重构了main.html的主函数

重构了大量的页面的sidebar
优化了代码结构

src/api/logs.py

@@ -4,9 +4,7 @@
 import os
 import sys
 import logging
-import yaml
 from datetime import datetime
-from logging.handlers import TimedRotatingFileHandler, RotatingFileHandler
 from flask import Blueprint, request, jsonify
 
 # 获取项目根目录
@@ -14,74 +12,40 @@ PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(_
 sys.path.insert(0, PROJECT_ROOT)
 
 # 加载配置
+import yaml
 CONFIG_PATH = os.path.join(PROJECT_ROOT, 'config.yaml')
 with open(CONFIG_PATH, 'r', encoding='utf-8') as f:
     CONFIG = yaml.safe_load(f)
 
-# 日志目录
-LOG_BASE_DIR = os.path.join(PROJECT_ROOT, 'logs')
+# 日志目录 - 使用与 main.py 相同的配置
+def get_log_base_dir():
+    """获取日志基础目录"""
+    # 1. 检查环境变量
+    if 'LOG_BASE_DIR' in os.environ:
+        return os.environ['LOG_BASE_DIR']
+    
+    # 2. 检查是否在容器环境中
+    mount_base = os.environ.get('MOUNT_BASE', '/app/base')
+    if os.path.exists(mount_base):
+        return os.path.join(mount_base, 'logs')
+    
+    # 3. 使用本地项目路径
+    return os.path.join(PROJECT_ROOT, 'logs')
+
+LOG_BASE_DIR = get_log_base_dir()
 
 # 创建蓝图
 logs_bp = Blueprint('logs', __name__, url_prefix='/api')
 
 
-def setup_logs_logger():
-    """配置日志系统，按日期分目录存储"""
-    today = datetime.now().strftime('%Y-%m-%d')
-    log_dir = os.path.join(LOG_BASE_DIR, today)
-    os.makedirs(log_dir, exist_ok=True)
-
-    logger = logging.getLogger('logs_api')
-    logger.setLevel(logging.DEBUG)
-    logger.handlers.clear()
-
-    # 全部日志处理器
-    all_log_path = os.path.join(log_dir, 'all.log')
-    all_handler = TimedRotatingFileHandler(all_log_path, when='midnight', interval=1, backupCount=30, encoding='utf-8')
-    all_handler.setLevel(logging.DEBUG)
-    all_handler.setFormatter(logging.Formatter(
-        '[%(asctime)s] %(levelname)s in %(module)s: %(message)s',
-        datefmt='%Y-%m-%d %H:%M:%S'
-    ))
-
-    # 控制台处理器
-    console_handler = logging.StreamHandler()
-    console_handler.setLevel(logging.INFO)
-    console_handler.setFormatter(logging.Formatter(
-        '[%(asctime)s] %(levelname)s: %(message)s',
-        datefmt='%H:%M:%S'
-    ))
-
-    logger.addHandler(all_handler)
-    logger.addHandler(console_handler)
-
-    return logger
+def get_logs_logger():
+    """从 main.py 获取日志记录器"""
+    return logging.getLogger('logs_api')
 
 
-logs_logger = setup_logs_logger()
-
-
-@logs_bp.route('/web-log', methods=['POST'])
-def receive_web_log():
-    """接收前端页面发送的日志"""
-    data = request.json
-    level = data.get('level', 'info')
-    message = data.get('message', '')
-    page = data.get('page', 'unknown')
-    timestamp = data.get('timestamp', datetime.now().isoformat())
-
-    log_message = f'[WEB-{page}] {message}'
-
-    if level == 'error':
-        logs_logger.error(log_message)
-    elif level == 'warning':
-        logs_logger.warning(log_message)
-    elif level == 'debug':
-        logs_logger.debug(log_message)
-    else:
-        logs_logger.info(log_message)
-
-    return jsonify({'code': 0, 'message': '日志接收成功'})
+def get_request_logger():
+    """获取请求日志记录器"""
+    return logging.getLogger('request')
 
 
 def format_file_size(size_bytes):
@@ -94,6 +58,30 @@ def format_file_size(size_bytes):
         return f'{size_bytes / (1024 * 1024):.1f} MB'
 
 
+@logs_bp.route('/web-log', methods=['POST'])
+def receive_web_log():
+    """接收前端页面发送的日志"""
+    data = request.json
+    level = data.get('level', 'info')
+    message = data.get('message', '')
+    page = data.get('page', 'unknown')
+    timestamp = data.get('timestamp', '')
+
+    log_message = f'[WEB-{page}] {message}'
+
+    logger = get_logs_logger()
+    if level == 'error':
+        logger.error(log_message)
+    elif level == 'warning':
+        logger.warning(log_message)
+    elif level == 'debug':
+        logger.debug(log_message)
+    else:
+        logger.info(log_message)
+
+    return jsonify({'code': 0, 'message': '日志接收成功'})
+
+
 @logs_bp.route('/log-files', methods=['GET'])
 def get_log_files():
     """获取指定日期的日志文件列表"""
@@ -113,7 +101,8 @@ def get_log_files():
         return jsonify({'code': 0, 'data': []})
 
     log_files = []
-    file_names = ['all.log', 'error.log', 'request.log']
+    # 定义日志文件的优先级顺序
+    file_names = ['all.log', 'api.log', 'error.log', 'request.log', 'train.log']
 
     for file_name in file_names:
         file_path = os.path.join(log_dir, file_name)
@@ -178,15 +167,13 @@ TRAINING_LOGS_BASE_DIR = '/app/base/logs'
 # 本地开发时的备用路径（Windows）
 LOCAL_TRAINING_LOGS_BASE_DIR = os.path.join(PROJECT_ROOT, 'logs')
 
-# 添加调试日志
-logs_logger.info(f"[DEBUG] TRAINING_LOGS_BASE_DIR: {TRAINING_LOGS_BASE_DIR}")
-logs_logger.info(f"[DEBUG] LOCAL_TRAINING_LOGS_BASE_DIR: {LOCAL_TRAINING_LOGS_BASE_DIR}")
-
 
 @logs_bp.route('/training-log-files', methods=['GET'])
 def get_training_log_files():
     """获取训练日志文件列表 - 从 logs/{日期} 目录下的 .log 文件"""
     try:
+        logs_logger = get_logs_logger()
+
         # 确定基础目录
         logs_base_dir = TRAINING_LOGS_BASE_DIR
         if not os.path.exists(logs_base_dir):
@@ -280,7 +267,7 @@ def get_training_log_files():
 
         return jsonify({'code': 0, 'data': log_files})
     except Exception as e:
-        logs_logger.error(f"[DEBUG] 获取训练日志列表失败: {e}")
+        get_logs_logger().error(f"[DEBUG] 获取训练日志列表失败: {e}")
         return jsonify({'code': 1, 'message': f'获取训练日志列表失败: {str(e)}'})
 
 
@@ -291,6 +278,7 @@ def get_training_log_content():
     if not file_name:
         return jsonify({'code': 1, 'message': '缺少文件参数'})
 
+    logs_logger = get_logs_logger()
     logs_logger.info(f"[DEBUG] ============ get_training_log_content ============")
     logs_logger.info(f"[DEBUG] file: {file_name}")
 

src/api/model_chat.py

@@ -8,8 +8,12 @@ import json
 import requests
 import concurrent.futures
 import subprocess
+import logging
 from flask import Blueprint, request, jsonify
 
+# 获取模块 logger（继承 main.py 的日志配置）
+logger = logging.getLogger(__name__)
+
 # 获取项目根目录
 PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
 
@@ -218,8 +222,6 @@ def preload_trained_model():
     import sys as sys_module
     import pymysql
     import yaml
-    import logging
-    logger = logging.getLogger(__name__)
 
     data = request.json
     model_name = data.get('model_name')  # 模型名称
@@ -572,3 +574,502 @@ if __name__ == "__main__":
         return jsonify({'code': 1, 'message': '推理超时，请稍后重试'})
     except Exception as e:
         return jsonify({'code': 1, 'message': f'推理异常: {str(e)}'})
+
+
+# ==================== Transformers 本地模型接口 ====================
+
+@model_chat_bp.route('/local/preload', methods=['POST'])
+def preload_local_model():
+    """预加载本地模型（使用 transformers）"""
+    import yaml
+    import subprocess
+    import sys as sys_module
+
+    data = request.json
+    model_path = data.get('model_path')  # 模型路径
+    model_name = data.get('model_name', '本地模型')  # 模型名称（用于显示）
+
+    if not model_path:
+        return jsonify({'code': 1, 'message': '缺少模型路径'})
+
+    logger.info(f"[PRELOAD_LOCAL] 开始预加载本地模型: {model_name}, 路径: {model_path}")
+
+    # 先生成唯一ID，避免并发冲突（必须在f-string之前定义）
+    import uuid as uuid_module
+    temp_id = uuid_module.uuid4().hex[:8]
+
+    # 获取项目根目录
+    PROJECT_ROOT = os.path.dirname(os.dirname(os.path.dirname(os.path.abspath(__file__))))
+    CONFIG_PATH = os.path.join(PROJECT_ROOT, 'config.yaml')
+
+    try:
+        with open(CONFIG_PATH, 'r', encoding='utf-8') as f:
+            CONFIG = yaml.safe_load(f)
+    except Exception as e:
+        return jsonify({'code': 1, 'message': f'读取配置失败: {str(e)}'})
+
+    # 构建 transformers 预加载脚本（不使用f-string避免import语句导致的模块shadow问题）
+    preload_script = '''# -*- coding: utf-8 -*-
+import sys
+import uuid
+import logging
+import os
+logging.basicConfig(level=logging.WARNING, format='%(message)s')
+
+# 生成唯一的临时文件路径，避免并发冲突
+temp_id = "TEMP_ID"
+log_file = "/app/base/logs/preload_local_{}.log".format(temp_id)
+
+# 设置环境变量
+os.environ["CUDA_VISIBLE_DEVICES"] = "0"
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
+
+try:
+    from transformers import AutoModelForCausalLM, AutoTokenizer
+    import torch
+
+    model_id = "MODEL_PATH"
+    with open(log_file, "w", encoding="utf-8") as f:
+        f.write("开始加载模型: {}\\n".format(model_id))
+
+    # 加载 tokenizer
+    tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("Tokenizer 加载完成\\n")
+
+    # 加载模型
+    model = AutoModelForCausalLM.from_pretrained(
+        model_id,
+        torch_dtype=torch.float16,
+        device_map="auto",
+        trust_remote_code=True
+    )
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("模型加载完成\\n")
+
+    # 测试推理
+    test_input = "你好"
+    inputs = tokenizer(test_input, return_tensors="pt").to(model.device)
+    outputs = model.generate(**inputs, max_new_tokens=10, do_sample=False)
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("测试推理成功: {}\\n".format(response))
+
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("SUCCESS\\n")
+
+except Exception as e:
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("ERROR: {}\\n".format(str(e)))
+    import traceback
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write(traceback.format_exc())
+    sys.exit(1)
+'''
+
+    # 使用 replace 替换变量（避免 f-string 中包含 import 语句）
+    preload_script = preload_script.replace('TEMP_ID', temp_id).replace('MODEL_PATH', model_path)
+
+    # 写入临时脚本 - 使用唯一文件名避免并发冲突
+    work_dir = '/app/base'
+    script_path = os.path.join(work_dir, f'temp_preload_local_{temp_id}.py')
+    log_path = os.path.join('/app/base/logs', f'preload_local_{temp_id}.log')
+
+    try:
+        # 确保logs目录存在
+        os.makedirs(os.path.dirname(log_path), exist_ok=True)
+
+        with open(script_path, 'w', encoding='utf-8') as f:
+            f.write(preload_script)
+
+        # 查找系统 Python（不使用虚拟环境）
+        def get_system_python():
+            # 尝试常见的系统 Python 路径
+            common_pythons = [
+                '/usr/bin/python3',
+                '/usr/bin/python',
+                '/usr/local/bin/python3',
+                '/usr/local/bin/python',
+            ]
+            for py in common_pythons:
+                if os.path.exists(py) and os.access(py, os.X_OK):
+                    return py
+            # 如果都没找到，使用系统 PATH 中的 python3
+            import shutil
+            py_path = shutil.which('python3')
+            if py_path:
+                return py_path
+            return sys_module.executable  # 兜底使用当前 Python
+
+        python_executable = get_system_python()
+        logger.info(f"[PRELOAD_LOCAL] 使用系统 Python: {python_executable}")
+
+        # 继承环境变量，但清除虚拟环境相关的变量
+        env = {**os.environ}
+        env['CUDA_VISIBLE_DEVICES'] = '0'
+        env['TOKENIZERS_PARALLELISM'] = 'false'
+        # 清除虚拟环境相关变量
+        env.pop('VIRTUAL_ENV', None)
+        env.pop('PYTHONHOME', None)
+
+        logger.info(f"[PRELOAD_LOCAL] 脚本路径: {script_path}")
+        logger.info(f"[PRELOAD_LOCAL] 日志路径: {log_path}")
+        logger.info(f"[PRELOAD_LOCAL] 工作目录: {work_dir}")
+        logger.info(f"[PRELOAD_LOCAL] Python: {python_executable}")
+        logger.info(f"[PRELOAD_LOCAL] 脚本是否存在: {os.path.exists(script_path)}")
+
+        # 执行预加载脚本
+        try:
+            result = subprocess.run(
+                [python_executable, script_path],
+                capture_output=True,
+                text=True,
+                timeout=600,  # 10分钟超时
+                cwd=work_dir,
+                env=env
+            )
+            logger.info(f"[PRELOAD_LOCAL] 返回码: {result.returncode}")
+            logger.info(f"[PRELOAD_LOCAL] stdout: {result.stdout[:500] if result.stdout else 'empty'}")
+            logger.info(f"[PRELOAD_LOCAL] stderr: {result.stderr[:500] if result.stderr else 'empty'}")
+        except Exception as sub_err:
+            logger.error(f"[PRELOAD_LOCAL] subprocess执行异常: {sub_err}")
+            return jsonify({'code': 1, 'message': f'执行异常: {str(sub_err)}'})
+
+        # 读取日志文件获取实际输出
+        try:
+            if os.path.exists(log_path):
+                with open(log_path, 'r', encoding='utf-8') as f:
+                    full_output = f.read()
+                logger.info(f"[PRELOAD_LOCAL] 日志文件内容: {full_output[:500]}")
+            else:
+                logger.warning(f"[PRELOAD_LOCAL] 日志文件不存在: {log_path}")
+                full_output = result.stdout + result.stderr
+        except Exception as read_err:
+            logger.error(f"[PRELOAD_LOCAL] 读取日志失败: {read_err}")
+            full_output = result.stdout + result.stderr
+
+        logger.info(f"[PRELOAD_LOCAL] 脚本输出: {full_output[:500] if len(full_output) > 500 else full_output}")
+
+        # 清理临时文件
+        try:
+            if os.path.exists(script_path):
+                os.remove(script_path)
+        except Exception:
+            pass
+        try:
+            if os.path.exists(log_path):
+                os.remove(log_path)
+        except Exception:
+            pass
+
+        if result.returncode == 0 and 'SUCCESS' in full_output:
+            logger.info(f"[PRELOAD_LOCAL] 模型预加载成功: {model_name}")
+            return jsonify({
+                'code': 0,
+                'message': '模型预加载成功',
+                'data': {'model_name': model_name}
+            })
+        else:
+            error_msg = '预加载失败'
+            for line in full_output.split('\n'):
+                if 'ERROR:' in line:
+                    error_msg = line.split('ERROR:')[1].strip()
+                    break
+            logger.error(f"[PRELOAD_LOCAL] 预加载失败: {error_msg}")
+            return jsonify({'code': 1, 'message': f'预加载失败: {error_msg}'})
+
+    except subprocess.TimeoutExpired:
+        # 清理临时文件
+        try:
+            if os.path.exists(script_path):
+                os.remove(script_path)
+        except Exception:
+            pass
+        try:
+            if os.path.exists(log_path):
+                os.remove(log_path)
+        except Exception:
+            pass
+        logger.error("[PRELOAD_LOCAL] 预加载超时")
+        return jsonify({'code': 1, 'message': '预加载超时，请确保模型路径正确且有足够显存'})
+    except Exception as e:
+        # 清理临时文件
+        try:
+            if os.path.exists(script_path):
+                os.remove(script_path)
+        except Exception:
+            pass
+        try:
+            if os.path.exists(log_path):
+                os.remove(log_path)
+        except Exception:
+            pass
+        logger.error(f"[PRELOAD_LOCAL] 预加载异常: {str(e)}")
+        return jsonify({'code': 1, 'message': f'预加载异常: {str(e)}'})
+
+
+@model_chat_bp.route('/local/chat', methods=['POST'])
+def chat_local_model():
+    """使用本地模型进行对话推理（使用 transformers）"""
+    import yaml
+    import subprocess
+    import sys as sys_module
+    import json
+
+    data = request.json
+    model_path = data.get('model_path')  # 模型路径
+    system_prompt = data.get('system_prompt', '')
+    user_question = data.get('user_question')
+    temperature = data.get('temperature', 0.7)
+    max_tokens = data.get('max_tokens', 2048)
+
+    if not model_path:
+        return jsonify({'code': 1, 'message': '缺少模型路径'})
+    if not user_question:
+        return jsonify({'code': 1, 'message': '缺少用户提问'})
+
+    logger.info(f"[CHAT_LOCAL] 开始对话推理, 模型路径: {model_path}")
+
+    # 构建消息
+    messages = []
+    if system_prompt:
+        messages.append({'role': 'system', 'content': system_prompt})
+    messages.append({'role': 'user', 'content': user_question})
+
+    # 先生成唯一ID，避免并发冲突（必须在f-string之前定义）
+    import uuid as uuid_module
+    temp_id = uuid_module.uuid4().hex[:8]
+
+    # 构建 transformers 推理脚本（不使用f-string避免import语句导致的模块shadow问题）
+    # 使用唯一占位符避免与其他内容冲突
+    import json as json_module
+    messages_json = json_module.dumps(messages, ensure_ascii=False)
+
+    inference_script = '''# -*- coding: utf-8 -*-
+import sys
+import os
+import uuid
+import logging
+import json
+logging.basicConfig(level=logging.WARNING, format='%(message)s')
+
+# 生成唯一的临时文件路径
+temp_id = "__TEMP_ID__"
+log_file = "/app/base/logs/chat_local_{}.log".format(temp_id)
+
+# 设置环境变量
+os.environ["CUDA_VISIBLE_DEVICES"] = "0"
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
+
+try:
+    from transformers import AutoModelForCausalLM, AutoTokenizer
+    import torch
+
+    model_id = "__MODEL_PATH__"
+    with open(log_file, "w", encoding="utf-8") as f:
+        f.write("正在加载模型: {}\\n".format(model_id))
+
+    # 加载 tokenizer
+    tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("Tokenizer 加载完成\\n")
+
+    # 加载模型
+    model = AutoModelForCausalLM.from_pretrained(
+        model_id,
+        torch_dtype=torch.float16,
+        device_map="auto",
+        trust_remote_code=True
+    )
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("模型加载完成\\n")
+
+    # 构建消息格式
+    messages = __MESSAGES_JSON__
+
+    # 应用 chat template
+    if tokenizer.chat_template:
+        text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+    else:
+        # 手动构建
+        text = ""
+        for msg in messages:
+            text += "<|im_start|>{}<|im_end|>\\n".format(msg['role'])
+            text += "{}<|im_end|>\\n".format(msg['content'])
+        text += "<|im_start|>assistant\\n"
+
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("构建输入完成\\n")
+
+    # 编码输入
+    inputs = tokenizer(text, return_tensors="pt").to(model.device)
+
+    # 生成回复
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=__MAX_TOKENS__,
+        temperature=__TEMPERATURE__,
+        do_sample=True,
+        top_p=0.9,
+        pad_token_id=tokenizer.eos_token_id
+    )
+
+    # 解码输出
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+    # 提取 assistant 的回复
+    if "assistant\\n" in response:
+        response = response.split("assistant\\n")[-1]
+    elif "<|im_start|>assistant" in response:
+        response = response.split("<|im_start|>assistant")[-1]
+    response = response.strip()
+
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("{}\\n".format(response))
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("SUCCESS\\n")
+
+except Exception as e:
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write("ERROR: {}\\n".format(str(e)))
+    import traceback
+    with open(log_file, "a", encoding="utf-8") as f:
+        f.write(traceback.format_exc())
+    sys.exit(1)
+'''
+
+    # 使用 replace 替换变量（避免 f-string 中包含 import 语句）
+    inference_script = inference_script.replace('__TEMP_ID__', temp_id)
+    inference_script = inference_script.replace('__MODEL_PATH__', model_path)
+    inference_script = inference_script.replace('__MESSAGES_JSON__', messages_json)
+    inference_script = inference_script.replace('__MAX_TOKENS__', str(max_tokens))
+    inference_script = inference_script.replace('__TEMPERATURE__', str(temperature))
+
+    # 写入临时脚本 - 使用唯一文件名避免并发冲突
+    work_dir = '/app/base'
+    script_path = os.path.join(work_dir, f'temp_chat_local_{temp_id}.py')
+    log_path = os.path.join('/app/base/logs', f'chat_local_{temp_id}.log')
+
+    try:
+        # 确保logs目录存在
+        os.makedirs(os.path.dirname(log_path), exist_ok=True)
+
+        with open(script_path, 'w', encoding='utf-8') as f:
+            f.write(inference_script)
+
+        # 查找系统 Python（不使用虚拟环境）
+        def get_system_python():
+            # 尝试常见的系统 Python 路径
+            common_pythons = [
+                '/usr/bin/python3',
+                '/usr/bin/python',
+                '/usr/local/bin/python3',
+                '/usr/local/bin/python',
+            ]
+            for py in common_pythons:
+                if os.path.exists(py) and os.access(py, os.X_OK):
+                    return py
+            # 如果都没找到，使用系统 PATH 中的 python3
+            import shutil
+            py_path = shutil.which('python3')
+            if py_path:
+                return py_path
+            return sys_module.executable  # 兜底使用当前 Python
+
+        python_executable = get_system_python()
+        logger.info(f"[CHAT_LOCAL] 使用系统 Python: {python_executable}")
+
+        # 继承环境变量，但清除虚拟环境相关的变量
+        env = {**os.environ}
+        env['CUDA_VISIBLE_DEVICES'] = '0'
+        env['TOKENIZERS_PARALLELISM'] = 'false'
+        # 清除虚拟环境相关变量
+        env.pop('VIRTUAL_ENV', None)
+        env.pop('PYTHONHOME', None)
+
+        # 执行推理脚本
+        result = subprocess.run(
+            [python_executable, script_path],
+            capture_output=True,
+            text=True,
+            timeout=600,  # 10分钟超时
+            cwd=work_dir,
+            env=env
+        )
+
+        # 读取日志文件获取实际输出
+        try:
+            if os.path.exists(log_path):
+                with open(log_path, 'r', encoding='utf-8') as f:
+                    full_output = f.read()
+            else:
+                full_output = result.stdout + result.stderr
+        except Exception as read_err:
+            logger.error(f"[CHAT_LOCAL] 读取日志失败: {read_err}")
+            full_output = result.stdout + result.stderr
+
+        logger.info(f"[CHAT_LOCAL] 脚本输出: {full_output[:500] if len(full_output) > 500 else full_output}")
+
+        # 清理临时文件
+        try:
+            if os.path.exists(script_path):
+                os.remove(script_path)
+        except Exception:
+            pass
+        try:
+            if os.path.exists(log_path):
+                os.remove(log_path)
+        except Exception:
+            pass
+
+        if result.returncode == 0 and 'SUCCESS' in full_output:
+            # 提取实际回复（去掉最后的 SUCCESS）
+            lines = full_output.strip().split('\n')
+            response_lines = [line for line in lines if line.strip() and line.strip() != 'SUCCESS']
+            assistant_content = '\n'.join(response_lines).strip()
+
+            logger.info(f"[CHAT_LOCAL] 对话成功, 回复长度: {len(assistant_content)}")
+            return jsonify({
+                'code': 0,
+                'data': {
+                    'model_path': model_path,
+                    'response': assistant_content
+                }
+            })
+        else:
+            error_msg = '推理失败'
+            for line in full_output.split('\n'):
+                if 'ERROR:' in line:
+                    error_msg = line.split('ERROR:')[1].strip()
+                    break
+            logger.error(f"[CHAT_LOCAL] 推理失败: {error_msg}")
+            return jsonify({'code': 1, 'message': f'推理失败: {error_msg}'})
+
+    except subprocess.TimeoutExpired:
+        # 清理临时文件
+        try:
+            if os.path.exists(script_path):
+                os.remove(script_path)
+        except Exception:
+            pass
+        try:
+            if os.path.exists(log_path):
+                os.remove(log_path)
+        except Exception:
+            pass
+        logger.error("[CHAT_LOCAL] 推理超时")
+        return jsonify({'code': 1, 'message': '推理超时，请减少生成token数量或调整参数'})
+    except Exception as e:
+        # 清理临时文件
+        try:
+            if os.path.exists(script_path):
+                os.remove(script_path)
+        except Exception:
+            pass
+        try:
+            if os.path.exists(log_path):
+                os.remove(log_path)
+        except Exception:
+            pass
+        logger.error(f"[CHAT_LOCAL] 推理异常: {str(e)}")
+        return jsonify({'code': 1, 'message': f'推理异常: {str(e)}'})

src/api/model_manage.py

@@ -4,8 +4,12 @@
 import os
 import pymysql
 import yaml
+import logging
 from flask import Blueprint, request, jsonify
 
+# 获取模块 logger（继承 main.py 的日志配置）
+logger = logging.getLogger(__name__)
+
 # 获取项目根目录 - 优先使用环境变量，否则从文件路径计算
 MOUNT_BASE = os.environ.get('MOUNT_BASE', '/app/base')
 PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
@@ -47,8 +51,6 @@ def generic_get_all(table_name, order_by='create_time DESC'):
 
 def get_model_path_by_name(model_name):
     """根据模型名称查询模型路径（用于获取基座模型路径）"""
-    import logging
-    logger = logging.getLogger(__name__)
     logger.info(f"[DEBUG get_model_path_by_name] 查询模型: {model_name}")
 
     try:
@@ -165,6 +167,23 @@ def get_model_manage_by_id(id):
     return jsonify({'code': 1, 'message': '模型不存在'})
 
 
+@model_manage_bp.route('/name/<model_name>', methods=['GET'])
+def get_model_manage_by_name(model_name):
+    """根据名称获取模型"""
+    logger.info(f"[DEBUG] 按名称查询模型: {model_name}")
+
+    conn = get_db_connection()
+    cursor = conn.cursor()
+    cursor.execute("SELECT * FROM model_manage WHERE name = %s LIMIT 1", (model_name,))
+    model = cursor.fetchone()
+    cursor.close()
+    conn.close()
+
+    if model:
+        return jsonify({'code': 0, 'data': model})
+    return jsonify({'code': 1, 'message': '模型不存在'})
+
+
 @model_manage_bp.route('', methods=['POST'])
 def create_model_manage():
     """创建模型"""
@@ -575,25 +594,57 @@ def merge_model():
 
 @model_manage_bp.route('/trained-models/<model_name>', methods=['DELETE'])
 def delete_trained_model(model_name):
-    """删除已训练模型（从local_trained_models目录）"""
+    """删除已训练模型
+    type=merged: 删除合并模型（local_trained_models目录）
+    type=lora: 删除权重（saves目录下的lora等权重文件）
+    """
     import shutil
     import logging
     logger = logging.getLogger(__name__)
 
+    # 获取删除类型参数
+    delete_type = request.args.get('type', 'merged')  # 默认删除合并模型
+
     try:
-        # 删除 local_trained_models 目录下的模型
-        model_path = os.path.join(PROJECT_ROOT, 'local_trained_models', model_name)
+        if delete_type == 'lora':
+            # 删除权重：删除 saves 目录下的权重
+            saves_path = os.path.join(PROJECT_ROOT, 'saves')
+            train_methods = ['lora', 'full', 'qlora', 'dpo', 'cpt', 'prefix', 'adapter', 'peft']
 
-        if not os.path.exists(model_path):
-            return jsonify({'code': 1, 'message': f'模型不存在: {model_name}'})
+            deleted = False
+            for method in train_methods:
+                weight_path = os.path.join(saves_path, method, model_name)
+                if os.path.exists(weight_path):
+                    shutil.rmtree(weight_path)
+                    logger.info(f"[DELETE] 已删除权重: {weight_path}")
+                    deleted = True
 
-        # 删除目录
-        shutil.rmtree(model_path)
-        logger.info(f"[DELETE] 已删除模型: {model_path}")
+            if not deleted:
+                # 也可能是老结构，直接在 saves 下的 model_name 目录
+                old_path = os.path.join(saves_path, model_name)
+                if os.path.exists(old_path):
+                    shutil.rmtree(old_path)
+                    logger.info(f"[DELETE] 已删除老结构权重: {old_path}")
+                    deleted = True
 
-        return jsonify({'code': 0, 'message': '删除成功'})
+            if deleted:
+                return jsonify({'code': 0, 'message': '权重已删除'})
+            else:
+                return jsonify({'code': 1, 'message': f'权重不存在: {model_name}'})
+        else:
+            # 默认删除合并模型（local_trained_models目录）
+            model_path = os.path.join(PROJECT_ROOT, 'local_trained_models', model_name)
+
+            if not os.path.exists(model_path):
+                return jsonify({'code': 1, 'message': f'合并模型不存在: {model_name}'})
+
+            # 删除目录
+            shutil.rmtree(model_path)
+            logger.info(f"[DELETE] 已删除合并模型: {model_path}")
+
+            return jsonify({'code': 0, 'message': '合并模型已删除'})
     except Exception as e:
-        logger.error(f"[DELETE] 删除模型失败: {str(e)}")
+        logger.error(f"[DELETE] 删除失败: {str(e)}")
         return jsonify({'code': 1, 'message': f'删除失败: {str(e)}'})
 
 

src/main.py

@@ -37,7 +37,25 @@ CONFIG = load_config()
 TRAINING_LOGS_DIR = CONFIG.get('training_logs_path', '/app/base/training_logs')
 
 # ============ 日志系统配置 ============
-LOG_BASE_DIR = os.path.join(PROJECT_ROOT, 'logs')
+# 日志目录逻辑：
+# 1. 优先使用环境变量 LOG_BASE_DIR
+# 2. 如果是容器环境（存在 /app/base），使用 /app/base/logs
+# 3. 否则使用本地项目路径 PROJECT_ROOT/logs
+def get_log_base_dir():
+    """获取日志基础目录"""
+    # 1. 检查环境变量
+    if 'LOG_BASE_DIR' in os.environ:
+        return os.environ['LOG_BASE_DIR']
+    
+    # 2. 检查是否在容器环境中
+    mount_base = os.environ.get('MOUNT_BASE', '/app/base')
+    if os.path.exists(mount_base):
+        return os.path.join(mount_base, 'logs')
+    
+    # 3. 使用本地项目路径
+    return os.path.join(PROJECT_ROOT, 'logs')
+
+LOG_BASE_DIR = get_log_base_dir()
 
 
 def setup_logger(name='app'):
@@ -98,6 +116,15 @@ def setup_logger(name='app'):
         datefmt='%Y-%m-%d %H:%M:%S'
     ))
 
+    # 6. API日志处理器 - 专门记录API相关日志
+    api_log_path = os.path.join(log_dir, 'api.log')
+    api_handler = RotatingFileHandler(api_log_path, maxBytes=10*1024*1024, backupCount=10, encoding='utf-8')
+    api_handler.setLevel(logging.DEBUG)
+    api_handler.setFormatter(logging.Formatter(
+        '[%(asctime)s] %(levelname)s: %(message)s',
+        datefmt='%Y-%m-%d %H:%M:%S'
+    ))
+
     # 添加处理器到 logger
     logger.addHandler(all_handler)
     logger.addHandler(error_handler)
@@ -117,6 +144,13 @@ def setup_logger(name='app'):
     train_logger.addHandler(train_handler)
     train_logger.addHandler(console_handler)
 
+    # 为 logs_api 创建单独的 logger (供 src/api/logs.py 使用)
+    logs_api_logger = logging.getLogger('logs_api')
+    logs_api_logger.setLevel(logging.DEBUG)
+    logs_api_logger.handlers.clear()
+    logs_api_logger.addHandler(api_handler)
+    logs_api_logger.addHandler(console_handler)
+
     return logger
 
 
@@ -588,6 +622,8 @@ def system_info():
 
 
 # ============ 通用 CRUD 操作 ============
+import json
+
 def generic_get_all(table_name, order_by='create_time DESC'):
     """通用查询所有"""
     conn = get_db_connection()
@@ -596,6 +632,19 @@ def generic_get_all(table_name, order_by='create_time DESC'):
     result = cursor.fetchall()
     cursor.close()
     conn.close()
+    # 自动解析 JSON 字段
+    for row in result:
+        for key, value in row.items():
+            if isinstance(value, str) and value.startswith('[') and value.endswith(']'):
+                try:
+                    row[key] = json.loads(value)
+                except:
+                    pass
+            elif isinstance(value, str) and value.startswith('{') and value.endswith('}'):
+                try:
+                    row[key] = json.loads(value)
+                except:
+                    pass
     return result
 
 
@@ -607,6 +656,19 @@ def generic_get_by_id(table_name, id_val):
     result = cursor.fetchone()
     cursor.close()
     conn.close()
+    # 自动解析 JSON 字段
+    if result:
+        for key, value in result.items():
+            if isinstance(value, str) and value.startswith('[') and value.endswith(']'):
+                try:
+                    result[key] = json.loads(value)
+                except:
+                    pass
+            elif isinstance(value, str) and value.startswith('{') and value.endswith('}'):
+                try:
+                    result[key] = json.loads(value)
+                except:
+                    pass
     return result
 
 
@@ -1053,36 +1115,19 @@ def delete_model_deploy(id):
     return jsonify({'code': 0, 'message': '删除成功'})
 
 
-# ============ 模型管理接口 ============
-@app.route('/api/model-manage', methods=['GET'])
-def get_model_manage():
-    return jsonify({'code': 0, 'data': generic_get_all('model_manage')})
-
-
-@app.route('/api/model-manage', methods=['POST'])
-def create_model_manage():
-    data = request.json
-    new_id = generic_create('model_manage', data)
-    return jsonify({'code': 0, 'message': '创建成功', 'id': new_id})
-
-
-@app.route('/api/model-manage/<int:id>', methods=['PUT'])
-def update_model_manage(id):
-    data = request.json
-    generic_update('model_manage', id, data)
-    return jsonify({'code': 0, 'message': '更新成功'})
-
-
-@app.route('/api/model-manage/<int:id>', methods=['DELETE'])
-def delete_model_manage(id):
-    generic_delete('model_manage', id)
-    return jsonify({'code': 0, 'message': '删除成功'})
-
-
 # ============ 模型对比接口 ============
 @app.route('/api/model-compare', methods=['GET'])
 def get_model_compare():
-    return jsonify({'code': 0, 'data': generic_get_all('model_compare')})
+    result = generic_get_all('model_compare')
+    # 确保 models 字段被正确解析为 JSON
+    for row in result:
+        if 'models' in row and isinstance(row['models'], str):
+            try:
+                row['models'] = json.loads(row['models'])
+                logger.debug(f"[model-compare] 解析 models 字段成功: {row['models']}")
+            except Exception as e:
+                logger.error(f"[model-compare] 解析 models 字段失败: {e}, 原始值: {row['models']}")
+    return jsonify({'code': 0, 'data': result})
 
 
 @app.route('/api/model-compare/<int:id>', methods=['GET'])
@@ -1090,13 +1135,25 @@ def get_model_compare_by_id(id):
     """获取单个模型对比任务"""
     result = generic_get_by_id('model_compare', id)
     if result:
+        # 确保 models 字段被正确解析为 JSON
+        if 'models' in result and isinstance(result['models'], str):
+            try:
+                result['models'] = json.loads(result['models'])
+            except Exception as e:
+                logger.error(f"[model-compare] 解析 models 字段失败: {e}")
         return jsonify({'code': 0, 'data': result})
     return jsonify({'code': 1, 'message': '任务不存在'})
 
 
 @app.route('/api/model-compare', methods=['POST'])
 def create_model_compare():
-    data = request.json
+    data = request.json.copy()
+    # 字段映射: name -> model_name
+    if 'name' in data:
+        data['model_name'] = data.pop('name')
+    # 设置默认加载状态
+    if 'status' not in data:
+        data['status'] = 'pending'
     new_id = generic_create('model_compare', data)
     return jsonify({'code': 0, 'message': '创建成功', 'id': new_id})
 
@@ -1110,10 +1167,278 @@ def update_model_compare(id):
 
 @app.route('/api/model-compare/<int:id>', methods=['DELETE'])
 def delete_model_compare(id):
+    # 先停止加载的模型服务
+    try:
+        conn = get_db_connection()
+        cursor = conn.cursor()
+        cursor.execute("SELECT models, load_status FROM model_compare WHERE id = %s", (id,))
+        row = cursor.fetchone()
+        cursor.close()
+        conn.close()
+
+        if row and row[1] == 'loaded':
+            # 停止模型服务
+            from subprocess import Popen
+            import signal
+            load_status = json.loads(row[0]) if isinstance(row[0], str) else row[0]
+            if isinstance(load_status, dict) and 'processes' in load_status:
+                for proc in load_status['processes']:
+                    if 'pid' in proc:
+                        try:
+                            os.kill(proc['pid'], signal.SIGTERM)
+                        except:
+                            pass
+    except Exception as e:
+        logger.error(f"[model-compare] 停止模型服务失败: {e}")
+
     generic_delete('model_compare', id)
     return jsonify({'code': 0, 'message': '删除成功'})
 
 
+# ============ 模型加载接口 ============
+import subprocess
+import signal
+
+# 存储加载状态 (生产环境应使用数据库或Redis)
+model_compare_processes = {}
+
+
+@app.route('/api/model-compare/<int:id>/load', methods=['POST'])
+def load_model_compare(id):
+    """加载模型 - 启动模型服务"""
+    try:
+        # 获取对比任务
+        conn = get_db_connection()
+        cursor = conn.cursor()
+        cursor.execute("SELECT models FROM model_compare WHERE id = %s", (id,))
+        row = cursor.fetchone()
+        cursor.close()
+        conn.close()
+
+        if not row:
+            return jsonify({'code': 1, 'message': '任务不存在'})
+
+        models = json.loads(row[0]) if isinstance(row[0], str) else row[0]
+        if not models or len(models) < 2:
+            return jsonify({'code': 1, 'message': '模型配置无效'})
+
+        # 更新状态为 loading
+        generic_update('model_compare', id, {'status': 'loading'})
+
+        # 启动模型服务
+        processes = []
+        loaded_models = []
+
+        for i, model in enumerate(models):
+            model_path = model.get('model_path', '')
+            gpu_id = model.get('gpu_id', 0)
+            port = model.get('port', 7862 + i * 10)
+
+            if not model_path:
+                continue
+
+            # 构建启动命令
+            # 获取模型名称和模板
+            model_name = model.get('model_name', '')
+            template = detect_template(model_path)
+
+            cmd = [
+                sys.executable,
+                '-m', 'llamafactory.api',
+                '--model_name_or_path', model_path,
+                '--template', template or 'default',
+                '--port', str(port),
+                '--gpu_ids', str(gpu_id)
+            ]
+
+            logger.info(f"[model-compare] 启动模型服务: {' '.join(cmd)}")
+
+            # 启动进程
+            proc = subprocess.Popen(
+                cmd,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE,
+                cwd=PROJECT_ROOT
+            )
+
+            processes.append({
+                'pid': proc.pid,
+                'port': port,
+                'model_name': model_name,
+                'model_path': model_path
+            })
+
+            loaded_models.append({
+                'port': port,
+                'model_name': model_name,
+                'status': 'starting'
+            })
+
+        # 保存进程信息
+        model_compare_processes[id] = {
+            'processes': processes,
+            'loaded_models': loaded_models,
+            'start_time': datetime.now().timestamp()
+        }
+
+        # 更新数据库中的加载状态
+        load_status = {
+            'processes': processes,
+            'loaded_models': loaded_models,
+            'started_at': datetime.now().isoformat()
+        }
+        generic_update('model_compare', id, {
+            'status': 'loading',
+            'load_status': json.dumps(load_status, ensure_ascii=False)
+        })
+
+        return jsonify({
+            'code': 0,
+            'message': '正在加载模型...',
+            'data': {
+                'models': loaded_models,
+                'check_url': f'/api/model-compare/{id}/load-status'
+            }
+        })
+
+    except Exception as e:
+        logger.error(f"[model-compare] 加载模型失败: {e}")
+        generic_update('model_compare', id, {'status': 'pending'})
+        return jsonify({'code': 1, 'message': f'加载失败: {str(e)}'})
+
+
+def detect_template(model_path):
+    """根据模型路径检测模板类型"""
+    model_lower = model_path.lower()
+    if 'qwen' in model_lower:
+        return 'qwen'
+    elif 'llama' in model_lower or 'llama' in model_lower:
+        return 'llama'
+    elif 'chatglm' in model_lower:
+        return 'chatglm'
+    elif 'baichuan' in model_lower:
+        return 'baichuan'
+    elif 'mistral' in model_lower:
+        return 'mistral'
+    elif 'yi' in model_lower:
+        return 'yi'
+    elif 'deepseek' in model_lower:
+        return 'deepseek'
+    return 'default'
+
+
+@app.route('/api/model-compare/<int:id>/load-status', methods=['GET'])
+def get_load_status(id):
+    """获取模型加载状态"""
+    try:
+        conn = get_db_connection()
+        cursor = conn.cursor()
+        cursor.execute("SELECT models, load_status FROM model_compare WHERE id = %s", (id,))
+        row = cursor.fetchone()
+        cursor.close()
+        conn.close()
+
+        if not row:
+            return jsonify({'code': 1, 'message': '任务不存在'})
+
+        models = json.loads(row[0]) if isinstance(row[0], str) else row[0]
+        load_status = json.loads(row[1]) if row[1] and isinstance(row[1], str) else {}
+
+        loaded_models = load_status.get('loaded_models', [])
+        processes = load_status.get('processes', [])
+
+        # 检查每个模型服务是否就绪
+        all_ready = True
+        for i, model in enumerate(loaded_models):
+            port = model.get('port')
+            try:
+                import requests
+                resp = requests.get(f'http://localhost:{port}/health', timeout=2)
+                if resp.status_code == 200:
+                    model['status'] = 'ready'
+                else:
+                    model['status'] = 'starting'
+                    all_ready = False
+            except:
+                # 检查进程是否还在运行
+                if 'pid' in processes[i] if i < len(processes) else False:
+                    model['status'] = 'starting'
+                    all_ready = False
+                else:
+                    model['status'] = 'failed'
+                    all_ready = False
+
+        # 更新状态
+        if all_ready:
+            generic_update('model_compare', id, {'status': 'loaded'})
+        else:
+            generic_update('model_compare', id, {'status': 'loading'})
+
+        # 更新加载状态
+        load_status['loaded_models'] = loaded_models
+        load_status['all_ready'] = all_ready
+        generic_update('model_compare', id, {
+            'load_status': json.dumps(load_status, ensure_ascii=False)
+        })
+
+        return jsonify({
+            'code': 0,
+            'data': {
+                'status': 'loaded' if all_ready else 'loading',
+                'models': loaded_models,
+                'all_ready': all_ready
+            }
+        })
+
+    except Exception as e:
+        logger.error(f"[model-compare] 获取加载状态失败: {e}")
+        return jsonify({'code': 1, 'message': str(e)})
+
+
+@app.route('/api/model-compare/<int:id>/unload', methods=['POST'])
+def unload_model_compare(id):
+    """停止加载的模型服务"""
+    try:
+        conn = get_db_connection()
+        cursor = conn.cursor()
+        cursor.execute("SELECT load_status FROM model_compare WHERE id = %s", (id,))
+        row = cursor.fetchone()
+        cursor.close()
+        conn.close()
+
+        if row and row[0]:
+            load_status = json.loads(row[0]) if isinstance(row[0], str) else row[0]
+            processes = load_status.get('processes', [])
+
+            # 停止所有进程
+            for proc in processes:
+                pid = proc.get('pid')
+                if pid:
+                    try:
+                        os.kill(pid, signal.SIGTERM)
+                        logger.info(f"[model-compare] 已停止进程 {pid}")
+                    except ProcessLookupError:
+                        pass
+                    except Exception as e:
+                        logger.warning(f"[model-compare] 停止进程 {pid} 失败: {e}")
+
+        # 清理内存
+        if id in model_compare_processes:
+            del model_compare_processes[id]
+
+        # 更新状态
+        generic_update('model_compare', id, {
+            'status': 'pending',
+            'load_status': None
+        })
+
+        return jsonify({'code': 0, 'message': '已停止模型服务'})
+
+    except Exception as e:
+        logger.error(f"[model-compare] 停止模型服务失败: {e}")
+        return jsonify({'code': 1, 'message': str(e)})
+
+
 # ============ 数据生成接口 ============
 @app.route('/api/data-generate', methods=['GET'])
 def get_data_generate():