Merge remote-tracking branch 'origin/3.0AI添加' into 3.0AI添加
This commit is contained in:
@ -90,6 +90,17 @@ def create_app():
|
||||
except ImportError as e:
|
||||
print(f"❌ 错误: Upload 模块导入失败: {e}")
|
||||
|
||||
# -----------------------------------------------------
|
||||
# 2.4 注册以图搜图模块 (Image Search)
|
||||
# -----------------------------------------------------
|
||||
try:
|
||||
from app.api.v1.common.image_search import image_search_bp
|
||||
app.register_blueprint(image_search_bp, url_prefix='/api/v1/common')
|
||||
app.register_blueprint(image_search_bp, url_prefix='/api/common', name='image_search_legacy')
|
||||
print("✅ Image Search 模块注册成功")
|
||||
except ImportError as e:
|
||||
print(f"❌ 错误: Image Search 模块导入失败: {e}")
|
||||
|
||||
# -----------------------------------------------------
|
||||
# 2.4 注册业务操作模块 (Transactions - 借还/维修/报废)
|
||||
# -----------------------------------------------------
|
||||
|
||||
153
inventory-backend/app/api/v1/common/image_search.py
Normal file
153
inventory-backend/app/api/v1/common/image_search.py
Normal file
@ -0,0 +1,153 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
"""
|
||||
以图搜图 API - CLIP Vision Embedding + pgvector 余弦距离检索
|
||||
"""
|
||||
|
||||
import os
|
||||
import uuid
|
||||
import json
|
||||
from flask import Blueprint, request, jsonify
|
||||
from sqlalchemy import text
|
||||
from app.extensions import db
|
||||
from app.utils.ai_vision import load_clip_model, get_image_embedding
|
||||
|
||||
# 注册蓝图
|
||||
image_search_bp = Blueprint('image_search', __name__)
|
||||
|
||||
|
||||
# ============================================================================
|
||||
# POST /api/v1/common/image-search
|
||||
# 以图搜图：上传图片 → CLIP embedding → pgvector 余弦相似度检索
|
||||
# ============================================================================
|
||||
|
||||
@image_search_bp.route('/image-search', methods=['POST'])
|
||||
def image_search():
|
||||
# ---------------------------------------------------------
|
||||
# 1. 检查文件
|
||||
# ---------------------------------------------------------
|
||||
if 'file' not in request.files:
|
||||
return jsonify({"code": 400, "msg": "未找到图片文件"}), 400
|
||||
|
||||
file = request.files['file']
|
||||
if file.filename == '':
|
||||
return jsonify({"code": 400, "msg": "未选择文件"}), 400
|
||||
|
||||
# ---------------------------------------------------------
|
||||
# 2. 安全保存临时文件
|
||||
# ---------------------------------------------------------
|
||||
ext = file.filename.rsplit('.', 1)[-1].lower()
|
||||
if ext not in {'png', 'jpg', 'jpeg', 'gif', 'bmp', 'webp'}:
|
||||
return jsonify({"code": 400, "msg": "不支持的图片格式"}), 400
|
||||
|
||||
tmp_filename = f"{uuid.uuid4().hex}.{ext}"
|
||||
tmp_dir = os.path.join(os.path.dirname(__file__), '..', '..', '..', 'uploads')
|
||||
os.makedirs(tmp_dir, exist_ok=True)
|
||||
tmp_path = os.path.join(tmp_dir, tmp_filename)
|
||||
|
||||
try:
|
||||
file.save(tmp_path)
|
||||
print(f"💾 [ImageSearch] 临时文件已保存: {tmp_path}")
|
||||
|
||||
# ---------------------------------------------------------
|
||||
# 3. 提取 CLIP embedding
|
||||
# ---------------------------------------------------------
|
||||
load_clip_model()
|
||||
embedding = get_image_embedding(tmp_path)
|
||||
print(f"✅ [ImageSearch] Embedding 提取成功，维度: {len(embedding)}")
|
||||
|
||||
except Exception as e:
|
||||
print(f"❌ [ImageSearch] 图像处理失败: {e}")
|
||||
return jsonify({"code": 500, "msg": f"图像处理失败: {str(e)}"}), 500
|
||||
|
||||
finally:
|
||||
# ---------------------------------------------------------
|
||||
# 4. 无论成功与否，都删除临时文件
|
||||
# ---------------------------------------------------------
|
||||
if os.path.exists(tmp_path):
|
||||
try:
|
||||
os.remove(tmp_path)
|
||||
print(f"🗑️ [ImageSearch] 临时文件已清理: {tmp_path}")
|
||||
except Exception as e:
|
||||
print(f"⚠️ [ImageSearch] 临时文件删除失败: {e}")
|
||||
|
||||
# ---------------------------------------------------------
|
||||
# 5. pgvector 余弦相似度检索（跨表联合检索）
|
||||
# ---------------------------------------------------------
|
||||
try:
|
||||
query_vector_str = '[' + ','.join(str(v) for v in embedding) + ']'
|
||||
|
||||
sql = text("""
|
||||
SELECT id, name, spec_model, image_url,
|
||||
(1 - (vec <=> :query_vector)) AS similarity
|
||||
FROM (
|
||||
SELECT id,
|
||||
COALESCE(name, '') AS name,
|
||||
COALESCE(spec, '') AS spec_model,
|
||||
COALESCE(product_image, '') AS image_url,
|
||||
img_embedding AS vec
|
||||
FROM material_base
|
||||
WHERE img_embedding IS NOT NULL
|
||||
|
||||
UNION ALL
|
||||
|
||||
SELECT id,
|
||||
'采购入库' AS name,
|
||||
'到货照片' AS spec_model,
|
||||
COALESCE(arrival_photo, '') AS image_url,
|
||||
arrival_image_embedding AS vec
|
||||
FROM stock_buy
|
||||
WHERE arrival_image_embedding IS NOT NULL
|
||||
|
||||
UNION ALL
|
||||
|
||||
SELECT id,
|
||||
'采购入库' AS name,
|
||||
'质检报告' AS spec_model,
|
||||
COALESCE(qc_report, '') AS image_url,
|
||||
qc_report_image_embedding AS vec
|
||||
FROM stock_buy
|
||||
WHERE qc_report_image_embedding IS NOT NULL
|
||||
) AS combined
|
||||
ORDER BY vec <=> :query_vector
|
||||
LIMIT 10
|
||||
""")
|
||||
|
||||
result = db.session.execute(sql, {"query_vector": query_vector_str})
|
||||
rows = result.fetchall()
|
||||
|
||||
results = []
|
||||
for row in rows:
|
||||
item_id = row[0]
|
||||
item_name = row[1] or ""
|
||||
spec_model = row[2] or ""
|
||||
raw_image = row[3]
|
||||
|
||||
# 解析图片 URL 列表，取第一张
|
||||
image_url = ""
|
||||
if raw_image:
|
||||
try:
|
||||
image_list = json.loads(raw_image)
|
||||
if image_list and len(image_list) > 0:
|
||||
image_url = image_list[0]
|
||||
except Exception:
|
||||
# 纯字符串直接使用
|
||||
image_url = str(raw_image)
|
||||
|
||||
results.append({
|
||||
"id": item_id,
|
||||
"name": item_name,
|
||||
"spec_model": spec_model,
|
||||
"image_url": image_url,
|
||||
"similarity": round(float(row[4]), 4)
|
||||
})
|
||||
|
||||
print(f"✅ [ImageSearch] 跨表检索完成，命中 {len(results)} 条结果")
|
||||
return jsonify({
|
||||
"code": 200,
|
||||
"msg": "检索成功",
|
||||
"data": results
|
||||
})
|
||||
|
||||
except Exception as e:
|
||||
print(f"❌ [ImageSearch] 数据库检索失败: {e}")
|
||||
return jsonify({"code": 500, "msg": f"检索失败: {str(e)}"}), 500
|
||||
@ -11,6 +11,8 @@ Dify 智能客服权限服务层
|
||||
- 跨模块越权查询：直接阻断，返回角色专属的错误信息给大模型
|
||||
"""
|
||||
|
||||
from typing import Optional
|
||||
|
||||
from flask import g, current_app
|
||||
from flask_jwt_extended import decode_token
|
||||
from app.models.system import SysRolePermission
|
||||
@ -185,7 +187,7 @@ class DifyPermissionService:
|
||||
返回:
|
||||
{
|
||||
'blocked': bool, # 是否被拦截
|
||||
'message': str | None, # AI 应返回给用户的错误信息（如果有）
|
||||
'message': Optional[str], # AI 应返回给用户的错误信息（如果有）
|
||||
}
|
||||
"""
|
||||
if DifyPermissionService.is_super_admin(role):
|
||||
|
||||
@ -20,6 +20,8 @@ import logging
|
||||
from threading import Thread
|
||||
from datetime import datetime
|
||||
|
||||
from typing import Optional
|
||||
|
||||
from openpyxl import Workbook
|
||||
from openpyxl.styles import Font, PatternFill, Alignment, Border, Side
|
||||
|
||||
@ -346,7 +348,7 @@ def get_task_status(task_id: str) -> dict:
|
||||
# 获取导出文件路径（供下载接口调用）
|
||||
# =============================================================================
|
||||
|
||||
def get_export_filepath(task_id: str) -> str | None:
|
||||
def get_export_filepath(task_id: str) -> Optional[str]:
|
||||
"""
|
||||
根据 task_id 返回已生成文件的完整路径。
|
||||
未完成或不存在返回 None。
|
||||
|
||||
132
inventory-backend/app/utils/ai_vision.py
Normal file
132
inventory-backend/app/utils/ai_vision.py
Normal file
@ -0,0 +1,132 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
"""
|
||||
AI Vision 模块 - CLIP Vision Encoder ONNX 推理
|
||||
"""
|
||||
|
||||
import os
|
||||
import numpy as np
|
||||
from PIL import Image
|
||||
import onnxruntime as ort
|
||||
|
||||
# ============================================================================
|
||||
# 全局模型单例（项目启动时加载一次）
|
||||
# ============================================================================
|
||||
|
||||
MODEL_PATH = os.path.join(os.path.dirname(__file__), '..', '..', 'models', 'clip_vision.onnx')
|
||||
|
||||
# 加载选项：CPU 推理，禁用依赖库的启动开销
|
||||
_session_options = ort.SessionOptions()
|
||||
_session_options.graph_optimization_level = ort.GraphOptimizationLevel.ORT_ENABLE_ALL
|
||||
|
||||
ort_session: ort.InferenceSession = None
|
||||
|
||||
|
||||
def load_clip_model():
|
||||
"""启动时调用：全局加载 CLIP Vision 模型"""
|
||||
global ort_session
|
||||
if ort_session is not None:
|
||||
return ort_session
|
||||
|
||||
if not os.path.exists(MODEL_PATH):
|
||||
raise FileNotFoundError(f"CLIP Vision 模型未找到: {MODEL_PATH}")
|
||||
|
||||
ort_session = ort.InferenceSession(MODEL_PATH, sess_options=_session_options, providers=['CPUExecutionProvider'])
|
||||
print(f"✅ [AI Vision] CLIP 模型加载成功: {MODEL_PATH}")
|
||||
return ort_session
|
||||
|
||||
|
||||
# ============================================================================
|
||||
# CLIP 预处理常量
|
||||
# ============================================================================
|
||||
|
||||
# ImageNet 标准归一化（CLIP 官方）
|
||||
IMAGENET_MEAN = [0.485, 0.456, 0.406]
|
||||
IMAGENET_STD = [0.229, 0.224, 0.225]
|
||||
|
||||
# 模型输入尺寸
|
||||
INPUT_SIZE = 224
|
||||
|
||||
|
||||
def _center_crop_and_resize(image: Image.Image) -> Image.Image:
|
||||
"""
|
||||
CLIP 官方预处理：中心裁剪抗干扰
|
||||
- 将图片最短边缩放到 224
|
||||
- 从正中间切取 224x224 区域
|
||||
"""
|
||||
w, h = image.size
|
||||
|
||||
# 计算缩放后的目标尺寸
|
||||
if w < h:
|
||||
new_w = INPUT_SIZE
|
||||
new_h = int(h * INPUT_SIZE / w)
|
||||
else:
|
||||
new_h = INPUT_SIZE
|
||||
new_w = int(w * INPUT_SIZE / h)
|
||||
|
||||
# 缩放
|
||||
image = image.resize((new_w, new_h), Image.BILINEAR)
|
||||
|
||||
# 中心裁剪
|
||||
left = (new_w - INPUT_SIZE) // 2
|
||||
top = (new_h - INPUT_SIZE) // 2
|
||||
right = left + INPUT_SIZE
|
||||
bottom = top + INPUT_SIZE
|
||||
|
||||
return image.crop((left, top, right, bottom))
|
||||
|
||||
|
||||
def _normalize(image_np: np.ndarray) -> np.ndarray:
|
||||
"""
|
||||
对 224x224x3 图像进行 CLIP 标准归一化
|
||||
image_np: shape (H, W, C), dtype uint8, 值域 [0, 255]
|
||||
返回: shape (C, H, W), dtype float32, 值域 [0, 1]
|
||||
"""
|
||||
# HWC -> CHW
|
||||
image_np = image_np.transpose(2, 0, 1).astype(np.float32) / 255.0
|
||||
|
||||
# 归一化
|
||||
for i, (mean, std) in enumerate(zip(IMAGENET_MEAN, IMAGENET_STD)):
|
||||
image_np[i] = (image_np[i] - mean) / std
|
||||
|
||||
return image_np
|
||||
|
||||
|
||||
# ============================================================================
|
||||
# 主函数：提取图像 embedding
|
||||
# ============================================================================
|
||||
|
||||
def get_image_embedding(image_path: str) -> list:
|
||||
"""
|
||||
提取图像的 512 维 CLIP embedding 向量
|
||||
|
||||
参数:
|
||||
image_path: 图像文件路径
|
||||
|
||||
返回:
|
||||
list: 512 维浮点向量
|
||||
"""
|
||||
if ort_session is None:
|
||||
load_clip_model()
|
||||
|
||||
# 1. 图片预处理
|
||||
image = Image.open(image_path).convert('RGB')
|
||||
image = _center_crop_and_resize(image)
|
||||
input_data = _normalize(np.array(image))
|
||||
input_data = np.expand_dims(input_data, axis=0) # [1, 3, 224, 224]
|
||||
|
||||
# 2. 构造占位符输入 (关键修复)
|
||||
dummy_ids = np.zeros((1, 77), dtype=np.int64)
|
||||
dummy_mask = np.zeros((1, 77), dtype=np.int64)
|
||||
|
||||
# 3. 传入模型进行推理
|
||||
# 注意: 模型输入名在你的模型里必须叫 'pixel_values', 'input_ids', 'attention_mask'
|
||||
# 如果报错找不到输入名，请打印 ort_session.get_inputs()[0].name 确认
|
||||
outputs = ort_session.run(
|
||||
['image_embeds'],
|
||||
{
|
||||
'input_ids': dummy_ids,
|
||||
'pixel_values': input_data.astype(np.float32),
|
||||
'attention_mask': dummy_mask
|
||||
}
|
||||
)
|
||||
return outputs[0][0].tolist()
|
||||
@ -10,6 +10,10 @@ flask-cors==4.0.0
|
||||
redis==5.0.1
|
||||
# 图片处理核心库
|
||||
Pillow>=10.0.0
|
||||
# ONNX 模型本地 CPU 推理
|
||||
onnxruntime>=1.16.0
|
||||
# 数值计算（ONNX 推理依赖）
|
||||
numpy>=1.24.0
|
||||
# [旧] 条形码生成库 (建议保留，防止旧代码报错)
|
||||
python-barcode>=0.14.0
|
||||
# [新增] 二维码生成库 (标签打印必需，包含PIL支持)
|
||||
|
||||
220
inventory-backend/scripts/init_all_vectors.py
Normal file
220
inventory-backend/scripts/init_all_vectors.py
Normal file
@ -0,0 +1,220 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
from __future__ import annotations
|
||||
|
||||
"""
|
||||
全量历史图片向量初始化脚本
|
||||
|
||||
功能：遍历配置表中所有历史图片字段，批量提取 CLIP 512 维向量并存回数据库。
|
||||
用法：python scripts/init_all_vectors.py
|
||||
"""
|
||||
|
||||
import os
|
||||
import json
|
||||
import sys
|
||||
from datetime import datetime
|
||||
from typing import List, Optional
|
||||
|
||||
# 将项目根目录加入 Python 路径
|
||||
sys.path.insert(0, os.path.join(os.path.dirname(__file__), '..'))
|
||||
|
||||
from tqdm import tqdm
|
||||
from sqlalchemy import text
|
||||
|
||||
# Flask 应用环境
|
||||
from app import create_app
|
||||
from app.extensions import db
|
||||
from app.utils.ai_vision import get_image_embedding, load_clip_model
|
||||
|
||||
# ============================================================================
|
||||
# 业务配置：表 → 图片字段 → 向量字段 映射
|
||||
# ============================================================================
|
||||
TARGET_TABLES = [
|
||||
# 基础物料
|
||||
{"table": "material_base", "img_col": "product_image", "vec_col": "img_embedding"},
|
||||
|
||||
# 采购入库
|
||||
{"table": "stock_buy", "img_col": "arrival_photo", "vec_col": "arrival_image_embedding"},
|
||||
{"table": "stock_buy", "img_col": "qc_report", "vec_col": "qc_report_image_embedding"},
|
||||
]
|
||||
|
||||
# 物理图片根目录（相对于 app 目录的相对路径 ../uploads/）
|
||||
APP_DIR = os.path.join(os.path.dirname(__file__), '..', 'app')
|
||||
UPLOADS_ROOT = os.path.abspath(os.path.join(APP_DIR, '..', 'uploads'))
|
||||
|
||||
|
||||
# ============================================================================
|
||||
# 核心工具函数
|
||||
# ============================================================================
|
||||
|
||||
def parse_img_field(raw_value: str) -> List[str]:
|
||||
"""
|
||||
健壮解析图片字段，支持以下格式：
|
||||
- JSON 数组字符串: ["a.jpg", "b.jpg"]
|
||||
- 纯字符串单图片: "a.jpg"
|
||||
- 带 /api/v1/files/ 前缀: ["/api/v1/files/a.jpg"]
|
||||
返回: 提取出的文件名列表
|
||||
"""
|
||||
if not raw_value or (isinstance(raw_value, str) and not raw_value.strip()):
|
||||
return []
|
||||
|
||||
try:
|
||||
# 先尝试按 JSON 解析（处理 JSON 数组字符串）
|
||||
parsed = json.loads(raw_value)
|
||||
if isinstance(parsed, list):
|
||||
items = parsed
|
||||
else:
|
||||
items = [parsed]
|
||||
except (json.JSONDecodeError, TypeError):
|
||||
# JSON 解析失败，说明是纯字符串，直接按单图片处理
|
||||
items = [raw_value.strip()]
|
||||
|
||||
filenames = []
|
||||
for item in items:
|
||||
if not item or not isinstance(item, str):
|
||||
continue
|
||||
item = item.strip()
|
||||
if not item:
|
||||
continue
|
||||
# 去掉可能的 /api/v1/files/ 前缀
|
||||
filename = os.path.basename(item)
|
||||
filenames.append(filename)
|
||||
|
||||
return filenames
|
||||
|
||||
|
||||
def build_local_path(filename: str) -> str:
|
||||
"""
|
||||
将文件名拼装成本地绝对路径
|
||||
"""
|
||||
return os.path.join(UPLOADS_ROOT, filename)
|
||||
|
||||
|
||||
def extract_first_valid_vector(raw_img_field: str, table_name: str, img_col: str) -> Optional[str]:
|
||||
"""
|
||||
读取图片字段，从第一条有效图片提取向量，返回写入 DB 的 JSON 字符串。
|
||||
如果所有图片均失败，返回 None。
|
||||
"""
|
||||
filenames = parse_img_field(raw_img_field)
|
||||
if not filenames:
|
||||
return None
|
||||
|
||||
for filename in filenames:
|
||||
local_path = build_local_path(filename)
|
||||
|
||||
if not os.path.exists(local_path):
|
||||
print(f"\033[91m[WARN] {table_name}.{img_col} | 文件不存在: {local_path}\033[0m")
|
||||
continue
|
||||
|
||||
try:
|
||||
vec = get_image_embedding(local_path)
|
||||
if vec is not None:
|
||||
return json.dumps(vec)
|
||||
except Exception as e:
|
||||
print(f"\033[91m[WARN] {table_name}.{img_col} | 推理异常 [{filename}]: {type(e).__name__}: {e}\033[0m")
|
||||
continue
|
||||
|
||||
return None
|
||||
|
||||
|
||||
# ============================================================================
|
||||
# 主入口
|
||||
# ============================================================================
|
||||
|
||||
def main():
|
||||
start = datetime.now()
|
||||
total_success = 0
|
||||
total_skip = 0
|
||||
|
||||
print("=" * 60)
|
||||
print("📦 全量历史图片向量初始化")
|
||||
print("=" * 60)
|
||||
print(f"图片目录: {UPLOADS_ROOT}")
|
||||
print(f"待处理表数: {len(TARGET_TABLES)}")
|
||||
print()
|
||||
|
||||
# 1. 初始化 Flask 应用上下文（加载 CLIP 模型）
|
||||
app = create_app()
|
||||
with app.app_context():
|
||||
load_clip_model()
|
||||
print("✅ CLIP 模型加载完成")
|
||||
print()
|
||||
|
||||
# 2. 遍历目标表
|
||||
for config in TARGET_TABLES:
|
||||
table_name = config["table"]
|
||||
img_col = config["img_col"]
|
||||
vec_col = config["vec_col"]
|
||||
|
||||
print(f"正在处理表: {table_name}, 字段: {img_col}")
|
||||
|
||||
# 3. 查询待清洗记录（只选未处理过的）
|
||||
sql = text(f"""
|
||||
SELECT id, {img_col}
|
||||
FROM {table_name}
|
||||
WHERE {img_col} IS NOT NULL
|
||||
AND {img_col} != '[]'
|
||||
AND ({vec_col} IS NULL)
|
||||
""")
|
||||
rows = db.session.execute(sql).fetchall()
|
||||
|
||||
if not rows:
|
||||
print(f"[{table_name}/{img_col}] ⏭ 无待处理记录")
|
||||
continue
|
||||
|
||||
print(f"\n[{table_name}/{img_col}] 📋 待处理: {len(rows)} 条")
|
||||
|
||||
# 4. 逐条处理
|
||||
processed = 0
|
||||
success_count = 0
|
||||
|
||||
for row in tqdm(rows, desc=f"{table_name}/{img_col}", unit="条"):
|
||||
record_id = row[0]
|
||||
raw_img = row[1]
|
||||
|
||||
try:
|
||||
vec_json = extract_first_valid_vector(raw_img, table_name, img_col)
|
||||
if vec_json is None:
|
||||
total_skip += 1
|
||||
continue
|
||||
|
||||
# 更新向量字段
|
||||
update_sql = text(f"""
|
||||
UPDATE {table_name} SET {vec_col} = :vec_str WHERE id = :id
|
||||
""")
|
||||
db.session.execute(update_sql, {"vec_str": vec_json, "id": record_id})
|
||||
success_count += 1
|
||||
|
||||
# 每 50 条提交一次
|
||||
if processed > 0 and processed % 50 == 0:
|
||||
db.session.commit()
|
||||
print(f"\n ✅ 已提交 {processed} 条")
|
||||
|
||||
except Exception as e:
|
||||
print(f"\n\033[91m[WARN] {table_name}/{img_col} | ID={record_id} 处理异常: {type(e).__name__}: {e}\033[0m")
|
||||
# 关键：任何异常都不中断，只 continue 下一条
|
||||
db.session.rollback()
|
||||
continue
|
||||
finally:
|
||||
processed += 1
|
||||
|
||||
# 循环结束后补一次 commit（处理未凑满50条的剩余数据）
|
||||
try:
|
||||
db.session.commit()
|
||||
except Exception:
|
||||
db.session.rollback()
|
||||
|
||||
total_success += success_count
|
||||
print(f"[{table_name}/{img_col}] ✅ 完成，成功 {success_count} 条 / 跳过 {len(rows) - success_count} 条")
|
||||
|
||||
# 5. 汇总报告
|
||||
elapsed = (datetime.now() - start).total_seconds()
|
||||
print()
|
||||
print("=" * 60)
|
||||
print(f"🏁 全部完成！总计耗时 {elapsed:.1f} 秒")
|
||||
print(f" ✅ 成功写入向量: {total_success} 条")
|
||||
print(f" ⏭ 无有效图片（跳过）: {total_skip} 条")
|
||||
print("=" * 60)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
Reference in New Issue
Block a user