diff --git a/README.md b/README.md
index 9812709..b421136 100644
--- a/README.md
+++ b/README.md
@@ -145,15 +145,19 @@ Common commands: `/help`, `/clear`, `/history`, `/stats`, `/exit`
 - ✅ **Agent Multi-round Execution Loop**: Complete tool calling pipeline
 - ✅ **Basic Tool Set**: Read / Write / Edit files + Bash commands
 - ✅ **Session Note Tool**: Agent actively records and retrieves session highlights ⭐
-- ✅ **Claude Skills Integration**: 20+ professional skills (documentation, design, testing, development) ⭐💡 🆕
-- ✅ **MCP Tool Integration**: Memory (knowledge graph) + MiniMax Search (web search) ⭐ 🆕
-- ✅ **MiniMax M2 Model**: Through Anthropic-compatible endpoint
+- ✅ **Claude Skills Integration**: 20+ professional skills (documentation, design, testing, development) ⭐💡
+- ✅ **MCP Tool Integration**: Memory (knowledge graph) + MiniMax Search (web search) ⭐
+- ✅ **MiniMax M2 Model**: Through Anthropic-compatible endpoint with extended thinking support 🆕
 
 ### Advanced Features ⭐
+- ✅ **Extended Thinking**: Full support for M2's thinking capability, preserves reasoning in context 🆕
+- ✅ **Intelligent Context Management**: Smart message history summarization (80K token limit) 🆕
+- ✅ **Precise Token Calculation**: Using tiktoken for accurate token counting 🆕
+- ✅ **Complete Run Logging**: Detailed logs of all requests, responses, and tool executions 🆕
 - ✅ **Persistent Notes**: Agent maintains context across sessions and execution chains
 - ✅ **Intelligent Recording**: Agent autonomously determines what information needs to be recorded
-- ✅ **Multi-round Sessions**: Supports session management, history clearing, statistics, etc. 🆕
-- ✅ **Beautiful Interaction**: Colorful terminal output, clear session interface 🆕
+- ✅ **Multi-round Sessions**: Supports session management, history clearing, statistics, etc.
+- ✅ **Beautiful Interaction**: Colorful terminal output, clear session interface
 - ✅ **Simple yet Complete**: Showcases core functionality, avoids excessive complexity
 
 ## Project Structure
@@ -172,15 +176,17 @@ mini-agent/
 │   ├── config-example.yaml # API configuration example
 │   ├── agent.py          # Core Agent
 │   ├── llm.py            # LLM Client (Anthropic compatible)
-│   ├── config.py         # Configuration loader 🆕
+│   ├── logger.py         # Agent run logger 🆕
+│   ├── retry.py          # Retry mechanism
+│   ├── config.py         # Configuration loader
 │   └── tools/
 │       ├── base.py       # Tool base class
 │       ├── file_tools.py # File tools
 │       ├── bash_tool.py  # Bash tool
 │       ├── note_tool.py  # Session Note tool ⭐
 │       ├── mcp_loader.py # MCP loader (supports external servers) ⭐
-│       ├── skill_loader.py # Skill loader 🆕
-│       └── skill_tool.py # Skill tool 🆕
+│       ├── skill_loader.py # Skill loader
+│       └── skill_tool.py # Skill tool
 ├── tests/
 │   ├── test_agent.py     # Agent integration tests
 │   ├── test_llm.py       # LLM tests
@@ -212,10 +218,27 @@ External MCP Servers:
 ```python
 # Simplified core loop (from agent.py)
 async def run(self) -> str:
+    # Initialize run logging
+    self.logger.start_new_run()
+    
     step = 0
     while step < self.max_steps:
+        # Check and summarize message history if needed
+        await self._summarize_messages()
+        
+        # Log LLM request
+        self.logger.log_request(messages=self.messages, tools=tool_schemas)
+        
         # 1. Call LLM
         response = await self.llm.generate(messages, tools)
+        
+        # Log LLM response
+        self.logger.log_response(
+            content=response.content,
+            thinking=response.thinking,
+            tool_calls=response.tool_calls,
+            finish_reason=response.finish_reason
+        )
 
         # 2. If no tool calls, task complete
         if not response.tool_calls:
@@ -224,12 +247,92 @@ async def run(self) -> str:
         # 3. Execute tool calls
         for tool_call in response.tool_calls:
             result = await tool.execute(**arguments)
+            
+            # Log tool result
+            self.logger.log_tool_result(
+                tool_name=function_name,
+                arguments=arguments,
+                result_success=result.success,
+                result_content=result.content
+            )
+            
             self.messages.append(tool_result_message)
 
         step += 1
 ```
 
-### 2. Session Note Tool - Session Note Recording ⭐
+### 2. Extended Thinking Support 🆕
+
+The agent fully supports MiniMax M2's extended thinking capability:
+
+**Key Features**:
+- **Thinking Preservation**: Thinking content is preserved in message history for context continuity
+- **Display Control**: Thinking is displayed separately from response content in the terminal
+- **Context Injection**: When replaying conversation history, thinking blocks are properly injected
+
+**Message Structure**:
+```python
+class Message(BaseModel):
+    role: str
+    content: str | List[Dict[str, Any]]
+    thinking: str | None = None  # Extended thinking content
+    tool_calls: List[Dict[str, Any]] | None = None
+```
+
+**API Format Conversion**:
+```python
+# When sending to API, thinking is injected as content block
+content_blocks = []
+if msg.thinking:
+    content_blocks.append({"type": "thinking", "thinking": msg.thinking})
+if msg.content:
+    content_blocks.append({"type": "text", "text": msg.content})
+```
+
+### 3. Intelligent Context Management 🆕
+
+The agent uses smart summarization instead of simple truncation to manage long conversations:
+
+**Token Limit**: 80,000 tokens (configurable)
+
+**Summarization Strategy** (Agent Mode):
+1. **Precise Token Calculation**: Uses tiktoken (cl100k_base encoder) for accurate token counting
+2. **Trigger Condition**: Automatically summarizes when token count exceeds the limit
+3. **Per-Round Summarization**: Summarizes agent execution between each pair of user messages
+4. **User Intent Preservation**: All user messages are preserved (never summarized)
+5. **In-Progress Handling**: Even the current executing round gets summarized
+6. **LLM-Generated Summary**: Uses LLM to create concise summaries (≤300 words) of each execution round
+
+**Message Structure After Summary**:
+```
+system → user1 → summary1 → user2 → summary2 → user3 → summary3 (if executing)
+```
+
+**Benefits**:
+- ✅ Maintains long-term context without token overflow
+- ✅ Preserves all user task intents (never lost)
+- ✅ Compresses only agent execution process (assistant/tool messages)
+- ✅ Each round independently summarized for clarity
+- ✅ Continuous context across long conversations
+
+### 4. Complete Run Logging 🆕
+
+Every agent run is fully logged with detailed information:
+
+**Log File Format**: `workspace/agent_run_YYYYMMDD_HHMMSS.log`
+
+**Logged Information**:
+- **[N] REQUEST**: LLM requests with all messages and available tools
+- **[N+1] RESPONSE**: LLM responses with content, thinking, and tool calls
+- **[N+2] TOOL_RESULT**: Tool execution results with arguments and outputs
+
+**Benefits**:
+- ✅ Complete audit trail of all interactions
+- ✅ Easy debugging and troubleshooting
+- ✅ Sequential indexing for event tracking
+- ✅ Separate log file per run
+
+### 5. Session Note Tool - Session Note Recording ⭐
 
 This is one of the **core highlights** of this demo, showcasing a simple and efficient session memory management approach.
 
@@ -298,7 +401,7 @@ Notes are stored in JSON format at `workspace/.agent_memory.json`:
 ]
 ```
 
-### 3. MiniMax Search - Web Search and Browse ⭐
+### 6. MiniMax Search - Web Search and Browse ⭐
 
 This is an **independent MCP Server** integrated into the Agent via `mcp.json`.
 
@@ -357,7 +460,7 @@ Agent: (calls browse tool)
 
 ---
 
-### 4. Tool Definition
+### 7. Tool Definition
 
 Each tool inherits from the `Tool` base class:
 
@@ -428,18 +531,26 @@ pytest tests/test_agent.py tests/test_note_tool.py -v
 
 ## Summary
 
-This project is an **educational-friendly** yet **technically complete** Agent implementation:
+This project is an **educational-friendly** yet **production-ready** Agent implementation:
+
+✅ **Simple Enough**: Clean code structure, easy to understand
+✅ **Complete Enough**: Includes all core Agent functionality with advanced features
+✅ **Production Features**: Extended thinking, smart context management, complete logging
+✅ **Shows the Gap**: Clearly contrasts differences between Demo and production
 
-✅ **Simple Enough**: Minimal code, easy to understand
-✅ **Complete Enough**: Includes core functionality and Session Note Tool
-✅ **Shows the Gap**: Clearly contrasts the huge difference between Demo and production
+**Key Highlights**:
+- 🧠 **Extended Thinking Support**: Full M2 thinking capability with context preservation
+- 📊 **Smart Context Management**: LLM-based summarization for long conversations (80K tokens)
+- 📝 **Complete Logging**: Detailed audit trail of all interactions
+- 🎯 **Precise Token Counting**: Using tiktoken for accurate calculation
 
 Suitable for:
 - 🎓 Learning Agent architecture and working principles
 - 🧪 Rapid experimentation and prototype validation
 - 📚 Understanding production environment complexity
+- 🚀 Building production-grade Agent applications
 
-**Not suitable** for direct production use.
+This implementation demonstrates both simplicity and completeness, bridging the gap between educational demos and production systems.
 
 ## 📚 Related Documentation
 
diff --git a/README_CN.md b/README_CN.md
index fe89401..cce91a8 100644
--- a/README_CN.md
+++ b/README_CN.md
@@ -147,15 +147,19 @@ uv run python main.py
 - ✅ **Agent 多轮执行循环**: 完整的工具调用链路
 - ✅ **基础工具集**: Read / Write / Edit 文件 + Bash 命令
 - ✅ **Session Note Tool**: Agent 主动记录和检索会话要点 ⭐
-- ✅ **Claude Skills 集成**: 20+ 专业技能（文档、设计、测试、开发）⭐💡 🆕
-- ✅ **MCP 工具集成**: Memory（知识图谱）+ MiniMax Search（网页搜索）⭐ 🆕
-- ✅ **MiniMax M2 模型**: 通过 Anthropic 兼容端点
+- ✅ **Claude Skills 集成**: 20+ 专业技能（文档、设计、测试、开发）⭐💡
+- ✅ **MCP 工具集成**: Memory（知识图谱）+ MiniMax Search（网页搜索）⭐
+- ✅ **MiniMax M2 模型**: 通过 Anthropic 兼容端点，支持扩展思维 🆕
 
 ### 进阶特性 ⭐
+- ✅ **扩展思维支持**: 完整支持 M2 的 thinking 能力，在上下文中保留推理过程 🆕
+- ✅ **智能上下文管理**: 智能消息历史摘要（80K token 限制）🆕
+- ✅ **精确 Token 计算**: 使用 tiktoken 进行准确的 token 统计 🆕
+- ✅ **完整运行日志**: 详细记录所有请求、响应和工具执行 🆕
 - ✅ **持久化笔记**: Agent 跨会话和执行链路保持上下文
 - ✅ **智能记录**: Agent 自主判断什么信息需要记录
-- ✅ **多轮会话**: 支持会话管理、历史清除、统计等功能 🆕
-- ✅ **美化交互**: 彩色终端输出，清晰的会话界面 🆕
+- ✅ **多轮会话**: 支持会话管理、历史清除、统计等功能
+- ✅ **美化交互**: 彩色终端输出，清晰的会话界面
 - ✅ **简洁但完整**: 展示核心功能，避免过度复杂
 
 ## 项目结构
@@ -174,15 +178,17 @@ mini-agent/
 │   ├── config-example.yaml # API 配置示例
 │   ├── agent.py          # 核心 Agent
 │   ├── llm.py            # LLM 客户端 (Anthropic 兼容)
-│   ├── config.py         # 配置加载器 🆕
+│   ├── logger.py         # Agent 运行日志记录器 🆕
+│   ├── retry.py          # 重试机制
+│   ├── config.py         # 配置加载器
 │   └── tools/
 │       ├── base.py       # 工具基类
 │       ├── file_tools.py # 文件工具
 │       ├── bash_tool.py  # Bash 工具
 │       ├── note_tool.py  # Session Note 工具 ⭐
 │       ├── mcp_loader.py # MCP 加载器（支持外部服务器）⭐
-│       ├── skill_loader.py # Skill 加载器 🆕
-│       └── skill_tool.py # Skill 工具 🆕
+│       ├── skill_loader.py # Skill 加载器
+│       └── skill_tool.py # Skill 工具
 ├── tests/
 │   ├── test_agent.py     # Agent 集成测试
 │   ├── test_llm.py       # LLM 测试
@@ -214,10 +220,27 @@ mini-agent/
 ```python
 # 简化的核心循环（来自 agent.py）
 async def run(self) -> str:
+    # 初始化运行日志
+    self.logger.start_new_run()
+    
     step = 0
     while step < self.max_steps:
+        # 检查并在需要时摘要消息历史
+        await self._summarize_messages()
+        
+        # 记录 LLM 请求日志
+        self.logger.log_request(messages=self.messages, tools=tool_schemas)
+        
         # 1. 调用 LLM
         response = await self.llm.generate(messages, tools)
+        
+        # 记录 LLM 响应日志
+        self.logger.log_response(
+            content=response.content,
+            thinking=response.thinking,
+            tool_calls=response.tool_calls,
+            finish_reason=response.finish_reason
+        )
 
         # 2. 如果没有工具调用，任务完成
         if not response.tool_calls:
@@ -226,12 +249,92 @@ async def run(self) -> str:
         # 3. 执行工具调用
         for tool_call in response.tool_calls:
             result = await tool.execute(**arguments)
+            
+            # 记录工具结果日志
+            self.logger.log_tool_result(
+                tool_name=function_name,
+                arguments=arguments,
+                result_success=result.success,
+                result_content=result.content
+            )
+            
             self.messages.append(tool_result_message)
 
         step += 1
 ```
 
-### 2. Session Note Tool - 会话笔记记录 ⭐
+### 2. 扩展思维支持 🆕
+
+Agent 完整支持 MiniMax M2 的扩展思维能力：
+
+**核心特性**：
+- **思维保留**：thinking 内容在消息历史中保留，确保上下文连续性
+- **展示控制**：thinking 在终端中与响应内容分开显示
+- **上下文注入**：回放对话历史时，thinking 块被正确注入
+
+**消息结构**：
+```python
+class Message(BaseModel):
+    role: str
+    content: str | List[Dict[str, Any]]
+    thinking: str | None = None  # 扩展思维内容
+    tool_calls: List[Dict[str, Any]] | None = None
+```
+
+**API 格式转换**：
+```python
+# 发送到 API 时，thinking 作为 content block 注入
+content_blocks = []
+if msg.thinking:
+    content_blocks.append({"type": "thinking", "thinking": msg.thinking})
+if msg.content:
+    content_blocks.append({"type": "text", "text": msg.content})
+```
+
+### 3. 智能上下文管理 🆕
+
+Agent 使用智能摘要而非简单截断来管理长对话：
+
+**Token 限制**：80,000 tokens（可配置）
+
+**摘要策略**（Agent 模式）：
+1. **精确 Token 计算**：使用 tiktoken（cl100k_base 编码器）进行准确的 token 统计
+2. **触发条件**：当 token 数量超过限制时自动摘要
+3. **逐轮摘要**：对每一轮 user-user 之间的 agent 执行过程进行摘要
+4. **保留用户意图**：所有 user 消息都被保留（永不被摘要）
+5. **执行中处理**：即使当前正在执行的轮次也会被摘要
+6. **LLM 生成摘要**：使用 LLM 为每轮执行创建简洁摘要（≤300 字）
+
+**摘要后的消息结构**：
+```
+system → user1 → summary1 → user2 → summary2 → user3 → summary3 (如果执行中)
+```
+
+**优势**：
+- ✅ 维持长期上下文而不会 token 溢出
+- ✅ 保留所有用户任务意图（永不丢失）
+- ✅ 仅压缩 agent 执行过程（assistant/tool 消息）
+- ✅ 每轮独立摘要，清晰明了
+- ✅ 长对话中的持续上下文
+
+### 4. 完整运行日志 🆕
+
+每次 agent 运行都有完整的日志记录：
+
+**日志文件格式**：`workspace/agent_run_YYYYMMDD_HHMMSS.log`
+
+**记录内容**：
+- **[N] REQUEST**：LLM 请求，包含所有消息和可用工具
+- **[N+1] RESPONSE**：LLM 响应，包含内容、thinking 和工具调用
+- **[N+2] TOOL_RESULT**：工具执行结果，包含参数和输出
+
+**优势**：
+- ✅ 完整的交互审计轨迹
+- ✅ 便于调试和故障排查
+- ✅ 顺序索引用于事件追踪
+- ✅ 每次运行独立的日志文件
+
+### 5. Session Note Tool - 会话笔记记录 ⭐
 
 这是本 demo 的**核心亮点**之一，展示了一种简洁高效的会话记忆管理方式。
 
@@ -300,7 +403,7 @@ Agent: (主动调用 recall_notes)
 ]
 ```
 
-### 3. MiniMax Search - 网页搜索和浏览 ⭐
+### 6. MiniMax Search - 网页搜索和浏览 ⭐
 
 这是一个**独立的 MCP Server**，通过 `mcp.json` 集成到 Agent 中。
 
@@ -360,7 +463,7 @@ Agent: (调用 browse 工具)
 
 ---
 
-### 4. 工具定义
+### 7. 工具定义
 
 每个工具继承自 `Tool` 基类：
 
@@ -431,18 +534,26 @@ pytest tests/test_agent.py tests/test_note_tool.py -v
 
 ## 总结
 
-本项目是一个**教学友好**但**技术完整**的 Agent 实现：
+本项目是一个**教学友好**且**生产级**的 Agent 实现：
+
+✅ **足够简单**: 代码结构清晰，易于理解
+✅ **足够完整**: 包含所有核心 Agent 功能和高级特性
+✅ **生产级特性**: 扩展思维、智能上下文管理、完整日志
+✅ **展示鸿沟**: 清晰对比 Demo 和生产环境的差异
 
-✅ **足够简单**: 代码量少，易于理解
-✅ **足够完整**: 包含核心功能和 Session Note Tool
-✅ **展示鸿沟**: 清晰对比 Demo 和生产环境的巨大差异
+**核心亮点**：
+- 🧠 **扩展思维支持**：完整的 M2 thinking 能力，保留上下文
+- 📊 **智能上下文管理**：基于 LLM 的长对话摘要（80K tokens）
+- 📝 **完整日志记录**：所有交互的详细审计轨迹
+- 🎯 **精确 Token 统计**：使用 tiktoken 进行准确计算
 
 适合用于：
 - 🎓 学习 Agent 架构和工作原理
 - 🧪 快速实验和原型验证
 - 📚 理解生产环境的复杂性
+- 🚀 构建生产级 Agent 应用
 
-**不适合**直接用于生产环境。
+本实现展示了简洁性和完整性的平衡，架起了教学演示和生产系统之间的桥梁。
 
 ## 📚 相关文档
 
diff --git a/mini_agent/agent.py b/mini_agent/agent.py
index 83d4e28..7224ebc 100644
--- a/mini_agent/agent.py
+++ b/mini_agent/agent.py
@@ -2,9 +2,12 @@
 
 import json
 from pathlib import Path
-from typing import Dict, List
+from typing import List
 
-from .llm import LLMClient, LLMResponse, Message
+import tiktoken
+
+from .llm import LLMClient, Message
+from .logger import AgentLogger
 from .tools.base import Tool, ToolResult
 
 
@@ -45,13 +48,13 @@ def __init__(
         tools: List[Tool],
         max_steps: int = 50,
         workspace_dir: str = "./workspace",
-        max_messages: int = 20,  # 最大消息数，防止 context 超限
+        token_limit: int = 80000,  # Token 超过此值时触发 summary
     ):
         self.llm = llm_client
         self.system_prompt = system_prompt
         self.tools = {tool.name: tool for tool in tools}
         self.max_steps = max_steps
-        self.max_messages = max_messages
+        self.token_limit = token_limit
         self.workspace_dir = Path(workspace_dir)
 
         # Ensure workspace exists
@@ -60,31 +63,225 @@ def __init__(
         # Initialize message history
         self.messages: List[Message] = [Message(role="system", content=system_prompt)]
 
+        # 初始化日志记录器
+        self.logger = AgentLogger(self.workspace_dir)
+
     def add_user_message(self, content: str):
         """Add a user message to history."""
         self.messages.append(Message(role="user", content=content))
 
-    def _truncate_messages(self):
-        """简单的消息截断：保留 system prompt + 最近的消息
+    def _estimate_tokens(self) -> int:
+        """使用 tiktoken 精确计算消息历史的 token 数量
 
-        策略：
-        - 永远保留第一条 system prompt
-        - 保留最近的 (max_messages - 1) 条消息
-        - 当消息数超过 max_messages 时触发截断
+        使用 cl100k_base 编码器（GPT-4/Claude 兼容）
+        """
+        try:
+            # 使用 cl100k_base 编码器（GPT-4 和大多数现代模型使用）
+            encoding = tiktoken.get_encoding("cl100k_base")
+        except Exception:
+            # Fallback: 如果 tiktoken 初始化失败，使用简单估算
+            return self._estimate_tokens_fallback()
+
+        total_tokens = 0
+
+        for msg in self.messages:
+            # 统计文本内容
+            if isinstance(msg.content, str):
+                total_tokens += len(encoding.encode(msg.content))
+            elif isinstance(msg.content, list):
+                for block in msg.content:
+                    if isinstance(block, dict):
+                        # 将字典转为字符串计算
+                        total_tokens += len(encoding.encode(str(block)))
+
+            # 统计 thinking
+            if msg.thinking:
+                total_tokens += len(encoding.encode(msg.thinking))
+
+            # 统计 tool_calls
+            if msg.tool_calls:
+                total_tokens += len(encoding.encode(str(msg.tool_calls)))
+
+            # 每条消息的元数据开销（约 4 tokens）
+            total_tokens += 4
+
+        return total_tokens
+
+    def _estimate_tokens_fallback(self) -> int:
+        """Fallback token 估算方法（当 tiktoken 不可用时）"""
+        total_chars = 0
+        for msg in self.messages:
+            if isinstance(msg.content, str):
+                total_chars += len(msg.content)
+            elif isinstance(msg.content, list):
+                for block in msg.content:
+                    if isinstance(block, dict):
+                        total_chars += len(str(block))
+
+            if msg.thinking:
+                total_chars += len(msg.thinking)
+
+            if msg.tool_calls:
+                total_chars += len(str(msg.tool_calls))
+
+        # 粗略估算：平均 2.5 个字符 = 1 token
+        return int(total_chars / 2.5)
+
+    async def _summarize_messages(self):
+        """消息历史摘要：当 token 超限时，对每一轮 user-user 之间的对话进行摘要
+
+        策略 (Agent 模式)：
+        - 保留所有 user 消息（这是用户的意图）
+        - 对每一轮 user-user 之间的内容（agent 执行过程）进行 summary
+        - 最后一轮如果还在执行中（有 agent/tool 消息但没有下一个 user），也要 summary
+        - 结构：system -> user1 -> summary1 -> user2 -> summary2 -> user3 -> summary3(如果执行中)
+        """
+        estimated_tokens = self._estimate_tokens()
+
+        # 如果未超限，不需要 summary
+        if estimated_tokens <= self.token_limit:
+            return
+
+        print(
+            f"\n{Colors.BRIGHT_YELLOW}📊 Token 估算值: {estimated_tokens}/{self.token_limit}{Colors.RESET}"
+        )
+        print(f"{Colors.BRIGHT_YELLOW}🔄 触发消息历史摘要...{Colors.RESET}")
+
+        # 找到所有 user 消息的索引（跳过 system prompt）
+        user_indices = [
+            i for i, msg in enumerate(self.messages) if msg.role == "user" and i > 0
+        ]
+
+        # 至少需要 1 个 user 消息才能进行 summary
+        if len(user_indices) < 1:
+            print(f"{Colors.BRIGHT_YELLOW}⚠️  消息不足，无法进行摘要{Colors.RESET}")
+            return
+
+        # 构建新的消息列表
+        new_messages = [self.messages[0]]  # 保留 system prompt
+        summary_count = 0
+
+        # 遍历每个 user 消息，对其后面的执行过程进行 summary
+        for i, user_idx in enumerate(user_indices):
+            # 添加当前 user 消息
+            new_messages.append(self.messages[user_idx])
+
+            # 确定要 summary 的消息范围
+            # 如果是最后一个 user，则到消息列表末尾；否则到下一个 user 之前
+            if i < len(user_indices) - 1:
+                next_user_idx = user_indices[i + 1]
+            else:
+                next_user_idx = len(self.messages)
+
+            # 提取这一轮的执行消息
+            execution_messages = self.messages[user_idx + 1 : next_user_idx]
+
+            # 如果这一轮有执行消息，进行 summary
+            if execution_messages:
+                summary_text = await self._create_summary(
+                    execution_messages, user_idx, i + 1
+                )
+                if summary_text:
+                    summary_message = Message(
+                        role="user", content=f"[执行摘要]\n\n{summary_text}"
+                    )
+                    new_messages.append(summary_message)
+                    summary_count += 1
+
+        # 替换消息列表
+        self.messages = new_messages
+
+        new_tokens = self._estimate_tokens()
+        print(
+            f"{Colors.BRIGHT_GREEN}✓ 摘要完成，Token 从 {estimated_tokens} 降至 {new_tokens}{Colors.RESET}"
+        )
+        print(
+            f"{Colors.DIM}  结构: system + {len(user_indices)} 个 user 消息 + {summary_count} 个 summary{Colors.RESET}"
+        )
+
+    async def _create_summary(
+        self, messages: List[Message], user_idx: int, round_num: int
+    ) -> str:
+        """为一轮执行创建摘要
+
+        Args:
+            messages: 要总结的消息列表
+            user_idx: 用户消息的索引
+            round_num: 轮次编号
+
+        Returns:
+            摘要文本
         """
-        if len(self.messages) > self.max_messages:
-            self.messages = [
-                self.messages[0],  # system prompt
-                *self.messages[-(self.max_messages - 1) :],  # 最近的消息
-            ]
+        if not messages:
+            return ""
+
+        # 构建摘要内容
+        summary_content = f"第 {round_num} 轮执行过程：\n\n"
+        for msg in messages:
+            if msg.role == "assistant":
+                content_text = (
+                    msg.content if isinstance(msg.content, str) else str(msg.content)
+                )
+                summary_content += f"助手: {content_text[:200]}\n"
+                if msg.tool_calls:
+                    tool_names = [tc["function"]["name"] for tc in msg.tool_calls]
+                    summary_content += f"  → 调用工具: {', '.join(tool_names)}\n"
+            elif msg.role == "tool":
+                result_preview = (
+                    msg.content[:100]
+                    if isinstance(msg.content, str)
+                    else str(msg.content)[:100]
+                )
+                summary_content += f"  ← 工具返回: {result_preview}...\n"
+
+        # 调用 LLM 生成简洁的摘要
+        try:
+            summary_prompt = f"""请将以下 Agent 执行过程进行简洁总结：
+
+{summary_content}
+
+要求：
+1. 重点记录完成了什么任务、调用了哪些工具
+2. 保留关键的执行结果和重要发现
+3. 简洁明了，控制在 300 字以内
+4. 使用中文
+5. 不要包含"用户"相关内容，只总结 Agent 的执行过程"""
+
+            summary_msg = Message(role="user", content=summary_prompt)
+            response = await self.llm.generate(
+                messages=[
+                    Message(
+                        role="system",
+                        content="你是一个擅长总结 Agent 执行过程的助手。",
+                    ),
+                    summary_msg,
+                ]
+            )
+
+            summary_text = response.content
+            print(f"{Colors.BRIGHT_GREEN}✓ 第 {round_num} 轮摘要生成完成{Colors.RESET}")
+            return summary_text
+
+        except Exception as e:
+            print(
+                f"{Colors.BRIGHT_RED}✗ 第 {round_num} 轮摘要生成失败: {e}{Colors.RESET}"
+            )
+            # 失败时使用简单的文本摘要
+            return summary_content
 
     async def run(self) -> str:
         """Execute agent loop until task is complete or max steps reached."""
+        # 开始新的运行，初始化日志文件
+        self.logger.start_new_run()
+        print(
+            f"{Colors.DIM}📝 日志文件: {self.logger.get_log_file_path()}{Colors.RESET}"
+        )
+
         step = 0
 
         while step < self.max_steps:
-            # 截断消息历史，防止 context 超限
-            self._truncate_messages()
+            # 检查并摘要消息历史，防止 context 超限
+            await self._summarize_messages()
 
             # 步骤标题
             print(f"\n{Colors.DIM}╭{'─' * 58}╮{Colors.RESET}")
@@ -96,6 +293,9 @@ async def run(self) -> str:
             # Get tool schemas
             tool_schemas = [tool.to_schema() for tool in self.tools.values()]
 
+            # 记录 LLM 请求日志
+            self.logger.log_request(messages=self.messages, tools=tool_schemas)
+
             # Call LLM
             try:
                 response = await self.llm.generate(
@@ -118,10 +318,19 @@ async def run(self) -> str:
                     print(f"\n{Colors.BRIGHT_RED}❌ Error:{Colors.RESET} {error_msg}")
                 return error_msg
 
+            # 记录 LLM 响应日志
+            self.logger.log_response(
+                content=response.content,
+                thinking=response.thinking,
+                tool_calls=response.tool_calls,
+                finish_reason=response.finish_reason,
+            )
+
             # Add assistant message
             assistant_msg = Message(
                 role="assistant",
                 content=response.content,
+                thinking=response.thinking,
                 tool_calls=response.tool_calls,
             )
             self.messages.append(assistant_msg)
@@ -170,8 +379,29 @@ async def run(self) -> str:
                         error=f"Unknown tool: {function_name}",
                     )
                 else:
-                    tool = self.tools[function_name]
-                    result = await tool.execute(**arguments)
+                    try:
+                        tool = self.tools[function_name]
+                        result = await tool.execute(**arguments)
+                    except Exception as e:
+                        # 捕获工具执行中的所有异常，转换为失败的 ToolResult
+                        import traceback
+
+                        error_detail = f"{type(e).__name__}: {str(e)}"
+                        error_trace = traceback.format_exc()
+                        result = ToolResult(
+                            success=False,
+                            content="",
+                            error=f"Tool execution failed: {error_detail}\n\nTraceback:\n{error_trace}",
+                        )
+
+                # 记录工具执行结果日志
+                self.logger.log_tool_result(
+                    tool_name=function_name,
+                    arguments=arguments,
+                    result_success=result.success,
+                    result_content=result.content if result.success else None,
+                    result_error=result.error if not result.success else None,
+                )
 
                 # Print result
                 if result.success:
diff --git a/mini_agent/llm.py b/mini_agent/llm.py
index a63a508..2bfa9e8 100644
--- a/mini_agent/llm.py
+++ b/mini_agent/llm.py
@@ -18,6 +18,7 @@ class Message(BaseModel):
 
     role: str  # "system", "user", "assistant", "tool"
     content: str | List[Dict[str, Any]]  # Can be string or list of content blocks
+    thinking: str | None = None  # Extended thinking content for assistant messages
     tool_calls: List[Dict[str, Any]] | None = None
     tool_call_id: str | None = None
     name: str | None = None  # For tool role
@@ -129,23 +130,34 @@ async def generate(
 
             # For user and assistant messages
             if msg.role in ["user", "assistant"]:
-                # Handle tool results embedded in assistant messages
-                if msg.role == "assistant" and msg.tool_calls:
-                    # Build content blocks for assistant with tool calls
+                # Handle assistant messages with thinking or tool calls
+                if msg.role == "assistant" and (msg.thinking or msg.tool_calls):
+                    # Build content blocks for assistant with thinking and/or tool calls
                     content_blocks = []
+
+                    # Add thinking block if present
+                    if msg.thinking:
+                        content_blocks.append(
+                            {"type": "thinking", "thinking": msg.thinking}
+                        )
+
+                    # Add text content if present
                     if msg.content:
                         content_blocks.append({"type": "text", "text": msg.content})
 
                     # Add tool use blocks
-                    for tool_call in msg.tool_calls:
-                        content_blocks.append(
-                            {
-                                "type": "tool_use",
-                                "id": tool_call["id"],
-                                "name": tool_call["function"]["name"],
-                                "input": json.loads(tool_call["function"]["arguments"]),
-                            }
-                        )
+                    if msg.tool_calls:
+                        for tool_call in msg.tool_calls:
+                            content_blocks.append(
+                                {
+                                    "type": "tool_use",
+                                    "id": tool_call["id"],
+                                    "name": tool_call["function"]["name"],
+                                    "input": json.loads(
+                                        tool_call["function"]["arguments"]
+                                    ),
+                                }
+                            )
 
                     api_messages.append(
                         {"role": "assistant", "content": content_blocks}
diff --git a/mini_agent/logger.py b/mini_agent/logger.py
new file mode 100644
index 0000000..903542f
--- /dev/null
+++ b/mini_agent/logger.py
@@ -0,0 +1,179 @@
+"""Agent 运行日志记录器"""
+
+import json
+from datetime import datetime
+from pathlib import Path
+from typing import Any, Dict
+
+
+class AgentLogger:
+    """Agent 运行日志记录器
+
+    负责记录每次 agent 运行的完整交互过程，包括：
+    - LLM 请求和响应
+    - 工具调用和结果
+    """
+
+    def __init__(self, workspace_dir: Path):
+        """初始化日志记录器
+
+        Args:
+            workspace_dir: 工作目录路径
+        """
+        self.workspace_dir = workspace_dir
+        self.log_file = None
+        self.log_index = 0
+
+    def start_new_run(self):
+        """开始新的运行，创建新的日志文件"""
+        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        log_filename = f"agent_run_{timestamp}.log"
+        self.log_file = self.workspace_dir / log_filename
+        self.log_index = 0
+
+        # 写入日志头部
+        with open(self.log_file, "w", encoding="utf-8") as f:
+            f.write("=" * 80 + "\n")
+            f.write(f"Agent Run Log - {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n")
+            f.write("=" * 80 + "\n\n")
+
+    def log_request(self, messages: list, tools: list = None):
+        """记录 LLM 请求
+
+        Args:
+            messages: 消息列表
+            tools: 工具列表（可选）
+        """
+        self.log_index += 1
+
+        # 构建完整的请求数据结构
+        request_data = {
+            "messages": [],
+            "tools": [],
+        }
+
+        # 转换消息为 JSON 可序列化格式
+        for msg in messages:
+            msg_dict = {
+                "role": msg.role,
+                "content": msg.content,
+            }
+            if msg.thinking:
+                msg_dict["thinking"] = msg.thinking
+            if msg.tool_calls:
+                msg_dict["tool_calls"] = msg.tool_calls
+            if msg.tool_call_id:
+                msg_dict["tool_call_id"] = msg.tool_call_id
+            if msg.name:
+                msg_dict["name"] = msg.name
+
+            request_data["messages"].append(msg_dict)
+
+        # 只记录工具名称
+        if tools:
+            request_data["tools"] = [
+                tool.get("function", {}).get("name", "unknown") for tool in tools
+            ]
+
+        # 格式化为 JSON
+        content = "LLM Request:\n\n"
+        content += json.dumps(request_data, indent=2, ensure_ascii=False)
+
+        self._write_log("REQUEST", content)
+
+    def log_response(
+        self,
+        content: str,
+        thinking: str = None,
+        tool_calls: list = None,
+        finish_reason: str = None,
+    ):
+        """记录 LLM 响应
+
+        Args:
+            content: 响应内容
+            thinking: 思考内容（可选）
+            tool_calls: 工具调用列表（可选）
+            finish_reason: 完成原因（可选）
+        """
+        self.log_index += 1
+
+        # 构建完整的响应数据结构
+        response_data = {
+            "content": content,
+        }
+
+        if thinking:
+            response_data["thinking"] = thinking
+
+        if tool_calls:
+            response_data["tool_calls"] = tool_calls
+
+        if finish_reason:
+            response_data["finish_reason"] = finish_reason
+
+        # 格式化为 JSON
+        log_content = "LLM Response:\n\n"
+        log_content += json.dumps(response_data, indent=2, ensure_ascii=False)
+
+        self._write_log("RESPONSE", log_content)
+
+    def log_tool_result(
+        self,
+        tool_name: str,
+        arguments: Dict[str, Any],
+        result_success: bool,
+        result_content: str = None,
+        result_error: str = None,
+    ):
+        """记录工具执行结果
+
+        Args:
+            tool_name: 工具名称
+            arguments: 工具参数
+            result_success: 是否成功
+            result_content: 结果内容（成功时）
+            result_error: 错误信息（失败时）
+        """
+        self.log_index += 1
+
+        # 构建完整的工具执行结果数据结构
+        tool_result_data = {
+            "tool_name": tool_name,
+            "arguments": arguments,
+            "success": result_success,
+        }
+
+        if result_success:
+            tool_result_data["result"] = result_content
+        else:
+            tool_result_data["error"] = result_error
+
+        # 格式化为 JSON
+        content = "Tool Execution:\n\n"
+        content += json.dumps(tool_result_data, indent=2, ensure_ascii=False)
+
+        self._write_log("TOOL_RESULT", content)
+
+    def _write_log(self, log_type: str, content: str):
+        """写入日志条目
+
+        Args:
+            log_type: 日志类型（REQUEST, RESPONSE, TOOL_RESULT）
+            content: 日志内容
+        """
+        if self.log_file is None:
+            return
+
+        with open(self.log_file, "a", encoding="utf-8") as f:
+            f.write("\n" + "-" * 80 + "\n")
+            f.write(f"[{self.log_index}] {log_type}\n")
+            f.write(
+                f"Timestamp: {datetime.now().strftime('%Y-%m-%d %H:%M:%S.%f')[:-3]}\n"
+            )
+            f.write("-" * 80 + "\n")
+            f.write(content + "\n")
+
+    def get_log_file_path(self) -> Path:
+        """获取当前日志文件路径"""
+        return self.log_file
diff --git a/pyproject.toml b/pyproject.toml
index bfdcde6..206c7bc 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -11,6 +11,7 @@ dependencies = [
     "mcp>=1.0.0",
     "pytest>=8.4.2",
     "requests>=2.31.0",
+    "tiktoken>=0.5.0",
 ]
 
 [project.optional-dependencies]
diff --git a/system_prompt.txt b/system_prompt.txt
index f1e8ab6..8596a36 100644
--- a/system_prompt.txt
+++ b/system_prompt.txt
@@ -1,32 +1,64 @@
-You are a helpful AI assistant that can execute tasks by using tools.
+You are a versatile AI assistant powered by Claude, capable of executing complex tasks through a rich toolset and specialized skills.
 
-You have access to the following capabilities:
-- Read, write, and edit files
-- Execute bash commands
-- Use additional tools loaded from MCP servers (if configured)
+## Core Capabilities
 
-## Guidelines:
+### 1. **Basic Tools**
+- **File Operations**: Read, write, edit files with full path support
+- **Bash Execution**: Run commands, manage git, packages, and system operations
+- **MCP Tools**: Access additional tools from configured MCP servers
 
-1. **Task Execution**: Break down user requests into clear steps and execute them using the available tools.
+### 2. **Claude Skills**
+You have access to 15+ specialized skills that provide expert guidance for specific tasks:
 
-2. **File Operations**:
-   - Always use absolute paths or paths relative to the workspace directory
-   - Check if files exist before reading or editing them
-   - Create necessary directories before writing files
+**Available Skills:**
+- `list_skills` - 列出所有可用的专业技能
+- `get_skill <skill_name>` - 获取某个技能的详细指导
+- `use_skill <skill_name> <task>` - 应用技能来执行特定任务
 
-3. **Bash Commands**:
-   - Use bash tool for system operations, git commands, package management, etc.
-   - Be careful with destructive commands - explain what you're doing
-   - Check command outputs for errors
+**Skill Categories:**
+- 🎨 **Creative & Design**: algorithmic-art, canvas-design, slack-gif-creator
+- 💻 **Development**: artifacts-builder, mcp-builder, webapp-testing  
+- 📄 **Documents**: docx, pdf, pptx, xlsx (create/edit/analyze)
+- 🏢 **Enterprise**: brand-guidelines, internal-comms, theme-factory
+- 🛠️ **Meta**: skill-creator, template-skill
 
-4. **Communication**:
-   - Be concise and clear in your responses
-   - Explain what you're doing before executing tools
-   - Report errors clearly and suggest solutions
+**When to Use Skills:**
+- When facing specialized tasks (e.g., creating PDFs, building web apps)
+- For following best practices in specific domains
+- When you need expert patterns and guidelines
 
-5. **Task Completion**:
-   - When the task is complete, provide a summary of what was accomplished
-   - Do not call tools unnecessarily
-   - Stop when the user's request has been fulfilled
+## Working Guidelines
 
-Remember: You are working in a workspace directory. All file operations should be relative to this directory unless specified otherwise.
+### Task Execution
+1. **Analyze** the request and identify if a skill can help
+2. **Break down** complex tasks into clear, executable steps
+3. **Use skills** when appropriate for specialized guidance
+4. **Execute** tools systematically and check results
+5. **Report** progress and any issues encountered
+
+### File Operations
+- Use absolute paths or workspace-relative paths
+- Verify file existence before reading/editing
+- Create parent directories before writing files
+- Handle errors gracefully with clear messages
+
+### Bash Commands
+- Explain destructive operations before execution
+- Check command outputs for errors
+- Use appropriate error handling
+- Prefer specialized tools over raw commands when available
+
+### Communication
+- Be concise but thorough in responses
+- Explain your approach before tool execution
+- Report errors with context and solutions
+- Summarize accomplishments when complete
+
+### Best Practices
+- **Don't guess** - use tools to discover missing information
+- **Be proactive** - infer intent and take reasonable actions
+- **Stay focused** - stop when the task is fulfilled
+- **Use skills** - leverage specialized knowledge when relevant
+
+## Workspace Context
+You are working in a workspace directory. All operations are relative to this context unless absolute paths are specified.
diff --git a/uv.lock b/uv.lock
index 458dbde..f683b7e 100644
--- a/uv.lock
+++ b/uv.lock
@@ -2,46 +2,6 @@ version = 1
 revision = 1
 requires-python = ">=3.10"
 
-[[package]]
-name = "mini-agent"
-version = "0.1.0"
-source = { editable = "." }
-dependencies = [
-    { name = "httpx" },
-    { name = "mcp" },
-    { name = "pydantic" },
-    { name = "pytest" },
-    { name = "pyyaml" },
-    { name = "requests" },
-]
-
-[package.optional-dependencies]
-dev = [
-    { name = "pytest" },
-    { name = "pytest-asyncio" },
-]
-
-[package.dev-dependencies]
-dev = [
-    { name = "pytest-asyncio" },
-]
-
-[package.metadata]
-requires-dist = [
-    { name = "httpx", specifier = ">=0.27.0" },
-    { name = "mcp", specifier = ">=1.0.0" },
-    { name = "pydantic", specifier = ">=2.0.0" },
-    { name = "pytest", specifier = ">=8.4.2" },
-    { name = "pytest", marker = "extra == 'dev'", specifier = ">=7.0.0" },
-    { name = "pytest-asyncio", marker = "extra == 'dev'", specifier = ">=0.21.0" },
-    { name = "pyyaml", specifier = ">=6.0.0" },
-    { name = "requests", specifier = ">=2.31.0" },
-]
-provides-extras = ["dev"]
-
-[package.metadata.requires-dev]
-dev = [{ name = "pytest-asyncio", specifier = ">=1.2.0" }]
-
 [[package]]
 name = "annotated-types"
 version = "0.7.0"
@@ -328,6 +288,48 @@ wheels = [
     { url = "https://pypi.tuna.tsinghua.edu.cn/packages/ce/a3/3e71a875a08b6a830b88c40bc413bff01f1650f1efe8a054b5e90a9d4f56/mcp-1.19.0-py3-none-any.whl", hash = "sha256:f5907fe1c0167255f916718f376d05f09a830a215327a3ccdd5ec8a519f2e572", size = 170105 },
 ]
 
+[[package]]
+name = "mini-agent"
+version = "0.1.0"
+source = { editable = "." }
+dependencies = [
+    { name = "httpx" },
+    { name = "mcp" },
+    { name = "pydantic" },
+    { name = "pytest" },
+    { name = "pyyaml" },
+    { name = "requests" },
+    { name = "tiktoken" },
+]
+
+[package.optional-dependencies]
+dev = [
+    { name = "pytest" },
+    { name = "pytest-asyncio" },
+]
+
+[package.dev-dependencies]
+dev = [
+    { name = "pytest-asyncio" },
+]
+
+[package.metadata]
+requires-dist = [
+    { name = "httpx", specifier = ">=0.27.0" },
+    { name = "mcp", specifier = ">=1.0.0" },
+    { name = "pydantic", specifier = ">=2.0.0" },
+    { name = "pytest", specifier = ">=8.4.2" },
+    { name = "pytest", marker = "extra == 'dev'", specifier = ">=7.0.0" },
+    { name = "pytest-asyncio", marker = "extra == 'dev'", specifier = ">=0.21.0" },
+    { name = "pyyaml", specifier = ">=6.0.0" },
+    { name = "requests", specifier = ">=2.31.0" },
+    { name = "tiktoken", specifier = ">=0.5.0" },
+]
+provides-extras = ["dev"]
+
+[package.metadata.requires-dev]
+dev = [{ name = "pytest-asyncio", specifier = ">=1.2.0" }]
+
 [[package]]
 name = "packaging"
 version = "25.0"
@@ -648,6 +650,113 @@ wheels = [
     { url = "https://pypi.tuna.tsinghua.edu.cn/packages/2c/58/ca301544e1fa93ed4f80d724bf5b194f6e4b945841c5bfd555878eea9fcb/referencing-0.37.0-py3-none-any.whl", hash = "sha256:381329a9f99628c9069361716891d34ad94af76e461dcb0335825aecc7692231", size = 26766 },
 ]
 
+[[package]]
+name = "regex"
+version = "2025.10.23"
+source = { registry = "https://pypi.tuna.tsinghua.edu.cn/simple" }
+sdist = { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f8/c8/1d2160d36b11fbe0a61acb7c3c81ab032d9ec8ad888ac9e0a61b85ab99dd/regex-2025.10.23.tar.gz", hash = "sha256:8cbaf8ceb88f96ae2356d01b9adf5e6306fa42fa6f7eab6b97794e37c959ac26", size = 401266 }
+wheels = [
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/88/11/849d5d23633a77047465eaae4cc0cbf24ded7aa496c02e8b9710e28b1687/regex-2025.10.23-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:17bbcde374bef1c5fad9b131f0e28a6a24856dd90368d8c0201e2b5a69533daa", size = 487957 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/87/12/5985386e7e3200a0d6a6417026d2c758d783a932428a5efc0a42ca1ddf74/regex-2025.10.23-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:b4e10434279cc8567f99ca6e018e9025d14f2fded2a603380b6be2090f476426", size = 290419 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/67/cf/a8615923f962f8fdc41a3a6093a48726955e8b1993f4614b26a41d249f9b/regex-2025.10.23-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:9c9bb421cbe7012c744a5a56cf4d6c80829c72edb1a2991677299c988d6339c8", size = 288285 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/4e/3d/6a3a1e12c86354cd0b3cbf8c3dd6acbe853609ee3b39d47ecd3ce95caf84/regex-2025.10.23-cp310-cp310-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:275cd1c2ed8c4a78ebfa489618d7aee762e8b4732da73573c3e38236ec5f65de", size = 781458 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/46/47/76a8da004489f2700361754859e373b87a53d043de8c47f4d1583fd39d78/regex-2025.10.23-cp310-cp310-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:7b426ae7952f3dc1e73a86056d520bd4e5f021397484a6835902fc5648bcacce", size = 850605 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/67/05/fa886461f97d45a6f4b209699cb994dc6d6212d6e219d29444dac5005775/regex-2025.10.23-cp310-cp310-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:c5cdaf5b6d37c7da1967dbe729d819461aab6a98a072feef65bbcff0a6e60649", size = 898563 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/2d/db/3ddd8d01455f23cabad7499f4199de0df92f5e96d39633203ff9d0b592dc/regex-2025.10.23-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:3bfeff0b08f296ab28b4332a7e03ca31c437ee78b541ebc874bbf540e5932f8d", size = 791535 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/7c/ae/0fa5cbf41ca92b6ec3370222fcb6c68b240d68ab10e803d086c03a19fd9e/regex-2025.10.23-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:5f97236a67307b775f30a74ef722b64b38b7ab7ba3bb4a2508518a5de545459c", size = 782461 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/d4/23/70af22a016df11af4def27870eb175c2c7235b72d411ecf75a4b4a422cb6/regex-2025.10.23-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:be19e7de499940cd72475fb8e46ab2ecb1cf5906bebdd18a89f9329afb1df82f", size = 774583 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/7a/ee/a54a6851f6905f33d3c4ed64e8737b1d85ed01b5724712530ddc0f9abdb1/regex-2025.10.23-cp310-cp310-musllinux_1_2_ppc64le.whl", hash = "sha256:883df76ee42d9ecb82b37ff8d01caea5895b3f49630a64d21111078bbf8ef64c", size = 845649 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/80/7d/c3ec1cae14e01fab00e38c41ed35f47a853359e95e9c023e9a4381bb122c/regex-2025.10.23-cp310-cp310-musllinux_1_2_s390x.whl", hash = "sha256:2e9117d1d35fc2addae6281019ecc70dc21c30014b0004f657558b91c6a8f1a7", size = 836037 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/15/ae/45771140dd43c4d67c87b54d3728078ed6a96599d9fc7ba6825086236782/regex-2025.10.23-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:0ff1307f531a5d8cf5c20ea517254551ff0a8dc722193aab66c656c5a900ea68", size = 779705 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/b8/95/074e2581760eafce7c816a352b7d3a322536e5b68c346d1a8bacd895545c/regex-2025.10.23-cp310-cp310-win32.whl", hash = "sha256:7888475787cbfee4a7cd32998eeffe9a28129fa44ae0f691b96cb3939183ef41", size = 265663 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f7/c7/a25f56a718847e34d3f1608c72eadeb67653bff1a0411da023dd8f4c647b/regex-2025.10.23-cp310-cp310-win_amd64.whl", hash = "sha256:ec41a905908496ce4906dab20fb103c814558db1d69afc12c2f384549c17936a", size = 277587 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/d3/e5/63eb17c6b5deaefd93c2bbb1feae7c0a8d2157da25883a6ca2569cf7a663/regex-2025.10.23-cp310-cp310-win_arm64.whl", hash = "sha256:b2b7f19a764d5e966d5a62bf2c28a8b4093cc864c6734510bdb4aeb840aec5e6", size = 269979 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/82/e5/74b7cd5cd76b4171f9793042045bb1726f7856dd56e582fc3e058a7a8a5e/regex-2025.10.23-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:6c531155bf9179345e85032052a1e5fe1a696a6abf9cea54b97e8baefff970fd", size = 487960 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/b9/08/854fa4b3b20471d1df1c71e831b6a1aa480281e37791e52a2df9641ec5c6/regex-2025.10.23-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:912e9df4e89d383681268d38ad8f5780d7cccd94ba0e9aa09ca7ab7ab4f8e7eb", size = 290425 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/ab/d3/6272b1dd3ca1271661e168762b234ad3e00dbdf4ef0c7b9b72d2d159efa7/regex-2025.10.23-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:4f375c61bfc3138b13e762fe0ae76e3bdca92497816936534a0177201666f44f", size = 288278 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/14/8f/c7b365dd9d9bc0a36e018cb96f2ffb60d2ba8deb589a712b437f67de2920/regex-2025.10.23-cp311-cp311-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:e248cc9446081119128ed002a3801f8031e0c219b5d3c64d3cc627da29ac0a33", size = 793289 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/d4/fb/b8fbe9aa16cf0c21f45ec5a6c74b4cecbf1a1c0deb7089d4a6f83a9c1caa/regex-2025.10.23-cp311-cp311-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:b52bf9282fdf401e4f4e721f0f61fc4b159b1307244517789702407dd74e38ca", size = 860321 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/b0/81/bf41405c772324926a9bd8a640dedaa42da0e929241834dfce0733070437/regex-2025.10.23-cp311-cp311-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:5c084889ab2c59765a0d5ac602fd1c3c244f9b3fcc9a65fdc7ba6b74c5287490", size = 907011 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/a4/fb/5ad6a8b92d3f88f3797b51bb4ef47499acc2d0b53d2fbe4487a892f37a73/regex-2025.10.23-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:d80e8eb79009bdb0936658c44ca06e2fbbca67792013e3818eea3f5f228971c2", size = 800312 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/42/48/b4efba0168a2b57f944205d823f8e8a3a1ae6211a34508f014ec2c712f4f/regex-2025.10.23-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:b6f259118ba87b814a8ec475380aee5f5ae97a75852a3507cf31d055b01b5b40", size = 782839 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/13/2a/c9efb4c6c535b0559c1fa8e431e0574d229707c9ca718600366fcfef6801/regex-2025.10.23-cp311-cp311-musllinux_1_2_ppc64le.whl", hash = "sha256:9b8c72a242683dcc72d37595c4f1278dfd7642b769e46700a8df11eab19dfd82", size = 854270 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/34/2d/68eecc1bdaee020e8ba549502291c9450d90d8590d0552247c9b543ebf7b/regex-2025.10.23-cp311-cp311-musllinux_1_2_s390x.whl", hash = "sha256:a8d7b7a0a3df9952f9965342159e0c1f05384c0f056a47ce8b61034f8cecbe83", size = 845771 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/a5/cd/a1ae499cf9b87afb47a67316bbf1037a7c681ffe447c510ed98c0aa2c01c/regex-2025.10.23-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:413bfea20a484c524858125e92b9ce6ffdd0a4b97d4ff96b5859aa119b0f1bdd", size = 788778 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/38/f9/70765e63f5ea7d43b2b6cd4ee9d3323f16267e530fb2a420d92d991cf0fc/regex-2025.10.23-cp311-cp311-win32.whl", hash = "sha256:f76deef1f1019a17dad98f408b8f7afc4bd007cbe835ae77b737e8c7f19ae575", size = 265666 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/9c/1a/18e9476ee1b63aaec3844d8e1cb21842dc19272c7e86d879bfc0dcc60db3/regex-2025.10.23-cp311-cp311-win_amd64.whl", hash = "sha256:59bba9f7125536f23fdab5deeea08da0c287a64c1d3acc1c7e99515809824de8", size = 277600 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/1d/1b/c019167b1f7a8ec77251457e3ff0339ed74ca8bce1ea13138dc98309c923/regex-2025.10.23-cp311-cp311-win_arm64.whl", hash = "sha256:b103a752b6f1632ca420225718d6ed83f6a6ced3016dd0a4ab9a6825312de566", size = 269974 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f6/57/eeb274d83ab189d02d778851b1ac478477522a92b52edfa6e2ae9ff84679/regex-2025.10.23-cp312-cp312-macosx_10_13_universal2.whl", hash = "sha256:7a44d9c00f7a0a02d3b777429281376370f3d13d2c75ae74eb94e11ebcf4a7fc", size = 489187 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/55/5c/7dad43a9b6ea88bf77e0b8b7729a4c36978e1043165034212fd2702880c6/regex-2025.10.23-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:b83601f84fde939ae3478bb32a3aef36f61b58c3208d825c7e8ce1a735f143f2", size = 291122 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/66/21/38b71e6f2818f0f4b281c8fba8d9d57cfca7b032a648fa59696e0a54376a/regex-2025.10.23-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:ec13647907bb9d15fd192bbfe89ff06612e098a5709e7d6ecabbdd8f7908fc45", size = 288797 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/be/95/888f069c89e7729732a6d7cca37f76b44bfb53a1e35dda8a2c7b65c1b992/regex-2025.10.23-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:78d76dd2957d62501084e7012ddafc5fcd406dd982b7a9ca1ea76e8eaaf73e7e", size = 798442 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/76/70/4f903c608faf786627a8ee17c06e0067b5acade473678b69c8094b248705/regex-2025.10.23-cp312-cp312-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:8668e5f067e31a47699ebb354f43aeb9c0ef136f915bd864243098524482ac43", size = 864039 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/62/19/2df67b526bf25756c7f447dde554fc10a220fd839cc642f50857d01e4a7b/regex-2025.10.23-cp312-cp312-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:a32433fe3deb4b2d8eda88790d2808fed0dc097e84f5e683b4cd4f42edef6cca", size = 912057 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/99/14/9a39b7c9e007968411bc3c843cc14cf15437510c0a9991f080cab654fd16/regex-2025.10.23-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:d97d73818c642c938db14c0668167f8d39520ca9d983604575ade3fda193afcc", size = 803374 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/d4/f7/3495151dd3ca79949599b6d069b72a61a2c5e24fc441dccc79dcaf708fe6/regex-2025.10.23-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:bca7feecc72ee33579e9f6ddf8babbe473045717a0e7dbc347099530f96e8b9a", size = 787714 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/28/65/ee882455e051131869957ee8597faea45188c9a98c0dad724cfb302d4580/regex-2025.10.23-cp312-cp312-musllinux_1_2_ppc64le.whl", hash = "sha256:7e24af51e907d7457cc4a72691ec458320b9ae67dc492f63209f01eecb09de32", size = 858392 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/53/25/9287fef5be97529ebd3ac79d256159cb709a07eb58d4be780d1ca3885da8/regex-2025.10.23-cp312-cp312-musllinux_1_2_s390x.whl", hash = "sha256:d10bcde58bbdf18146f3a69ec46dd03233b94a4a5632af97aa5378da3a47d288", size = 850484 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f3/b4/b49b88b4fea2f14dc73e5b5842755e782fc2e52f74423d6f4adc130d5880/regex-2025.10.23-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:44383bc0c933388516c2692c9a7503e1f4a67e982f20b9a29d2fb70c6494f147", size = 789634 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/b6/3c/2f8d199d0e84e78bcd6bdc2be9b62410624f6b796e2893d1837ae738b160/regex-2025.10.23-cp312-cp312-win32.whl", hash = "sha256:6040a86f95438a0114bba16e51dfe27f1bc004fd29fe725f54a586f6d522b079", size = 266060 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/d7/67/c35e80969f6ded306ad70b0698863310bdf36aca57ad792f45ddc0e2271f/regex-2025.10.23-cp312-cp312-win_amd64.whl", hash = "sha256:436b4c4352fe0762e3bfa34a5567079baa2ef22aa9c37cf4d128979ccfcad842", size = 276931 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f5/a1/4ed147de7d2b60174f758412c87fa51ada15cd3296a0ff047f4280aaa7ca/regex-2025.10.23-cp312-cp312-win_arm64.whl", hash = "sha256:f4b1b1991617055b46aff6f6db24888c1f05f4db9801349d23f09ed0714a9335", size = 270103 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/28/c6/195a6217a43719d5a6a12cc192a22d12c40290cecfa577f00f4fb822f07d/regex-2025.10.23-cp313-cp313-macosx_10_13_universal2.whl", hash = "sha256:b7690f95404a1293923a296981fd943cca12c31a41af9c21ba3edd06398fc193", size = 488956 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/4c/93/181070cd1aa2fa541ff2d3afcf763ceecd4937b34c615fa92765020a6c90/regex-2025.10.23-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:1a32d77aeaea58a13230100dd8797ac1a84c457f3af2fdf0d81ea689d5a9105b", size = 290997 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/b6/c5/9d37fbe3a40ed8dda78c23e1263002497540c0d1522ed75482ef6c2000f0/regex-2025.10.23-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:b24b29402f264f70a3c81f45974323b41764ff7159655360543b7cabb73e7d2f", size = 288686 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/5f/e7/db610ff9f10c2921f9b6ac0c8d8be4681b28ddd40fc0549429366967e61f/regex-2025.10.23-cp313-cp313-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:563824a08c7c03d96856d84b46fdb3bbb7cfbdf79da7ef68725cda2ce169c72a", size = 798466 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/90/10/aab883e1fa7fe2feb15ac663026e70ca0ae1411efa0c7a4a0342d9545015/regex-2025.10.23-cp313-cp313-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:a0ec8bdd88d2e2659c3518087ee34b37e20bd169419ffead4240a7004e8ed03b", size = 863996 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/a2/b0/8f686dd97a51f3b37d0238cd00a6d0f9ccabe701f05b56de1918571d0d61/regex-2025.10.23-cp313-cp313-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:b577601bfe1d33913fcd9276d7607bbac827c4798d9e14d04bf37d417a6c41cb", size = 912145 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/a3/ca/639f8cd5b08797bca38fc5e7e07f76641a428cf8c7fca05894caf045aa32/regex-2025.10.23-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:7c9f2c68ac6cb3de94eea08a437a75eaa2bd33f9e97c84836ca0b610a5804368", size = 803370 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/0d/1e/a40725bb76959eddf8abc42a967bed6f4851b39f5ac4f20e9794d7832aa5/regex-2025.10.23-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:89f8b9ea3830c79468e26b0e21c3585f69f105157c2154a36f6b7839f8afb351", size = 787767 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/3d/d8/8ee9858062936b0f99656dce390aa667c6e7fb0c357b1b9bf76fb5e2e708/regex-2025.10.23-cp313-cp313-musllinux_1_2_ppc64le.whl", hash = "sha256:98fd84c4e4ea185b3bb5bf065261ab45867d8875032f358a435647285c722673", size = 858335 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/d8/0a/ed5faaa63fa8e3064ab670e08061fbf09e3a10235b19630cf0cbb9e48c0a/regex-2025.10.23-cp313-cp313-musllinux_1_2_s390x.whl", hash = "sha256:1e11d3e5887b8b096f96b4154dfb902f29c723a9556639586cd140e77e28b313", size = 850402 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/79/14/d05f617342f4b2b4a23561da500ca2beab062bfcc408d60680e77ecaf04d/regex-2025.10.23-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:4f13450328a6634348d47a88367e06b64c9d84980ef6a748f717b13f8ce64e87", size = 789739 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f9/7b/e8ce8eef42a15f2c3461f8b3e6e924bbc86e9605cb534a393aadc8d3aff8/regex-2025.10.23-cp313-cp313-win32.whl", hash = "sha256:37be9296598a30c6a20236248cb8b2c07ffd54d095b75d3a2a2ee5babdc51df1", size = 266054 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/71/2d/55184ed6be6473187868d2f2e6a0708195fc58270e62a22cbf26028f2570/regex-2025.10.23-cp313-cp313-win_amd64.whl", hash = "sha256:ea7a3c283ce0f06fe789365841e9174ba05f8db16e2fd6ae00a02df9572c04c0", size = 276917 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/9c/d4/927eced0e2bd45c45839e556f987f8c8f8683268dd3c00ad327deb3b0172/regex-2025.10.23-cp313-cp313-win_arm64.whl", hash = "sha256:d9a4953575f300a7bab71afa4cd4ac061c7697c89590a2902b536783eeb49a4f", size = 270105 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/3e/b3/95b310605285573341fc062d1d30b19a54f857530e86c805f942c4ff7941/regex-2025.10.23-cp313-cp313t-macosx_10_13_universal2.whl", hash = "sha256:7d6606524fa77b3912c9ef52a42ef63c6cfbfc1077e9dc6296cd5da0da286044", size = 491850 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/a4/8f/207c2cec01e34e56db1eff606eef46644a60cf1739ecd474627db90ad90b/regex-2025.10.23-cp313-cp313t-macosx_10_13_x86_64.whl", hash = "sha256:c037aadf4d64bdc38af7db3dbd34877a057ce6524eefcb2914d6d41c56f968cc", size = 292537 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/98/3b/025240af4ada1dc0b5f10d73f3e5122d04ce7f8908ab8881e5d82b9d61b6/regex-2025.10.23-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:99018c331fb2529084a0c9b4c713dfa49fafb47c7712422e49467c13a636c656", size = 290904 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/81/8e/104ac14e2d3450c43db18ec03e1b96b445a94ae510b60138f00ce2cb7ca1/regex-2025.10.23-cp313-cp313t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:fd8aba965604d70306eb90a35528f776e59112a7114a5162824d43b76fa27f58", size = 807311 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/19/63/78aef90141b7ce0be8a18e1782f764f6997ad09de0e05251f0d2503a914a/regex-2025.10.23-cp313-cp313t-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:238e67264b4013e74136c49f883734f68656adf8257bfa13b515626b31b20f8e", size = 873241 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/b3/a8/80eb1201bb49ae4dba68a1b284b4211ed9daa8e74dc600018a10a90399fb/regex-2025.10.23-cp313-cp313t-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:b2eb48bd9848d66fd04826382f5e8491ae633de3233a3d64d58ceb4ecfa2113a", size = 914794 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f0/d5/1984b6ee93281f360a119a5ca1af6a8ca7d8417861671388bf750becc29b/regex-2025.10.23-cp313-cp313t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:d36591ce06d047d0c0fe2fc5f14bfbd5b4525d08a7b6a279379085e13f0e3d0e", size = 812581 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/c4/39/11ebdc6d9927172a64ae237d16763145db6bd45ebb4055c17b88edab72a7/regex-2025.10.23-cp313-cp313t-musllinux_1_2_aarch64.whl", hash = "sha256:b5d4ece8628d6e364302006366cea3ee887db397faebacc5dacf8ef19e064cf8", size = 795346 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/3b/b4/89a591bcc08b5e436af43315284bd233ba77daf0cf20e098d7af12f006c1/regex-2025.10.23-cp313-cp313t-musllinux_1_2_ppc64le.whl", hash = "sha256:39a7e8083959cb1c4ff74e483eecb5a65d3b3e1d821b256e54baf61782c906c6", size = 868214 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/3d/ff/58ba98409c1dbc8316cdb20dafbc63ed267380a07780cafecaf5012dabc9/regex-2025.10.23-cp313-cp313t-musllinux_1_2_s390x.whl", hash = "sha256:842d449a8fefe546f311656cf8c0d6729b08c09a185f1cad94c756210286d6a8", size = 854540 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/9a/f2/4a9e9338d67626e2071b643f828a482712ad15889d7268e11e9a63d6f7e9/regex-2025.10.23-cp313-cp313t-musllinux_1_2_x86_64.whl", hash = "sha256:d614986dc68506be8f00474f4f6960e03e4ca9883f7df47744800e7d7c08a494", size = 799346 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/63/be/543d35c46bebf6f7bf2be538cca74d6585f25714700c36f37f01b92df551/regex-2025.10.23-cp313-cp313t-win32.whl", hash = "sha256:a5b7a26b51a9df473ec16a1934d117443a775ceb7b39b78670b2e21893c330c9", size = 268657 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/14/9f/4dd6b7b612037158bb2c9bcaa710e6fb3c40ad54af441b9c53b3a137a9f1/regex-2025.10.23-cp313-cp313t-win_amd64.whl", hash = "sha256:ce81c5544a5453f61cb6f548ed358cfb111e3b23f3cd42d250a4077a6be2a7b6", size = 280075 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/81/7a/5bd0672aa65d38c8da6747c17c8b441bdb53d816c569e3261013af8e83cf/regex-2025.10.23-cp313-cp313t-win_arm64.whl", hash = "sha256:e9bf7f6699f490e4e43c44757aa179dab24d1960999c84ab5c3d5377714ed473", size = 271219 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/73/f6/0caf29fec943f201fbc8822879c99d31e59c1d51a983d9843ee5cf398539/regex-2025.10.23-cp314-cp314-macosx_10_13_universal2.whl", hash = "sha256:5b5cb5b6344c4c4c24b2dc87b0bfee78202b07ef7633385df70da7fcf6f7cec6", size = 488960 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/8e/7d/ebb7085b8fa31c24ce0355107cea2b92229d9050552a01c5d291c42aecea/regex-2025.10.23-cp314-cp314-macosx_10_13_x86_64.whl", hash = "sha256:a6ce7973384c37bdf0f371a843f95a6e6f4e1489e10e0cf57330198df72959c5", size = 290932 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/27/41/43906867287cbb5ca4cee671c3cc8081e15deef86a8189c3aad9ac9f6b4d/regex-2025.10.23-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:2ee3663f2c334959016b56e3bd0dd187cbc73f948e3a3af14c3caaa0c3035d10", size = 288766 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/ab/9e/ea66132776700fc77a39b1056e7a5f1308032fead94507e208dc6716b7cd/regex-2025.10.23-cp314-cp314-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:2003cc82a579107e70d013482acce8ba773293f2db534fb532738395c557ff34", size = 798884 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/d5/99/aed1453687ab63819a443930770db972c5c8064421f0d9f5da9ad029f26b/regex-2025.10.23-cp314-cp314-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:182c452279365a93a9f45874f7f191ec1c51e1f1eb41bf2b16563f1a40c1da3a", size = 864768 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/99/5d/732fe747a1304805eb3853ce6337eea16b169f7105a0d0dd9c6a5ffa9948/regex-2025.10.23-cp314-cp314-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:b1249e9ff581c5b658c8f0437f883b01f1edcf424a16388591e7c05e5e9e8b0c", size = 911394 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/5e/48/58a1f6623466522352a6efa153b9a3714fc559d9f930e9bc947b4a88a2c3/regex-2025.10.23-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:2b841698f93db3ccc36caa1900d2a3be281d9539b822dc012f08fc80b46a3224", size = 803145 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/ea/f6/7dea79be2681a5574ab3fc237aa53b2c1dfd6bd2b44d4640b6c76f33f4c1/regex-2025.10.23-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:956d89e0c92d471e8f7eee73f73fdff5ed345886378c45a43175a77538a1ffe4", size = 787831 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/3a/ad/07b76950fbbe65f88120ca2d8d845047c401450f607c99ed38862904671d/regex-2025.10.23-cp314-cp314-musllinux_1_2_ppc64le.whl", hash = "sha256:5c259cb363299a0d90d63b5c0d7568ee98419861618a95ee9d91a41cb9954462", size = 859162 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/41/87/374f3b2021b22aa6a4fc0b750d63f9721e53d1631a238f7a1c343c1cd288/regex-2025.10.23-cp314-cp314-musllinux_1_2_s390x.whl", hash = "sha256:185d2b18c062820b3a40d8fefa223a83f10b20a674bf6e8c4a432e8dfd844627", size = 849899 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/12/4a/7f7bb17c5a5a9747249807210e348450dab9212a46ae6d23ebce86ba6a2b/regex-2025.10.23-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:281d87fa790049c2b7c1b4253121edd80b392b19b5a3d28dc2a77579cb2a58ec", size = 789372 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/c9/dd/9c7728ff544fea09bbc8635e4c9e7c423b11c24f1a7a14e6ac4831466709/regex-2025.10.23-cp314-cp314-win32.whl", hash = "sha256:63b81eef3656072e4ca87c58084c7a9c2b81d41a300b157be635a8a675aacfb8", size = 271451 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/48/f8/ef7837ff858eb74079c4804c10b0403c0b740762e6eedba41062225f7117/regex-2025.10.23-cp314-cp314-win_amd64.whl", hash = "sha256:0967c5b86f274800a34a4ed862dfab56928144d03cb18821c5153f8777947796", size = 280173 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/8e/d0/d576e1dbd9885bfcd83d0e90762beea48d9373a6f7ed39170f44ed22e336/regex-2025.10.23-cp314-cp314-win_arm64.whl", hash = "sha256:c70dfe58b0a00b36aa04cdb0f798bf3e0adc31747641f69e191109fd8572c9a9", size = 273206 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/a6/d0/2025268315e8b2b7b660039824cb7765a41623e97d4cd421510925400487/regex-2025.10.23-cp314-cp314t-macosx_10_13_universal2.whl", hash = "sha256:1f5799ea1787aa6de6c150377d11afad39a38afd033f0c5247aecb997978c422", size = 491854 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/44/35/5681c2fec5e8b33454390af209c4353dfc44606bf06d714b0b8bd0454ffe/regex-2025.10.23-cp314-cp314t-macosx_10_13_x86_64.whl", hash = "sha256:a9639ab7540cfea45ef57d16dcbea2e22de351998d614c3ad2f9778fa3bdd788", size = 292542 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/5d/17/184eed05543b724132e4a18149e900f5189001fcfe2d64edaae4fbaf36b4/regex-2025.10.23-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:08f52122c352eb44c3421dab78b9b73a8a77a282cc8314ae576fcaa92b780d10", size = 290903 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/25/d0/5e3347aa0db0de382dddfa133a7b0ae72f24b4344f3989398980b44a3924/regex-2025.10.23-cp314-cp314t-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl", hash = "sha256:ebf1baebef1c4088ad5a5623decec6b52950f0e4d7a0ae4d48f0a99f8c9cb7d7", size = 807546 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/d2/bb/40c589bbdce1be0c55e9f8159789d58d47a22014f2f820cf2b517a5cd193/regex-2025.10.23-cp314-cp314t-manylinux2014_ppc64le.manylinux_2_17_ppc64le.manylinux_2_28_ppc64le.whl", hash = "sha256:16b0f1c2e2d566c562d5c384c2b492646be0a19798532fdc1fdedacc66e3223f", size = 873322 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/fe/56/a7e40c01575ac93360e606278d359f91829781a9f7fb6e5aa435039edbda/regex-2025.10.23-cp314-cp314t-manylinux2014_s390x.manylinux_2_17_s390x.manylinux_2_28_s390x.whl", hash = "sha256:f7ada5d9dceafaab92646aa00c10a9efd9b09942dd9b0d7c5a4b73db92cc7e61", size = 914855 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/5c/4b/d55587b192763db3163c3f508b3b67b31bb6f5e7a0e08b83013d0a59500a/regex-2025.10.23-cp314-cp314t-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl", hash = "sha256:3a36b4005770044bf08edecc798f0e41a75795b9e7c9c12fe29da8d792ef870c", size = 812724 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/33/20/18bac334955fbe99d17229f4f8e98d05e4a501ac03a442be8facbb37c304/regex-2025.10.23-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:af7b2661dcc032da1fae82069b5ebf2ac1dfcd5359ef8b35e1367bfc92181432", size = 795439 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/67/46/c57266be9df8549c7d85deb4cb82280cb0019e46fff677534c5fa1badfa4/regex-2025.10.23-cp314-cp314t-musllinux_1_2_ppc64le.whl", hash = "sha256:1cb976810ac1416a67562c2e5ba0accf6f928932320fef302e08100ed681b38e", size = 868336 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/b8/f3/bd5879e41ef8187fec5e678e94b526a93f99e7bbe0437b0f2b47f9101694/regex-2025.10.23-cp314-cp314t-musllinux_1_2_s390x.whl", hash = "sha256:1a56a54be3897d62f54290190fbcd754bff6932934529fbf5b29933da28fcd43", size = 854567 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/e6/57/2b6bbdbd2f24dfed5b028033aa17ad8f7d86bb28f1a892cac8b3bc89d059/regex-2025.10.23-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:8f3e6d202fb52c2153f532043bbcf618fd177df47b0b306741eb9b60ba96edc3", size = 799565 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/c7/ba/a6168f542ba73b151ed81237adf6b869c7b2f7f8d51618111296674e20ee/regex-2025.10.23-cp314-cp314t-win32.whl", hash = "sha256:1fa1186966b2621b1769fd467c7b22e317e6ba2d2cdcecc42ea3089ef04a8521", size = 274428 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/ef/a0/c84475e14a2829e9b0864ebf77c3f7da909df9d8acfe2bb540ff0072047c/regex-2025.10.23-cp314-cp314t-win_amd64.whl", hash = "sha256:08a15d40ce28362eac3e78e83d75475147869c1ff86bc93285f43b4f4431a741", size = 284140 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/51/33/6a08ade0eee5b8ba79386869fa6f77afeb835b60510f3525db987e2fffc4/regex-2025.10.23-cp314-cp314t-win_arm64.whl", hash = "sha256:a93e97338e1c8ea2649e130dcfbe8cd69bba5e1e163834752ab64dcb4de6d5ed", size = 274497 },
+]
+
 [[package]]
 name = "requests"
 version = "2.32.5"
@@ -819,6 +928,67 @@ wheels = [
     { url = "https://pypi.tuna.tsinghua.edu.cn/packages/be/72/2db2f49247d0a18b4f1bb9a5a39a0162869acf235f3a96418363947b3d46/starlette-0.48.0-py3-none-any.whl", hash = "sha256:0764ca97b097582558ecb498132ed0c7d942f233f365b86ba37770e026510659", size = 73736 },
 ]
 
+[[package]]
+name = "tiktoken"
+version = "0.12.0"
+source = { registry = "https://pypi.tuna.tsinghua.edu.cn/simple" }
+dependencies = [
+    { name = "regex" },
+    { name = "requests" },
+]
+sdist = { url = "https://pypi.tuna.tsinghua.edu.cn/packages/7d/ab/4d017d0f76ec3171d469d80fc03dfbb4e48a4bcaddaa831b31d526f05edc/tiktoken-0.12.0.tar.gz", hash = "sha256:b18ba7ee2b093863978fcb14f74b3707cdc8d4d4d3836853ce7ec60772139931", size = 37806 }
+wheels = [
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/89/b3/2cb7c17b6c4cf8ca983204255d3f1d95eda7213e247e6947a0ee2c747a2c/tiktoken-0.12.0-cp310-cp310-macosx_10_12_x86_64.whl", hash = "sha256:3de02f5a491cfd179aec916eddb70331814bd6bf764075d39e21d5862e533970", size = 1051991 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/27/0f/df139f1df5f6167194ee5ab24634582ba9a1b62c6b996472b0277ec80f66/tiktoken-0.12.0-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:b6cfb6d9b7b54d20af21a912bfe63a2727d9cfa8fbda642fd8322c70340aad16", size = 995798 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/ef/5d/26a691f28ab220d5edc09b9b787399b130f24327ef824de15e5d85ef21aa/tiktoken-0.12.0-cp310-cp310-manylinux_2_28_aarch64.whl", hash = "sha256:cde24cdb1b8a08368f709124f15b36ab5524aac5fa830cc3fdce9c03d4fb8030", size = 1129865 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/b2/94/443fab3d4e5ebecac895712abd3849b8da93b7b7dec61c7db5c9c7ebe40c/tiktoken-0.12.0-cp310-cp310-manylinux_2_28_x86_64.whl", hash = "sha256:6de0da39f605992649b9cfa6f84071e3f9ef2cec458d08c5feb1b6f0ff62e134", size = 1152856 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/54/35/388f941251b2521c70dd4c5958e598ea6d2c88e28445d2fb8189eecc1dfc/tiktoken-0.12.0-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:6faa0534e0eefbcafaccb75927a4a380463a2eaa7e26000f0173b920e98b720a", size = 1195308 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f8/00/c6681c7f833dd410576183715a530437a9873fa910265817081f65f9105f/tiktoken-0.12.0-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:82991e04fc860afb933efb63957affc7ad54f83e2216fe7d319007dab1ba5892", size = 1255697 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/5f/d2/82e795a6a9bafa034bf26a58e68fe9a89eeaaa610d51dbeb22106ba04f0a/tiktoken-0.12.0-cp310-cp310-win_amd64.whl", hash = "sha256:6fb2995b487c2e31acf0a9e17647e3b242235a20832642bb7a9d1a181c0c1bb1", size = 879375 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/de/46/21ea696b21f1d6d1efec8639c204bdf20fde8bafb351e1355c72c5d7de52/tiktoken-0.12.0-cp311-cp311-macosx_10_12_x86_64.whl", hash = "sha256:6e227c7f96925003487c33b1b32265fad2fbcec2b7cf4817afb76d416f40f6bb", size = 1051565 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/c9/d9/35c5d2d9e22bb2a5f74ba48266fb56c63d76ae6f66e02feb628671c0283e/tiktoken-0.12.0-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:c06cf0fcc24c2cb2adb5e185c7082a82cba29c17575e828518c2f11a01f445aa", size = 995284 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/01/84/961106c37b8e49b9fdcf33fe007bb3a8fdcc380c528b20cc7fbba80578b8/tiktoken-0.12.0-cp311-cp311-manylinux_2_28_aarch64.whl", hash = "sha256:f18f249b041851954217e9fd8e5c00b024ab2315ffda5ed77665a05fa91f42dc", size = 1129201 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/6a/d0/3d9275198e067f8b65076a68894bb52fd253875f3644f0a321a720277b8a/tiktoken-0.12.0-cp311-cp311-manylinux_2_28_x86_64.whl", hash = "sha256:47a5bc270b8c3db00bb46ece01ef34ad050e364b51d406b6f9730b64ac28eded", size = 1152444 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/78/db/a58e09687c1698a7c592e1038e01c206569b86a0377828d51635561f8ebf/tiktoken-0.12.0-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:508fa71810c0efdcd1b898fda574889ee62852989f7c1667414736bcb2b9a4bd", size = 1195080 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/9e/1b/a9e4d2bf91d515c0f74afc526fd773a812232dd6cda33ebea7f531202325/tiktoken-0.12.0-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:a1af81a6c44f008cba48494089dd98cccb8b313f55e961a52f5b222d1e507967", size = 1255240 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/9d/15/963819345f1b1fb0809070a79e9dd96938d4ca41297367d471733e79c76c/tiktoken-0.12.0-cp311-cp311-win_amd64.whl", hash = "sha256:3e68e3e593637b53e56f7237be560f7a394451cb8c11079755e80ae64b9e6def", size = 879422 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/a4/85/be65d39d6b647c79800fd9d29241d081d4eeb06271f383bb87200d74cf76/tiktoken-0.12.0-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:b97f74aca0d78a1ff21b8cd9e9925714c15a9236d6ceacf5c7327c117e6e21e8", size = 1050728 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/4a/42/6573e9129bc55c9bf7300b3a35bef2c6b9117018acca0dc760ac2d93dffe/tiktoken-0.12.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:2b90f5ad190a4bb7c3eb30c5fa32e1e182ca1ca79f05e49b448438c3e225a49b", size = 994049 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/66/c5/ed88504d2f4a5fd6856990b230b56d85a777feab84e6129af0822f5d0f70/tiktoken-0.12.0-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:65b26c7a780e2139e73acc193e5c63ac754021f160df919add909c1492c0fb37", size = 1129008 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f4/90/3dae6cc5436137ebd38944d396b5849e167896fc2073da643a49f372dc4f/tiktoken-0.12.0-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:edde1ec917dfd21c1f2f8046b86348b0f54a2c0547f68149d8600859598769ad", size = 1152665 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/a3/fe/26df24ce53ffde419a42f5f53d755b995c9318908288c17ec3f3448313a3/tiktoken-0.12.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:35a2f8ddd3824608b3d650a000c1ef71f730d0c56486845705a8248da00f9fe5", size = 1194230 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/20/cc/b064cae1a0e9fac84b0d2c46b89f4e57051a5f41324e385d10225a984c24/tiktoken-0.12.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:83d16643edb7fa2c99eff2ab7733508aae1eebb03d5dfc46f5565862810f24e3", size = 1254688 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/81/10/b8523105c590c5b8349f2587e2fdfe51a69544bd5a76295fc20f2374f470/tiktoken-0.12.0-cp312-cp312-win_amd64.whl", hash = "sha256:ffc5288f34a8bc02e1ea7047b8d041104791d2ddbf42d1e5fa07822cbffe16bd", size = 878694 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/00/61/441588ee21e6b5cdf59d6870f86beb9789e532ee9718c251b391b70c68d6/tiktoken-0.12.0-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:775c2c55de2310cc1bc9a3ad8826761cbdc87770e586fd7b6da7d4589e13dab3", size = 1050802 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/1f/05/dcf94486d5c5c8d34496abe271ac76c5b785507c8eae71b3708f1ad9b45a/tiktoken-0.12.0-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:a01b12f69052fbe4b080a2cfb867c4de12c704b56178edf1d1d7b273561db160", size = 993995 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/a0/70/5163fe5359b943f8db9946b62f19be2305de8c3d78a16f629d4165e2f40e/tiktoken-0.12.0-cp313-cp313-manylinux_2_28_aarch64.whl", hash = "sha256:01d99484dc93b129cd0964f9d34eee953f2737301f18b3c7257bf368d7615baa", size = 1128948 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/0c/da/c028aa0babf77315e1cef357d4d768800c5f8a6de04d0eac0f377cb619fa/tiktoken-0.12.0-cp313-cp313-manylinux_2_28_x86_64.whl", hash = "sha256:4a1a4fcd021f022bfc81904a911d3df0f6543b9e7627b51411da75ff2fe7a1be", size = 1151986 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/a0/5a/886b108b766aa53e295f7216b509be95eb7d60b166049ce2c58416b25f2a/tiktoken-0.12.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:981a81e39812d57031efdc9ec59fa32b2a5a5524d20d4776574c4b4bd2e9014a", size = 1194222 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f4/f8/4db272048397636ac7a078d22773dd2795b1becee7bc4922fe6207288d57/tiktoken-0.12.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:9baf52f84a3f42eef3ff4e754a0db79a13a27921b457ca9832cf944c6be4f8f3", size = 1255097 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/8e/32/45d02e2e0ea2be3a9ed22afc47d93741247e75018aac967b713b2941f8ea/tiktoken-0.12.0-cp313-cp313-win_amd64.whl", hash = "sha256:b8a0cd0c789a61f31bf44851defbd609e8dd1e2c8589c614cc1060940ef1f697", size = 879117 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/ce/76/994fc868f88e016e6d05b0da5ac24582a14c47893f4474c3e9744283f1d5/tiktoken-0.12.0-cp313-cp313t-macosx_10_13_x86_64.whl", hash = "sha256:d5f89ea5680066b68bcb797ae85219c72916c922ef0fcdd3480c7d2315ffff16", size = 1050309 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f6/b8/57ef1456504c43a849821920d582a738a461b76a047f352f18c0b26c6516/tiktoken-0.12.0-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:b4e7ed1c6a7a8a60a3230965bdedba8cc58f68926b835e519341413370e0399a", size = 993712 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/72/90/13da56f664286ffbae9dbcfadcc625439142675845baa62715e49b87b68b/tiktoken-0.12.0-cp313-cp313t-manylinux_2_28_aarch64.whl", hash = "sha256:fc530a28591a2d74bce821d10b418b26a094bf33839e69042a6e86ddb7a7fb27", size = 1128725 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/05/df/4f80030d44682235bdaecd7346c90f67ae87ec8f3df4a3442cb53834f7e4/tiktoken-0.12.0-cp313-cp313t-manylinux_2_28_x86_64.whl", hash = "sha256:06a9f4f49884139013b138920a4c393aa6556b2f8f536345f11819389c703ebb", size = 1151875 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/22/1f/ae535223a8c4ef4c0c1192e3f9b82da660be9eb66b9279e95c99288e9dab/tiktoken-0.12.0-cp313-cp313t-musllinux_1_2_aarch64.whl", hash = "sha256:04f0e6a985d95913cabc96a741c5ffec525a2c72e9df086ff17ebe35985c800e", size = 1194451 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/78/a7/f8ead382fce0243cb625c4f266e66c27f65ae65ee9e77f59ea1653b6d730/tiktoken-0.12.0-cp313-cp313t-musllinux_1_2_x86_64.whl", hash = "sha256:0ee8f9ae00c41770b5f9b0bb1235474768884ae157de3beb5439ca0fd70f3e25", size = 1253794 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/93/e0/6cc82a562bc6365785a3ff0af27a2a092d57c47d7a81d9e2295d8c36f011/tiktoken-0.12.0-cp313-cp313t-win_amd64.whl", hash = "sha256:dc2dd125a62cb2b3d858484d6c614d136b5b848976794edfb63688d539b8b93f", size = 878777 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/72/05/3abc1db5d2c9aadc4d2c76fa5640134e475e58d9fbb82b5c535dc0de9b01/tiktoken-0.12.0-cp314-cp314-macosx_10_13_x86_64.whl", hash = "sha256:a90388128df3b3abeb2bfd1895b0681412a8d7dc644142519e6f0a97c2111646", size = 1050188 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/e3/7b/50c2f060412202d6c95f32b20755c7a6273543b125c0985d6fa9465105af/tiktoken-0.12.0-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:da900aa0ad52247d8794e307d6446bd3cdea8e192769b56276695d34d2c9aa88", size = 993978 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/14/27/bf795595a2b897e271771cd31cb847d479073497344c637966bdf2853da1/tiktoken-0.12.0-cp314-cp314-manylinux_2_28_aarch64.whl", hash = "sha256:285ba9d73ea0d6171e7f9407039a290ca77efcdb026be7769dccc01d2c8d7fff", size = 1129271 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f5/de/9341a6d7a8f1b448573bbf3425fa57669ac58258a667eb48a25dfe916d70/tiktoken-0.12.0-cp314-cp314-manylinux_2_28_x86_64.whl", hash = "sha256:d186a5c60c6a0213f04a7a802264083dea1bbde92a2d4c7069e1a56630aef830", size = 1151216 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/75/0d/881866647b8d1be4d67cb24e50d0c26f9f807f994aa1510cb9ba2fe5f612/tiktoken-0.12.0-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:604831189bd05480f2b885ecd2d1986dc7686f609de48208ebbbddeea071fc0b", size = 1194860 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/b3/1e/b651ec3059474dab649b8d5b69f5c65cd8fcd8918568c1935bd4136c9392/tiktoken-0.12.0-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:8f317e8530bb3a222547b85a58583238c8f74fd7a7408305f9f63246d1a0958b", size = 1254567 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/80/57/ce64fd16ac390fafde001268c364d559447ba09b509181b2808622420eec/tiktoken-0.12.0-cp314-cp314-win_amd64.whl", hash = "sha256:399c3dd672a6406719d84442299a490420b458c44d3ae65516302a99675888f3", size = 921067 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/ac/a4/72eed53e8976a099539cdd5eb36f241987212c29629d0a52c305173e0a68/tiktoken-0.12.0-cp314-cp314t-macosx_10_13_x86_64.whl", hash = "sha256:c2c714c72bc00a38ca969dae79e8266ddec999c7ceccd603cc4f0d04ccd76365", size = 1050473 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/e6/d7/0110b8f54c008466b19672c615f2168896b83706a6611ba6e47313dbc6e9/tiktoken-0.12.0-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:cbb9a3ba275165a2cb0f9a83f5d7025afe6b9d0ab01a22b50f0e74fee2ad253e", size = 993855 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/5f/77/4f268c41a3957c418b084dd576ea2fad2e95da0d8e1ab705372892c2ca22/tiktoken-0.12.0-cp314-cp314t-manylinux_2_28_aarch64.whl", hash = "sha256:dfdfaa5ffff8993a3af94d1125870b1d27aed7cb97aa7eb8c1cefdbc87dbee63", size = 1129022 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/4e/2b/fc46c90fe5028bd094cd6ee25a7db321cb91d45dc87531e2bdbb26b4867a/tiktoken-0.12.0-cp314-cp314t-manylinux_2_28_x86_64.whl", hash = "sha256:584c3ad3d0c74f5269906eb8a659c8bfc6144a52895d9261cdaf90a0ae5f4de0", size = 1150736 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/28/c0/3c7a39ff68022ddfd7d93f3337ad90389a342f761c4d71de99a3ccc57857/tiktoken-0.12.0-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:54c891b416a0e36b8e2045b12b33dd66fb34a4fe7965565f1b482da50da3e86a", size = 1194908 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/ab/0d/c1ad6f4016a3968c048545f5d9b8ffebf577774b2ede3e2e352553b685fe/tiktoken-0.12.0-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:5edb8743b88d5be814b1a8a8854494719080c28faaa1ccbef02e87354fe71ef0", size = 1253706 },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/af/df/c7891ef9d2712ad774777271d39fdef63941ffba0a9d59b7ad1fd2765e57/tiktoken-0.12.0-cp314-cp314t-win_amd64.whl", hash = "sha256:f61c0aea5565ac82e2ec50a05e02a6c44734e91b51c10510b084ea1b8e633a71", size = 920667 },
+]
+
 [[package]]
 name = "tomli"
 version = "2.3.0"