reason) !== "stop") {
+                    return;
+                }
+                const standardChunk = {
+                    id: requestId,
+                    object: "chat.completion.chunk",
+                    created: created,
+                    model: modelName,
+                    log_id: requestId,
+                    error: "",
+                    code: 0,
+                    choices: [
+                        {
+                            index: 0,
+                            message: {
+                                id: requestId,
+                                role: "assistant",
+                                type: "answer",
+                                content: deltaContent,
+                                reasoning_content: "",
+                            },
+                            finish_reason: (choice === null || choice === void 0 ? void 0 : choice.finish_reason) || null,
+                        },
+                    ],
+                    usage: {
+                        prompt_tokens: ((_c = chunk.usage) === null || _c === void 0 ? void 0 : _c.input_tokens) || 0,
+                        completion_tokens: ((_d = chunk.usage) === null || _d === void 0 ? void 0 : _d.output_tokens) || 0,
+                        knowledge_tokens: 0,
+                        reasoning_tokens: 0,
+                        total_tokens: ((_e = chunk.usage) === null || _e === void 0 ? void 0 : _e.total_tokens) || 0,
+                    },
+                };
+                controller.enqueue(standardChunk);
+            },
+        }));
+    }
+    convertToQwenDocTurboRequestOptions(data, stream) {
+        const clamp = (value, min, max, defaultValue) => value !== undefined ? Math.max(min, Math.min(max, value)) : defaultValue;
+        const messages = (data.messages || []).map((msg) => {
+            var _a;
+            const role = msg.role;
+            if (role === "system") {
+                return {
+                    role,
+                    content: Array.isArray(msg.content) ? ((_a = msg.content.find((c) => c.type === "text")) === null || _a === void 0 ? void 0 : _a.text) || "" : msg.content,
+                };
+            }
+            return {
+                role,
+                content: Array.isArray(msg.content) ? (0, model_type_1.filterContentByTypes)(msg.content, ["text", "doc_url"]) : msg.content,
+            };
+        });
+        const parameters = Object.assign({ stream, max_tokens: clamp(data.max_tokens, 1, 8000, 2000), temperature: clamp(data.temperature, 0, 2, 1.0), top_p: clamp(data.top_p, 0, 1, 1.0), frequency_penalty: clamp(data.frequency_penalty, -2, 2, 0.0), presence_penalty: clamp(data.presence_penalty, -2, 2, 0.0), n: clamp(data.n, 1, 10, 1), incremental_output: false }, ((data === null || data === void 0 ? void 0 : data.parameters) || {}));
+        return {
+            model: this.modelName,
+            input: {
+                messages,
+            },
+            parameters,
+        };
+    }
+}
+exports.QwenDocTurbo = QwenDocTurbo;
+QwenDocTurbo.BASE_URL = "https://dashscope.aliyuncs.com";
+QwenDocTurbo.SUB_URL = "api/v1/services/aigc/text-generation/generation";

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

@@ -1,18 +1,11 @@
 import type { IAbstractRequest } from "@vectorx/ai-types";
-import type { DoGenerateOutput, DoStreamOutput, ModelRequestOptions, ReqOptions } from "../../model-type";
+import type { DoGenerateOutput, DoStreamOutput, ImageUrlContent, ModelRequestOptions, ReqOptions, TextContent } from "../../model-type";
 import type { TokenManager } from "../../tokenManager";
 import { SimpleChatModel } from "../Chat";
 import type { MultiModalModelName } from "../index";
 export interface QwenVlMaxMessage {
     role: "system" | "user" | "assistant";
-    content: string | Array<{
-        type: "text" | "image_url";
-        text?: string;
-        image_url?: {
-            url: string;
-            detail?: "auto" | "low" | "high";
-        };
-    }>;
+    content: string | Array<TextContent | ImageUrlContent>;
 }
 export interface QwenVlMaxParameters {
     seed?: number;

package/lib/models/QwenVlMax/index.js

@@ -10,6 +10,7 @@ var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, ge
 };
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.QwenVlMax = void 0;
+const model_type_1 = require("../../model-type");
 const stream_1 = require("../../stream");
 const utils_1 = require("../../utils");
 const Chat_1 = require("../Chat");
@@ -99,7 +100,7 @@ class QwenVlMax extends Chat_1.SimpleChatModel {
         const clamp = (value, min, max, defaultValue) => value !== undefined ? Math.max(min, Math.min(max, value)) : defaultValue;
         const messages = (data.messages || []).map((msg) => ({
             role: msg.role,
-            content: msg.content,
+            content: Array.isArray(msg.content) ? (0, model_type_1.filterContentByTypes)(msg.content, ["text", "image_url"]) : msg.content,
         }));
         return {
             model: this.modelName,

package/lib/models/index.d.ts

@@ -1,4 +1,5 @@
 import { DefaultSimpleModel } from "./Default";
+import { QwenDocTurbo } from "./QwenDocTurbo";
 import { QwenImageModel } from "./QwenImage";
 import { WanxSketchToImageLiteModel } from "./QwenSketchToImage";
 import { QwenStyleRepaintV1Model } from "./QwenStyleRepaintV1";
@@ -17,6 +18,7 @@ export declare enum ModelName {
 export declare enum MultiModalModelName {
     QvqMaxLatest = "qvq-max-latest",
     QwenVlMax = "qwen-vl-max",
+    QwenDocTurbo = "qwen-doc-turbo",
     QwenOmniTurboRealtime = "qwen-omni-turbo-realtime",
     Wanx21T2iPlus = "wanx2.1-t2i-plus",
     Wanx21T2iTurbo = "wanx2.1-t2i-turbo",
@@ -34,6 +36,7 @@ export declare const modelName: {
     "qwen-vl-ocr": string;
     "qvq-max-latest": string;
     "qwen-vl-max": string;
+    "qwen-doc-turbo": string;
     "qwen-omni-turbo-realtime": string;
     "wanx2.1-t2i-plus": string;
     "wanx2.1-t2i-turbo": string;
@@ -44,4 +47,4 @@ export declare const modelName: {
 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, ReActModel, toolMap, WanxSketchToImageLiteModel, QwenImageModel, QwenVlMax, QwenStyleRepaintV1Model, };
+export { DefaultSimpleModel, QwenDocTurbo, ReActModel, toolMap, WanxSketchToImageLiteModel, QwenImageModel, QwenVlMax, QwenStyleRepaintV1Model, };

package/lib/models/index.js

@@ -1,8 +1,10 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.QwenStyleRepaintV1Model = exports.QwenVlMax = exports.QwenImageModel = exports.WanxSketchToImageLiteModel = exports.toolMap = exports.ReActModel = exports.DefaultSimpleModel = exports.isMultiModalModel = exports.isValidModel = exports.modelName = exports.MultiModalModelName = exports.ModelName = void 0;
+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;
 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 QwenSketchToImage_1 = require("./QwenSketchToImage");
@@ -28,6 +30,7 @@ var MultiModalModelName;
 (function (MultiModalModelName) {
     MultiModalModelName["QvqMaxLatest"] = "qvq-max-latest";
     MultiModalModelName["QwenVlMax"] = "qwen-vl-max";
+    MultiModalModelName["QwenDocTurbo"] = "qwen-doc-turbo";
     MultiModalModelName["QwenOmniTurboRealtime"] = "qwen-omni-turbo-realtime";
     MultiModalModelName["Wanx21T2iPlus"] = "wanx2.1-t2i-plus";
     MultiModalModelName["Wanx21T2iTurbo"] = "wanx2.1-t2i-turbo";
@@ -45,6 +48,7 @@ exports.modelName = {
     [ModelName.QwenVlOcr]: "qwen-vl-ocr",
     [MultiModalModelName.QvqMaxLatest]: "qvq-max-latest",
     [MultiModalModelName.QwenVlMax]: "qwen-vl-max",
+    [MultiModalModelName.QwenDocTurbo]: "qwen-doc-turbo",
     [MultiModalModelName.QwenOmniTurboRealtime]: "qwen-omni-turbo-realtime",
     [MultiModalModelName.Wanx21T2iPlus]: "wanx2.1-t2i-plus",
     [MultiModalModelName.Wanx21T2iTurbo]: "wanx2.1-t2i-turbo",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vectorx/ai-sdk",
-  "version": "0.3.0",
+  "version": "0.4.0",
   "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": "0.3.0",
+    "@vectorx/ai-types": "0.4.0",
     "langfuse": "^3.38.4",
     "openai": "^4.103.0",
     "text-encoding-shim": "^1.0.5",