main.py

from fastapi import FastAPI, HTTPException, Request
from fastapi.staticfiles import StaticFiles
from fastapi.responses import FileResponse
from fastapi.responses import JSONResponse
from pydantic import BaseModel
from playwright.async_api import async_playwright
import os
import time
from urllib.parse import urlparse
from typing import Optional, Dict, List
import logging
import json
import base64
from io import BytesIO
import aiohttp
import traceback
import requests
from openai import OpenAI
from starlette.middleware.base import BaseHTTPMiddleware
import uvicorn
from collections import defaultdict
from PIL import Image
from Crypto.PublicKey import RSA
from Crypto.Cipher import PKCS1_v1_5
from datetime import datetime
from pymongo.mongo_client import MongoClient
from pymongo.server_api import ServerApi


# mongodb uri
URI = os.getenv("URI")
# Create a new client and connect to the server
client = MongoClient(URI, server_api=ServerApi('1'))

app = FastAPI()

# 设置最大连接数
MAX_CONNECTIONS = 100
current_connections = 0

# 设置优化设计接口的访问限制
optimize_design_requests = defaultdict(int)  # 记录每个IP的请求次数
optimize_design_timestamps = defaultdict(float)  # 记录每个IP的首次请求时间
white_list = eval(os.getenv("WHITELIST"))
print(logging.INFO, white_list)

class ConnectionLimitMiddleware(BaseHTTPMiddleware):
    async def dispatch(self, request: Request, call_next):
        global current_connections
        if current_connections >= MAX_CONNECTIONS:
            return JSONResponse(
                status_code=503,
                content={"detail": "已超过最大链接数，请稍后重试"}
            )
        current_connections += 1
        try:
            response = await call_next(request)
            return response
        finally:
            current_connections -= 1

# 添加中间件
app.add_middleware(ConnectionLimitMiddleware)

# 确保缓存目录存在
CACHE_DIR = "cache"
os.makedirs(CACHE_DIR, exist_ok=True)

# 挂载静态文件目录
app.mount("/screenshots", StaticFiles(directory=CACHE_DIR), name="screenshots")
app.mount("/static", StaticFiles(directory="static"), name="static")

# API Keys
SECRET_KEY = os.getenv("SECRET_KEY", "wangyue")
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
OPENAI_API_IMAGE_EDIT_KEY = os.getenv("OPENAI_API_IMAGE_EDIT_KEY")
GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
SEARCH_ENGINE_ID = os.getenv("SEARCH_ENGINE_ID", "27acb0d55ad504716")

class ScreenshotRequest(BaseModel):
    url: str
    width: Optional[int] = 1024
    height: Optional[int] = 768
    format: Optional[str] = "png"
    custom_headers: Optional[Dict[str, str]] = {}

class AnalysisRequest(BaseModel):
    text: str
    image_data: Optional[str] = None
    request_model_id: str = 'gpt-4.1-mini'

class OptimizationRequest(BaseModel):
    text: str
    image_data: str
    suggestions: List[str]
    request_model_id: str = 'gpt-image-1'
    openai_key: str = ''
    user_key: str = ''

class CaseStudyRequest(BaseModel):
    user_input: str
    request_model_id: str = 'gpt-4.1-mini'

class TextOptimizationRequest(BaseModel):
    original_feedback: str
    user_input: str
    request_model_id: str = 'gpt-4.1-mini'

class SearchRequest(BaseModel):
    query: str
    num_results: Optional[int] = 2

@app.post("/capture")
async def capture_screenshot(request: ScreenshotRequest):
    try:
        if not request.url:
            raise HTTPException(status_code=400, detail="需要提供URL参数")

        # 生成唯一的文件名
        domain = urlparse(request.url).netloc.replace(".", "_")
        timestamp = int(time.time() * 1000)
        filename = f"{domain}_{timestamp}.{request.format}"
        filepath = os.path.join(CACHE_DIR, filename)

        print(f"开始为 {request.url} 生成截图...")

        # 默认请求头
        default_headers = {
            "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/96.0.4664.110 Safari/537.36",
            "Accept-Language": "zh-CN,zh;q=0.9,en;q=0.8",
            "Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,image/apng,*/*;q=0.8",
            "Connection": "keep-alive",
            "Cache-Control": "max-age=0",
            "Sec-Fetch-Dest": "document",
            "Sec-Fetch-Mode": "navigate",
            "Sec-Fetch-Site": "none",
            "Sec-Fetch-User": "?1",
            "Upgrade-Insecure-Requests": "1"
        }

        # 合并默认请求头和自定义请求头
        headers = {**default_headers, **(request.custom_headers or {})}

        async with async_playwright() as p:
            browser = await p.chromium.launch()
            page = await browser.new_page()
            
            # 设置视口大小
            await page.set_viewport_size({
                "width": request.width,
                "height": request.height
            })

            # 设置请求头
            await page.set_extra_http_headers(headers)

            # 访问页面
            await page.goto(request.url, wait_until="networkidle")
            
            # 生成截图
            await page.screenshot(path=filepath, type=request.format)
            
            await browser.close()

        print(f"截图完成: {filepath}")

        # 返回截图URL
        screenshot_url = f"/screenshots/{filename}"
        
        return JSONResponse({
            "success": True,
            "imageUrl": screenshot_url,
            "filename": filename
        })

    except Exception as e:
        logging.error(f"截图过程中出错: {str(e)}")
        raise HTTPException(status_code=500, detail=f"截图生成失败: {str(e)}")


@app.get("/health")
async def health_check():
    return {
        "status": "ok",
        "current_connections": current_connections,
        "max_connections": MAX_CONNECTIONS
    }


@app.get("/")
def index() -> FileResponse:
    return FileResponse(path="static/index.html", media_type="text/html")


@app.post("/api/analyze")
async def analyze_feedback(request: AnalysisRequest):
    try:
        context = "这是用户上传的设计图。" if request.image_data else ''
        context += "以下是老板对设计的反馈内容：\n" + request.text
        context += '\n请根据老板的反馈分析情绪值（用emoji表示），并结合给出的设计图稿，给出三个具体的修改建议，最后给出合适的搜索关键词，以获取合适的参考设计案例。每个建议应该包含一个标题和详细描述。首先你需要对老板的情绪进行解读，使用"情绪值："开头并分为五类：1. 非常满意-😊😊😊 2. 比较满意-🙂🙂🙂 3. 一般般-😐😐😐 4. 不太满意-🙁🙁🙁 5. 非常不满意-😠😠😠，然后在下一行用一句话分析老板的情绪，以"情绪分析："开头。随后，请以"修改建议：\n"开头，并以有序列表分三行说明三个具体建议，并以"\n\n"分隔，比如："1. 提高对比度：xxx\n\n 2. ...\n\n 3. ...\n\n"。记得结合图片进行分析和提出修改建议。最后，你需要使用网页搜索来获取合适的参考UI设计案例，来为修改当前案例提供参考。请在新的一行以"搜索内容："开头，给出合适的搜索内容，以获取合适的参考设计案例，注意，你只能搜索UI界面或虚拟形象设计相关的案例。'

        # 调用OpenAI API进行分析
        response = await call_openai_api(
            system_prompt='你是一位专业的设计顾问，擅长分析客户反馈，提取关键信息，并提供专业建议，你需要严格遵循用户给出的输出格式要求。',
            user_content=[
                *([{"type": "input_image", "image_url": request.image_data}] if request.image_data else []),
                {"type": "input_text", "text": context}
            ],
            request_model_id=request.request_model_id
        )
        return JSONResponse(response)
    except Exception as e:
        logging.error(f'Error: {e}, traceback: {traceback.format_exc()}')
        raise HTTPException(status_code=500, detail=f'Error: {e}, traceback: {traceback.format_exc()}')


def decrypt_user_key(encrypted_key: str) -> dict:
    try:
        key = RSA.import_key(SECRET_KEY)
        cipher = PKCS1_v1_5.new(key)

        # Decode the base64 encrypted text
        encrypted_bytes = base64.b64decode(encrypted_key)

        # Decrypt the message
        decrypted_bytes = cipher.decrypt(encrypted_bytes, None)

        # Convert bytes to string
        decrypted_text = decrypted_bytes.decode('utf-8')

        # Try to parse as JSON for pretty printing
        try:
            data = json.loads(decrypted_text)
            return data
        except Exception as e:
            logging.error(f'Error: {e}, traceback: {traceback.format_exc()}')
            return {}
    except Exception as e:
        logging.error(f'Error decrypting user key: {e}')
        return {}

def search_user_key(user_key: str) -> dict:
    try:
        # Create a new client and connect to the server
        client = MongoClient(URI, server_api=ServerApi('1'))
        
        # Get the user_info database and user_keys collection
        user_info_db = client.user_info
        user_keys_collection = user_info_db['user_keys']
        
        # Search for the user key
        search_res = user_keys_collection.aggregate([
            {
                '$search': {
                    'index': "user_key",
                    'text': {
                        'query': user_key,
                        'path': {
                            'wildcard': "*"
                        }
                    }
                }
            },
            {
                '$limit': 3
            },
            {
                '$project': {
                    '_id': 0,
                    'user_key': 1,
                    'user_name': 1,
                    'credit': 1,
                    'expiration': 1
                }
            }
        ])
        
        # Get the first result
        result = next(search_res, None)
        
        if result:
            return result
        else:
            return None
            
    except Exception as e:
        logging.error(f'Error searching user key: {e}')
        return None

@app.post("/api/optimize-design")
async def optimize_design(request: OptimizationRequest):
    try:
        # 检查用户密钥
        user_key = None
        if request.user_key:
            user_info = decrypt_user_key(request.user_key)
            if not user_info:
                raise HTTPException(
                    status_code=400,
                    detail="无效的用户密钥"
                )
            
            credit_data = {
                "user_key": request.user_key,
                "user_name": user_info['user_id'],
                "credit": user_info['credit'],
                "expiration": datetime.fromisoformat(user_info['expiration'])
            }
            
            query_res = search_user_key(request.user_key)
            user_info_db = client.user_info
            user_keys_collection = user_info_db['user_keys']
            if not query_res:
                user_keys_collection.insert_many([credit_data])
            else:
                credit_data = query_res
            
            if credit_data['credit'] < 1:
                raise HTTPException(
                    status_code=503,
                    detail="当前user-key额度已用尽。"
                )
        
        # 查看客户是否提供了openai_key
        if not request.openai_key and not user_key:
            raise HTTPException(
                status_code=503,
                detail="当前用户无生图权限，请点击'想使用自己的OpenAI API Key？'输入您的OpenAI API Key或联系@wangyue161并添加白名单user-key后重试。"
            )
    
        # 提取设计类型
        design_type = f"设计类型：{request.text.split()[0]}\n" if len(request.text.split()) > 1 else ""
        
        # 构建图像生成提示词
        prompt = f"{design_type}基于以下设计反馈优化UI设计: {', '.join(request.suggestions)}"
        
        # 处理图片数据
        image_data = request.image_data
        
        # 调用OpenAI图像编辑API
        response = await call_openai_image_api(
            image_data=image_data,
            prompt=prompt,
            request_model_id=request.request_model_id,
            openai_key=request.openai_key
        )

        if user_key:
            # Update credit count by decrementing it by 1
            user_info_db = client.user_info
            user_keys_collection = user_info_db['user_keys']
            user_keys_collection.update_one(
                {"user_key": request.user_key},
                {"$inc": {"credit": -1}}
            )
        
        return JSONResponse(response)
    except HTTPException as he:
        raise he
    except Exception as e:
        logging.error(f'Error: {e}, traceback: {traceback.format_exc()}')
        raise HTTPException(status_code=500, detail=f'Error: {e}, traceback: {traceback.format_exc()}')

@app.post("/api/optimize-text")
async def optimize_text(request: TextOptimizationRequest):
    try:
        response = await call_openai_api(
            system_prompt="你是一个专业的文案优化助手，擅长将简单直接的反馈转换为礼貌、专业且保持原意的表达方式。",
            user_content=[{
                "type": "input_text",
                "text": f"原始反馈内容：{request.original_feedback}\n\n我想回复：{request.user_input}\n\n请优化我的回复内容，使其更加礼貌、专业，同时保持原始意思，增加一些共情和专业术语。"
            }],
            request_model_id=request.request_model_id
        )
        
        return JSONResponse(response)
    except Exception as e:
        logging.error(f'Error: {e}, traceback: {traceback.format_exc()}')
        raise HTTPException(status_code=500, detail=f'Error: {e}, traceback: {traceback.format_exc()}')


@app.post("/api/analyze-case-study")
async def analyze_case_study(request: CaseStudyRequest):
    try:
        response = await call_openai_api(
            system_prompt="你是一个专业的案例分析助手，你需要根据用户需求进行案例分析。",
            user_content=[{
                "type": "input_text",
                "text": request.user_input
            }],
            request_model_id=request.request_model_id
        )

        return JSONResponse(response)
    except Exception as e:
        logging.error(f'Error: {e}, traceback: {traceback.format_exc()}')
        raise HTTPException(status_code=500, detail=f'Error: {e}, traceback: {traceback.format_exc()}')

@app.post("/api/search")
async def search_design_examples(request: SearchRequest):
    try:
        # 构建搜索查询
        search_query = f"{request.query} UI设计"
        
        # 调用Google Custom Search API
        async with aiohttp.ClientSession() as session:
            async with session.get(
                "https://customsearch.googleapis.com/customsearch/v1",
                params={
                    "key": GOOGLE_API_KEY,
                    "q": search_query,
                    "cx": SEARCH_ENGINE_ID,
                    "num": request.num_results
                }
            ) as response:
                if response.status != 200:
                    raise HTTPException(status_code=response.status, detail="Google Search API调用失败")
                
                search_data = await response.json()
                
                if not search_data.get("items"):
                    return JSONResponse({"items": []})
                
                # 处理搜索结果
                results = []
                for item in search_data["items"]:
                    result = {
                        "title": item["title"].replace("</?b>", ""),
                        "link": item["link"],
                        "snippet": item.get("snippet", ""),
                        "image": None
                    }
                    
                    # 尝试获取图片URL
                    if "pagemap" in item:
                        if "cse_image" in item["pagemap"]:
                            result["image"] = item["pagemap"]["cse_image"][0]["src"]
                        elif "cse_thumbnail" in item["pagemap"]:
                            result["image"] = item["pagemap"]["cse_thumbnail"][0]["src"]
                    
                    # 如果没有图片，使用截图服务
                    if not result["image"]:
                        try:
                            screenshot_response = await capture_screenshot(ScreenshotRequest(
                                url=result["link"],
                                width=1024,
                                height=768,
                                format="png"
                            ))
                            if isinstance(screenshot_response, dict) and "imageUrl" in screenshot_response:
                                result["image"] = screenshot_response["imageUrl"]
                        except Exception as e:
                            print(f"获取截图失败: {str(e)}")
                            # 使用默认图片
                            result["image"] = "https://img.freepik.com/free-vector/gradient-ui-ux-background_23-2149052117.jpg"
                    
                    results.append(result)
                
                return JSONResponse({"items": results})
                
    except Exception as e:
        logging.error(f'Error: {e}, traceback: {traceback.format_exc()}')
        raise HTTPException(status_code=500, detail=f'Error: {e}, traceback: {traceback.format_exc()}')

async def call_openai_api(system_prompt: str, user_content: List[Dict], request_model_id='gpt-4.1-nano'):
    headers = {
        "Authorization": f"Bearer {OPENAI_API_KEY}",
        "Content-Type": "application/json"
    }
    
    data = {
        "model": request_model_id,
        "input": [
            {
                "role": "system",
                "content": [{"type": "input_text", "text": system_prompt}]
            },
            {
                "role": "user",
                "content": user_content
            }
        ]
    }
    
    async with aiohttp.ClientSession() as session:
        async with session.post("https://api.openai.com/v1/responses", headers=headers, json=data) as response:
            if response.status != 200:
                resp = await response.json()
                logging.error(f'response: {resp}')
                raise HTTPException(status_code=response.status, detail=f"OpenAI API调用失败, response: {resp}")
            return await response.json()

async def call_openai_image_api(image_data: str, prompt: str, request_model_id='gpt-image-1', openai_key=''):
    try:
        # 从base64字符串中提取纯base64数据（如果包含前缀）
        if image_data and 'base64,' in image_data:
            image_data = image_data.split('base64,')[1]
        
        logging.log(logging.DEBUG, f"Processing image data (first 100 chars): {image_data[:100]}")
        
        # 将base64图片数据转换为文件对象
        image_bytes = base64.b64decode(image_data)
        image_file = BytesIO(image_bytes)
        
        # 如果是dall-e-2模型，需要将图片调整为800x800
        if request_model_id == 'dall-e-2':
            # 打开图片
            img = Image.open(image_file)
            # 调整图片大小为800x800，使用LANCZOS重采样方法以获得更好的质量
            img = img.resize((800, 800), Image.Resampling.LANCZOS)
            # 创建新的BytesIO对象
            image_file = BytesIO()
            # 保存调整后的图片
            img.save(image_file, format='PNG')
            # 将文件指针移到开始位置
            image_file.seek(0)
        
        image_file.name = "original-design.png"  # 设置文件名，与JS代码一致
        
        # 创建OpenAI客户端
        client = OpenAI(api_key=openai_key if openai_key else OPENAI_API_IMAGE_EDIT_KEY)
        
        # 调用图像编辑API
        response = client.images.edit(
            model=request_model_id,
            image=image_file,
            prompt=prompt
        )
        
        # 获取生成的图片数据
        if not response.data or len(response.data) == 0:
            raise ValueError("No image data returned from API")
            
        image_result = response.data[0]
        
        # 返回与JS代码一致的格式
        return {
            "data": [{
                "url": f"data:image/png;base64,{image_result.b64_json}",
                "b64_json": image_result.b64_json
            }]
        }
        
    except Exception as e:
        logging.error(f'Error in call_openai_image_api: {e}, traceback: {traceback.format_exc()}')
        raise HTTPException(
            status_code=500, 
            detail=f'Error processing image: {str(e)}'
        )

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(
        app, 
        host="0.0.0.0", 
        port=7860,
        limit_concurrency=MAX_CONNECTIONS
    )