myagent-ai 1.16.10 → 1.16.12

package/agents/main_agent.py

@@ -634,17 +634,30 @@ class MainAgent(BaseAgent):
                 ))
                 all_tool_outputs = ""
             else:
-                messages.append(Message(
-                    role="user",
-                    content=context.user_message or "请处理上述上下文。"
-                ))
+                # [v1.16.12] 支持多模态消息（图片+文本）
+                user_images = context.metadata.get("user_images", [])
+                if user_images:
+                    # OpenAI Vision 格式: [{type: "text"}, {type: "image_url"}]
+                    multimodal_content = [{"type": "text", "text": context.user_message or "请描述这些图片。"}]
+                    for img in user_images:
+                        if img.get("url"):
+                            multimodal_content.append({
+                                "type": "image_url",
+                                "image_url": {"url": img["url"]}
+                            })
+                    messages.append(Message(role="user", content=multimodal_content))
+                else:
+                    messages.append(Message(
+                        role="user",
+                        content=context.user_message or "请处理上述上下文。"
+                    ))
 
             # [v1.15.52] 保存完整的 LLM 输入消息（system prompt + user message + 工具结果回调）
             # 用于 Raw 查看器完整回溯 LLM 交互过程
             if self.memory:
                 _input_parts = []
                 for _msg in messages:
-                    _input_parts.append(f"=== {_msg.role.upper()} ===\n{_msg.content}")
+                    _input_parts.append(f"=== {_msg.role.upper()} ===\n{_msg.get_text_content()}")
                 _llm_input_text = "\n\n".join(_input_parts)
                 self.memory.add_session(
                     session_id=context.session_id,

package/core/llm.py

@@ -17,7 +17,7 @@ import json
 import time
 import asyncio
 from typing import (
-    Optional, Dict, Any, List, Generator, AsyncGenerator,
+    Optional, Dict, Any, List, Generator, AsyncGenerator, Union,
 )
 from dataclasses import dataclass, field
 
@@ -33,9 +33,16 @@ logger = get_logger("myagent.llm")
 
 @dataclass
 class Message:
-    """聊天消息"""
+    """聊天消息
+
+    [v1.16.12] content 支持 str | list 类型:
+      - str: 纯文本消息 (兼容旧行为)
+      - list: 多模态内容，OpenAI Vision 格式:
+        [\n          {"type": "text", "text": "描述这张图片"},
+          {"type": "image_url", "image_url": {"url": "data:image/png;base64,..."}}\n        ]
+    """
     role: str              # system | user | assistant | tool
-    content: str = ""
+    content: Any = ""      # str | list (多模态内容)
     name: str = ""         # 消息发送者标识
     tool_call_id: str = "" # 工具调用ID
     tool_calls: List[Dict] = field(default_factory=list)
@@ -67,6 +74,18 @@ class Message:
             result["tool_calls"] = formatted
         return result
 
+    def get_text_content(self) -> str:
+        """[v1.16.12] 提取纯文本内容（忽略图片等非文本部分）"""
+        if isinstance(self.content, str):
+            return self.content
+        if isinstance(self.content, list):
+            parts = []
+            for item in self.content:
+                if isinstance(item, dict) and item.get("type") == "text":
+                    parts.append(item.get("text", ""))
+            return "\n".join(parts)
+        return str(self.content) if self.content else ""
+
     @classmethod
     def from_dict(cls, data: dict) -> "Message":
         return cls(

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "myagent-ai",
-  "version": "1.16.10",
+  "version": "1.16.12",
   "description": "本地桌面端执行型AI助手 - Open Interpreter 风格 | Local Desktop Execution-Oriented AI Assistant",
   "main": "main.py",
   "bin": {

package/web/api_server.py

@@ -490,6 +490,112 @@ class ApiServer:
             self._task_persistence.initialize()
         return self._task_persistence
 
+    @staticmethod
+    def _extract_text_from_file(filename: str, mime_type: str, data: bytes) -> str:
+        """[v1.16.12] 从文件中提取文本内容，支持 txt/pdf/csv/md/json/py/js/html 等格式。
+        
+        Args:
+            filename: 文件名
+            mime_type: MIME 类型
+            data: 文件二进制数据
+            
+        Returns:
+            提取的文本内容，最多 50000 字符
+        """
+        import io
+        
+        fname_lower = filename.lower()
+        
+        # 纯文本格式：直接读取
+        _text_exts = {'.txt', '.md', '.csv', '.tsv', '.log', '.yaml', '.yml',
+                      '.toml', '.ini', '.cfg', '.conf', '.env', '.sh', '.bash',
+                      '.zsh', '.fish', '.ps1', '.bat', '.cmd'}
+        _code_exts = {'.py', '.js', '.ts', '.jsx', '.tsx', '.java', '.c', '.cpp',
+                      '.h', '.hpp', '.cs', '.go', '.rs', '.rb', '.php', '.swift',
+                      '.kt', '.scala', '.r', '.sql', '.lua', '.vim', '.el'}
+        _markup_exts = {'.html', '.htm', '.xml', '.svg', '.css', '.scss', '.less'}
+        _data_exts = {'.json', '.jsonl'}
+        
+        if (fname_lower.endswith(tuple(_text_exts | _code_exts | _markup_exts | _data_exts))):
+            # 尝试 UTF-8 解码，失败则尝试其他编码
+            for encoding in ('utf-8', 'gbk', 'gb2312', 'latin-1'):
+                try:
+                    return data.decode(encoding)[:50000]
+                except (UnicodeDecodeError, LookupError):
+                    continue
+            return f"[无法解码文件 {filename}]"
+        
+        # PDF 格式
+        if fname_lower.endswith('.pdf') or 'pdf' in mime_type:
+            try:
+                import subprocess
+                # 使用 pdftotext 提取（系统工具，大多数 Linux 环境自带）
+                result = subprocess.run(
+                    ['pdftotext', '-', '-'],
+                    input=data, capture_output=True, text=True, timeout=30,
+                )
+                if result.returncode == 0 and result.stdout.strip():
+                    return result.stdout.strip()[:50000]
+            except Exception:
+                pass
+            # 尝试 PyPDF2
+            try:
+                import io as _io
+                from PyPDF2 import PdfReader
+                reader = PdfReader(_io.BytesIO(data))
+                texts = []
+                for page in reader.pages:
+                    text = page.extract_text()
+                    if text:
+                        texts.append(text)
+                if texts:
+                    return "\n".join(texts)[:50000]
+            except ImportError:
+                pass
+            except Exception:
+                pass
+            return f"[无法提取 PDF 文件 {filename} 的文本内容，请安装 poppler-utils (apt install poppler-utils)]"
+        
+        # Excel 格式
+        if fname_lower.endswith(('.xlsx', '.xls')) or 'excel' in mime_type or 'spreadsheet' in mime_type:
+            try:
+                from openpyxl import load_workbook
+                wb = load_workbook(io.BytesIO(data), read_only=True, data_only=True)
+                texts = []
+                for ws in wb.worksheets:
+                    rows = []
+                    for row in ws.iter_rows(values_only=True):
+                        row_text = "\t".join(str(c) if c is not None else "" for c in row)
+                        rows.append(row_text)
+                    if rows:
+                        texts.append(f"[Sheet: {ws.title}]\n" + "\n".join(rows[:500]))
+                wb.close()
+                if texts:
+                    return "\n".join(texts)[:50000]
+            except ImportError:
+                return f"[无法读取 Excel 文件，需安装 openpyxl: pip install openpyxl]"
+            except Exception as e:
+                return f"[Excel 文件读取失败: {e}]"
+        
+        # Word 文档
+        if fname_lower.endswith('.docx') or 'wordprocessingml' in mime_type:
+            try:
+                from docx import Document
+                doc = Document(io.BytesIO(data))
+                text = "\n".join(p.text for p in doc.paragraphs if p.text)
+                return text[:50000] if text else "[Word 文档内容为空]"
+            except ImportError:
+                return f"[无法读取 Word 文件，需安装 python-docx: pip install python-docx]"
+            except Exception as e:
+                return f"[Word 文件读取失败: {e}]"
+        
+        # 图片格式 — 返回描述提示（图片由 Vision API 直接处理）
+        _image_exts = {'.png', '.jpg', '.jpeg', '.gif', '.webp', '.bmp', '.ico'}
+        if fname_lower.endswith(_image_exts) or 'image/' in mime_type:
+            return f"[图片文件: {filename}]"
+        
+        return f"[不支持的文件格式: {filename} ({mime_type})]"
+
     # --- Execution Progress ---
     async def handle_execution_progress(self, request):
         """GET /api/execution/progress - 返回当前执行进度"""
@@ -620,7 +726,15 @@ class ApiServer:
         print(f"[STREAM_MESSAGE] 函数被调用, PYTHONIOENCODING={getattr(sys, 'stdout', None)}", flush=True, file=sys.stdout)
 
         message = data.get("message", "").strip()
-        if not message:
+        # [v1.16.12] 支持图片附件
+        user_images = data.get("images", [])  # [{"type": "image/png", "data": "base64..."}, ...]
+        # 支持文件附件（文档等，提取文本后作为上下文）
+        user_files = data.get("files", [])    # [{"name": "xxx.pdf", "type": "application/pdf", "data": "base64..."}, ...]
+        
+        # 如果消息为空但有图片，设置默认提示
+        if not message and user_images:
+            message = "请查看这些图片"
+        if not message and not user_images and not user_files:
             return web.Response(text="data: " + json.dumps({"error": "message is required"}) + "\n\n", content_type="text/event-stream")
 
         agent_path = data.get("agent_path", data.get("agent_name", "default")) or "default"
@@ -744,6 +858,7 @@ class ApiServer:
                         clean_message, session_id, proxy,
                         agent_path=agent_path, agent_system_prompt=agent_system_prompt,
                         chat_mode=chat_mode, voice_text=voice_text,
+                        user_images=user_images, user_files=user_files,
                     )
                 else:
                     full_response = await self.core.process_message(clean_message, session_id)
@@ -1140,6 +1255,8 @@ class ApiServer:
 
                 wav_buf.seek(0)
                 segments, info = whisper_model.transcribe(wav_buf, beam_size=1,
+                                                          language="zh",
+                                                          initial_prompt="以下是普通话的句子",
                                                           vad_filter=True, vad_parameters=dict(
                                                               min_silence_duration_ms=300))
                 text = "".join(seg.text for seg in segments).strip()
@@ -1188,6 +1305,8 @@ class ApiServer:
                                 return web.json_response({"error": "音频格式不支持"}, status=400)
                         wav_buf.seek(0)
                         segments, info = whisper_model.transcribe(wav_buf, beam_size=1,
+                                                                  language="zh",
+                                                                  initial_prompt="以下是普通话的句子",
                                                                   vad_filter=True, vad_parameters=dict(
                                                                       min_silence_duration_ms=300))
                         text = "".join(seg.text for seg in segments).strip()
@@ -3670,7 +3789,7 @@ class ApiServer:
 
     async def _stream_process_message(self, user_message, session_id, stream_response,
                                       agent_path=None, agent_system_prompt=None, chat_mode="",
-                                      voice_text=""):
+                                      voice_text="", user_images=None, user_files=None):
         """使用流式LLM调用处理消息，支持完整的agent循环（工具调用/操作执行）+ 实时流式输出
 
         核心改进:
@@ -3678,8 +3797,7 @@ class ApiServer:
         - 使用 frequency_penalty 减少大模型重复输出
         - 最终保存时使用累积文本而非 final_response，确保完整内容不丢失
 
-        实现与 MainAgent._process_inner() 相同的计划-执行-反思循环，
-        但将 LLM 的文本响应逐 token 流式推送到 SSE。
+        [v1.16.12] 新增 user_images/user_files 参数支持多模态消息
         """
         logger.info(f"[{session_id}] _stream_process_message 开始处理，chat_mode={chat_mode}")
         if not self.core.main_agent or not self.core.llm:
@@ -3698,6 +3816,47 @@ class ApiServer:
         context.metadata["chat_mode"] = chat_mode
         context.metadata["user_voice_text"] = voice_text  # 语音输入原始文本（用于 usersays_correct）
 
+        # [v1.16.12] 处理用户图片附件 — 转换为 data URI 传给 LLM Vision API
+        if user_images:
+            _processed_images = []
+            for img in user_images:
+                mime = img.get("type", "image/png")
+                b64 = img.get("data", "")
+                if b64:
+                    _processed_images.append({
+                        "url": f"data:{mime};base64,{b64}",
+                        "type": mime,
+                        "name": img.get("name", ""),
+                    })
+            if _processed_images:
+                context.metadata["user_images"] = _processed_images
+                logger.info(f"[{session_id}] 用户发送了 {len(_processed_images)} 张图片")
+
+        # [v1.16.12] 处理用户文件附件 — 提取文本内容附加到消息中
+        if user_files:
+            _file_texts = []
+            for f in user_files:
+                fname = f.get("name", "unknown")
+                ftype = f.get("type", "")
+                fdata_b64 = f.get("data", "")
+                if not fdata_b64:
+                    continue
+                try:
+                    import base64 as _b64mod
+                    fbytes = _b64mod.b64decode(fdata_b64)
+                    _text = self._extract_text_from_file(fname, ftype, fbytes)
+                    if _text:
+                        _file_texts.append(f"--- 文件: {fname} ---\n{_text}")
+                except Exception as _fe:
+                    logger.warning(f"[{session_id}] 文件 {fname} 提取失败: {_fe}")
+            if _file_texts:
+                file_context = "\n\n".join(_file_texts)
+                context.metadata["user_file_texts"] = file_context
+                # 将文件内容附加到 user_message 中
+                user_message = f"{user_message}\n\n[附件内容]\n{file_context}" if user_message else f"[附件内容]\n{file_context}"
+                context.user_message = user_message
+                logger.info(f"[{session_id}] 用户发送了 {len(user_files)} 个文件，提取文本 {len(file_context)} 字符")
+
         # ── 根据 Agent 配置设置执行引擎参数（execution_mode 等）──
         agent_cfg_for_exec = self._read_agent_config(agent_path)
         _original_exec_mode = None

package/web/ui/chat/chat.css

@@ -570,7 +570,7 @@ input,textarea,select{font:inherit}
 .text-input-area {
   display: flex;
   align-items: flex-end;
-  gap: 10px;
+  gap: 8px;
   width: 100%;
   flex: 1;
 }
@@ -581,6 +581,82 @@ input,textarea,select{font:inherit}
 }
 .input-box textarea::placeholder{color:var(--text3)}
 
+/* [v1.16.12] 附件按钮 */
+.attach-buttons {
+  display: flex;
+  flex-direction: column;
+  gap: 4px;
+  flex-shrink: 0;
+  padding-bottom: 2px;
+}
+.attach-btn {
+  width:30px;height:30px;border:none;background:transparent;color:var(--text3);
+  display:grid;place-items:center;border-radius:var(--radius-xs);
+  cursor:pointer;transition:var(--transition);
+}
+.attach-btn:hover{color:var(--accent);background:var(--accent-light)}
+.attach-btn svg{width:18px;height:18px}
+
+/* [v1.16.12] 拖拽高亮 */
+.input-box.drag-over{border-color:var(--accent) !important;box-shadow:0 0 0 2px rgba(79,70,229,.2)}
+
+/* [v1.16.12] 附件预览 */
+.attachment-preview {
+  display:flex;flex-wrap:wrap;gap:8px;padding:8px 0;
+  align-items:center;
+}
+.attachment-preview:empty{display:none}
+.attachment-thumb {
+  position:relative;border-radius:var(--radius-sm);overflow:hidden;
+  border:1px solid var(--bg4);cursor:pointer;transition:var(--transition);
+}
+.attachment-thumb:hover{border-color:var(--accent)}
+.attachment-thumb-image {
+  width:80px;height:80px;
+}
+.attachment-thumb-image img {
+  width:100%;height:100%;object-fit:cover;display:block;
+}
+.attachment-thumb-file {
+  display:flex;flex-direction:column;align-items:center;justify-content:center;
+  width:80px;height:80px;background:var(--bg3);padding:6px;gap:4px;
+}
+.attachment-file-icon{font-size:22px;line-height:1}
+.attachment-file-name{
+  font-size:10px;color:var(--text2);text-align:center;
+  max-width:70px;overflow:hidden;text-overflow:ellipsis;white-space:nowrap;
+}
+.attachment-remove {
+  position:absolute;top:-2px;right:-2px;
+  width:18px;height:18px;border-radius:50%;
+  background:var(--danger);color:#fff;border:none;
+  font-size:13px;line-height:1;cursor:pointer;
+  display:none;align-items:center;justify-content:center;
+  padding:0;
+}
+.attachment-thumb:hover .attachment-remove{display:flex}
+
+/* [v1.16.12] 消息气泡中的附件 */
+.msg-attachments {
+  display:flex;flex-wrap:wrap;gap:8px;margin-bottom:8px;
+}
+.msg-image-wrapper {
+  max-width:300px;border-radius:var(--radius-sm);overflow:hidden;
+  border:1px solid var(--bg4);cursor:pointer;
+}
+.msg-image {
+  width:100%;height:auto;display:block;max-height:300px;object-fit:cover;
+}
+.msg-image-wrapper:hover .msg-image{max-height:none}
+.msg-file-item {
+  display:flex;align-items:center;gap:8px;padding:6px 10px;
+  background:var(--bg3);border-radius:var(--radius-sm);
+  font-size:13px;color:var(--text2);max-width:240px;
+}
+.msg-file-icon{font-size:18px;flex-shrink:0}
+.msg-file-name{overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
+.msg-file-size{font-size:11px;color:var(--text3);flex-shrink:0}
+
 .send-btn{
   width:36px;height:36px;border-radius:var(--radius-sm);
   background:var(--accent);color:#fff;

package/web/ui/chat/chat_container.html

@@ -157,9 +157,18 @@
             <span class="lock-text" id="lockText"></span>
           </div>
         </div>
-        <div class="input-box" id="inputBox">
+        <div class="input-box" id="inputBox" ondragover="event.preventDefault();this.classList.add('drag-over')" ondragleave="this.classList.remove('drag-over')" ondrop="handleDropEvent(event);this.classList.remove('drag-over')">
           <div class="text-input-area" id="textInputArea">
-            <textarea id="userInput" placeholder="输入消息... (Enter 发送, Shift+Enter 换行)" rows="1" onkeydown="handleKeyDown(event)" oninput="autoResize(this)"></textarea>
+            <textarea id="userInput" placeholder="输入消息... (Enter 发送, Shift+Enter 换行, 可粘贴图片)" rows="1" onkeydown="handleKeyDown(event)" oninput="autoResize(this);updateSendBtnState()" onpaste="handlePasteEvent(event)"></textarea>
+            <!-- [v1.16.12] 附件按钮 -->
+            <div class="attach-buttons">
+              <button class="attach-btn" id="attachImageBtn" onclick="document.getElementById('imageFileInput').click()" title="上传图片">
+                <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><rect x="3" y="3" width="18" height="18" rx="2" ry="2"/><circle cx="8.5" cy="8.5" r="1.5"/><polyline points="21 15 16 10 5 21"/></svg>
+              </button>
+              <button class="attach-btn" id="attachFileBtn" onclick="document.getElementById('docFileInput').click()" title="上传文件">
+                <svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><path d="M21.44 11.05l-9.19 9.19a6 6 0 0 1-8.49-8.49l9.19-9.19a4 4 0 0 1 5.66 5.66l-9.2 9.19a2 2 0 0 1-2.83-2.83l8.49-8.48"/></svg>
+              </button>
+            </div>
             <button class="send-btn" id="sendBtn" onclick="sendMessage()" disabled>
               <svg viewBox="0 0 24 24" fill="currentColor"><path d="M2.01 21L23 12 2.01 3 2 10l15 2-15 2z"/></svg>
             </button>

package/web/ui/chat/chat_main.js

@@ -328,6 +328,8 @@ async function initChat() {
   initTheme();
   // Restore persisted UI state
   StatePersistence.restoreUIState();
+  // [v1.16.12] 初始化附件上传 UI
+  initAttachmentUI();
 
   // URL 参数处理: ?agent=xxx&mode=exec&session=xxx&popout=1
   const urlParams = new URLSearchParams(window.location.search);
@@ -2498,6 +2500,31 @@ function _renderMessagesInner() {
 
     const avatar = isUser ? '<span style="font-size:18px">👤</span>' : avatarHtml({avatar_image: state.currentAgent?.avatar_image, avatar_emoji: botEmoji, avatar_color: state.currentAgent?.avatar_color, name: state.currentAgent?.name}, 32, 'border-radius:8px;');
     const content = renderMarkdown(msg.content);
+    // [v1.16.12] 渲染图片和文件附件
+    const attachmentHtml = (() => {
+      if (!isUser) return '';
+      let parts = [];
+      // 图片
+      if (msg.images && msg.images.length > 0) {
+        for (const img of msg.images) {
+          const dataUrl = 'data:' + (img.type || 'image/png') + ';base64,' + (img.data || '');
+          parts.push('<div class="msg-image-wrapper"><img src="' + dataUrl + '" class="msg-image" onclick="window.open(this.src)" loading="lazy" alt="' + escapeHtml(img.name || 'image') + '" /></div>');
+        }
+      }
+      // 文件
+      if (msg.files && msg.files.length > 0) {
+        for (const f of msg.files) {
+          const sizeStr = f.size ? formatFileSize(f.size) : '';
+          const icon = _getFileIcon(f.name || f.type || '');
+          parts.push('<div class="msg-file-item">' +
+            '<span class="msg-file-icon">' + icon + '</span>' +
+            '<span class="msg-file-name" title="' + escapeHtml(f.name) + '">' + escapeHtml(f.name) + '</span>' +
+            (sizeStr ? '<span class="msg-file-size">' + sizeStr + '</span>' : '') +
+            '</div>');
+        }
+      }
+      return parts.length > 0 ? '<div class="msg-attachments">' + parts.join('') + '</div>' : '';
+    })();
     const thoughtHtml = msg.thought ? (() => {
       const isStreaming = !!msg.streaming;
       return `<details class="thought-block ${isStreaming ? 'streaming' : ''}" ${isStreaming ? 'open' : ''}>
@@ -2603,6 +2630,7 @@ function _renderMessagesInner() {
           ${finishReasonHtml}
           ${timelineHtml}
           ${singleBubbleHtml}
+          ${attachmentHtml}
           ${streamingIndicator}
           ${execEventsHtml}
           ${msg.time ? `<div class="message-time">${formatTime(msg.time)}</div>` : ''}
@@ -2719,6 +2747,203 @@ function renderMarkdown(text) {
 }
 
 // 高效的 HTML 转义（不创建 DOM 元素，避免大文本时性能问题）
+// ── [v1.16.12] 文件上传辅助函数 ──
+function formatFileSize(bytes) {
+  if (bytes < 1024) return bytes + ' B';
+  if (bytes < 1048576) return (bytes / 1024).toFixed(1) + ' KB';
+  return (bytes / 1048576).toFixed(1) + ' MB';
+}
+
+function _getFileIcon(name) {
+  var ext = (name || '').split('.').pop().toLowerCase();
+  var map = {
+    'pdf': '📄', 'doc': '📝', 'docx': '📝', 'xls': '📊', 'xlsx': '📊',
+    'ppt': '📽️', 'pptx': '📽️', 'csv': '📋', 'json': '🔧', 'xml': '🔧',
+    'html': '🌐', 'css': '🎨', 'js': '⚡', 'ts': '⚡', 'py': '🐍',
+    'java': '☕', 'c': '⚙️', 'cpp': '⚙️', 'go': '🔵', 'rs': '🦀',
+    'md': '📝', 'txt': '📃', 'log': '📃', 'yaml': '⚙️', 'yml': '⚙️',
+    'zip': '📦', 'rar': '📦', '7z': '📦', 'tar': '📦', 'gz': '📦',
+    'png': '🖼️', 'jpg': '🖼️', 'jpeg': '🖼️', 'gif': '🖼️', 'webp': '🖼️',
+    'svg': '🖼️', 'mp3': '🎵', 'mp4': '🎬', 'wav': '🎵',
+  };
+  return map[ext] || '📎';
+}
+
+// ── 附件上传系统 ──
+var _attachState = { images: [], files: [] };
+
+function initAttachmentUI() {
+  window._pendingImages = [];
+  window._pendingFiles = [];
+  
+  // 创建隐藏的文件输入
+  if (!document.getElementById('imageFileInput')) {
+    var imgInput = document.createElement('input');
+    imgInput.type = 'file';
+    imgInput.id = 'imageFileInput';
+    imgInput.accept = 'image/*';
+    imgInput.multiple = true;
+    imgInput.style.display = 'none';
+    imgInput.onchange = function() { handleFileSelect(this, 'image'); this.value = ''; };
+    document.body.appendChild(imgInput);
+  }
+  if (!document.getElementById('docFileInput')) {
+    var docInput = document.createElement('input');
+    docInput.type = 'file';
+    docInput.id = 'docFileInput';
+    docInput.accept = '.txt,.md,.csv,.json,.py,.js,.ts,.java,.c,.cpp,.go,.rs,.rb,.php,.html,.css,.xml,.yaml,.yml,.toml,.ini,.sh,.log,.pdf,.docx,.xlsx,.pptx';
+    docInput.multiple = true;
+    docInput.style.display = 'none';
+    docInput.onchange = function() { handleFileSelect(this, 'file'); this.value = ''; };
+    document.body.appendChild(docInput);
+  }
+}
+
+function handleFileSelect(input, type) {
+  var files = input.files;
+  if (!files || files.length === 0) return;
+  
+  for (var i = 0; i < files.length; i++) {
+    (function(file) {
+      var reader = new FileReader();
+      reader.onload = function(e) {
+        var base64 = e.target.result.split(',')[1];
+        if (type === 'image' && file.type.startsWith('image/')) {
+          window._pendingImages.push({
+            type: file.type,
+            data: base64,
+            name: file.name,
+            size: file.size,
+          });
+        } else {
+          window._pendingFiles.push({
+            type: file.type,
+            data: base64,
+            name: file.name,
+            size: file.size,
+          });
+        }
+        renderAttachmentPreview();
+        updateSendBtnState();
+      };
+      reader.readAsDataURL(file);
+    })(files[i]);
+  }
+}
+
+function handlePasteEvent(e) {
+  var items = e.clipboardData && e.clipboardData.items;
+  if (!items) return;
+  var hasImage = false;
+  for (var i = 0; i < items.length; i++) {
+    if (items[i].type.startsWith('image/')) {
+      e.preventDefault();
+      hasImage = true;
+      var file = items[i].getAsFile();
+      if (file) {
+        (function(f) {
+          var reader = new FileReader();
+          reader.onload = function(ev) {
+            var base64 = ev.target.result.split(',')[1];
+            window._pendingImages.push({
+              type: f.type,
+              data: base64,
+              name: 'paste-' + Date.now() + '.png',
+              size: f.size,
+            });
+            renderAttachmentPreview();
+            updateSendBtnState();
+          };
+          reader.readAsDataURL(f);
+        })(file);
+      }
+    }
+  }
+}
+
+function handleDropEvent(e) {
+  e.preventDefault();
+  e.stopPropagation();
+  var dt = e.dataTransfer;
+  if (!dt || !dt.files || dt.files.length === 0) return;
+  for (var i = 0; i < dt.files.length; i++) {
+    (function(file) {
+      var reader = new FileReader();
+      reader.onload = function(ev) {
+        var base64 = ev.target.result.split(',')[1];
+        if (file.type.startsWith('image/')) {
+          window._pendingImages.push({ type: file.type, data: base64, name: file.name, size: file.size });
+        } else {
+          window._pendingFiles.push({ type: file.type, data: base64, name: file.name, size: file.size });
+        }
+        renderAttachmentPreview();
+        updateSendBtnState();
+      };
+      reader.readAsDataURL(file);
+    })(dt.files[i]);
+  }
+}
+
+function renderAttachmentPreview() {
+  var container = document.getElementById('attachmentPreview');
+  if (!container) {
+    // 创建预览容器（在 input-box 之前）
+    var inputBox = document.getElementById('inputBox');
+    if (!inputBox) return;
+    container = document.createElement('div');
+    container.id = 'attachmentPreview';
+    container.className = 'attachment-preview';
+    inputBox.parentNode.insertBefore(container, inputBox);
+  }
+  
+  var html = '';
+  // 图片预览
+  var imgs = window._pendingImages || [];
+  for (var i = 0; i < imgs.length; i++) {
+    var img = imgs[i];
+    html += '<div class="attachment-thumb attachment-thumb-image" onclick="removeAttachment(\'image\',' + i + ')">' +
+      '<img src="data:' + (img.type || 'image/png') + ';base64,' + img.data + '" />' +
+      '<button class="attachment-remove" onclick="event.stopPropagation();removeAttachment(\'image\',' + i + ')">×</button>' +
+      '</div>';
+  }
+  // 文件预览
+  var files = window._pendingFiles || [];
+  for (var j = 0; j < files.length; j++) {
+    var f = files[j];
+    html += '<div class="attachment-thumb attachment-thumb-file" onclick="removeAttachment(\'file\',' + j + ')">' +
+      '<span class="attachment-file-icon">' + _getFileIcon(f.name) + '</span>' +
+      '<span class="attachment-file-name">' + escapeHtml(f.name) + '</span>' +
+      '<button class="attachment-remove" onclick="event.stopPropagation();removeAttachment(\'file\',' + j + ')">×</button>' +
+      '</div>';
+  }
+  container.innerHTML = html;
+  container.style.display = html ? 'flex' : 'none';
+}
+
+function removeAttachment(type, index) {
+  if (type === 'image') {
+    (window._pendingImages || []).splice(index, 1);
+  } else {
+    (window._pendingFiles || []).splice(index, 1);
+  }
+  renderAttachmentPreview();
+  updateSendBtnState();
+}
+
+function clearAttachmentPreview() {
+  var container = document.getElementById('attachmentPreview');
+  if (container) { container.innerHTML = ''; container.style.display = 'none'; }
+}
+
+function updateSendBtnState() {
+  var input = document.getElementById('userInput');
+  var btn = document.getElementById('sendBtn');
+  if (!btn) return;
+  var hasContent = input && input.value.trim().length > 0;
+  var hasAttach = (window._pendingImages || []).length > 0 || (window._pendingFiles || []).length > 0;
+  btn.disabled = !hasContent && !hasAttach;
+}
+
 function escapeHtml(text) {
   if (!text) return '';
   return text

package/web/ui/chat/flow_engine.js

@@ -1354,7 +1354,29 @@ async function sendMessage(opts) {
   }
 
   // Add user message
-  state.messages.push({ role: 'user', content: text, time: new Date().toISOString(), _voiceText: voiceText });
+  // [v1.16.12] 支持图片和文件附件
+  var _pendingImages = window._pendingImages || [];   // [{type, data, name}]
+  var _pendingFiles = window._pendingFiles || [];      // [{type, data, name}]
+  var _msgImages = _pendingImages.slice();
+  var _msgFiles = _pendingFiles.slice();
+  
+  var userMsgObj = { 
+    role: 'user', 
+    content: text, 
+    time: new Date().toISOString(), 
+    _voiceText: voiceText 
+  };
+  if (_msgImages.length > 0) {
+    userMsgObj.images = _msgImages.map(function(img) { 
+      return { type: img.type, data: img.data, name: img.name };
+    });
+  }
+  if (_msgFiles.length > 0) {
+    userMsgObj.files = _msgFiles.map(function(f) {
+      return { type: f.type, name: f.name, size: f.size };
+    });
+  }
+  state.messages.push(userMsgObj);
   renderMessages();
 
   // Clear input
@@ -1362,6 +1384,12 @@ async function sendMessage(opts) {
   input.style.height = 'auto';
   document.getElementById('sendBtn').disabled = true;
   clearDraft();
+  // [v1.16.12] 清除附件
+  window._pendingImages = [];
+  window._pendingFiles = [];
+  var _attachPreview = document.getElementById('attachmentPreview');
+  if (_attachPreview) _attachPreview.innerHTML = '';
+  if (typeof clearAttachmentPreview === 'function') clearAttachmentPreview();
 
   // 用户发消息后，强制滚到底部
   _userScrollLocked = false;
@@ -1394,6 +1422,8 @@ async function sendMessage(opts) {
         mode: state.chatMode,
         escalated: state.escalated,
         voice_text: voiceText,  // 语音转文字原始文本（用于后端 usersays_correct）
+        images: _msgImages.map(function(img) { return { type: img.type, data: img.data, name: img.name }; }),
+        files: _msgFiles.map(function(f) { return { type: f.type, data: f.data, name: f.name }; }),
       }),
       signal: state.abortController.signal,
     });