更新DeepSeek模型配置，新增DeepSeek-V3和DeepSeek推理模型，优化API调用逻辑，支持多模态模型的图像分析，改进前端按钮显示逻辑，提升用户体验

config/models.json

@@ -38,15 +38,23 @@
             "provider": "openai",
             "supportsMultimodal": false,
             "isReasoning": true,
-            "version": "不可用",
+            "version": "2025-01-31",
             "description": "OpenAI的o3-mini模型，支持图像理解和思考过程"
         },
-        "deepseek-r1": {
+        "deepseek-chat": {
+            "name": "DeepSeek-V3",
+            "provider": "deepseek",
+            "supportsMultimodal": false,
+            "isReasoning": false,
+            "version": "2025-01",
+            "description": "DeepSeek最新大模型，671B MoE模型，支持60 tokens/秒的高速生成"
+        },
+        "deepseek-reasoner": {
             "name": "DeepSeek-R1",
             "provider": "deepseek",
             "supportsMultimodal": false,
             "isReasoning": true,
-            "version": "不可用",
+            "version": "latest",
             "description": "DeepSeek推理模型，提供详细思考过程（仅支持文本）"
         }
     }

models/deepseek.py

@@ -1,10 +1,15 @@
 import json
 import requests
+import os
 from typing import Generator
 from openai import OpenAI
 from .base import BaseModel
 
 class DeepSeekModel(BaseModel):
+    def __init__(self, api_key: str, temperature: float = 0.7, system_prompt: str = None, language: str = None, model_name: str = "deepseek-reasoner"):
+        super().__init__(api_key, temperature, system_prompt, language)
+        self.model_name = model_name
+
     def get_default_system_prompt(self) -> str:
         return """You are an expert at analyzing questions and providing detailed solutions. When presented with an image of a question:
 1. First read and understand the question carefully
@@ -14,6 +19,11 @@ class DeepSeekModel(BaseModel):
 5. If there are multiple approaches, explain the most efficient one first"""
 
     def get_model_identifier(self) -> str:
+        """根据模型名称返回正确的API标识符"""
+        # 通过模型名称来确定实际的API调用标识符
+        if self.model_name == "deepseek-chat":
+            return "deepseek-chat"
+        # deepseek-reasoner是默认的推理模型名称
         return "deepseek-reasoner"
 
     def analyze_text(self, text: str, proxies: dict = None) -> Generator[dict, None, None]:
@@ -22,142 +32,365 @@ class DeepSeekModel(BaseModel):
             # Initial status
             yield {"status": "started", "content": ""}
 
-            # Configure client with proxy if needed
-            client_args = {
-                "api_key": self.api_key,
-                "base_url": "https://api.deepseek.com"
+            # 保存原始环境变量
+            original_env = {
+                'http_proxy': os.environ.get('http_proxy'),
+                'https_proxy': os.environ.get('https_proxy')
             }
-            
-            if proxies:
-                session = requests.Session()
-                session.proxies = proxies
-                client_args["http_client"] = session
 
-            client = OpenAI(**client_args)
+            try:
+                # 如果提供了代理设置，通过环境变量设置
+                if proxies:
+                    if 'http' in proxies:
+                        os.environ['http_proxy'] = proxies['http']
+                    if 'https' in proxies:
+                        os.environ['https_proxy'] = proxies['https']
 
-            response = client.chat.completions.create(
-                model=self.get_model_identifier(),
-                messages=[{
-                    'role': 'system',
-                    'content': self.system_prompt
-                }, {
-                    'role': 'user',
-                    'content': text
-                }],
-                stream=True
-            )
+                # 初始化DeepSeek客户端，不再使用session对象
+                client = OpenAI(
+                    api_key=self.api_key,
+                    base_url="https://api.deepseek.com"
+                )
 
-            for chunk in response:
-                try:
-                    if hasattr(chunk.choices[0].delta, 'reasoning_content'):
-                        content = chunk.choices[0].delta.reasoning_content
-                        if content:
-                            yield {
-                                "status": "streaming",
-                                "content": content
-                            }
-                    elif hasattr(chunk.choices[0].delta, 'content'):
-                        content = chunk.choices[0].delta.content
-                        if content:
-                            yield {
-                                "status": "streaming",
-                                "content": content
-                            }
+                # 添加系统语言指令
+                system_prompt = self.system_prompt
+                language = self.language or '中文'
+                if not any(phrase in system_prompt for phrase in ['Please respond in', '请用', '使用', '回答']):
+                    system_prompt = f"{system_prompt}\n\n请务必使用{language}回答。"
 
-                except Exception as e:
-                    print(f"Chunk processing error: {str(e)}")
-                    continue
+                # 构建请求参数
+                params = {
+                    "model": self.get_model_identifier(),
+                    "messages": [
+                        {
+                            'role': 'system',
+                            'content': system_prompt
+                        }, 
+                        {
+                            'role': 'user',
+                            'content': text
+                        }
+                    ],
+                    "stream": True
+                }
+                
+                # 只有非推理模型才设置temperature参数
+                if not self.model_name.endswith('reasoner') and self.temperature is not None:
+                    params["temperature"] = self.temperature
+                    
+                print(f"调用DeepSeek API: {self.get_model_identifier()}, 是否设置温度: {not self.model_name.endswith('reasoner')}")
 
-            # Send completion status
-            yield {
-                "status": "completed",
-                "content": ""
-            }
+                response = client.chat.completions.create(**params)
+                
+                # 使用两个缓冲区，分别用于常规内容和思考内容
+                response_buffer = ""
+                thinking_buffer = ""
+                
+                for chunk in response:
+                    # 打印chunk以调试
+                    try:
+                        print(f"DeepSeek API返回chunk: {chunk}")
+                    except:
+                        print("无法打印chunk")
+                        
+                    try:
+                        # 处理推理模型的思考内容
+                        if hasattr(chunk.choices[0].delta, 'reasoning_content'):
+                            content = chunk.choices[0].delta.reasoning_content
+                            if content:
+                                # 累积思考内容
+                                thinking_buffer += content
+                                
+                                # 只在积累一定数量的字符或遇到句子结束标记时才发送
+                                if len(content) >= 20 or content.endswith(('.', '!', '?', '。', '！', '？', '\n')):
+                                    yield {
+                                        "status": "thinking",
+                                        "content": thinking_buffer
+                                    }
+                                    
+                        # 处理常规内容
+                        elif hasattr(chunk.choices[0].delta, 'content'):
+                            content = chunk.choices[0].delta.content
+                            if content:
+                                # 累积响应内容
+                                response_buffer += content
+                                print(f"累积响应内容: '{content}', 当前buffer: '{response_buffer}'")
+                                
+                                # 只在积累一定数量的字符或遇到句子结束标记时才发送
+                                if len(content) >= 10 or content.endswith(('.', '!', '?', '。', '！', '？', '\n')):
+                                    yield {
+                                        "status": "streaming",
+                                        "content": response_buffer
+                                    }
+                        # 尝试直接从message内容获取
+                        elif hasattr(chunk.choices[0], 'message') and hasattr(chunk.choices[0].message, 'content'):
+                            content = chunk.choices[0].message.content
+                            if content:
+                                response_buffer += content
+                                print(f"从message获取内容: '{content}'")
+                                yield {
+                                    "status": "streaming",
+                                    "content": response_buffer
+                                }
+                        # 检查是否有finish_reason，表示生成结束
+                        elif hasattr(chunk.choices[0], 'finish_reason') and chunk.choices[0].finish_reason:
+                            print(f"生成结束，原因: {chunk.choices[0].finish_reason}")
+                            
+                            # 如果没有内容但有思考内容，把思考内容作为正文显示
+                            if not response_buffer and thinking_buffer:
+                                response_buffer = thinking_buffer
+                                yield {
+                                    "status": "streaming",
+                                    "content": response_buffer
+                                }
+                    except Exception as e:
+                        print(f"解析响应chunk时出错: {str(e)}")
+                        continue
+
+                # 确保发送最终的缓冲内容
+                if thinking_buffer:
+                    yield {
+                        "status": "thinking_complete",
+                        "content": thinking_buffer
+                    }
+                
+                # 如果推理完成后没有正文内容，则使用思考内容作为最终响应
+                if not response_buffer and thinking_buffer:
+                    response_buffer = thinking_buffer
+                
+                # 发送最终响应内容
+                if response_buffer:
+                    yield {
+                        "status": "streaming",
+                        "content": response_buffer
+                    }
+
+                # 发送完成状态
+                yield {
+                    "status": "completed",
+                    "content": response_buffer
+                }
+                
+            except Exception as e:
+                error_msg = str(e)
+                print(f"DeepSeek API调用出错: {error_msg}")
+                
+                # 提供具体的错误信息
+                if "invalid_api_key" in error_msg.lower():
+                    error_msg = "DeepSeek API密钥无效，请检查您的API密钥"
+                elif "rate_limit" in error_msg.lower():
+                    error_msg = "DeepSeek API请求频率超限，请稍后再试"
+                elif "quota_exceeded" in error_msg.lower():
+                    error_msg = "DeepSeek API配额已用完，请续费或等待下个计费周期"
+                
+                yield {
+                    "status": "error",
+                    "error": f"DeepSeek API错误: {error_msg}"
+                }
+            finally:
+                # 恢复原始环境变量
+                for key, value in original_env.items():
+                    if value is None:
+                        if key in os.environ:
+                            del os.environ[key]
+                    else:
+                        os.environ[key] = value
 
         except Exception as e:
             error_msg = str(e)
+            print(f"调用DeepSeek模型时发生错误: {error_msg}")
+            
             if "invalid_api_key" in error_msg.lower():
-                error_msg = "Invalid API key provided"
+                error_msg = "API密钥无效，请检查设置"
             elif "rate_limit" in error_msg.lower():
-                error_msg = "Rate limit exceeded. Please try again later."
+                error_msg = "API请求频率超限，请稍后再试"
             
             yield {
                 "status": "error",
-                "error": f"DeepSeek API error: {error_msg}"
+                "error": f"DeepSeek API错误: {error_msg}"
             }
 
     def analyze_image(self, image_data: str, proxies: dict = None) -> Generator[dict, None, None]:
         """Stream DeepSeek's response for image analysis"""
         try:
+            # 检查我们是否有支持图像的模型
+            if self.model_name == "deepseek-chat" or self.model_name == "deepseek-reasoner":
+                yield {
+                    "status": "error",
+                    "error": "当前DeepSeek模型不支持图像分析，请使用Anthropic或OpenAI的多模态模型"
+                }
+                return
+                
             # Initial status
             yield {"status": "started", "content": ""}
 
-            # Configure client with proxy if needed
-            client_args = {
-                "api_key": self.api_key,
-                "base_url": "https://api.deepseek.com"
+            # 保存原始环境变量
+            original_env = {
+                'http_proxy': os.environ.get('http_proxy'),
+                'https_proxy': os.environ.get('https_proxy')
             }
-            
-            if proxies:
-                session = requests.Session()
-                session.proxies = proxies
-                client_args["http_client"] = session
 
-            client = OpenAI(**client_args)
+            try:
+                # 如果提供了代理设置，通过环境变量设置
+                if proxies:
+                    if 'http' in proxies:
+                        os.environ['http_proxy'] = proxies['http']
+                    if 'https' in proxies:
+                        os.environ['https_proxy'] = proxies['https']
 
-            # 检查系统提示词是否已包含语言设置指令
-            system_prompt = self.system_prompt
-            language = self.language or '中文'
-            if not any(phrase in system_prompt for phrase in ['Please respond in', '请用', '使用', '回答']):
-                system_prompt = f"{system_prompt}\n\n请务必使用{language}回答，无论问题是什么语言。即使在分析图像时也请使用{language}回答。"
+                # 初始化DeepSeek客户端，不再使用session对象
+                client = OpenAI(
+                    api_key=self.api_key,
+                    base_url="https://api.deepseek.com"
+                )
 
-            response = client.chat.completions.create(
-                model=self.get_model_identifier(),
-                messages=[{
-                    'role': 'system',
-                    'content': system_prompt
-                }, {
-                    'role': 'user',
-                    'content': f"Here's an image of a question to analyze: data:image/png;base64,{image_data}"
-                }],
-                stream=True
-            )
+                # 检查系统提示词是否已包含语言设置指令
+                system_prompt = self.system_prompt
+                language = self.language or '中文'
+                if not any(phrase in system_prompt for phrase in ['Please respond in', '请用', '使用', '回答']):
+                    system_prompt = f"{system_prompt}\n\n请务必使用{language}回答，无论问题是什么语言。即使在分析图像时也请使用{language}回答。"
 
-            for chunk in response:
-                try:
-                    if hasattr(chunk.choices[0].delta, 'reasoning_content'):
-                        content = chunk.choices[0].delta.reasoning_content
-                        if content:
-                            yield {
-                                "status": "streaming",
-                                "content": content
-                            }
-                    elif hasattr(chunk.choices[0].delta, 'content'):
-                        content = chunk.choices[0].delta.content
-                        if content:
-                            yield {
-                                "status": "streaming",
-                                "content": content
-                            }
+                # 构建请求参数
+                params = {
+                    "model": self.get_model_identifier(),
+                    "messages": [
+                        {
+                            'role': 'system',
+                            'content': system_prompt
+                        }, 
+                        {
+                            'role': 'user',
+                            'content': f"Here's an image of a question to analyze: data:image/png;base64,{image_data}"
+                        }
+                    ],
+                    "stream": True
+                }
+                
+                # 只有非推理模型才设置temperature参数
+                if not self.model_name.endswith('reasoner') and self.temperature is not None:
+                    params["temperature"] = self.temperature
 
-                except Exception as e:
-                    print(f"Chunk processing error: {str(e)}")
-                    continue
+                response = client.chat.completions.create(**params)
+                
+                # 使用两个缓冲区，分别用于常规内容和思考内容
+                response_buffer = ""
+                thinking_buffer = ""
+                
+                for chunk in response:
+                    # 打印chunk以调试
+                    try:
+                        print(f"DeepSeek图像API返回chunk: {chunk}")
+                    except:
+                        print("无法打印chunk")
+                
+                    try:
+                        # 处理推理模型的思考内容
+                        if hasattr(chunk.choices[0].delta, 'reasoning_content'):
+                            content = chunk.choices[0].delta.reasoning_content
+                            if content:
+                                # 累积思考内容
+                                thinking_buffer += content
+                                
+                                # 只在积累一定数量的字符或遇到句子结束标记时才发送
+                                if len(content) >= 20 or content.endswith(('.', '!', '?', '。', '！', '？', '\n')):
+                                    yield {
+                                        "status": "thinking",
+                                        "content": thinking_buffer
+                                    }
+                        # 处理常规内容
+                        elif hasattr(chunk.choices[0].delta, 'content'):
+                            content = chunk.choices[0].delta.content
+                            if content:
+                                # 累积响应内容
+                                response_buffer += content
+                                print(f"累积图像响应内容: '{content}', 当前buffer: '{response_buffer}'")
+                                
+                                # 只在积累一定数量的字符或遇到句子结束标记时才发送
+                                if len(content) >= 10 or content.endswith(('.', '!', '?', '。', '！', '？', '\n')):
+                                    yield {
+                                        "status": "streaming",
+                                        "content": response_buffer
+                                    }
+                        # 尝试直接从message内容获取
+                        elif hasattr(chunk.choices[0], 'message') and hasattr(chunk.choices[0].message, 'content'):
+                            content = chunk.choices[0].message.content
+                            if content:
+                                response_buffer += content
+                                print(f"从message获取图像内容: '{content}'")
+                                yield {
+                                    "status": "streaming",
+                                    "content": response_buffer
+                                }
+                        # 检查是否有finish_reason，表示生成结束
+                        elif hasattr(chunk.choices[0], 'finish_reason') and chunk.choices[0].finish_reason:
+                            print(f"图像生成结束，原因: {chunk.choices[0].finish_reason}")
+                            
+                            # 如果没有内容但有思考内容，把思考内容作为正文显示
+                            if not response_buffer and thinking_buffer:
+                                response_buffer = thinking_buffer
+                                yield {
+                                    "status": "streaming",
+                                    "content": response_buffer
+                                }
+                    except Exception as e:
+                        print(f"解析图像响应chunk时出错: {str(e)}")
+                        continue
 
-            # Send completion status
-            yield {
-                "status": "completed",
-                "content": ""
-            }
+                # 确保发送最终的缓冲内容
+                if thinking_buffer:
+                    yield {
+                        "status": "thinking_complete",
+                        "content": thinking_buffer
+                    }
+                
+                # 如果推理完成后没有正文内容，则使用思考内容作为最终响应
+                if not response_buffer and thinking_buffer:
+                    response_buffer = thinking_buffer
+                
+                # 发送最终响应内容
+                if response_buffer:
+                    yield {
+                        "status": "streaming",
+                        "content": response_buffer
+                    }
+
+                # 发送完成状态
+                yield {
+                    "status": "completed",
+                    "content": response_buffer
+                }
+                
+            except Exception as e:
+                error_msg = str(e)
+                print(f"DeepSeek API调用出错: {error_msg}")
+                
+                # 提供具体的错误信息
+                if "invalid_api_key" in error_msg.lower():
+                    error_msg = "DeepSeek API密钥无效，请检查您的API密钥"
+                elif "rate_limit" in error_msg.lower():
+                    error_msg = "DeepSeek API请求频率超限，请稍后再试"
+                
+                yield {
+                    "status": "error",
+                    "error": f"DeepSeek API错误: {error_msg}"
+                }
+            finally:
+                # 恢复原始环境变量
+                for key, value in original_env.items():
+                    if value is None:
+                        if key in os.environ:
+                            del os.environ[key]
+                    else:
+                        os.environ[key] = value
 
         except Exception as e:
             error_msg = str(e)
             if "invalid_api_key" in error_msg.lower():
-                error_msg = "Invalid API key provided"
+                error_msg = "API密钥无效，请检查设置"
             elif "rate_limit" in error_msg.lower():
-                error_msg = "Rate limit exceeded. Please try again later."
+                error_msg = "API请求频率超限，请稍后再试"
             
             yield {
                 "status": "error",
-                "error": f"DeepSeek API error: {error_msg}"
+                "error": f"DeepSeek API错误: {error_msg}"
             }

models/factory.py

@@ -126,6 +126,15 @@ class ModelFactory:
                 temperature=temperature,
                 system_prompt=system_prompt
             )
+        # 对于DeepSeek模型，传递model_name参数
+        elif "deepseek" in model_name.lower():
+            return model_class(
+                api_key=api_key,
+                temperature=temperature,
+                system_prompt=system_prompt,
+                language=language,
+                model_name=model_name
+            )
         else:
             # 对于其他模型，传递所有参数
             return model_class(

static/js/main.js

@@ -698,8 +698,10 @@ class SnapSolver {
                     this.screenshotImg.src = this.croppedImage;
                     this.imagePreview.classList.remove('hidden');
                     this.cropBtn.classList.remove('hidden');
-                    this.sendToClaudeBtn.classList.remove('hidden');
-                    this.extractTextBtn.classList.remove('hidden');
+                    
+                    // 根据当前选择的模型类型决定显示哪些按钮
+                    this.updateImageActionButtons();
+                    
                     window.uiManager.showToast('Cropping successful');
                 } catch (error) {
                     console.error('Cropping error details:', {
@@ -725,8 +727,8 @@ class SnapSolver {
                 this.cropper = null;
             }
             this.cropContainer.classList.add('hidden');
-            this.sendToClaudeBtn.classList.add('hidden');
-            this.extractTextBtn.classList.add('hidden');
+            // 取消裁剪时隐藏图像预览和相关按钮
+            this.imagePreview.classList.add('hidden');
             document.querySelector('.crop-area').innerHTML = '';
         });
     }
@@ -858,6 +860,17 @@ class SnapSolver {
             if (this.sendToClaudeBtn.disabled) return;
             this.sendToClaudeBtn.disabled = true;
             
+            // 获取当前模型设置
+            const settings = window.settingsManager.getSettings();
+            const isMultimodalModel = settings.modelInfo?.supportsMultimodal || false;
+            const modelName = settings.model || '未知';
+            
+            if (!isMultimodalModel) {
+                window.uiManager.showToast(`当前选择的模型 ${modelName} 不支持图像分析。请先提取文本或切换到支持多模态的模型。`, 'error');
+                this.sendToClaudeBtn.disabled = false;
+                return;
+            }
+            
             if (this.croppedImage) {
                 try {
                     // 清空之前的结果
@@ -1068,6 +1081,12 @@ class SnapSolver {
         // 更新图像操作按钮
         this.updateImageActionButtons();
         
+        // 确保DOM完全加载后再次更新按钮状态（以防设置未完全加载）
+        setTimeout(() => {
+            this.updateImageActionButtons();
+            console.log('延时更新图像操作按钮完成');
+        }, 1000);
+        
         console.log('SnapSolver initialization complete');
     }
 
@@ -1187,25 +1206,40 @@ class SnapSolver {
 
     // 新增方法：根据所选模型更新图像操作按钮
     updateImageActionButtons() {
-        if (!window.settingsManager) return;
+        if (!window.settingsManager) {
+            console.error('Settings manager not available');
+            return;
+        }
         
         const settings = window.settingsManager.getSettings();
         const isMultimodalModel = settings.modelInfo?.supportsMultimodal || false;
+        const modelName = settings.model || '未知';
+        
+        console.log(`更新图像操作按钮 - 当前模型: ${modelName}, 是否支持多模态: ${isMultimodalModel}`);
         
         // 对于截图后的操作按钮显示逻辑
         if (this.sendToClaudeBtn && this.extractTextBtn) {
             if (!isMultimodalModel) {
                 // 非多模态模型：只显示提取文本按钮，隐藏发送到AI按钮
+                console.log('非多模态模型：隐藏"发送图片至AI"按钮');
                 this.sendToClaudeBtn.classList.add('hidden');
                 this.extractTextBtn.classList.remove('hidden');
             } else {
                 // 多模态模型：显示两个按钮
                 if (!this.imagePreview.classList.contains('hidden')) {
                     // 只有在有图像时才显示按钮
+                    console.log('多模态模型：显示全部按钮');
                     this.sendToClaudeBtn.classList.remove('hidden');
                     this.extractTextBtn.classList.remove('hidden');
+                } else {
+                    // 无图像时隐藏所有按钮
+                    console.log('无图像：隐藏所有按钮');
+                    this.sendToClaudeBtn.classList.add('hidden');
+                    this.extractTextBtn.classList.add('hidden');
                 }
             }
+        } else {
+            console.warn('按钮元素不可用');
         }
     }
 }

static/js/settings.js

@@ -434,6 +434,11 @@ class SettingsManager {
             this.updateUIBasedOnModelType();
             this.updateModelVersionDisplay(e.target.value);
             this.saveSettings();
+            
+            // 通知应用更新图像操作按钮
+            if (window.app && typeof window.app.updateImageActionButtons === 'function') {
+                window.app.updateImageActionButtons();
+            }
         });
 
         this.temperatureInput.addEventListener('input', (e) => {