添加模型下载功能并更新依赖说明，优化云端部署配置

Author: Viper373

.gitignore

@@ -3,4 +3,5 @@ venv/
 logs/
 .env.local
 .env
-checkpoints
+checkpoints
+.vercel

README.md

@@ -115,7 +115,12 @@ base_model:
    cd LOL-DeepWinPredictor
    ```
 2. **安装依赖（建议虚拟环境）**
-   - 先取消requirements.txt中的`torch`相关注释（该注释是为了避免云端部署依赖安装的错误）
+   - 先在requirements.txt中加上如下依赖（该操作是为了避免云端部署依赖安装的错误）
+     ```txt
+     torch==2.3.0+cu121
+     torch_geometric==2.5.3
+     ```
+   - 创建虚拟环境
 
    ```bash
    python -m venv venv

api/app.py

@@ -1,5 +1,6 @@
 import json
 import os
+import requests
 
 import numpy as np
 import torch
@@ -34,6 +35,20 @@
     rich_logger.error(f"文件不存在错误: {e}")
     hero_list, hero_win_rate, team_list = [], [], []
 
+# 下载模型文件（如果不存在）
+MODEL_LOCAL_PATH = os.path.join(STATIC_DIR, 'saved_model', 'BILSTM_Att.pt')
+MODEL_REMOTE_URL = 'https://mori.teracloud.jp:443/v2/dav/BILSTM_Att.pt'
+if not os.path.exists(MODEL_LOCAL_PATH):
+    os.makedirs(os.path.dirname(MODEL_LOCAL_PATH), exist_ok=True)
+    rich_logger.info(f"模型文件不存在，正在从云端下载: {MODEL_REMOTE_URL}")
+    r = requests.get(MODEL_REMOTE_URL, stream=True)
+    r.raise_for_status()
+    with open(MODEL_LOCAL_PATH, 'wb') as f:
+        for chunk in r.iter_content(chunk_size=8192):
+            if chunk:
+                f.write(chunk)
+    rich_logger.info("模型下载完成！")
+
 # 加载LSTM模型（添加异常处理）
 try:
     model = BiLSTMModelWithAttention(input_size=32, hidden_size=1024, num_layers=2, output_size=1)
@@ -42,7 +57,7 @@
     :param MODEL_PATH: 模型文件路径
     :return: None
     """
-    model.load_state_dict(torch.load(env.MODEL_PATH, map_location=torch.device('cpu'), weights_only=True))  # 确保在CPU上加载，且只加载权重
+    model.load_state_dict(torch.load(MODEL_LOCAL_PATH, map_location=torch.device('cpu'), weights_only=True))  # 确保在CPU上加载，且只加载权重
     model.eval()
 except Exception as e:
     rich_logger.error(f"模型加载错误: {e}")

requirements.txt

@@ -1,7 +1,5 @@
 aiohttp>=3.11.16
 beautifulsoup4>=4.12.2
-boto3==1.37.37
-botocore==1.37.37
 fake_useragent==1.5.1
 Flask==3.0.3
 flask_cors==5.0.1
@@ -11,13 +9,8 @@ orjson==3.10.16
 pandas==2.2.3
 pymongo==4.6.1
 pymysql==1.1.1
-pyspark==3.5.1
 python-dotenv>=1.0.1
 Requests>=2.32.3
 rich>=14.0.0
-rocketmq==0.4.4
 scikit_learn==1.4.2
-# torch==2.3.0+cu121
-# torch_geometric==2.5.3
-tqdm==4.67.1
 urllib3>=2.3.0

tool_utils/file_utils.py

@@ -5,197 +5,13 @@
 # @Author    :Zhangjinzhao
 # @Software  :PyCharm
 
-import mimetypes
-import os
-import subprocess
 from pathlib import Path
 
-import boto3
-from botocore.exceptions import ClientError
-
 from tool_utils.log_utils import RichLogger
 
 rich_logger = RichLogger()
 
 
-class S3Utils:
-    def __init__(self):
-        s3endpoint = os.getenv('S3_ENDPOINT')  # 请填入控制台 “Bucket 设置” 页面底部的 “Endpoint” 标签中的信息
-        s3region = os.getenv('S3_REGION')
-        s3accessKeyId = os.getenv('S3_ACCESS_KEY')  # 请到控制台创建子账户，并为子账户创建相应 accessKey
-        s3SecretKeyId = os.getenv('S3_SECRET_KEY')  # ！！切记，创建子账户时，需要手动为其分配具体权限！！
-        self.bucket = os.getenv('S3_BUCKET')  # 请填入控制台 “Bucket 列表” 页面的 “Bucket 名称”
-        self.s3_client = boto3.client(
-            's3',
-            aws_access_key_id=s3accessKeyId,
-            aws_secret_access_key=s3SecretKeyId,
-            endpoint_url=s3endpoint,
-            region_name=s3region
-        )
-
-    @rich_logger
-    def check_s3_file_exists(self, file_path):
-        """
-        检查文件在 S3 上是否已存在
-        :param file_path: 本地文件路径
-        :return: 存在返回 True，不存在返回 False
-        """
-        try:
-            # 将路径替换为统一的单斜杠
-            unified_path = file_path.replace("\\", "/")
-            # 分割路径
-            parts = unified_path.split("/")
-            xovideos_indices = [i for i, part in enumerate(parts) if part == "XOVideos"]
-
-            videos_index = xovideos_indices[1]  # 获取第二个 "XOVideos" 的索引
-            s3_key = "/".join(parts[videos_index:])
-
-            # 检查文件是否存在
-            try:
-                self.s3_client.head_object(Bucket=self.bucket, Key=s3_key)
-                rich_logger.info(f"文件已存在于 S3: {s3_key}")
-                return True
-            except ClientError as e:
-                error_code = e.response['Error'].get('Code', 'Unknown')
-                if error_code == '404':
-                    rich_logger.info(f"文件不存在于 S3: {s3_key}")
-                    return False
-                else:
-                    rich_logger.error(f"检查 S3 文件时出错 ({error_code}): {e}")
-                    return False
-
-        except Exception as e:
-            rich_logger.error(f"检查 S3 文件时发生未知错误: {e}")
-            return False
-
-    @rich_logger
-    def s4_upload_file(self, file_path, delete_on_success=True, delete_on_failure=False):
-
-        try:
-            # 路径合规性检查
-            if not os.path.isfile(file_path):
-                rich_logger.error(f"文件不存在或不可读: {file_path}")
-                return False
-
-            # 将路径替换为统一的单斜杠
-            unified_path = file_path.replace("\\", "/")
-            # 分割路径
-            parts = unified_path.split("/")
-            xovideos_indices = [i for i, part in enumerate(parts) if part == "XOVideos"]
-
-            videos_index = xovideos_indices[1]  # 获取第二个 "XOVideos" 的索引
-            s3_key = "/".join(parts[videos_index:])
-
-            # 检查文件是否存在
-            try:
-                self.s3_client.head_object(Bucket=self.bucket, Key=s3_key)
-                rich_logger.info(f"S4文件已存在，跳过上传: {s3_key}")
-
-                if delete_on_success:
-                    try:
-                        os.remove(file_path)
-                    except OSError as e:
-                        rich_logger.error(f"删除本地文件失败: {e}")
-                return True
-            except ClientError as e:
-                error_code = e.response['Error'].get('Code', 'Unknown')
-                if error_code != '404':
-                    raise
-
-            # 动态检测 MIME 类型
-            content_type, _ = mimetypes.guess_type(file_path)
-            extra_args = {
-                "ContentType": content_type or "application/octet-stream",
-                "ContentDisposition": "inline"
-            }
-
-            # 记录文件大小
-            file_size = os.path.getsize(file_path)
-            size_units = ['B', 'KB', 'MB', 'GB', 'TB']
-            size_index = 0
-            while file_size >= 1024 and size_index < len(size_units) - 1:
-                file_size /= 1024.0
-                size_index += 1
-            size_str = f"{int(file_size)} {size_units[size_index]}" if size_index == 0 else f"{file_size:.2f} {size_units[size_index]}".rstrip('0').rstrip('.')
-            rich_logger.info(f"S4开始上传: {s3_key}，大小: {size_str}")
-
-            # 执行上传
-            self.s3_client.upload_file(
-                Filename=file_path,
-                Bucket=self.bucket,
-                Key=s3_key,
-                ExtraArgs=extra_args
-            )
-            rich_logger.info(f"S4上传成功: {s3_key}")
-
-            # 上传成功后删除本地文件
-            if delete_on_success:
-                try:
-                    os.remove(file_path)
-                except OSError as e:
-                    rich_logger.error(f"删除本地文件失败: {e}")
-            return True
-
-        except ClientError as e:
-            error_code = e.response['Error'].get('Code', 'Unknown')
-            rich_logger.error(f"S4上传失败 ({error_code}): {e}")
-            if delete_on_failure:
-                try:
-                    os.remove(file_path)
-                except OSError as err:
-                    rich_logger.error(f"删除本地文件失败: {err}")
-            return False
-        except Exception as e:
-            rich_logger.exception(f"S4未知错误: {e}")
-            if delete_on_failure:
-                try:
-                    os.remove(file_path)
-                except OSError as err:
-                    rich_logger.error(f"删除本地文件失败: {err}")
-            return False
-
-    @staticmethod
-    @rich_logger
-    def ffmpeg_video_streaming(input_file):
-        """
-        使用FFmpeg将视频转换为H.264格式，并进行流优化处理，使其能够流式传输。
-
-        参数:
-        - input_file: 输入的视频文件路径
-        - output_file: 输出的优化后的视频文件路径
-
-        返回:
-        - 成功: 返回输出文件的路径
-        - 失败: 返回 None
-        """
-        output_file = input_file.replace('.mp4', 'h264.mp4')
-        command = [
-            'ffmpeg',
-            '-i', input_file,
-            '-c:v', 'libx264',  # 使用H.264编解码器
-            '-c:a', 'copy',  # 保留原始音频
-            '-movflags', 'faststart',  # 使视频文件头放在文件开始处，优化流媒体播放
-            output_file
-        ]
-
-        try:
-            # 使用 subprocess.run 执行 FFmpeg 命令，并捕获 stdout 和 stderr 以便调试
-            subprocess.run(command, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
-
-            rich_logger.info(f"视频优化和H.264转换完成: {output_file}")
-            os.remove(input_file)  # 删除原视频，保留流式优化视频
-            os.rename(output_file, input_file)
-            return input_file
-        except subprocess.CalledProcessError as e:
-            # 捕获 FFmpeg 命令的返回码和输出
-            rich_logger.exception(f"视频优化失败: {output_file}\n错误信息: {e.stderr.decode('utf-8')}")
-            return None
-        except Exception as e:
-            # 捕获其他意外错误
-            rich_logger.error(f"未知错误: {e}")
-            return None
-
-
 class ProjectRootFinder:
     PROJECT_MARKERS = {
         "Python": ["setup.py", "requirements.txt", "pyproject.toml", "manage.py", "Pipfile", "app.py"],