multi-modal-mcp 0.0.2 → 0.0.3

package/dist/config.js

@@ -20,5 +20,5 @@ export const config = {
     /** MCP 服务器名称 */
     serverName: 'multi-modal-mcp',
     /** MCP 服务器版本 */
-    serverVersion: '1.0.0',
+    serverVersion: '0.0.3',
 };

package/dist/tools/ImageGenerationTool.js

@@ -3,7 +3,7 @@
  * @Project_description: 智谱AI图片生成工具
  * @Description: 代码是我抄的，不会也是真的
  */
-import { MCPTool, logger } from 'mcp-framework';
+import { MCPTool } from 'mcp-framework';
 import { z } from 'zod';
 import { http } from '../utils/http.js';
 import { config } from '../config.js';
@@ -33,7 +33,6 @@ class ImageGenerationTool extends MCPTool {
      * 执行图片生成
      */
     async execute(input) {
-        logger.info(`开始执行图片生成，描述: ${input.prompt}，尺寸: ${input.size || '1024x1024'}`);
         try {
             const requestData = {
                 model: config.imageModel,
@@ -41,17 +40,14 @@ class ImageGenerationTool extends MCPTool {
                 size: input.size || '1024x1024',
                 watermark_enabled: input.watermark_enabled ?? false,
             };
-            logger.info('调用智谱AI图片生成API');
             const apiResponse = (await http.post('/images/generations', requestData));
             const imageUrl = apiResponse.data?.[0]?.url;
             if (!imageUrl)
                 throw new Error('生成图片失败，未返回图片URL');
-            logger.info('图片生成成功');
             return imageUrl;
         }
         catch (error) {
-            logger.error(`图片生成失败: ${error}`);
-            throw new Error(`生成图片时发生错误: ${error}`);
+            return `生成图片时发生错误: ${error}`;
         }
     }
 }

package/dist/tools/MultiModalUnderstandingTool.js

@@ -3,7 +3,7 @@
  * @Project_description: 智谱AI多模态理解工具
  * @Description: 代码是我抄的，不会也是真的
  */
-import { MCPTool, logger } from 'mcp-framework';
+import { MCPTool } from 'mcp-framework';
 import { z } from 'zod';
 import { http } from '../utils/http.js';
 import { config } from '../config.js';
@@ -17,29 +17,12 @@ class MultiModalUnderstandingTool extends MCPTool {
     description = '基于GLM-4.6V-Flash模型的多模态理解工具，支持图片、视频、文档等多种媒体类型的智能分析与理解。可执行OCR文字识别、表格解析、内容分析、缺陷检测、图像转提示词、视频标签提取、关键帧提取、时间线生成、脚本生成、视频问答、文档问答、文档对比等多种任务。支持同时处理多种媒体类型。';
     /** 参数模式定义 */
     schema = z.object({
-        content: z
+        media_content: z
             .array(z.object({
             type: z
                 .enum(['image_url', 'video_url', 'file_url'])
                 .describe('媒体类型：image_url-图片，video_url-视频，file_url-文档'),
-            image_url: z
-                .object({
-                url: z.string().describe('图片的URL地址或Base64编码'),
-            })
-                .optional()
-                .describe('图片URL对象，当type为image_url时必填'),
-            video_url: z
-                .object({
-                url: z.string().describe('视频的URL地址'),
-            })
-                .optional()
-                .describe('视频URL对象，当type为video_url时必填'),
-            file_url: z
-                .object({
-                url: z.string().describe('文件的URL地址'),
-            })
-                .optional()
-                .describe('文件URL对象，当type为file_url时必填'),
+            url: z.string().describe('图片、视频、文档的URL地址'),
         }))
             .describe('媒体内容列表，支持混合图片、视频、文档'),
         question: z
@@ -58,9 +41,8 @@ class MultiModalUnderstandingTool extends MCPTool {
      * 执行多模态理解
      */
     async execute(input) {
-        logger.info(`开始执行多模态理解，问题: ${input.question}，媒体数量: ${input.content.length}`);
         try {
-            const messageContent = this.buildMessageContent(input.content, input.question);
+            const messageContent = this.buildMessageContent(input.media_content, input.question);
             const requestData = {
                 model: config.visualModel,
                 messages: [
@@ -75,7 +57,6 @@ class MultiModalUnderstandingTool extends MCPTool {
                 temperature: input.temperature,
                 stream: false,
             };
-            logger.info('调用智谱AI多模态理解API');
             const apiResponse = (await http.post('/chat/completions', requestData));
             const messageContentResult = apiResponse.choices?.[0]?.message?.content;
             let generatedText = '';
@@ -86,12 +67,10 @@ class MultiModalUnderstandingTool extends MCPTool {
                 const textItem = messageContentResult.find((item) => item.type === 'text');
                 generatedText = textItem?.text || '';
             }
-            logger.info('多模态理解成功');
             return generatedText;
         }
         catch (error) {
-            logger.error(`多模态理解失败: ${error}`);
-            throw new Error(`多模态理解时发生错误: ${error}`);
+            return `多模态理解时发生错误: ${error}`;
         }
     }
     /**
@@ -100,27 +79,27 @@ class MultiModalUnderstandingTool extends MCPTool {
     buildMessageContent(content, question) {
         const result = [];
         for (const item of content) {
-            if (item.type === 'image_url' && item.image_url) {
+            if (item.type === 'file_url') {
                 result.push({
-                    type: 'image_url',
-                    image_url: {
-                        url: item.image_url.url,
+                    type: item.type,
+                    file_url: {
+                        url: item.url,
                     },
                 });
             }
-            else if (item.type === 'video_url' && item.video_url) {
+            if (item.type === 'image_url') {
                 result.push({
-                    type: 'video_url',
-                    video_url: {
-                        url: item.video_url.url,
+                    type: item.type,
+                    image_url: {
+                        url: item.url,
                     },
                 });
             }
-            else if (item.type === 'file_url' && item.file_url) {
+            if (item.type === 'video_url') {
                 result.push({
-                    type: 'file_url',
-                    file_url: {
-                        url: item.file_url.url,
+                    type: item.type,
+                    video_url: {
+                        url: item.url,
                     },
                 });
             }

package/dist/tools/TextGenerationTool.js

@@ -3,7 +3,7 @@
  * @Project_description: 智谱AI文本生成工具
  * @Description: 代码是我抄的，不会也是真的
  */
-import { MCPTool, logger } from 'mcp-framework';
+import { MCPTool } from 'mcp-framework';
 import { z } from 'zod';
 import { http } from '../utils/http.js';
 import { config } from '../config.js';
@@ -31,7 +31,6 @@ class TextGenerationTool extends MCPTool {
      * 执行文本生成
      */
     async execute(input) {
-        logger.info(`开始执行文本生成，输入: ${input.messages}`);
         try {
             // 构建请求参数
             const requestData = {
@@ -48,7 +47,6 @@ class TextGenerationTool extends MCPTool {
                 temperature: input.temperature,
                 stream: false, // 确保非流式返回
             };
-            logger.info('调用智谱AI文本生成API');
             // 调用智谱AI API
             const apiResponse = (await http.post('/chat/completions', requestData));
             // 提取生成的文本
@@ -63,12 +61,10 @@ class TextGenerationTool extends MCPTool {
                 const textItem = messageContent.find((item) => item.type === 'text');
                 generatedText = textItem?.text || '';
             }
-            logger.info('文本生成成功');
             return generatedText;
         }
         catch (error) {
-            logger.error(`文本生成失败: ${error}`);
-            throw new Error(`生成文本时发生错误: ${error}`);
+            return `生成文本时发生错误: ${error}`;
         }
     }
 }

package/dist/tools/VideoGenerationTool.js

@@ -3,7 +3,7 @@
  * @Project_description: 智谱AI视频生成工具
  * @Description: 代码是我抄的，不会也是真的
  */
-import { MCPTool, logger } from 'mcp-framework';
+import { MCPTool } from 'mcp-framework';
 import { z } from 'zod';
 import { http } from '../utils/http.js';
 import { config } from '../config.js';
@@ -63,7 +63,6 @@ class VideoGenerationTool extends MCPTool {
      * 执行视频生成
      */
     async execute(input) {
-        logger.info(`开始执行视频生成，描述: ${input.prompt}，尺寸: ${input.size || '1024x1024'}，帧率: ${input.fps || 30}`);
         try {
             const requestData = {
                 model: config.videoModel,
@@ -75,19 +74,15 @@ class VideoGenerationTool extends MCPTool {
                 size: input.size,
                 fps: input.fps || 30,
             };
-            logger.info('调用智谱AI视频生成API，创建任务');
             const apiResponse = (await http.post('/videos/generations', requestData));
             const taskId = apiResponse.id;
             if (!taskId)
                 throw new Error('视频生成任务创建失败，未返回任务ID');
-            logger.info(`视频生成任务创建成功，任务ID: ${taskId}，开始轮询任务状态`);
             const videoUrl = await this.pollTaskStatus(taskId);
-            logger.info('视频生成成功');
             return videoUrl;
         }
         catch (error) {
-            logger.error(`视频生成失败: ${error}`);
-            throw new Error(`视频生成时发生错误: ${error}`);
+            return `视频生成时发生错误: ${error}`;
         }
     }
     /**
@@ -96,25 +91,17 @@ class VideoGenerationTool extends MCPTool {
     async pollTaskStatus(taskId) {
         const startTime = Date.now();
         const timeout = 60000;
-        logger.info(`开始轮询任务状态，任务ID: ${taskId}`);
         while (true) {
             const elapsedTime = Date.now() - startTime;
             if (elapsedTime >= timeout) {
-                logger.error(`视频生成超时，超过60秒未完成，任务ID: ${taskId}`);
                 throw new Error('视频生成超时，超过60秒未完成');
             }
             const apiResponse = (await http.get(`/async-result/${taskId}`));
             const taskData = apiResponse;
             if (taskData.task_status === 'SUCCESS' &&
                 taskData.video_result?.[0]?.url) {
-                logger.info(`任务完成，任务ID: ${taskId}，状态: ${taskData.task_status}`);
                 return taskData.video_result[0].url;
             }
-            if (taskData.task_status === 'FAILED') {
-                logger.error(`任务失败，任务ID: ${taskId}，状态: ${taskData.task_status}`);
-                throw new Error('视频生成失败');
-            }
-            logger.debug(`任务进行中，任务ID: ${taskId}，状态: ${taskData.task_status}，已等待: ${Math.floor(elapsedTime / 1000)}秒`);
             await new Promise((resolve) => setTimeout(resolve, 1000));
         }
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "multi-modal-mcp",
-  "version": "0.0.2",
+  "version": "0.0.3",
   "description": "multi-modal MCP server",
   "type": "module",
   "bin": {
@@ -48,4 +48,4 @@
   ],
   "author": "橘子",
   "license": "MIT"
-}
+}