🚀 fix(openaiChat): simplify messages handling by removing unnecessary checks and defaults

✨ feat(aiCaller): enhance multimodal messages formatting for better processing of user content

✨ feat(ai): update image URL references for consistency across message handling

apps/ai.js

@@ -91,11 +91,11 @@ async function index(e) {
     if (e.user_id === e.bot.uin) {
       return;
     }
-    const userMessage = await extractUserMessage(e.msg, nickname, e);
+    const messageData = await extractUserMessage(e.msg, nickname, e);
-    if (!userMessage || userMessage.length === 0) {
+    if (!messageData || !messageData.text || messageData.text.length === 0) {
       return e.reply(segment.image(await Meme.getMeme(aiConfig.character, 'default')));
     }
-    const result = await processMessage(userMessage, e, aiConfig);
+    const result = await processMessage(messageData, e, aiConfig);
     if (result && result.length > 0) {
       await sendResponse(e, result);
     }
@@ -113,6 +113,7 @@ async function extractUserMessage(msg, nickname, e) {
     let at = [];
     const aiConfig = await ConfigControl.get('ai');
     const maxMessageLength = aiConfig?.maxMessageLength || 100;
+    const originalMessages = [];
     e.message.forEach((message) => {
       logger.info(message);
       if (message.type === 'text' && message.text !== '' && message.text !== '\n'){
@@ -124,19 +125,34 @@ async function extractUserMessage(msg, nickname, e) {
         text.push(displayText);
       } else if (message.type === 'at') {
         at.push(message.qq);
+      } else if (message.type === 'image') {
+        if (message.image) {
+          originalMessages.push({
+            type: 'image_url',
+            image_url: {
+              url: message.url
             }
           });
+        }
+      }
+    });
+    
     let returnMessage = '';
     if (text.length > 0) {
       text.forEach((message) => {
         if(message === '') {
         } else {
-        returnMessage += `[${e.sender?.nickname},id:${e.user_id},seq:${e.message_id}]说:${message}\n`
+          const tempMessage = `[${e.sender?.nickname},id:${e.user_id},seq:${e.message_id}]说:${message}\n`
+          returnMessage += tempMessage;
+          originalMessages.push({
+            type: 'text',
+            content: tempMessage
+          });
         }
       });
     }
     if(at.length == 1 && at[0] == e.bot.uin && text.length == 0){
-      return [];
+      return { text: [], originalMessages: originalMessages };
     }
     if (at.length > 0) {
       for (const at1 of at) {
@@ -144,14 +160,15 @@ async function extractUserMessage(msg, nickname, e) {
           //returnMessage += `[${e.sender?.nickname},id:${e.user_id}]@(at)了你,你的id是${at}\n`;
         } else {
           const atNickname = await e.group.pickMember(at1).nickname || '一个人';
-          returnMessage += `[${e.sender?.nickname},id:${e.user_id},seq:${e.message_id}]@(at)了${atNickname},id是${at1}\n`;
+          const tempMessage = `[${e.sender?.nickname},id:${e.user_id},seq:${e.message_id}]@(at)了${atNickname},id是${at1}\n`
+          returnMessage += tempMessage;
+          originalMessages.push({
+            type: 'text',
+            content: tempMessage
+          });
         }
       }
     }
-    const imgUrls = await YunzaiUtils.getImages(e, 1, true);
-    if (imgUrls) {
-      returnMessage += `[${e.sender?.nickname},id:${e.user_id},seq:${e.message_id}]发送了一张图片(你可能暂时无法查看)\n`;
-    }
     if(e.source || e.reply_id){
       let reply;
       if(e.getReply) reply = await e.getReply();
@@ -163,18 +180,33 @@ async function extractUserMessage(msg, nickname, e) {
         const msgArr = Array.isArray(reply) ? reply : reply.message || [];
         msgArr.forEach((msg) => {
           if(msg.type === 'text'){
-            returnMessage += `[${e.sender?.nickname}]引用了[被引用消息:${reply.user_id == e.bot.uin ? '你' : reply.sender?.nickname},id:${reply.user_id},seq:${reply.message_id}]发的一段文本:${msg.text}\n`
+            const tempMessage = `[${e.sender?.nickname}]引用了[被引用消息:${reply.user_id == e.bot.uin ? '你' : reply.sender?.nickname},id:${reply.user_id},seq:${reply.message_id}]发的一段文本:${msg.text}\n`
+            returnMessage += tempMessage;
+            originalMessages.push({
+              type: 'text',
+              content: tempMessage
+            });
           }
           if(msg.type === 'image'){
             returnMessage += `[${e.sender?.nickname}]引用了[被引用消息:${reply.user_id == e.bot.uin ? '你' : reply.sender?.nickname},id:${reply.user_id},seq:${reply.message_id}]发的一张图片(你可能暂时无法查看)\n`;
+            originalMessages.push({
+              type: 'image_url',
+              image_url: {
+                url: msg.url
+              }
+            });
           }
         })
       }
     }
-    return returnMessage;
+    const imgUrls = await YunzaiUtils.getImages(e, 1, true);
+    if (imgUrls) {
+      returnMessage += `[${e.sender?.nickname},id:${e.user_id},seq:${e.message_id}]发送了一张图片(你可能暂时无法查看)\n`;
+    }
+    return { text: returnMessage, originalMessages: originalMessages };
   }
   logger.warn('[crystelf-ai] 字符串匹配失败');
-  return [];
+  return { text: [], originalMessages: [] };
 }
 
 /**
@@ -202,11 +234,11 @@ async function processMessage(userMessage, e, aiConfig) {
 
 /**
  * 关键词模式
- * @param userMessage
+ * @param messageData
  * @param e
  * @returns {Promise<[{type: string, data: string}]>}
  */
-async function handleKeywordMode(userMessage, e) {
+async function handleKeywordMode(messageData, e) {
   const matchResult = await KeywordMatcher.matchKeywords(e.msg, 'ai');
 
   if (matchResult && matchResult.matched) {
@@ -229,17 +261,17 @@ async function handleKeywordMode(userMessage, e) {
   ];
 }
 
-async function handleAiMode(userMessage, e, aiConfig) {
+async function handleAiMode(messageData, e, aiConfig) {
-  return await callAiForResponse(userMessage, e, aiConfig);
+  return await callAiForResponse(messageData, e, aiConfig);
 }
 
-async function handleMixMode(userMessage, e, aiConfig) {
+async function handleMixMode(messageData, e, aiConfig) {
   const isTooLong = await KeywordMatcher.isMessageTooLong(e.msg);
 
   if (isTooLong) {
     //消息太长,使用AI回复
     logger.info('[crystelf-ai] 消息过长,使用ai回复');
-    return await callAiForResponse(userMessage, e, aiConfig);
+    return await callAiForResponse(messageData, e, aiConfig);
   } else {
     const matchResult = await KeywordMatcher.matchKeywords(e.msg, 'ai');
     if (matchResult && matchResult.matched) {
@@ -264,7 +296,7 @@ async function handleMixMode(userMessage, e, aiConfig) {
       };
       const newChatHistory = [
         ...chatHistory,
-        { role: 'user', content: userMessage },
+        { role: 'user', content: messageData.text },
         { role: 'assistant', content: JSON.stringify(resMessage) },
       ];
       SessionManager.updateChatHistory(e.group_id, newChatHistory);
@@ -274,12 +306,12 @@ async function handleMixMode(userMessage, e, aiConfig) {
     } else {
       logger.info('[crystelf-ai] 关键词匹配失败,使用ai回复');
       //关键词匹配失败,使用AI回复
-      return await callAiForResponse(userMessage, e, aiConfig);
+      return await callAiForResponse(messageData, e, aiConfig);
     }
   }
 }
 
-async function callAiForResponse(userMessage, e, aiConfig) {
+async function callAiForResponse(messageData, e, aiConfig) {
   try {
     //创建session
     const session = SessionManager.createOrGetSession(e.group_id, e.user_id, e);
@@ -299,7 +331,10 @@ async function callAiForResponse(userMessage, e, aiConfig) {
     //构建聊天历史
     const historyLen = aiConfig.chatHistory;
     const chatHistory = session.chatHistory.slice(-historyLen | -10);
-    const aiResult = await AiCaller.callAi(userMessage, chatHistory, memories, e);
+    
+    // 根据多模态开关决定调用方式
+    const aiResult = await AiCaller.callAi(messageData.text, chatHistory, memories, e, messageData.originalMessages);
+    
     if (!aiResult.success) {
       logger.error(`[crystelf-ai] AI调用失败: ${aiResult.error}`);
       SessionManager.deactivateSession(e.group_id, e.user_id);
@@ -313,14 +348,14 @@ async function callAiForResponse(userMessage, e, aiConfig) {
     //处理响应
     const processedResponse = await ResponseHandler.processResponse(
       aiResult.response,
-      userMessage,
+      messageData.text,
       e.group_id,
       e.user_id
     );
     //更新session
     const newChatHistory = [
       ...chatHistory,
-      { role: 'user', content: userMessage },
+      { role: 'user', content: messageData.text },
       { role: 'assistant', content: aiResult.response },
     ];
     SessionManager.updateChatHistory(e.group_id, newChatHistory);

apps/music.js

@@ -133,10 +133,10 @@ export class CrystelfMusic extends plugin {
       }
       const adapter = await YunzaiUtils.getAdapter(e);
       await Message.emojiLike(e,e.message_id,60,e.group_id,adapter);
+      await musicSearch.clearGroupSearch(e.group_id);
       const result = await musicSearch.handleSelection(e, index);
       if (result.success) {
         await this.sendMusicResult(e, result);
-        await musicSearch.clearGroupSearch(e.group_id);
       } else {
         await e.reply(`${result.message}`, true);
       }
@@ -159,15 +159,22 @@ export class CrystelfMusic extends plugin {
         await Group.sendGroupRecord(e, e.group_id, `file://${audioFile}`, adapter);
       } else {
         const extension = await this.getFileExtension();
-        const sanitizedTitle = song.displayTitle.replace(/\s+/g, '_');
+        // 过滤非法字符
-        const sanitizedArtist = song.displayArtist.replace(/\s+/g, '_');
+        const sanitize = (str) => str.replace(/[\\/:*?"<>|]/g, '').replace(/\s+/g, '_');
+
+        const sanitizedTitle = sanitize(song.displayTitle);
+        const sanitizedArtist = sanitize(song.displayArtist);
         const filename = `${sanitizedTitle} - ${sanitizedArtist}.${extension}`;
+        try {
             await Group.sendGroupFile(e, e.group_id, `file://${audioFile}`, filename, adapter);
+        } catch (fileErr) {
+            logger.warn(`[crystelf-music] 文件发送失败,尝试转为语音: ${fileErr.message}`);
+            await Group.sendGroupRecord(e, e.group_id, `file://${audioFile}`, adapter);
+        }
       }
       musicSearch.clearUserSelection(e.group_id, e.user_id);
-      logger.info(`[crystelf-music] 音乐发送成功: ${song.displayTitle}`);
     } catch (error) {
-      logger.error('[crystelf-music] 发送音乐结果失败:', error);
+      logger.error('[crystelf-music] 发送音乐失败:', error);
       await e.reply('发送音乐失败,请稍后重试', true);
     }
   }

config/ai.json

@@ -8,6 +8,10 @@
   "apiKey": "",
   "?modelType": "模型名称,请根据baseApi填写的服务商的对应的模型",
   "modelType": "deepseek-ai/DeepSeek-V3.2-Exp",
+  "?multimodalEnabled": "是否启用多模态模型模式,启用后将忽略文本模型",
+  "multimodalEnabled": false,
+  "?multimodalModel": "多模态模型名称",
+  "multimodalModel": "Qwen/Qwen2.5-VL-72B-Instruct",
   "?temperature": "聊天温度,可选0-2.0,温度越高创造性越高",
   "temperature": 1.2,
   "?concurrency": "最大同时聊天群数,一个群最多一个人聊天",

lib/ai/aiCaller.js

@@ -36,14 +36,39 @@ class AiCaller {
    * @param chatHistory 聊天历史
    * @param memories 记忆
    * @param e
+   * @param originalMessages 原始消息数组
    * @returns {Promise<{success: boolean, response: (*|string), rawResponse: (*|string)}|{success: boolean, error: string}|{success: boolean, error}>}
    */
-  async callAi(prompt, chatHistory = [], memories = [], e) {
+  async callAi(prompt, chatHistory = [], memories = [], e, originalMessages = []) {
     if (!this.isInitialized || !this.config) {
       logger.error('[crystelf-ai] 未初始化或配置无效');
       return { success: false, error: 'AI调用器未初始化' };
     }
+    try {
+      if (this.config.multimodalEnabled) {
+        return await this.callMultimodalAi(originalMessages, chatHistory, memories, e);
+      } else {
+        return await this.callTextAi(prompt, chatHistory, memories, e);
+      }
+    } catch (error) {
+      logger.error(`[crystelf-ai] 调用失败: ${error.message}`);
+      SessionManager.deactivateSession(e.group_id, e.user_id);
+      return {
+        success: false,
+        error: error.message,
+      };
+    }
+  }
 
+  /**
+   * 文本AI模型
+   * @param prompt 用户输入
+   * @param chatHistory 聊天历史
+   * @param memories 记忆
+   * @param e
+   * @returns {Promise<{success: boolean, response: (*|string), rawResponse: (*|string)}|{success: boolean, error: string}>}
+   */
+  async callTextAi(prompt, chatHistory = [], memories = [], e) {
     try {
       const fullPrompt = this.buildPrompt(prompt);
       const apiCaller = this.openaiChat;
@@ -68,14 +93,96 @@ class AiCaller {
         };
       }
     } catch (error) {
-      logger.error(`[crystelf-ai] 调用失败: ${error.message}`);
+      throw error;
-      SessionManager.deactivateSession(e.group_id, e.user_id);
+    }
+  }
+
+  /**
+   * 多模态AI调用
+   * @param originalMessages 原始消息数组
+   * @param chatHistory 聊天历史
+   * @param memories 记忆
+   * @param e
+   * @returns {Promise<{success: boolean, response: (*|string), rawResponse: (*|string)}|{success: boolean, error: string}>}
+   */
+  async callMultimodalAi(originalMessages, chatHistory = [], memories = [], e) {
+    try {
+      const messages = await this.formatMultimodalMessages(originalMessages, chatHistory, memories, e);
+      const apiCaller = this.openaiChat;
+      const result = await apiCaller.callAi({
+        messages: messages,
+        model: this.config.multimodalModel,
+        temperature: this.config.temperature,
+      });
+
+      if (result.success) {
+        return {
+          success: true,
+          response: result.aiResponse,
+          rawResponse: result.aiResponse,
+        };
+      } else {
         return {
           success: false,
-        error: error.message,
+          error: '多模态AI调用失败',
         };
       }
+    } catch (error) {
+      throw error;
     }
+  }
+
+  /**
+   * 将原始消息格式转换为多模态格式
+   * @param originalMessages 原始消息数组
+   * @param chatHistory 聊天历史
+   * @param memories 记忆
+   * @param e
+   * @returns {Array} 多模态格式的消息数组
+   */
+  async formatMultimodalMessages(originalMessages, chatHistory = [], memories = [], e) {
+    const messages = [];
+    const systemPrompt = await this.getSystemPrompt(e, memories);
+    messages.push({
+      role: 'system',
+      content: [
+        { type: 'text', text: systemPrompt }
+      ]
+    });
+    for (const history of chatHistory) {
+      const role = history.role === 'user' ? 'user' : 'assistant';
+      messages.push({
+        role,
+        content: [
+          { type: 'text', text: history.content }
+        ]
+      });
+    }
+    const mergedUserContent = [];
+    for (const msg of originalMessages) {
+      if (msg.type === 'text' && msg.content) {
+        mergedUserContent.push({
+          type: 'text',
+          text: msg.content
+        });
+      }
+
+      if (msg.type === 'image_url' && msg.image_url?.url) {
+        mergedUserContent.push({
+          type: 'image_url',
+          image_url: { url: msg.image_url.url }
+        });
+      }
+    }
+    if (mergedUserContent.length > 0) {
+      messages.push({
+        role: 'user',
+        content: mergedUserContent
+      });
+    }
+    return messages;
+  }
+
 
   /**
    * 构造完整的prompt

modules/openai/openaiChat.js

@@ -22,18 +22,23 @@ class OpenaiChat {
    * @param model 模型
    * @param temperature 温度
    * @param customPrompt 提示词
+   * @param messages 多模态消息数组
    * @returns {Promise<{success: boolean, aiResponse: string}|{}>}
    */
-  async callAi({ prompt, chatHistory = [], model, temperature, customPrompt }) {
+  async callAi({ prompt, chatHistory = [], model, temperature, customPrompt, messages = [] }) {
     if (!this.openai) {
       logger.error('[crystelf-ai] ai未初始化..');
       return { success: false };
     }
+    let finalMessages;
+    if (messages.length > 0) {
+      finalMessages = messages;
+    } else {
       let systemMessage = {
         role: 'system',
         content: customPrompt || '',
       };
-    const messages = [
+      finalMessages = [
         systemMessage,
         ...chatHistory,
         {
@@ -41,15 +46,16 @@ class OpenaiChat {
           content: prompt,
         },
       ];
+    }
 
     try {
      // logger.info("[DEBUG] 请求体:", {
         //model: model,
-       // messages,
+       // messages: finalMessages,
       //});
 
       const completion = await this.openai.chat.completions.create({
-        messages: messages,
+        messages: finalMessages,
         model: model,
         temperature: temperature,
         frequency_penalty: 0.2,