auto-coder 1.0.0__py3-none-any.whl → 2.0.1__py3-none-any.whl

autocoder/common/conversations/get_conversation_manager.py

@@ -1,6 +1,6 @@
 import os
 import threading
-from typing import Optional
+from typing import Optional, List, Dict
 from .manager import PersistConversationManager
 from .config import ConversationManagerConfig
 
@@ -91,7 +91,8 @@ def get_conversation_manager(config: Optional[ConversationManagerConfig] = None)
         # 使用自定义配置（仅在首次调用时生效）
         config = ConversationManagerConfig(
             storage_path="./my_conversations",
-            max_cache_size=200
+            max_cache_size=200,
+            default_namespace="my_project"
         )
         manager = get_conversation_manager(config)
         
@@ -100,6 +101,10 @@ def get_conversation_manager(config: Optional[ConversationManagerConfig] = None)
             name="测试对话",
             description="这是一个测试对话"
         )
+        
+        # 使用命名空间管理当前对话
+        manager.set_current_conversation(conv_id, "my_project")
+        current_id = manager.get_current_conversation_id("my_project")
         ```
     """
     return ConversationManagerSingleton.get_instance(config)
@@ -137,7 +142,100 @@ def get_conversation_manager_config() -> Optional[ConversationManagerConfig]:
     return ConversationManagerSingleton.get_config()
 
 
+def get_current_conversation_id_with_namespace(namespace: Optional[str] = None) -> Optional[str]:
+    """
+    获取指定命名空间的当前对话ID
+    
+    Args:
+        namespace: 命名空间，None表示使用默认命名空间
+        
+    Returns:
+        当前对话ID，如果未设置返回None
+        
+    Example:
+        ```python
+        # 获取默认命名空间的当前对话ID
+        current_id = get_current_conversation_id_with_namespace()
+        
+        # 获取特定命名空间的当前对话ID
+        project_id = get_current_conversation_id_with_namespace("my_project")
+        ```
+    """
+    manager = get_conversation_manager()
+    return manager.get_current_conversation_id(namespace)
+
+
+def set_current_conversation_with_namespace(conversation_id: str, namespace: Optional[str] = None) -> bool:
+    """
+    设置指定命名空间的当前对话
+    
+    Args:
+        conversation_id: 对话ID
+        namespace: 命名空间，None表示使用默认命名空间
+        
+    Returns:
+        True if setting was successful
+        
+    Example:
+        ```python
+        # 设置默认命名空间的当前对话
+        set_current_conversation_with_namespace("conv-123")
+        
+        # 设置特定命名空间的当前对话
+        set_current_conversation_with_namespace("conv-456", "my_project")
+        ```
+    """
+    manager = get_conversation_manager()
+    return manager.set_current_conversation(conversation_id, namespace)
+
+
+def list_conversation_namespaces() -> List[Optional[str]]:
+    """
+    列出所有已使用的命名空间
+    
+    Returns:
+        命名空间列表，None 表示默认命名空间
+        
+    Example:
+        ```python
+        namespaces = list_conversation_namespaces()
+        for ns in namespaces:
+            if ns is None:
+                print("默认命名空间")
+            else:
+                print(f"命名空间: {ns}")
+        ```
+    """
+    manager = get_conversation_manager()
+    return manager.list_namespaces()
+
+
+def get_all_current_conversations_by_namespace() -> Dict[Optional[str], str]:
+    """
+    获取所有命名空间的当前对话ID
+    
+    Returns:
+        命名空间到对话ID的映射
+        
+    Example:
+        ```python
+        all_current = get_all_current_conversations_by_namespace()
+        for namespace, conv_id in all_current.items():
+            ns_name = "默认" if namespace is None else namespace
+            print(f"命名空间 '{ns_name}' 的当前对话: {conv_id}")
+        ```
+    """
+    manager = get_conversation_manager()
+    return manager.get_all_current_conversations()
+
+
 # 便捷别名
 get_manager = get_conversation_manager
 reset_manager = reset_conversation_manager
 get_manager_config = get_conversation_manager_config
+
+# 命名空间相关的便捷别名
+get_current_conv_id = get_current_conversation_id_with_namespace
+set_current_conv = set_current_conversation_with_namespace
+list_namespaces = list_conversation_namespaces
+get_all_current_convs = get_all_current_conversations_by_namespace

autocoder/common/conversations/llm_stats_models.py

@@ -0,0 +1,264 @@
+"""
+LLM 调用统计信息的 Pydantic 模型定义
+"""
+
+from typing import Optional, Dict, Any, Union
+from pydantic import BaseModel, Field, ConfigDict
+import time
+from loguru import logger
+
+
+class LLMUsageStats(BaseModel):
+    """LLM 使用统计信息模型"""
+    
+    input_tokens: Optional[int] = Field(default=None, description="输入token数量")
+    output_tokens: Optional[int] = Field(default=None, description="输出token数量") 
+    total_tokens: Optional[int] = Field(default=None, description="总token数量")
+    
+    # 缓存相关信息
+    cache_creation_input_tokens: Optional[int] = Field(default=None, description="缓存创建输入token数量")
+    cache_read_input_tokens: Optional[int] = Field(default=None, description="缓存读取输入token数量")
+    
+    # 成本信息
+    input_cost: Optional[float] = Field(default=None, description="输入成本")
+    output_cost: Optional[float] = Field(default=None, description="输出成本")
+    total_cost: Optional[float] = Field(default=None, description="总成本")
+    
+    # 模型信息
+    model_name: Optional[str] = Field(default=None, description="使用的模型名称")
+    
+    # 时间信息
+    request_time: Optional[float] = Field(default_factory=time.time, description="请求时间戳")
+    response_time: Optional[float] = Field(default=None, description="响应时间戳")
+    duration_ms: Optional[float] = Field(default=None, description="请求持续时间(毫秒)")
+    
+    model_config = ConfigDict(
+        extra="allow",  # 允许额外字段，以便支持未来扩展
+        validate_assignment=True  # 赋值时验证
+    )
+
+
+class LLMCacheInfo(BaseModel):
+    """LLM 缓存命中信息模型"""
+    
+    cache_hit: Optional[bool] = Field(default=None, description="是否命中缓存")
+    cache_type: Optional[str] = Field(default=None, description="缓存类型(如: claude_sonnet)")
+    cache_key: Optional[str] = Field(default=None, description="缓存键")
+    cache_creation_input_tokens: Optional[int] = Field(default=None, description="缓存创建使用的输入token数量")
+    cache_read_input_tokens: Optional[int] = Field(default=None, description="从缓存读取的输入token数量")
+    
+    model_config = ConfigDict(
+        extra="allow",
+        validate_assignment=True
+    )
+
+
+class LLMCallMetadata(BaseModel):
+    """LLM 调用的完整元数据模型"""
+    
+    usage_stats: Optional[LLMUsageStats] = Field(default=None, description="使用统计信息")
+    cache_info: Optional[LLMCacheInfo] = Field(default=None, description="缓存信息")
+    
+    # 请求相关信息
+    request_id: Optional[str] = Field(default=None, description="请求ID")
+    conversation_round: Optional[int] = Field(default=None, description="对话轮次")
+    
+    # 错误信息
+    error_message: Optional[str] = Field(default=None, description="错误信息")
+    error_code: Optional[str] = Field(default=None, description="错误代码")
+    
+    # 额外的原始数据
+    raw_response_metadata: Optional[Dict[str, Any]] = Field(default=None, description="原始响应元数据")
+    
+    model_config = ConfigDict(
+        extra="allow",
+        validate_assignment=True
+    )
+    
+    def to_dict(self) -> Dict[str, Any]:
+        """转换为字典格式"""
+        return self.model_dump(exclude_none=True)
+    
+    @classmethod
+    def from_dict(cls, data: Dict[str, Any]) -> "LLMCallMetadata":
+        """从字典创建实例"""
+        return cls(**data)
+    
+    @classmethod
+    def from_raw_metadata(cls, raw_metadata: Dict[str, Any], **extra_fields) -> "LLMCallMetadata":
+        """从原始元数据创建实例"""
+        # 解析常见的元数据字段
+        usage_stats = None
+        cache_info = None
+        
+        # 解析 usage 信息
+        usage_data = None
+        if "usage" in raw_metadata:
+            usage_data = raw_metadata["usage"]
+        elif any(key in raw_metadata for key in ["prompt_tokens", "completion_tokens", "input_tokens", "output_tokens", "total_tokens", 
+                                                "input_tokens_count", "generated_tokens_count", "total_tokens_count"]):
+            # 如果直接包含使用统计字段，则使用 raw_metadata 本身
+            usage_data = raw_metadata
+        
+        if usage_data:
+            usage_stats = cls._extract_usage_stats(usage_data, raw_metadata)
+        
+        # 解析缓存信息
+        cache_info = cls._extract_cache_info(raw_metadata)
+        
+        return cls(
+            usage_stats=usage_stats,
+            cache_info=cache_info,
+            raw_response_metadata=raw_metadata,
+            **extra_fields
+        )
+    
+    @classmethod
+    def _extract_usage_stats(cls, usage_data: Union[Dict[str, Any], object], raw_metadata: Dict[str, Any]) -> Optional[LLMUsageStats]:
+        """从usage数据中提取统计信息，支持dict和对象两种形态"""
+        def safe_get(data, *keys):
+            """安全获取值，支持dict和对象属性访问"""
+            for key in keys:
+                if isinstance(data, dict):
+                    value = data.get(key)
+                else:
+                    value = getattr(data, key, None)
+                # 检查值是否有效（不是Mock对象或其他无效类型）
+                if value is not None and not str(type(value)).startswith("<class 'unittest.mock."):
+                    # 验证基本数据类型
+                    if isinstance(value, (int, float, str, bool)):
+                        return value
+            return None
+        
+        # 提取input tokens (支持多种字段名)
+        input_tokens = safe_get(usage_data, "prompt_tokens", "input_tokens", "input_tokens_count")
+        
+        # 提取output tokens (支持多种字段名)
+        output_tokens = safe_get(usage_data, "completion_tokens", "output_tokens", "generated_tokens_count")
+        
+        # 提取total tokens，如果没有则计算
+        total_tokens = safe_get(usage_data, "total_tokens", "total_tokens_count")
+        if total_tokens is None and input_tokens is not None and output_tokens is not None:
+            total_tokens = input_tokens + output_tokens
+        
+        # 提取模型名称
+        model_name = safe_get(usage_data, "model", "model_name") or safe_get(raw_metadata, "model", "model_name")
+        
+        # 提取缓存相关token数量 (支持多种字段名变体)
+        cache_creation_input_tokens = safe_get(usage_data, "cache_creation_input_tokens", "cache_creation_tokens")
+        cache_read_input_tokens = safe_get(usage_data, "cache_read_input_tokens", "cache_read_tokens")
+        
+        # 提取成本信息
+        input_cost = safe_get(usage_data, "input_cost", "prompt_cost")
+        output_cost = safe_get(usage_data, "output_cost", "completion_cost")
+        total_cost = safe_get(usage_data, "total_cost")
+        if total_cost is None and input_cost is not None and output_cost is not None:
+            total_cost = input_cost + output_cost
+        
+        # 提取时间信息
+        duration_ms = safe_get(usage_data, "duration_ms", "duration", "response_time_ms")
+        
+        # 如果没有任何有效的token信息，返回None
+        if input_tokens is None and output_tokens is None and total_tokens is None:
+            return None
+        
+        return LLMUsageStats(
+            input_tokens=input_tokens,
+            output_tokens=output_tokens,
+            total_tokens=total_tokens,
+            cache_creation_input_tokens=cache_creation_input_tokens,
+            cache_read_input_tokens=cache_read_input_tokens,
+            input_cost=input_cost,
+            output_cost=output_cost,
+            total_cost=total_cost,
+            model_name=model_name,
+            duration_ms=duration_ms
+        )
+    
+    @classmethod
+    def _extract_cache_info(cls, raw_metadata: Dict[str, Any]) -> Optional[LLMCacheInfo]:
+        """从原始元数据中提取缓存信息"""
+        # 检查是否有缓存相关字段
+        cache_keys = [key for key in raw_metadata.keys() if key.startswith("cache")]
+        if not cache_keys:
+            return None
+        
+        def safe_get(data, *keys):
+            """安全获取值"""
+            for key in keys:
+                value = data.get(key) if isinstance(data, dict) else getattr(data, key, None)
+                if value is not None:
+                    return value
+            return None
+        
+        # 提取缓存信息 (支持多种字段名变体)
+        cache_hit = safe_get(raw_metadata, "cache_hit", "is_cache_hit")
+        cache_type = safe_get(raw_metadata, "cache_type", "cache_provider")
+        cache_key = safe_get(raw_metadata, "cache_key", "cache_id")
+        cache_creation_input_tokens = safe_get(raw_metadata, "cache_creation_input_tokens", "cache_creation_tokens")
+        cache_read_input_tokens = safe_get(raw_metadata, "cache_read_input_tokens", "cache_read_tokens")
+        
+        return LLMCacheInfo(
+            cache_hit=cache_hit,
+            cache_type=cache_type,
+            cache_key=cache_key,
+            cache_creation_input_tokens=cache_creation_input_tokens,
+            cache_read_input_tokens=cache_read_input_tokens
+        )
+
+
+def create_llm_metadata_from_token_usage_event(
+    token_usage_event: Any,
+    conversation_round: Optional[int] = None,
+    request_id: Optional[str] = None
+) -> Optional[LLMCallMetadata]:
+    """
+    从 TokenUsageEvent 创建 LLM 元数据
+    
+    Args:
+        token_usage_event: TokenUsageEvent 实例
+        conversation_round: 对话轮次
+        request_id: 请求ID
+        
+    Returns:
+        LLMCallMetadata 实例或 None
+    """
+    try:
+        if not hasattr(token_usage_event, 'usage') or not token_usage_event.usage:
+            return None
+        
+        usage_data = token_usage_event.usage
+        
+        # 如果usage是对象而不是dict，将其转换为dict以便处理
+        if not isinstance(usage_data, dict):
+            # 尝试将对象转换为dict
+            raw_metadata = {}
+            
+            # 获取对象的所有属性
+            for attr_name in dir(usage_data):
+                if not attr_name.startswith('_') and not callable(getattr(usage_data, attr_name, None)):  # 跳过私有属性和方法
+                    try:
+                        attr_value = getattr(usage_data, attr_name)
+                        # 只保存基本数据类型
+                        if isinstance(attr_value, (int, float, str, bool, type(None))):
+                            raw_metadata[attr_name] = attr_value
+                    except Exception:
+                        continue  # 跳过无法访问的属性
+        else:
+            raw_metadata = usage_data
+        
+        # 如果还有其他有用的属性，也加入到metadata中
+        if hasattr(token_usage_event, 'model'):
+            raw_metadata['model'] = token_usage_event.model
+        if hasattr(token_usage_event, 'request_id'):
+            raw_metadata['request_id'] = token_usage_event.request_id
+        
+        return LLMCallMetadata.from_raw_metadata(
+            raw_metadata=raw_metadata,
+            conversation_round=conversation_round,
+            request_id=request_id
+        )
+    except Exception as e:
+        # 记录错误但不抛出异常，确保程序继续运行
+        logger.warning(f"Failed to create LLM metadata from token usage event: {e}")
+        return None