feat: 添加图片搜索功能和 Qwen 模型支持

图片搜索功能（以图搜图）：
- Chatwoot webhook 检测图片搜索消息 (content_type="search_image")
- 从 content_attributes.url 提取图片 URL
- 调用 Mall API 图片搜索接口 (/mall/api/spu?searchImageUrl=...)
- 支持嵌套和顶层 URL 位置提取
- Product Agent 添加 fast path 直接调用图片搜索工具
- 防止无限循环（使用后清除 context.image_search_url）

Qwen 模型支持：
- 添加 LLM provider 选择（zhipu/qwen）
- 实现 QwenLLMClient 类（基于 DashScope SDK）
- 添加 dashscope>=1.14.0 依赖
- 修复 API key 设置（直接设置 dashscope.api_key）
- 更新 .env.example 和 docker-compose.yml 配置

其他优化：
- 重构 Chatwoot 集成代码（删除冗余）
- 优化 Product Agent prompt
- 增强 Customer Service Agent 多语言支持

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

agent/agents/customer_service.py

@@ -322,16 +322,51 @@ async def customer_service_agent(state: AgentState) -> AgentState:
             return state
 
         except json.JSONDecodeError as e:
-            # JSON parsing failed
+            # JSON parsing failed - try alternative format: "tool_name\n{args}"
             logger.error(
                 "Failed to parse LLM response as JSON",
                 error=str(e),
                 raw_content=content[:500],
                 conversation_id=state["conversation_id"]
             )
-            # Don't use raw content as response - use fallback instead
-            state = set_response(state, "抱歉，我无法理解您的请求。请尝试重新表述或联系人工客服。")
-            return state
+
+            # Handle non-JSON format: "tool_name\n{args}"
+            if '\n' in content and not content.startswith('{'):
+                lines = content.split('\n', 1)
+                tool_name = lines[0].strip()
+                args_json = lines[1].strip() if len(lines) > 1 else '{}'
+
+                try:
+                    arguments = json.loads(args_json) if args_json else {}
+                    logger.info(
+                        "Customer service agent calling tool (alternative format)",
+                        tool_name=tool_name,
+                        arguments=arguments,
+                        conversation_id=state["conversation_id"]
+                    )
+
+                    state = add_tool_call(
+                        state,
+                        tool_name=tool_name,
+                        arguments=arguments,
+                        server="strapi"
+                    )
+                    state["state"] = ConversationState.TOOL_CALLING.value
+                    return state
+                except json.JSONDecodeError:
+                    # Args parsing also failed
+                    logger.warning(
+                        "Failed to parse tool arguments",
+                        tool_name=tool_name,
+                        args_json=args_json[:200],
+                        conversation_id=state["conversation_id"]
+                    )
+                    state = set_response(state, "抱歉，我无法理解您的请求。请尝试重新表述或联系人工客服。")
+                    return state
+            else:
+                # Not a recognized format
+                state = set_response(state, "抱歉，我无法理解您的请求。请尝试重新表述或联系人工客服。")
+                return state
 
     except Exception as e:
         logger.error("Customer service agent failed", error=str(e), exc_info=True)
@@ -342,11 +377,59 @@ async def customer_service_agent(state: AgentState) -> AgentState:
 async def _generate_response_from_results(state: AgentState) -> AgentState:
     """Generate response based on tool results"""
 
-    # Build context from tool results
+    # Build context from tool results - extract only essential info to reduce prompt size
     tool_context = []
     for result in state["tool_results"]:
         if result["success"]:
-            tool_context.append(f"Tool {result['tool_name']} returned:\n{json.dumps(result['data'], ensure_ascii=False, indent=2)}")
+            tool_name = result['tool_name']
+            data = result['data']
+
+            # Extract only essential information based on tool type
+            if tool_name == "get_company_info":
+                # Extract key contact info only
+                contact = data.get('contact', {})
+                emails = contact.get('email', [])
+                if isinstance(emails, list) and emails:
+                    email_str = ", ".join(emails[:3])  # Max 3 emails
+                else:
+                    email_str = str(emails) if emails else "N/A"
+
+                phones = contact.get('phone', [])
+                if isinstance(phones, list) and phones:
+                    phone_str = ", ".join(phones[:2])  # Max 2 phones
+                else:
+                    phone_str = str(phones) if phones else "N/A"
+
+                address = contact.get('address', {})
+                address_str = f"{address.get('city', '')}, {address.get('country', '')}".strip(', ')
+
+                summary = f"Contact Information: Emails: {email_str} | Phones: {phone_str} | Address: {address_str} | Working hours: {contact.get('working_hours', 'N/A')}"
+                tool_context.append(summary)
+
+            elif tool_name == "query_faq" or tool_name == "search_knowledge_base":
+                # Extract FAQ items summary
+                faqs = data.get('faqs', []) if isinstance(data, dict) else []
+                if faqs:
+                    faq_summaries = [f"- Q: {faq.get('question', '')[:50]}... A: {faq.get('answer', '')[:50]}..." for faq in faqs[:3]]
+                    summary = f"Found {len(faqs)} FAQ items:\n" + "\n".join(faq_summaries)
+                    tool_context.append(summary)
+                else:
+                    tool_context.append("No FAQ items found")
+
+            elif tool_name == "get_categories":
+                # Extract category names only
+                categories = data.get('categories', []) if isinstance(data, dict) else []
+                category_names = [cat.get('name', '') for cat in categories[:5] if cat.get('name')]
+                summary = f"Available categories: {', '.join(category_names)}"
+                if len(categories) > 5:
+                    summary += f" (and {len(categories) - 5} more)"
+                tool_context.append(summary)
+
+            else:
+                # For other tools, include concise summary (limit to 200 chars)
+                data_str = json.dumps(data, ensure_ascii=False)[:200]
+                tool_context.append(f"Tool {tool_name} returned: {data_str}...")
+
         else:
             tool_context.append(f"Tool {result['tool_name']} failed: {result['error']}")
 
@@ -357,7 +440,8 @@ User question: {state["current_message"]}
 Tool returned information:
 {chr(10).join(tool_context)}
 
-Please generate a friendly and professional response. If the tool did not return useful information, honestly inform the user and suggest other ways to get help.
+Please generate a friendly and professional response in Chinese. Keep it concise but informative.
+If the tool did not return useful information, honestly inform the user and suggest other ways to get help.
 Return only the response content, do not return JSON."""
 
     messages = [
@@ -367,11 +451,12 @@ Return only the response content, do not return JSON."""
 
     try:
         llm = get_llm_client()
-        response = await llm.chat(messages, temperature=0.7)
+        # Lower temperature for faster response
+        response = await llm.chat(messages, temperature=0.3)
         state = set_response(state, response.content)
         return state
 
     except Exception as e:
         logger.error("Response generation failed", error=str(e))
-        state = set_response(state, "Sorry, there was a problem processing your request. Please try again later or contact customer support.")
+        state = set_response(state, "抱歉，处理您的请求时出现问题。请稍后重试或联系人工客服。")
         return state