app.py

import os
import sys
import turtle
import requests
import json
import gradio as gr
import logging
from typing import List, Optional, Dict, Tuple, Any
from dataclasses import dataclass
from dotenv import load_dotenv
import base64
import threading
import time
from logging.handlers import RotatingFileHandler
import functools
from PIL import Image
import io
from urllib.parse import urlparse
import tempfile

# Load environment variables
load_dotenv()

# Configure logging with rotation
def setup_logging():
    log_file = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'chatbot.log')
    handler = RotatingFileHandler(
        log_file,
        maxBytes=10*1024*1024,  # 10MB
        backupCount=5,
        encoding='utf-8'
    )
    handler.setFormatter(logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s'))
    
    logger = logging.getLogger(__name__)
    logger.setLevel(logging.INFO)
    logger.addHandler(handler)
    
    # Also log to console
    console_handler = logging.StreamHandler()
    console_handler.setFormatter(logging.Formatter('%(levelname)s: %(message)s'))
    logger.addHandler(console_handler)
    
    return logger

logger = setup_logging()

# Configuration
class Config:
    # Default values as class attributes
    API_BASE_URL = os.getenv('OLLAMA_API_URL', 'http://localhost:11434')
    GRADIO_SERVER_PORT = int(os.getenv('GRADIO_SERVER_PORT', '7588'))
    OCR_TIMEOUT = int(os.getenv('OCR_TIMEOUT', '60'))
    CHAT_TIMEOUT = int(os.getenv('CHAT_TIMEOUT', '30'))
    MAX_HISTORY = int(os.getenv('MAX_HISTORY', '100'))
    RETRY_ATTEMPTS = int(os.getenv('RETRY_ATTEMPTS', '3'))
    REQUEST_RATE_LIMIT = int(os.getenv('REQUEST_RATE_LIMIT', '10'))
    CACHE_TIMEOUT = int(os.getenv('CACHE_TIMEOUT', '3600'))
    MAX_INPUT_LENGTH = int(os.getenv('MAX_INPUT_LENGTH', '1000'))

    API_GENERATE_URL = f"{API_BASE_URL}/api/generate"
    API_CHAT_URL = f"{API_BASE_URL}/api/chat"

    @classmethod
    def validate(cls):
        if not (1024 <= cls.GRADIO_SERVER_PORT <= 65535):
            raise ValueError("Port must be between 1024 and 65535")
        if cls.OCR_TIMEOUT < 5:
            logger.warning(f"OCR_TIMEOUT is too small, using 5s")
            cls.OCR_TIMEOUT = 5
        if cls.CHAT_TIMEOUT < 5:
            logger.warning(f"CHAT_TIMEOUT is too small, using 5s")
            cls.CHAT_TIMEOUT = 5
        try:
            result = urlparse(cls.API_BASE_URL)
            if not all([result.scheme, result.netloc]):
                raise ValueError("Invalid URL format")
        except Exception as e:
            raise ValueError(f"Invalid API URL configuration: {str(e)}")

# Validate configuration on module load
Config.validate()

class RateLimit:
    def __init__(self, max_requests: int, per_seconds: int = 60):
        self.max_requests = max_requests
        self.per_seconds = per_seconds
        self.requests = []
        self._lock = threading.Lock()

    def is_allowed(self) -> bool:
        now = time.time()
        with self._lock:
            # 清理过期的请求记录
            self.requests = [req_time for req_time in self.requests 
                           if now - req_time < self.per_seconds]
            
            if len(self.requests) >= self.max_requests:
                return False
                
            self.requests.append(now)
            return True

class Cache:
    def __init__(self, timeout: int = 3600):
        self.cache = {}
        self.timeout = timeout
        self._lock = threading.Lock()

    def get(self, key: str) -> Optional[str]:
        with self._lock:
            if key in self.cache:
                timestamp, value = self.cache[key]
                if time.time() - timestamp < self.timeout:
                    return value
                del self.cache[key]
            return None

    def set(self, key: str, value: str) -> None:
        with self._lock:
            self.cache[key] = (time.time(), value)

class Conversation:
    def __init__(self, history: List[str], max_length: int = Config.MAX_HISTORY):
        self.history = history
        self.max_length = max_length

    def add_message(self, message: str) -> None:
        self.history.append(message)
        if len(self.history) > self.max_length:
            self.history = self.history[-self.max_length:]

    def get_context(self) -> str:
        return "\n".join(self.history)

    def clear(self) -> None:
        self.history = []

class ResourceManager:
    def __init__(self, base_dir: str = None):
        if base_dir is None:
            base_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'temp')
        self.base_dir = base_dir
        self.temp_files = set()
        self._lock = threading.Lock()
        self._setup_temp_dir()
        
    def _setup_temp_dir(self):
        """初始化临时目录"""
        try:
            os.makedirs(self.base_dir, exist_ok=True)
            # 启动时清理遗留的临时文件
            self.cleanup_all()
        except Exception as e:
            logger.error(f"Failed to setup temp directory: {e}")
            raise
            
    def add_temp_file(self, filepath: str):
        """记录临时文件"""
        with self._lock:
            self.temp_files.add(filepath)
            
    def remove_temp_file(self, filepath: str):
        """删除临时文件"""
        with self._lock:
            try:
                if os.path.exists(filepath):
                    os.remove(filepath)
                self.temp_files.discard(filepath)
            except Exception as e:
                logger.error(f"Failed to remove temp file {filepath}: {e}")
                
    def cleanup_all(self):
        """清理所有临时文件"""
        with self._lock:
            for filepath in list(self.temp_files):
                self.remove_temp_file(filepath)
            # 清理可能遗留的其他临时文件
            for filename in os.listdir(self.base_dir):
                try:
                    filepath = os.path.join(self.base_dir, filename)
                    if os.path.isfile(filepath):
                        os.remove(filepath)
                except Exception as e:
                    logger.error(f"Failed to remove file {filepath}: {e}")
                    
    def create_temp_file(self, suffix: str = '') -> str:
        """创建临时文件"""
        with self._lock:
            try:
                fd, temp_path = tempfile.mkstemp(suffix=suffix, dir=self.base_dir)
                os.close(fd)
                self.add_temp_file(temp_path)
                return temp_path
            except Exception as e:
                logger.error(f"Failed to create temp file: {e}")
                raise
                
    def __del__(self):
        """确保在对象销毁时清理临时文件"""
        self.cleanup_all()

class ProcessState:
    """处理状态管理"""
    def __init__(self):
        self._state = {}
        self._lock = threading.Lock()
        
    def set_state(self, key: str, value: Any) -> None:
        """设置状态"""
        with self._lock:
            self._state[key] = value
            
    def get_state(self, key: str, default: Any = None) -> Any:
        """获取状态"""
        with self._lock:
            return self._state.get(key, default)
            
    def clear_state(self, key: str) -> None:
        """清除状态"""
        with self._lock:
            self._state.pop(key, None)
            
    def is_processing(self, key: str) -> bool:
        """检查是否正在处理"""
        return self.get_state(key) == 'processing'

def retry_on_exception(retries: int = 3, delay: float = 1.0):
    """重试装饰器"""
    def decorator(func):
        @functools.wraps(func)
        def wrapper(*args, **kwargs):
            for attempt in range(retries):
                try:
                    return func(*args, **kwargs)
                except Exception as e:
                    if attempt == retries - 1:
                        raise
                    logger.warning(f"Attempt {attempt + 1} failed: {e}")
                    time.sleep(delay * (attempt + 1))  # 指数退避
            return None
        return wrapper
    return decorator

class ChatBot:
    def __init__(self):
        self.conversation = Conversation([])
        self.session = requests.Session()
        self.current_image: Optional[str] = None
        self.current_ocr_result: Optional[str] = None
        self.rate_limit = RateLimit(Config.REQUEST_RATE_LIMIT)
        self.cache = Cache(Config.CACHE_TIMEOUT)
        self.resource_manager = ResourceManager()
        self.process_state = ProcessState()
        
        # 设置会话超时
        self.session.timeout = (Config.CHAT_TIMEOUT, Config.CHAT_TIMEOUT)
        
        # 设置重试策略
        retry_strategy = requests.adapters.Retry(
            total=3,
            backoff_factor=0.5,
            status_forcelist=[500, 502, 503, 504]
        )
        adapter = requests.adapters.HTTPAdapter(max_retries=retry_strategy)
        self.session.mount("http://", adapter)
        self.session.mount("https://", adapter)

    def check_ollama_service(self):
        return check_ollama_service()

    def _encode_image(self, image_path: str) -> str:
        """编码图片，支持压缩"""
        try:
            # 验证文件
            if not os.path.exists(image_path):
                raise ValueError("图片文件不存在")
                
            # 验证文件类型
            try:
                with Image.open(image_path) as img:
                    if img.format.upper() not in {'JPEG', 'PNG', 'GIF', 'WEBP'}:
                        raise ValueError("不支持的图片格式")
            except Exception as e:
                raise ValueError(f"无效的图片文件: {str(e)}")
                
            # 检查文件大小并在需要时压缩
            file_size = os.path.getsize(image_path)
            max_size = 10 * 1024 * 1024  # 10MB
            
            if file_size > max_size:
                compressed_path = self.resource_manager.create_temp_file(
                    suffix=os.path.splitext(image_path)[1]
                )
                image_path = self.compress_image(image_path, compressed_path, max_size)
            
            # 编码图片
            with open(image_path, "rb") as image_file:
                return base64.b64encode(image_file.read()).decode('utf-8')
                
        except Exception as e:
            logger.error(f"Image processing failed: {str(e)}")
            raise
            
    def compress_image(self, input_path: str, output_path: str, max_size: int) -> str:
        """压缩图片到指定大小以下"""
        try:
            with Image.open(input_path) as img:
                format = img.format or 'JPEG'
                quality = 95
                
                while True:
                    # 保存压缩后的图片
                    img.save(output_path, format=format, quality=quality)
                    size = os.path.getsize(output_path)
                    
                    if size <= max_size or quality <= 5:
                        break
                        
                    quality -= 5
                    
                logger.info(f"Image compressed from {os.path.getsize(input_path)} to {size} bytes")
                return output_path
                
        except Exception as e:
            logger.error(f"Image compression failed: {str(e)}")
            if os.path.exists(output_path):
                self.resource_manager.remove_temp_file(output_path)
            raise

    @retry_on_exception(retries=Config.RETRY_ATTEMPTS)
    def _validate_input(self, text: str) -> str:
        """验证并清理输入文本"""
        if not text or not text.strip():
            raise ValueError("输入不能为空")
        
        text = text.strip()
        if len(text) > Config.MAX_INPUT_LENGTH:
            raise ValueError(f"输入长度不能超过{Config.MAX_INPUT_LENGTH}字符")
            
        # 移除潜在的危险字符
        text = text.replace('<', '&lt;').replace('>', '&gt;')
        return text

    def get_model_list(self) -> Tuple[List[str], str]:
        """Returns (model_list, status_message)"""
        try:
            response = self.session.get(
                f"{Config.API_BASE_URL}/api/tags",
                timeout=Config.CHAT_TIMEOUT
            )
            response.raise_for_status()
            data = response.json()
            if not data or 'models' not in data or not data['models']:
                return [], "No models found. Please install at least one model using 'ollama pull model_name'"
            # Extract model names and check for vision capabilities
            models = []
            for model in data['models']:
                name = model['name']
                details = model.get('details', {})
                families = details.get('families', [])
                # Add a visual indicator for models with vision capabilities
                if 'clip' in families:
                    name = f"🖼️ {name}"
                models.append(name)
            return models, "Models refreshed successfully!"
        except requests.exceptions.ConnectionError:
            return [], "Error: Cannot connect to Ollama server. Is it running?"
        except requests.exceptions.Timeout:
            return [], "Error: Request timed out. Please try again."
        except Exception as e:
            logger.error(f"Error fetching model list: {e}")
            return [], f"Error: {str(e)}"

    def generate_response(self, prompt: str, model_name: str, image: Optional[str] = None, is_ocr: bool = False) -> str:
        """生成响应，支持OCR和问答两种模式"""
        # 生成唯一的处理ID
        process_id = f"{time.time()}_{model_name}"
        
        try:
            # 检查是否有其他处理正在进行
            if self.process_state.is_processing(model_name):
                return "Error: 模型正在处理其他请求，请稍后再试"
            
            # 设置处理状态
            self.process_state.set_state(model_name, 'processing')
            
            # 输入验证
            try:
                prompt = self._validate_input(prompt)
                model_name = self._validate_input(model_name)
            except ValueError as e:
                return f"Error: {str(e)}"

            # 请求频率限制
            if not self.rate_limit.is_allowed():
                return "Error: 请求过于频繁，请稍后再试"

            # 检查缓存
            cache_key = f"{prompt}_{model_name}_{image if image else ''}"
            cached_response = self.cache.get(cache_key)
            if cached_response:
                return cached_response

            # 原有的响应生成逻辑
            mode = "OCR" if is_ocr else "问答"
            for attempt in range(Config.RETRY_ATTEMPTS):
                try:
                    # 根据模式选择超时时间
                    timeout = Config.OCR_TIMEOUT if is_ocr else Config.CHAT_TIMEOUT
                    
                    headers = {"Content-Type": "application/json"}
                    payload = {
                        "model": model_name.replace("🖼️ ", ""),
                        "prompt": prompt,
                        "stream": True
                    }
                    
                    if image:
                        payload["images"] = [image]
                    
                    response = self.session.post(
                        Config.API_GENERATE_URL,
                        headers=headers,
                        json=payload,
                        stream=True,
                        timeout=timeout
                    )
                    
                    if response.status_code != 200:
                        error_msg = f"服务器返回错误 {response.status_code}"
                        try:
                            error_data = response.json()
                            if 'error' in error_data:
                                error_msg += f": {error_data['error']}"
                        except:
                            if response.text:
                                error_msg += f": {response.text}"
                        logger.error(f"{mode}请求失败: {error_msg}")
                        return f"Error: {error_msg}"
                    
                    # Handle streaming response
                    full_response = []
                    try:
                        for line in response.iter_lines(decode_unicode=True):
                            if line:
                                try:
                                    data = json.loads(line)
                                    if 'response' in data:
                                        full_response.append(data['response'])
                                except json.JSONDecodeError as e:
                                    logger.warning(f"{mode}响应解析错误: {str(e)}, line: {line}")
                                    continue
                    except requests.exceptions.ChunkedEncodingError as e:
                        error_msg = f"{mode}流式响应中断: {str(e)}"
                        logger.error(error_msg)
                        if not full_response:
                            return f"Error: {error_msg}"
                        # 如果已经有部分响应，继续处理
                        logger.warning(f"{mode}使用已接收的部分响应")
                    
                    result = "".join(full_response).strip()
                    if not result:
                        error_msg = f"{mode}生成的响应为空"
                        logger.error(error_msg)
                        return f"Error: {error_msg}"
                    
                    # 缓存结果
                    self.cache.set(cache_key, result)
                    return result
                    
                except requests.exceptions.Timeout:
                    error_msg = f"{mode}请求超时(尝试 {attempt + 1}/{Config.RETRY_ATTEMPTS})"
                    logger.error(error_msg)
                    if attempt == Config.RETRY_ATTEMPTS - 1:
                        return f"Error: {error_msg}"
                    continue
                    
                except requests.exceptions.RequestException as e:
                    error_msg = f"{mode}请求异常: {str(e)}"
                    logger.error(error_msg)
                    return f"Error: {error_msg}"
                    
                except Exception as e:
                    error_msg = f"{mode}处理过程出错: {str(e)}"
                    logger.error(error_msg)
                    return f"Error: {error_msg}"
                    
        finally:
            # 清理处理状态
            self.process_state.clear_state(model_name)

    def clear_conversation(self) -> str:
        self.conversation.clear()
        return "Conversation cleared."

    def __del__(self):
        """清理资源"""
        try:
            self.session.close()
            if hasattr(self, 'resource_manager'):
                self.resource_manager.cleanup_all()
        except Exception as e:
            logger.error(f"Cleanup failed: {e}")

def check_ollama_service(max_attempts: int = 3, wait_time: int = 2) -> bool:
    """检查Ollama服务是否可用，支持多次重试"""
    for attempt in range(max_attempts):
        try:
            response = requests.get(Config.API_BASE_URL, timeout=5)
            if response.status_code == 200:
                return True
            logger.warning(f"Ollama service check failed (attempt {attempt + 1}/{max_attempts}): Status code {response.status_code}")
        except requests.exceptions.ConnectionError:
            logger.warning(f"Ollama service connection failed (attempt {attempt + 1}/{max_attempts})")
        except requests.exceptions.Timeout:
            logger.warning(f"Ollama service timeout (attempt {attempt + 1}/{max_attempts})")
        except Exception as e:
            logger.error(f"Unexpected error checking Ollama service: {str(e)}")
            
        if attempt < max_attempts - 1:
            time.sleep(wait_time * (attempt + 1))  # 指数退避
            
    return False

def create_interface() -> None:
    try:
        chatbot = ChatBot()
        
        # 预先获取模型列表
        models, status = chatbot.get_model_list()
        if not models:
            logger.error(f"无法获取模型列表: {status}")
            raise ValueError(f"无法获取模型列表: {status}")
            
        # 分离视觉模型和普通模型
        vision_models = [m for m in models if "🖼️" in m]
        all_models = models
        
        if not vision_models:
            logger.error("未找到支持视觉功能的模型")
            raise ValueError("未找到支持视觉功能的模型，请确保安装了支持视觉功能的模型（如llava）")
        
        interface = gr.Blocks(
            title="智能文档分析系统",
            theme=gr.themes.Default(),
            css=".output-text {font-size: 16px;}"
        )

        with interface:
            # 标题
            with gr.Column(elem_classes="header"):
                gr.Markdown("# 智能文档分析系统")
            
            # OCR区域（可选）
            with gr.Column():
                gr.Markdown("### 图片识别（可选）")
                with gr.Row():
                    image_input = gr.Image(
                        type="filepath",
                        label="上传图片（可选）",
                        interactive=True
                    )
                    ocr_model_selector = gr.Dropdown(
                        choices=vision_models,
                        label="选择OCR模型（可选）",
                        value=None,
                        interactive=True
                    )
                ocr_output = gr.Textbox(
                    label="识别结果",
                    lines=4,
                    interactive=False,
                    elem_classes="output-text"
                )
            
            # 问答区域
            with gr.Column():
                gr.Markdown("### 问答")
                question_input = gr.Textbox(
                    label="输入问题",
                    lines=2,
                    interactive=True
                )
                with gr.Row():
                    chat_model_selector1 = gr.Dropdown(
                        choices=all_models,
                        label="模型1",
                        value=all_models[0] if all_models else None,
                        interactive=True
                    )
                    chat_model_selector2 = gr.Dropdown(
                        choices=all_models,
                        label="模型2（可选）",
                        value=None,
                        interactive=True
                    )
                    chat_model_selector3 = gr.Dropdown(
                        choices=all_models,
                        label="模型3（可选）",
                        value=None,
                        interactive=True
                    )
                with gr.Row():
                    chat_output1 = gr.Textbox(
                        label="模型1回答",
                        lines=4,
                        interactive=False,
                        elem_classes="output-text"
                    )
                    chat_output2 = gr.Textbox(
                        label="模型2回答",
                        lines=4,
                        interactive=False,
                        elem_classes="output-text"
                    )
                    chat_output3 = gr.Textbox(
                        label="模型3回答",
                        lines=4,
                        interactive=False,
                        elem_classes="output-text"
                    )
            
            # 状态显示
            status_text = gr.Textbox(
                label="状态",
                interactive=False,
                elem_classes="status-text"
            )
            
            # 处理按钮
            with gr.Row():
                process_button = gr.Button(
                    "开始处理",
                    variant="primary",
                    elem_classes="primary",
                    interactive=True
                )
                retry_button = gr.Button(
                    "重新问答",
                    variant="secondary",
                    elem_classes="secondary",
                    interactive=True
                )

            # 存储OCR结果用于重试
            ocr_result_store = gr.State(value=None)

            # 处理函数：一键处理所有流程
            def process_all(image_path, ocr_model, question, chat_model1, chat_model2, chat_model3, progress=gr.Progress()):
                """一键处理所有流程"""
                # 基本验证
                if not question.strip():
                    return (
                        "",
                        "",
                        "",
                        "",
                        "请输入问题",
                        None
                    )

                if not any([chat_model1, chat_model2, chat_model3]):
                    return (
                        "",
                        "",
                        "",
                        "",
                        "请至少选择一个问答模型",
                        None
                    )

                results = {
                    "ocr_output": "",
                    "chat_output1": "",
                    "chat_output2": "",
                    "chat_output3": "",
                    "status_text": "处理中...",
                    "ocr_result_store": None
                }

                try:
                    chatbot = ChatBot()
                    qa_prompt = question
                    
                    # OCR处理（完全可选）
                    if image_path and ocr_model:
                        try:
                            progress(0.2, desc="正在进行图片识别...")
                            base64_image = chatbot._encode_image(image_path)
                            ocr_prompt = "请识别这张图片中的文字内容，尽可能保持原有格式。"
                            ocr_result = chatbot.generate_response(ocr_prompt, ocr_model, base64_image, is_ocr=True)
                            
                            # 清理OCR结果格式
                            if ocr_result and not ocr_result.startswith("Error:"):
                                # 移除多余的空行,保留段落格式
                                ocr_result = "\n".join(line for line in ocr_result.splitlines() if line.strip())
                                results["ocr_output"] = ocr_result
                                results["ocr_result_store"] = ocr_result
                                # 将OCR结果作为补充信息添加到问题中
                                qa_prompt = f"{question}\n\n补充信息：\n{ocr_result}"
                            else:
                                # OCR失败,记录错误但继续问答
                                logger.error(f"OCR识别失败: {ocr_result}")
                                results["ocr_output"] = ocr_result
                                results["ocr_result_store"] = None
                                qa_prompt = question
                        except Exception as e:
                            error_msg = f"OCR处理出错: {str(e)}"
                            logger.error(error_msg)
                            results["ocr_output"] = f"错误: {error_msg}"
                            results["ocr_result_store"] = None
                            # OCR错误不影响问答继续进行
                            qa_prompt = question
                    
                    # 问答处理
                    progress(0.4, desc="准备进行问答...")
                    total_models = sum(1 for m in [chat_model1, chat_model2, chat_model3] if m)
                    current_model = 0
                    
                    if chat_model1:
                        current_model += 1
                        progress(0.4 + 0.6 * (current_model / total_models), desc=f"获取模型{current_model}回答...")
                        try:
                            results["chat_output1"] = chatbot.generate_response(qa_prompt, chat_model1)
                        except Exception as e:
                            logger.error(f"模型1回答出错: {str(e)}")
                            results["chat_output1"] = f"错误: {str(e)}"
                    
                    if chat_model2:
                        current_model += 1
                        progress(0.4 + 0.6 * (current_model / total_models), desc=f"获取模型{current_model}回答...")
                        try:
                            results["chat_output2"] = chatbot.generate_response(qa_prompt, chat_model2)
                        except Exception as e:
                            logger.error(f"模型2回答出错: {str(e)}")
                            results["chat_output2"] = f"错误: {str(e)}"
                    
                    if chat_model3:
                        current_model += 1
                        progress(0.4 + 0.6 * (current_model / total_models), desc=f"获取模型{current_model}回答...")
                        try:
                            results["chat_output3"] = chatbot.generate_response(qa_prompt, chat_model3)
                        except Exception as e:
                            logger.error(f"模型3回答出错: {str(e)}")
                            results["chat_output3"] = f"错误: {str(e)}"
                    
                    progress(1.0, desc="处理完成")
                    results["status_text"] = "处理完成"
                    return (
                        results["ocr_output"],
                        results["chat_output1"],
                        results["chat_output2"],
                        results["chat_output3"],
                        results["status_text"],
                        results["ocr_result_store"]
                    )
                    
                except Exception as e:
                    logger.error(f"处理过程出错: {str(e)}")
                    error_msg = f"错误: {str(e)}"
                    return (
                        "",
                        "",
                        "",
                        "",
                        error_msg,
                        None
                    )

            # 重试问答函数
            def retry_qa(question, chat_model1, chat_model2, chat_model3, ocr_result):
                """重新进行问答"""
                if not question.strip():
                    return ["请输入问题"] * 3

                if not any([chat_model1, chat_model2, chat_model3]):
                    return ["请至少选择一个问答模型"] * 3

                try:
                    chatbot = ChatBot()
                    # 构建提示词：如果有OCR结果就作为补充信息
                    qa_prompt = question
                    if ocr_result and ocr_result.strip():
                        qa_prompt = f"{question}\n\n补充信息：\n{ocr_result}"
                    
                    results = [""] * 3
                    if chat_model1:
                        try:
                            results[0] = chatbot.generate_response(qa_prompt, chat_model1)
                        except Exception as e:
                            logger.error(f"模型1重试出错: {str(e)}")
                            results[0] = f"错误: {str(e)}"
                    
                    if chat_model2:
                        try:
                            results[1] = chatbot.generate_response(qa_prompt, chat_model2)
                        except Exception as e:
                            logger.error(f"模型2重试出错: {str(e)}")
                            results[1] = f"错误: {str(e)}"
                    
                    if chat_model3:
                        try:
                            results[2] = chatbot.generate_response(qa_prompt, chat_model3)
                        except Exception as e:
                            logger.error(f"模型3重试出错: {str(e)}")
                            results[2] = f"错误: {str(e)}"
                    
                    return results
                except Exception as e:
                    logger.error(f"重试过程出错: {str(e)}")
                    return [f"错误: {str(e)}"] * 3

            # 处理按钮点击事件
            process_button.click(
                fn=process_all,
                inputs=[
                    image_input,
                    ocr_model_selector,
                    question_input,
                    chat_model_selector1,
                    chat_model_selector2,
                    chat_model_selector3
                ],
                outputs=[
                    ocr_output,
                    chat_output1,
                    chat_output2,
                    chat_output3,
                    status_text,
                    ocr_result_store
                ],
                api_name="process"
            )

            # 重试按钮点击事件
            retry_button.click(
                fn=retry_qa,
                inputs=[
                    question_input,
                    chat_model_selector1,
                    chat_model_selector2,
                    chat_model_selector3,
                    ocr_result_store
                ],
                outputs=[
                    chat_output1,
                    chat_output2,
                    chat_output3
                ],
                api_name="retry"
            )

            # 图片和模型选择变更事件
            def clear_outputs():
                return [""] * 4  # 清空4个输出框
                
            image_input.change(
                fn=clear_outputs,
                inputs=None,
                outputs=[ocr_output, chat_output1, chat_output2, chat_output3]
            )
            
            ocr_model_selector.change(
                fn=clear_outputs,
                inputs=None,
                outputs=[ocr_output, chat_output1, chat_output2, chat_output3]
            )

        interface.launch(
            server_name="0.0.0.0",
            server_port=7588,
            share=True,
            inbrowser=True
        )
        
    except Exception as e:
        logger.error(f"Failed to create interface: {e}")
        sys.exit(1)

if __name__ == "__main__":
    # 设置日志输出到文件
    log_file = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'chatbot.log')
    file_handler = logging.FileHandler(log_file, encoding='utf-8')
    file_handler.setFormatter(logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s'))
    logger.addHandler(file_handler)
    
    try:
        logger.info("Starting chatbot interface...")
        print("正在启动聊天机器人界面...")
        if not ChatBot().check_ollama_service():
            error_msg = "Error: Ollama service is not running. Please start Ollama first."
            logger.error(error_msg)
            print(error_msg)
            input("按回车键退出...")
            sys.exit(1)
            
        create_interface()  # 直接调用create_interface，它已经包含了launch逻辑
    except Exception as e:
        print(f"\nError: {str(e)}")
        logger.error(f"Application error: {str(e)}", exc_info=True)
        print("\nPress Enter to exit...")
        input()
        sys.exit(1)