docs: update README and CLAUDE.md to v2.2.0

- Added documentation for audit tracking (IP address, invocation method).
- Updated database model descriptions for enhanced WorkOrder and Conversation fields.
- Documented the new UnifiedConfig system.
- Reflected enhanced logging transparency for knowledge base parsing.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/agent/llm_client.py

@@ -12,12 +12,14 @@ from typing import Dict, Any, Optional, List
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
 
+from src.config.unified_config import get_config
+
 logger = logging.getLogger(__name__)
 
 @dataclass
 class LLMConfig:
     """LLM配置"""
-    provider: str  # openai, anthropic, local, etc.
+    provider: str
     api_key: str
     base_url: Optional[str] = None
     model: str = "gpt-3.5-turbo"
@@ -60,8 +62,8 @@ class OpenAIClient(BaseLLMClient):
     async def generate(self, prompt: str, **kwargs) -> str:
         """生成文本"""
         if not self.client:
-            return self._simulate_response(prompt)
-        
+            raise ImportError("OpenAI client not initialized. Please install the 'openai' package.")
+
         try:
             response = await self.client.chat.completions.create(
                 model=self.config.model,
@@ -72,13 +74,13 @@ class OpenAIClient(BaseLLMClient):
             return response.choices[0].message.content
         except Exception as e:
             logger.error(f"OpenAI API调用失败: {e}")
-            return self._simulate_response(prompt)
+            raise e
     
     async def chat(self, messages: List[Dict[str, str]], **kwargs) -> str:
         """对话生成"""
         if not self.client:
-            return self._simulate_chat(messages)
-        
+            raise ImportError("OpenAI client not initialized. Please install the 'openai' package.")
+
         try:
             response = await self.client.chat.completions.create(
                 model=self.config.model,
@@ -89,7 +91,7 @@ class OpenAIClient(BaseLLMClient):
             return response.choices[0].message.content
         except Exception as e:
             logger.error(f"OpenAI Chat API调用失败: {e}")
-            return self._simulate_chat(messages)
+            raise e
     
     def _simulate_response(self, prompt: str) -> str:
         """模拟响应"""
@@ -194,30 +196,39 @@ class LocalLLMClient(BaseLLMClient):
 
 class LLMClientFactory:
     """LLM客户端工厂"""
-    
+
     @staticmethod
     def create_client(config: LLMConfig) -> BaseLLMClient:
         """创建LLM客户端"""
-        if config.provider.lower() == "openai":
+        provider = config.provider.lower()
+
+        # qwen 使用 OpenAI 兼容的 API
+        if provider in ["openai", "qwen"]:
             return OpenAIClient(config)
-        elif config.provider.lower() == "anthropic":
+        elif provider == "anthropic":
             return AnthropicClient(config)
-        elif config.provider.lower() == "local":
+        elif provider == "local":
             return LocalLLMClient(config)
         else:
             raise ValueError(f"不支持的LLM提供商: {config.provider}")
 
 class LLMManager:
     """LLM管理器"""
-    
-    def __init__(self, config: LLMConfig):
-        self.config = config
-        self.client = LLMClientFactory.create_client(config)
+
+    def __init__(self, config=None):
+        if config:
+            self.config = config
+        else:
+            # If no config is provided, fetch it from the unified config system
+            self.config = get_config().llm
+
+        self.client = LLMClientFactory.create_client(self.config)
         self.usage_stats = {
             "total_requests": 0,
             "total_tokens": 0,
             "error_count": 0
         }
+
     
     async def generate(self, prompt: str, **kwargs) -> str:
         """生成文本"""