feat: Agent不支持视觉能力时上传图片提示用户

main_v2.py

@@ -3,7 +3,7 @@ AI对话系统 v2.0.0 - 主应用
 支持：大模型池、Agent管理、渠道独立绑定、思考功能开关
 """
 from fastapi import FastAPI, WebSocket, WebSocketDisconnect, Depends, HTTPException, Request
-from fastapi.responses import HTMLResponse
+from fastapi.responses import HTMLResponse, JSONResponse
 from fastapi.staticfiles import StaticFiles
 from fastapi.templating import Jinja2Templates
 from sqlalchemy.orm import Session
@@ -13,6 +13,9 @@ import json
 import logging
 from datetime import datetime
 import os
+import base64
+import uuid
+import time
 
 # 使用新的数据模型
 from models_v2 import (
@@ -34,7 +37,14 @@ app = FastAPI(title="AI对话系统 v2.0", version="2.0.0")
 
 # 静态文件和模板
 BASE_DIR = os.path.dirname(os.path.abspath(__file__))
+UPLOADS_DIR = os.path.join(BASE_DIR, "uploads", "images")
+
+# 确保上传目录存在
+os.makedirs(UPLOADS_DIR, exist_ok=True)
+
+# 静态文件服务
 app.mount("/static", StaticFiles(directory=os.path.join(BASE_DIR, "static")), name="static")
+app.mount("/uploads", StaticFiles(directory=os.path.join(BASE_DIR, "uploads")), name="uploads")
 templates = Jinja2Templates(directory=os.path.join(BASE_DIR, "templates"))
 
 # WebSocket连接管理
@@ -110,6 +120,8 @@ async def get_providers(db: Session = Depends(get_db)):
                 "default_model": p.default_model,
                 "supports_thinking": p.supports_thinking,
                 "thinking_model": p.thinking_model,
+                "supports_vision": p.supports_vision,
+                "vision_model": p.vision_model,
                 "max_tokens": p.max_tokens,
                 "temperature": p.temperature,
                 "is_active": p.is_active,
@@ -575,6 +587,56 @@ async def perform_search(data: dict, db: Session = Depends(get_db)):
     return {"success": False, "message": "不支持的搜索提供商"}
 
 
+# ==================== 图片上传 API ====================
+
+@app.post("/api/v2/upload-image")
+async def upload_image(data: dict):
+    """上传图片到服务器，返回文件路径"""
+    try:
+        image_data = data.get('image')
+        file_name = data.get('name', 'image.png')
+        
+        if not image_data:
+            return {"success": False, "message": "缺少图片数据"}
+        
+        # 解析 base64 数据
+        if image_data.startswith('data:image/'):
+            # 提取格式和base64内容
+            header, base64_content = image_data.split(',', 1)
+            # 从header中提取图片格式
+            format_match = header.split(':')[1].split(';')[0]  # 如 'image/png'
+            ext = format_match.split('/')[1] if '/' in format_match else 'png'
+        else:
+            base64_content = image_data
+            ext = 'png'
+        
+        # 生成唯一文件名
+        timestamp = int(time.time())
+        unique_id = uuid.uuid4().hex[:8]
+        safe_name = f"{timestamp}_{unique_id}.{ext}"
+        
+        # 保存文件
+        file_path = os.path.join(UPLOADS_DIR, safe_name)
+        image_bytes = base64.b64decode(base64_content)
+        
+        # 检查文件大小（限制10MB）
+        if len(image_bytes) > 10 * 1024 * 1024:
+            return {"success": False, "message": "图片大小超过10MB限制"}
+        
+        with open(file_path, 'wb') as f:
+            f.write(image_bytes)
+        
+        # 返回可访问的URL路径
+        url_path = f"/uploads/images/{safe_name}"
+        logger.info(f"图片已保存: {file_path}, URL: {url_path}")
+        
+        return {"success": True, "path": url_path, "name": safe_name}
+        
+    except Exception as e:
+        logger.error(f"图片上传失败: {e}")
+        return {"success": False, "message": str(e)}
+
+
 # ==================== 对话 API（保留原有） ====================
 
 @app.get("/api/conversations")
@@ -787,6 +849,7 @@ async def websocket_endpoint(websocket: WebSocket, user_id: str):
                     # 处理文件内容，添加到消息
                     image_contents = []  # 图片内容（用于视觉模型）
                     text_contents = []  # 文本文件内容
+                    image_paths = []  # 图片服务器路径（用于历史记录显示）
                     if files:
                         for f in files:
                             if f.get('type') and f['type'].startswith('image/'):
@@ -796,6 +859,13 @@ async def websocket_endpoint(websocket: WebSocket, user_id: str):
                                     'type': f['type'],
                                     'data': f.get('content', '')  # base64 数据
                                 })
+                                # 记录服务器路径（用于历史记录）
+                                if f.get('serverPath'):
+                                    image_paths.append({
+                                        'name': f['name'],
+                                        'type': f['type'],
+                                        'url': f['serverPath']  # 服务器文件路径
+                                    })
                                 # 不添加文件名文本，图片信息保存在 extra_data 中
                             elif f.get('content'):
                                 # 文本文件：直接添加内容，不带文件名前缀
@@ -808,10 +878,16 @@ async def websocket_endpoint(websocket: WebSocket, user_id: str):
                         for content in text_contents:
                             message += f"\n\n{content}"
                     
-                    # 保存图片信息到 extra_data（用于历史记录）
+                    # 保存图片和文件信息到 extra_data（用于历史记录）
                     extra_data_for_msg = None
-                    if image_contents:
-                        # 只保存图片 URL（不保存完整 base64）
+                    if image_paths:
+                        # 图片保存服务器路径URL，历史记录可以显示
+                        extra_data_for_msg = {
+                            'images': image_paths,
+                            'files': [{'name': f['name'], 'type': f['type']} for f in files if not f['type'].startswith('image/')]
+                        }
+                    elif image_contents:
+                        # 没有服务器路径但有问题（可能上传失败）
                         extra_data_for_msg = {
                             'images': [{'name': i['name'], 'type': i['type']} for i in image_contents],
                             'files': [{'name': f['name'], 'type': f['type']} for f in files if not f['type'].startswith('image/')]

models_v2.py

@@ -32,6 +32,10 @@ class LLMProvider(Base):
     supports_thinking = Column(Boolean, default=False)  # 是否原生支持思考
     thinking_model = Column(String(100), nullable=True)  # 思考模式模型名（如有单独模型）
     
+    # 视觉能力支持
+    supports_vision = Column(Boolean, default=False)  # 是否支持图片理解（多模态）
+    vision_model = Column(String(100), nullable=True)  # 视觉模型名（如与默认模型不同）
+    
     # 配额和限制
     max_tokens = Column(Integer, default=4096)
     temperature = Column(Float, default=0.7)

templates/admin_v2/index.html

@@ -58,8 +58,8 @@
                 </div>
                 <div class="card-body">
                     <table class="table">
-                        <thead><tr><th>名称</th><th>API地址</th><th>默认模型</th><th>思考支持</th><th>状态</th><th>操作</th></tr></thead>
-                        <tbody id="providers-list"><tr><td colspan="6" class="text-center">加载中...</td></tr></tbody>
+                        <thead><tr><th>名称</th><th>API地址</th><th>默认模型</th><th>思考</th><th>视觉</th><th>状态</th><th>操作</th></tr></thead>
+                        <tbody id="providers-list"><tr><td colspan="7" class="text-center">加载中...</td></tr></tbody>
                     </table>
                 </div>
             </div>
@@ -162,6 +162,8 @@
             <div class="mt-3 form-check"><input type="checkbox" class="form-check-input" id="provider-active" checked><label class="form-check-label">启用</label></div>
             <hr><h6>思考功能</h6>
             <div class="thinking-config"><div class="row"><div class="col-md-6 form-check"><input type="checkbox" class="form-check-input" id="provider-supports-thinking"><label class="form-check-label">支持原生思考</label></div><div class="col-md-6"><label class="form-label">思考模型名</label><input type="text" class="form-control" id="provider-thinking-model"></div></div></div>
+            <hr><h6>视觉能力</h6>
+            <div class="thinking-config"><div class="row"><div class="col-md-6 form-check"><input type="checkbox" class="form-check-input" id="provider-supports-vision"><label class="form-check-label">支持图片理解</label></div><div class="col-md-6"><label class="form-label">视觉模型名</label><input type="text" class="form-control" id="provider-vision-model" placeholder="留空则使用默认模型"></div></div><small class="text-muted mt-2 d-block">启用后可上传图片让AI识别分析内容</small></div>
             <div class="mt-3"><button type="button" class="btn btn-outline-primary" onclick="fetchProviderModels()"><i class="ri-refresh-line"></i> 获取模型</button><button type="button" class="btn btn-outline-secondary" onclick="testProviderConnection()"><i class="ri-link"></i> 测试连接</button></div>
             <div class="mt-2" id="provider-models-preview"></div><div class="mt-2" id="provider-test-result"></div>
         </form></div>
@@ -274,6 +276,7 @@
             tbody.innerHTML = providersData.map(p => `<tr>
                 <td><strong>${p.name}</strong></td><td><small>${p.api_base||'-'}</small></td><td>${p.default_model||'auto'}</td>
                 <td>${p.supports_thinking?'<span class="badge bg-success">支持</span>':'<span class="badge bg-secondary">不支持</span>'}</td>
+                <td>${p.supports_vision?'<span class="badge bg-info">支持</span>':'<span class="badge bg-secondary">不支持</span>'}</td>
                 <td>${p.is_active?'<span class="badge bg-success">启用</span>':'<span class="badge bg-secondary">禁用</span>'}</td>
                 <td><button class="btn btn-sm btn-outline-primary" onclick="editProvider(${p.id})"><i class="ri-edit-line"></i></button>
                     <button class="btn btn-sm btn-outline-danger" onclick="deleteProvider(${p.id},'${p.name}')"><i class="ri-delete-bin-line"></i></button></td>
@@ -290,6 +293,8 @@
             document.getElementById('provider-form').reset();
             document.getElementById('provider-id').value = '';
             document.getElementById('provider-active').checked = true;
+            document.getElementById('provider-supports-thinking').checked = false;
+            document.getElementById('provider-supports-vision').checked = false;
             document.getElementById('provider-models-preview').innerHTML = '';
             document.getElementById('provider-test-result').innerHTML = '';
             new bootstrap.Modal(document.getElementById('providerModal')).show();
@@ -310,6 +315,8 @@
             document.getElementById('provider-active').checked = p.is_active;
             document.getElementById('provider-supports-thinking').checked = p.supports_thinking;
             document.getElementById('provider-thinking-model').value = p.thinking_model || '';
+            document.getElementById('provider-supports-vision').checked = p.supports_vision;
+            document.getElementById('provider-vision-model').value = p.vision_model || '';
             new bootstrap.Modal(document.getElementById('providerModal')).show();
         }
 
@@ -326,7 +333,9 @@
                 description: document.getElementById('provider-description').value,
                 is_active: document.getElementById('provider-active').checked,
                 supports_thinking: document.getElementById('provider-supports-thinking').checked,
-                thinking_model: document.getElementById('provider-thinking-model').value
+                thinking_model: document.getElementById('provider-thinking-model').value,
+                supports_vision: document.getElementById('provider-supports-vision').checked,
+                vision_model: document.getElementById('provider-vision-model').value
             };
             const res = await fetch(id ? `/api/v2/providers/${id}` : '/api/v2/providers', { method: id ? 'PUT' : 'POST', headers: {'Content-Type':'application/json'}, body: JSON.stringify(data) });
             const result = await res.json();

templates/index.html

@@ -146,6 +146,17 @@
         .modal-buttons { display: flex; gap: 12px; justify-content: flex-end; }
         .modal-buttons button { padding: 8px 16px; border-radius: 8px; cursor: pointer; }
         
+        /* 图片放大弹窗 */
+        .image-lightbox { position: fixed; top: 0; left: 0; right: 0; bottom: 0; background: rgba(0,0,0,0.9); display: none; align-items: center; justify-content: center; z-index: 2000; cursor: zoom-out; }
+        .image-lightbox.show { display: flex; }
+        .image-lightbox img { max-width: 90%; max-height: 90%; border-radius: 8px; box-shadow: 0 0 30px rgba(255,255,255,0.2); }
+        .image-lightbox-close { position: absolute; top: 20px; right: 20px; width: 40px; height: 40px; background: rgba(255,255,255,0.2); border-radius: 50%; display: flex; align-items: center; justify-content: center; color: #fff; font-size: 20px; cursor: pointer; transition: background 0.2s; }
+        .image-lightbox-close:hover { background: rgba(255,255,255,0.3); }
+        
+        /* 对话中的图片可点击 */
+        .uploaded-image img { cursor: zoom-in; transition: transform 0.2s; }
+        .uploaded-image img:hover { transform: scale(1.02); }
+        
         .welcome { display: flex; flex-direction: column; align-items: center; justify-content: center; height: 100%; color: #666; }
         .welcome h2 { font-size: 28px; margin-bottom: 16px; color: #333; }
         
@@ -211,6 +222,34 @@
         </div>
     </div>
 
+    <!-- 图片放大弹窗 -->
+    <div class="image-lightbox" id="imageLightbox" onclick="closeImageLightbox()">
+        <div class="image-lightbox-close"><i class="ri-close-line"></i></div>
+        <img id="lightboxImage" src="" alt="放大图片">
+    </div>
+
+    <!-- 隐藏的图片上传API处理 -->
+    <script>
+        // 图片上传到服务器（保存文件）
+        async function uploadImageToServer(base64Data, fileName) {
+            try {
+                const response = await fetch('/api/v2/upload-image', {
+                    method: 'POST',
+                    headers: { 'Content-Type': 'application/json' },
+                    body: JSON.stringify({ image: base64Data, name: fileName })
+                });
+                const result = await response.json();
+                if (result.success) {
+                    return result.path;  // 返回服务器文件路径
+                }
+                return null;
+            } catch (e) {
+                console.error('图片上传失败:', e);
+                return null;
+            }
+        }
+    </script>
+
     <!-- Markdown渲染库 -->
     <script src="https://cdn.jsdelivr.net/npm/marked/marked.min.js"></script>
     <script>
@@ -221,6 +260,7 @@
         let currentConversationId = null;
         let currentAgentId = null;
         let agents = [];
+        let providers = [];  // 大模型池数据
         let quickPhrases = [];
         let lastSentMessage = null;  // 记录最后发送的消息
         let lastSentFiles = null;  // 记录发送的文件
@@ -233,6 +273,7 @@
         let messageVersions = {}; // 存储每个assistant消息的多个版本 { messageId: [{content, thinking}] }
         
         document.addEventListener('DOMContentLoaded', () => {
+            loadProviders();  // 加载大模型池
             loadAgents();
             loadQuickPhrases();
             connectWebSocket();
@@ -240,6 +281,23 @@
             setupTextarea();
         });
         
+        // 加载大模型池
+        async function loadProviders() {
+            try {
+                const res = await fetch('/api/v2/providers');
+                const data = await res.json();
+                providers = data.providers || [];
+            } catch (e) { console.error('加载Provider失败:', e); }
+        }
+        
+        // 检查当前Agent是否支持视觉
+        function checkAgentVisionSupport() {
+            const agent = agents.find(a => a.id === currentAgentId);
+            if (!agent) return false;
+            const provider = providers.find(p => p.id === agent.llm_provider_id);
+            return provider?.supports_vision || false;
+        }
+        
         // 加载Agent
         async function loadAgents() {
             try {
@@ -393,19 +451,50 @@
             html += '</div>';
             div.innerHTML = html;
             
-            // 如果是用户消息且有搜索结果，在设置innerHTML后追加
+            // 如果是用户消息且有额外数据（搜索结果、图片、文件），在设置innerHTML后追加
             if (role === 'user' && extraData) {
                 console.log('Processing extraData for user message:', extraData);
-                console.log('search_results exists:', extraData.search_results);
+                const bodyDiv = div.querySelector('.message-body');
+                
+                // 处理图片（如果有服务器URL，显示图片）
+                if (extraData.images && extraData.images.length > 0) {
+                    let imagesHtml = '<div class="history-images" style="margin-top:8px;display:flex;gap:8px;flex-wrap:wrap;">';
+                    for (const img of extraData.images) {
+                        if (img.url) {
+                            // 有服务器URL，显示真实图片
+                            imagesHtml += `<div class="history-image" style="display:inline-block;">
+                                <img src="${img.url}" style="max-width:300px;max-height:200px;border-radius:8px;cursor:zoom-in;" onclick="openImageLightbox('${img.url}')">
+                            </div>`;
+                        } else {
+                            // 没有URL，显示占位符
+                            imagesHtml += `<div class="history-image-placeholder" style="padding:8px 12px;background:#f0f0f0;border-radius:8px;display:flex;align-items:center;gap:6px;font-size:13px;color:#666;">
+                                <i class="ri-image-line" style="color:#10a37f;"></i>
+                                <span>${escapeHtml(img.name || '图片')}</span>
+                            </div>`;
+                        }
+                    }
+                    imagesHtml += '</div>';
+                    if (bodyDiv) bodyDiv.insertAdjacentHTML('beforeend', imagesHtml);
+                }
+                
+                // 处理文本文件
+                if (extraData.files && extraData.files.length > 0) {
+                    let filesHtml = '<div class="history-files" style="margin-top:8px;">';
+                    for (const f of extraData.files) {
+                        filesHtml += `<div class="history-file-placeholder" style="padding:6px 10px;background:#f5f5f5;border-radius:6px;margin-bottom:4px;display:flex;align-items:center;gap:6px;font-size:12px;color:#666;">
+                            <i class="ri-file-text-line" style="color:#10a37f;"></i>
+                            <span>${escapeHtml(f.name || '文件')}</span>
+                        </div>`;
+                    }
+                    filesHtml += '</div>';
+                    if (bodyDiv) bodyDiv.insertAdjacentHTML('beforeend', filesHtml);
+                }
+                
+                // 处理搜索结果
                 if (extraData.search_results && extraData.search_results.length > 0) {
                     console.log('Building search results HTML for', extraData.search_results.length, 'results');
                     const searchHtml = buildSearchResultsHtml(extraData.search_results, extraData.search_query || content);
-                    const bodyDiv = div.querySelector('.message-body');
-                    console.log('bodyDiv found:', bodyDiv != null);
-                    if (bodyDiv) {
-                        bodyDiv.insertAdjacentHTML('beforeend', searchHtml);
-                        console.log('Search results HTML inserted');
-                    }
+                    if (bodyDiv) bodyDiv.insertAdjacentHTML('beforeend', searchHtml);
                 }
             }
             
@@ -798,6 +887,16 @@
             // 如果没有消息且没有文件，不发送
             if (!msg && pendingFiles.length === 0) return;
             
+            // 检查是否有图片上传
+            const hasImages = pendingFiles.some(f => f.type && f.type.startsWith('image/'));
+            if (hasImages && !checkAgentVisionSupport()) {
+                const agent = agents.find(a => a.id === currentAgentId);
+                const agentName = agent?.display_name || agent?.name || '当前Agent';
+                alert(`⚠️ ${agentName} 不支持图片识别功能\n\n请选择支持视觉能力的Agent（如 vlm-agent），或者移除图片后再发送。`);
+                document.getElementById('sendBtn').disabled = false;
+                return;
+            }
+            
             document.getElementById('sendBtn').disabled = true;
             input.value = '';
             input.style.height = 'auto';
@@ -852,13 +951,15 @@
                 lastSentFiles = files.map(f => ({
                     name: f.name,
                     type: f.type,
-                    content: f.content
+                    content: f.content,
+                    serverPath: f.serverPath  // 服务器路径（用于历史记录）
                 }));
                 
                 for (const f of files) {
                     if (f.type.startsWith('image/')) {
-                        // 图片直接显示
-                        html += `<div class="uploaded-image" style="margin-bottom:8px"><img src="${f.content}" style="max-width:300px;border-radius:8px"></div>`;
+                        // 图片直接显示（用服务器路径或base64）
+                        const imgSrc = f.serverPath || f.content;
+                        html += `<div class="uploaded-image" style="margin-bottom:8px"><img src="${f.content}" style="max-width:300px;border-radius:8px" onclick="openImageLightbox('${imgSrc}')"></div>`;
                     } else {
                         // 文本文件显示名称和内容摘要
                         html += `<div class="uploaded-file" style="padding:8px;background:#f5f5f5;border-radius:6px;margin-bottom:8px">`;
@@ -883,7 +984,7 @@
         }
         
         // 文件上传处理
-        function handleFileUpload(event) {
+        async function handleFileUpload(event) {
             const files = event.target.files;
             const previewArea = document.getElementById('filePreviewArea');
             
@@ -892,13 +993,23 @@
                 
                 // 读取文件内容
                 const reader = new FileReader();
-                reader.onload = (e) => {
+                reader.onload = async (e) => {
+                    const base64Content = e.target.result;
+                    
+                    // 图片：先上传到服务器保存
+                    let serverPath = null;
+                    if (file.type.startsWith('image/')) {
+                        serverPath = await uploadImageToServer(base64Content, file.name);
+                        console.log('图片上传结果:', serverPath);
+                    }
+                    
                     const fileData = {
                         id: fileId,
                         name: file.name,
                         type: file.type,
                         size: file.size,
-                        content: e.target.result
+                        content: base64Content,  // base64数据（用于多模态模型）
+                        serverPath: serverPath   // 服务器路径（用于历史记录显示）
                     };
                     pendingFiles.push(fileData);
                     
@@ -909,8 +1020,9 @@
                     
                     if (file.type.startsWith('image/')) {
                         previewItem.classList.add('image-preview');
+                        // 预览用本地base64，显示更快
                         previewItem.innerHTML = `
-                            <img src="${e.target.result}" alt="${file.name}">
+                            <img src="${base64Content}" alt="${file.name}" style="cursor:pointer" onclick="openImageLightbox('${serverPath || base64Content}')">
                             <button class="file-remove" onclick="removeFile('${fileId}')"><i class="ri-close-line"></i></button>
                         `;
                     } else {
@@ -931,6 +1043,21 @@
                     }
                     
                     previewArea.appendChild(previewItem);
+                    
+                    // 如果上传的是图片且当前Agent不支持视觉，显示警告提示
+                    if (file.type.startsWith('image/') && !checkAgentVisionSupport()) {
+                        const agent = agents.find(a => a.id === currentAgentId);
+                        const agentName = agent?.display_name || agent?.name || '当前Agent';
+                        const warningDiv = document.createElement('div');
+                        warningDiv.className = 'vision-warning';
+                        warningDiv.id = 'vision-warning-tip';
+                        warningDiv.style.cssText = 'margin-top:8px;padding:8px 12px;background:#fff3cd;border:1px solid #ffc107;border-radius:6px;font-size:13px;color:#856404;';
+                        warningDiv.innerHTML = `<i class="ri-alert-line"></i> <strong>${agentName}</strong> 不支持图片识别，请选择支持视觉的Agent或移除图片`;
+                        // 避免重复添加
+                        if (!document.getElementById('vision-warning-tip')) {
+                            previewArea.appendChild(warningDiv);
+                        }
+                    }
                 };
                 
                 // 根据文件类型选择读取方式
@@ -952,6 +1079,13 @@
             pendingFiles = pendingFiles.filter(f => f.id !== fileId);
             const preview = document.getElementById(fileId + '-preview');
             if (preview) preview.remove();
+            
+            // 如果没有图片了，移除视觉警告提示
+            const hasImages = pendingFiles.some(f => f.type && f.type.startsWith('image/'));
+            if (!hasImages) {
+                const warning = document.getElementById('vision-warning-tip');
+                if (warning) warning.remove();
+            }
         }
         
         // 发送消息
@@ -1014,6 +1148,26 @@
         }
         
         document.getElementById('newPhraseInput').addEventListener('keydown', e => { if (e.key === 'Enter') addPhrase(); if (e.key === 'Escape') hidePhraseModal(); });
+        
+        // 图片放大弹窗
+        function openImageLightbox(imageSrc) {
+            const lightbox = document.getElementById('imageLightbox');
+            const lightboxImg = document.getElementById('lightboxImage');
+            lightboxImg.src = imageSrc;
+            lightbox.classList.add('show');
+        }
+        
+        function closeImageLightbox() {
+            const lightbox = document.getElementById('imageLightbox');
+            lightbox.classList.remove('show');
+        }
+        
+        // ESC键关闭图片弹窗
+        document.addEventListener('keydown', e => {
+            if (e.key === 'Escape') {
+                closeImageLightbox();
+            }
+        });
     </script>
 </body>
 </html>