LightAgent 0.2.6__tar.gz → 0.2.9__tar.gz

{lightagent-0.2.6 → lightagent-0.2.9}/LightAgent/la_core.py

@@ -11,13 +11,16 @@ import logging
 import os
 import httpx
 import importlib
+from openai.types.chat import ChatCompletionChunk
+import asyncio
 
 # 全局工具注册表
 _FUNCTION_MAPPINGS = {}  # 工具名称 -> 工具函数
+_FUNCTION_INFO = {}   # 工具名称 -> 工具info信息
 _OPENAI_FUNCTION_SCHEMAS = []  # OpenAI 格式的工具描述
 _PROMPT_FUNCTION_SCHEMAS = []  # prompt 格式的工具描述
 
-__version__ = "0.2.1"  # 你可以根据需要设置版本号
+__version__ = "0.2.9"  # 你可以根据需要设置版本号
 
 
 def register_tool_manually(tools: List[Union[str, Callable]]) -> bool:
@@ -32,6 +35,9 @@ def register_tool_manually(tools: List[Union[str, Callable]]) -> bool:
 
         tool_info = func.tool_info
         tool_name = tool_info["tool_name"]
+
+        # 注册到全局字典
+        _FUNCTION_INFO[tool_name] = tool_info
         _FUNCTION_MAPPINGS[tool_name] = func  # 注册工具
 
         # 构建 OpenAI 格式的工具描述
@@ -82,36 +88,41 @@ def load_tool(tool_name: str, tools_directory: str = "tools"):
             return tool_func
     raise AttributeError(f"Tool '{tool_name}' is not properly defined in {tool_path}")
 
+from typing import Dict, Any, Union, Generator, AsyncGenerator
+import inspect
+import traceback
 
-def dispatch_tool(tool_name: str, tool_params: Dict[str, Any]) -> Union[str, Generator[str, None, None]]:
+async def dispatch_tool(tool_name: str, tool_params: Dict[str, Any]) -> Union[str, Generator[str, None, None], AsyncGenerator[str, None]]:
     """
-    调用工具执行，支持流式输出。
-
-    :param tool_name: 工具名称
-    :param tool_params: 工具参数
-    :return: 如果工具是流式输出，返回生成器；否则返回字符串结果。
+    调用工具执行，支持同步/异步工具及流式输出。
     """
     if tool_name not in _FUNCTION_MAPPINGS:
         return f"Tool `{tool_name}` not found."
 
     tool_call = _FUNCTION_MAPPINGS[tool_name]
     try:
-        # print(f"Calling tool: {tool_name} with params: {tool_params}")  # 调试信息
-        result = tool_call(**tool_params)
+        # 区分同步/异步工具
+        if inspect.iscoroutinefunction(tool_call):
+            result = await tool_call(**tool_params)
+        else:
+            result = tool_call(**tool_params)
 
-        # 如果工具返回的是可迭代对象（如生成器），则逐块处理并返回生成器
-        if isinstance(result, Iterable) and not isinstance(result, (str, bytes)):
+        # 处理不同类型的流式输出
+        if inspect.isasyncgen(result):
+            return async_stream_generator(result)
+        elif inspect.isgenerator(result):
             return stream_generator(result)
-        # 否则，返回字符串结果
         else:
             return str(result)
     except Exception as e:
-        # print(f"Tool call failed: {e}")  # 调试信息
         return traceback.format_exc()
 
+async def async_stream_generator(async_gen: AsyncGenerator) -> AsyncGenerator[str, None]:
+    async for chunk in async_gen:
+        yield chunk
 
-def stream_generator(result):
-    for chunk in result:
+def stream_generator(sync_gen: Generator) -> Generator[str, None, None]:
+    for chunk in sync_gen:
         yield chunk
 
 
@@ -150,7 +161,7 @@ def get_tools_str() -> str:
 
 
 class LightAgent:
-    __version__ = "0.2.1"  # 将版本号放在类中
+    __version__ = "0.2.9"  # 将版本号放在类中
 
     def __init__(
             self,
@@ -159,11 +170,14 @@ class LightAgent:
             instructions: Optional[str] = None,  # 代理指令
             role: Optional[str] = None,
             model: str,
-            api_key: str,
+            api_key: str | None = None,
             base_url: str | httpx.URL | None = None,
             websocket_base_url: str | httpx.URL | None = None,
             memory=None,  # 支持外部传入记忆模块
             tree_of_thought: bool = False,  # 是否启用链式思考
+            tot_model: str | None = None,
+            tot_api_key: str | None = None,
+            tot_base_url: str | httpx.URL | None = None,
             self_learning: bool = False,  # 是否启用agent自我学习
             tools: List[Union[str, Callable]] = None,  # 支持混合输入
             debug: bool = False,  # 是否启用调试模式
@@ -182,6 +196,9 @@ class LightAgent:
         :param websocket_base_url: WebSocket 的基础 URL。
         :param memory: 外部传入的记忆模块，需实现 `retrieve` 和 `store` 方法。
         :param tree_of_thought: 是否启用思维链功能。
+        :param tot_model: 使用的模型名称。
+        :param tot_api_key: API 密钥。
+        :param tot_base_url: API 的基础 URL。
         :param tools: 工具列表，支持函数名称（字符串）或函数对象。
         :param debug: 是否启用调试模式。
         :param log_level: 日志级别（INFO, DEBUG, ERROR）。
@@ -236,9 +253,21 @@ class LightAgent:
             base_url = f"https://api.openai.com/v1"
 
         self.client = OpenAI(
-            base_url=base_url,
-            api_key=self.api_key
+            base_url = base_url,
+            api_key = self.api_key
         )
+        if self.tree_of_thought:
+            if tot_api_key is None:
+                tot_api_key = api_key
+            if tot_base_url is None:
+                tot_base_url = base_url
+            if not tot_model:
+                tot_model = "deepseek-r1"  # 默认思维推理模型为deepseek-r1
+            self.tot_model = tot_model
+            self.tot_client = OpenAI(
+                base_url = tot_base_url,
+                api_key = tot_api_key
+            )
 
     def get_tool(self, tool_name: str) -> Callable:
         """
@@ -272,6 +301,7 @@ class LightAgent:
                 # 注册工具函数
                 if hasattr(tool_func, "tool_info"):
                     tool_info = tool_func.tool_info
+                    _FUNCTION_INFO[tool_name] = tool_info  # 注册工具info信息
                     _FUNCTION_MAPPINGS[tool_name] = tool_func
 
                     # 构建 OpenAI 格式的工具描述
@@ -358,7 +388,7 @@ class LightAgent:
             query: str,
             light_swarm=None,
             stream: bool = False,
-            max_retry: int = 5,
+            max_retry: int = 10,
             user_id: str = "default_user",
             history: list = None,
             metadata: Optional[Dict] = None,
@@ -398,7 +428,6 @@ class LightAgent:
                 return result
 
         # 2. 正常处理任务
-
         now = datetime.now()
         current_date = now.strftime("%Y-%m-%d")
         current_time = now.strftime("%H:%M:%S")
@@ -456,20 +485,25 @@ class LightAgent:
                 output = ""
                 function_call_name = ""
                 tool_calls = response.choices[0].message.tool_calls
-                self.log("DEBUG", "tool_calls", {"tool_calls": tool_calls})
+                self.log("DEBUG", "non_stream tool_calls", {"tool_calls": tool_calls})
 
                 # 遍历所有工具调用
                 for tool_call in tool_calls:
                     function_call = tool_call.function
-                    self.log("DEBUG", "function_call", {"function_call": function_call.model_dump()})
 
+                    # 尝试自动修复常见转义问题
+                    fixed_args = function_call.arguments.replace('\\"', '"').replace('\\\\', '\\')
+                    self.log("DEBUG", "non_stream function_call", {"function_call": fixed_args})
+
+                    function_args = json.loads(fixed_args)
                     # 解析函数参数
-                    function_args = json.loads(function_call.arguments)
+                    # function_args = json.loads(function_call.arguments)
 
                     # 调用工具并获取响应
-                    tool_response = dispatch_tool(function_call.name, function_args)
+                    tool_response = asyncio.run(dispatch_tool(function_call.name, function_args))
                     function_call_name = function_call.name
                     combined_response = ""
+                    single_tool_response = ""
 
                     # 如果工具返回的是生成器（流式输出），则将所有 chunk 叠加
                     if isinstance(tool_response, Generator):
@@ -483,26 +517,48 @@ class LightAgent:
                                 combined_response += chunk  # 将每个 chunk 叠加
                         if combined_response == "":
                             combined_response = "".join(tool_response)
-                        tool_responses.append(combined_response)  # 将叠加后的完整响应添加到列表
+
+                        # 将 combined_response 解析为 JSON 对象（如果它是 JSON 字符串）
+                        try:
+                            combined_response = json.loads(combined_response)  # 解析 JSON
+                        except json.JSONDecodeError:
+                            pass  # 如果不是 JSON 字符串，保持原样
+
+                        # 将 JSON 对象中的 Unicode 编码转换为中文字符
+                        if isinstance(combined_response, dict):
+                            combined_response = json.dumps(combined_response, ensure_ascii=False)  # 确保中文字符不转义
+                        single_tool_response = combined_response  # 处理单个工具的方法
+
                     else:
                         # print(f"Non-streaming response from tool: {function_call.name}")
                         combined_response = tool_response
-                        tool_responses.append(combined_response)  # 直接添加普通响应
+                        # print("tool_response type:",type(combined_response))
+                        # 如果是 JSON 字符串，解析并转换为中文
+                        if isinstance(combined_response, str):
+                            try:
+                                combined_response = json.loads(combined_response)  # 解析 JSON
+                                combined_response = json.dumps(combined_response, ensure_ascii=False)  # 转换为中文
+                            except json.JSONDecodeError:
+                                combined_response = tool_response
+                                pass  # 如果不是 JSON 字符串，保持原样
+                        single_tool_response = combined_response  # 处理单个工具的方法
 
-                    self.log("INFO", "tool_response", {"tool_response": combined_response})
 
-                    # 将工具调用的结果添加到列表中
-                    tool_responses.append(combined_response)
+                    self.log("DEBUG", "non_stream single_tool_response", {"single_tool_response": single_tool_response})
+
+                    # 将单个工具的响应结果添加到列表中
+                    tool_responses.append(single_tool_response)
 
                 # 将所有工具调用的结果合并为一个字符串
+                self.log("DEBUG", "non_stream tool_responses", {"tool_responses": tool_responses})
+
                 combined_tool_response = "\n".join(tool_responses)
 
                 # 将工具调用和响应添加到消息列表中
                 params["messages"].append(
                     {
                         "role": "assistant",
-                        "content": json.dumps(
-                            [tool_call.function.model_dump() for tool_call in tool_calls]),
+                        "content": f"使用工具： \n {json.dumps([tool_call.function.model_dump() for tool_call in tool_calls],ensure_ascii=False)}\n",
                     }
                 )
                 params["messages"].append(
@@ -520,7 +576,13 @@ class LightAgent:
             # 更新响应
             if function_call_name == 'finish':
                 return  # 如果最后调用了finish工具，则结束生成器
-            response = self.client.chat.completions.create(**params)
+            # print("params:",params)
+            self.log("DEBUG", "chat-completions params", {"params": params})
+
+            try:
+                response = self.client.chat.completions.create(**params)
+            except Exception as e:
+                print(f"An error occurred: {e}")
 
         # 重试次数用尽
         self.log("ERROR", "max_retry_reached", {"message": "Failed to generate a valid response."})
@@ -583,6 +645,7 @@ class LightAgent:
                         if tool_call["name"]:  # 确保工具调用有名称
                             function_call = {
                                 "name": tool_call["name"],
+                                "title": _FUNCTION_INFO.get(tool_call["name"], {}).get('tool_title') or '',
                                 "arguments": tool_call["arguments"],
                             }
                             self.log("INFO", "tool_call", {"function_call": function_call})
@@ -602,24 +665,42 @@ class LightAgent:
 
                                 for json_obj in json_objects:
                                     function_args = json.loads(json_obj)
-                                    tool_response = dispatch_tool(function_call["name"], function_args)
+                                    # tool_response = dispatch_tool(function_call["name"], function_args)
+                                    tool_response = asyncio.run(dispatch_tool(function_call["name"], function_args))
                                     function_call_name = function_call["name"]
+                                    combined_response = ""
+                                    single_tool_response = ""
 
                                     # 如果工具返回的是生成器（流式输出），则将所有 chunk 叠加
                                     if isinstance(tool_response, Generator):
                                         # print(f"Streaming response from tool: {function_call['name']}")
-                                        combined_response = ""
                                         for chunk in tool_response:
-                                            yield chunk
+                                            # 将工具返回的数据继续流出
+                                            if isinstance(chunk, ChatCompletionChunk):
+                                                yield chunk
+                                            else:
+                                                tool_output = {
+                                                    "name": tool_call["name"],
+                                                    "title": _FUNCTION_INFO.get(tool_call["name"], {}).get(
+                                                        'tool_title') or '',
+                                                    "output": chunk,
+                                                }
+                                                self.log("INFO", "tool_call", {"tool_output": tool_output})
+                                                yield tool_output
+                                            # 将工具的调用信息推送给开发者
                                             if function_call_name == 'finish':
                                                 content = chunk.choices[0].delta.content or ""
                                                 combined_response += content  # 将每个 chunk 叠加
                                             else:
                                                 combined_response += chunk  # 将每个 chunk 叠加
-                                        tool_responses.append(combined_response)  # 将叠加后的完整响应添加到列表
+                                        single_tool_response = combined_response  # 处理单个工具的方法
                                     else:
-                                        # print(f"Non-streaming response from tool: {function_call['name']}")
-                                        tool_responses.append(tool_response)  # 直接添加普通响应
+                                        # print(f"Non-streaming response from tool: {tool_response}")
+                                        combined_response = tool_response
+                                        single_tool_response = combined_response  # 处理单个工具的方法
+                                    self.log("INFO", "stream single_tool_response", {"single_tool_response": single_tool_response})
+                                    # 将单个工具的响应结果添加到列表中
+                                    tool_responses.append(single_tool_response)
 
                             except json.JSONDecodeError as e:
                                 self.log("ERROR", "json_decode_error",
@@ -651,6 +732,7 @@ class LightAgent:
             # 更新响应
             if function_call_name == 'finish':
                 return  # 如果最后调用了finish工具，则结束生成器
+            self.log("DEBUG", "chat-completions params", {"params": params})
             response = self.client.chat.completions.create(**params)
 
         # 重试次数用尽
@@ -788,18 +870,21 @@ class LightAgent:
 
     def run_thought(self, query: str, stream=False, tools=None):
         """使用思维树的方式 让大模型先根据get_tools_str生成一个解答用户query的工具使用计划"""
-        tot_model = "deepseek-chat"  # self.model
+        tot_model = self.tot_model  # self.model
         tools = get_tools_str()
         if not isinstance(tools, str):
             tools = str(tools)  # 确保 tools 是字符串
-        system_prompt = f"""你是一个智能助手，请根据用户输入的问题，结合工具使用计划，生成一个思维树，并按照思维树依次调用工具步骤，最终生成一个最终回答。/n 工具列表: {tools}"""
+        now = datetime.now()
+        current_date = now.strftime("%Y-%m-%d")
+        current_time = now.strftime("%H:%M:%S")
+        system_prompt = f"""你是一个智能助手，请根据用户输入的问题，结合工具使用计划，生成一个思维树，并按照思维树依次调用工具步骤，最终生成一个最终回答。/n 今日的日期: {current_date} 当前时间: {current_time} /n 工具列表: {tools}"""
         self.log("DEBUG", "run_thought", {"system_prompt": system_prompt})
 
         # 第一次请求，生成初始的工具使用计划
         params = dict(model=tot_model,
                       messages=[{"role": "system", "content": system_prompt}, {"role": "user", "content": query}],
                       stream=False)
-        response = self.client.chat.completions.create(**params)
+        response = self.tot_client.chat.completions.create(**params)
         initial_content = response.choices[0].message.content
         self.log("DEBUG", "initial_response", {"response": initial_content})
 
@@ -812,7 +897,7 @@ class LightAgent:
         ], stream=False)
         self.log("DEBUG", "reflection_params", {"params": reflection_params})
 
-        reflection_response = self.client.chat.completions.create(**reflection_params)
+        reflection_response = self.tot_client.chat.completions.create(**reflection_params)
         refined_content = reflection_response.choices[0].message.content
         self.log("DEBUG", "refined_response", {"response": refined_content})
         return refined_content
@@ -951,6 +1036,7 @@ class LightAgent:
 # 在函数内部定义工具信息
 get_weather.tool_info = {
     "tool_name": "get_weather",
+    "tool_title": "天气查询",
     "tool_description": "获取指定城市的当前天气信息",
     "tool_params": [
         {"name": "city_name", "description": "要查询的城市名称", "type": "string", "required": True},
@@ -1033,4 +1119,4 @@ class LightSwarm:
 if __name__ == "__main__":
     # Example of registering and using a tool
     print("This is LightAgent")
-    # print(dispatch_tool("example_tool", {"param1": "test"}))
+    # print(dispatch_tool("example_tool", {"param1": "test"}))

{lightagent-0.2.6 → lightagent-0.2.9}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.1
 Name: LightAgent
-Version: 0.2.6
-Summary: **LightAgent** is an extremely lightweight active Agentic Framework with memory (`mem0`), tools (`Tools`), and a Tree of Thought (`ToT`). It supports swarm-like multi-agent collaboration, automated tool generation, and agent assessment, with underlying model support for OpenAI, Zhipu ChatGLM, Baichuan Large Model, DeepSeek, Qwen series large models, and more. At the same time, LightAgent supports OpenAI streaming format API service output, seamlessly integrating with major mainstream chat frameworks. 🌟
+Version: 0.2.9
+Summary: **LightAgent** is an extremely lightweight active Agentic Framework with memory (`mem0`), tools (`Tools`), and a Tree of Thought (`ToT`). It supports swarm-like multi-agent collaboration, automated tool generation, and agent assessment, with underlying model support for OpenAI, Zhipu ChatGLM, Baichuan Large Model, DeepSeek R1, Qwen series large models, and more. At the same time, LightAgent supports OpenAI streaming format API service output, seamlessly integrating with major mainstream chat frameworks. 🌟
 Home-page: https://github.com/wxai-space/LightAgent
 License: Apache 2.0
 Author: caiweige
@@ -54,10 +54,12 @@ Description-Content-Type: text/markdown
   <h1>LightAgent🚀 (Next Generation Agentic AI Framework)</h1>
 </div>
 
-**LightAgent** is an extremely lightweight active Agentic Framework with memory (`mem0`), tools (`Tools`), and a Tree of Thought (`ToT`). It supports swarm-like multi-agent collaboration, automated tool generation, and agent assessment, with underlying model support for OpenAI, Zhipu ChatGLM, Baichuan Large Model, DeepSeek, Qwen series large models, and more. At the same time, LightAgent supports OpenAI streaming format API service output, seamlessly integrating with major mainstream chat frameworks. 🌟
+**LightAgent** is an extremely lightweight active Agentic Framework with memory (`mem0`), tools (`Tools`), and a Tree of Thought (`ToT`). It supports swarm-like multi-agent collaboration, automated tool generation, and agent assessment, with underlying model support for OpenAI, Zhipu ChatGLM, Baichuan Large Model, DeepSeek, Qwen, stepfun and more. At the same time, LightAgent supports OpenAI streaming format API service output, seamlessly integrating with major mainstream chat frameworks. 🌟
 
 ---
 
+![lightswarm_demo_en.png](docs%2Fimages%2Flightswarm_demo_en.png)
+
 ## ✨ Features
 
 - **Lightweight and Efficient** 🚀: Minimalist design, quick deployment, suitable for various application scenarios. (No LangChain, No LlamaIndex) 100% Python implementation, no additional dependencies, core code is only 1000 lines, fully open source. 
@@ -72,6 +74,15 @@ Description-Content-Type: text/markdown
 - **Tool Generator** 🚀: Just provide your API documentation to the [Tool Generator], which will automatically create exclusive tools for you, allowing you to quickly build hundreds of personalized custom tools in just 1 hour to improve efficiency and unleash your creative potential.
 - **Agent Self-Learning** 🧠️: Each agent has its own scene memory capabilities and the ability to self-learn from user conversations.
 
+---
+## News
+- <img src="https://img.alicdn.com/imgextra/i3/O1CN01SFL0Gu26nrQBFKXFR_!!6000000007707-2-tps-500-500.png" alt="new" width="30" height="30"/>**[2025-02-19]** LightAgent v0.2.7  supports deepseek-r1 model for tot now.Significantly enhances the multi-tool planning capability for complex tasks.
+- **[2025-02-06]** LightAgent version 0.2.5 is released now.
+- **[2025-01-20]** LightAgent version 0.2.0 is released now.
+- **[2025-01-05]** LightAgent version 0.1.0 is released now.
+
+---
+
 ## 🚧 Coming Soon
 
 - **Adaptive Tool Mechanism** 🛠️: Supports adding an unlimited number of tools, allowing the large model to first select a candidate tool set from thousands of tools, filtering irrelevant tools before submitting context to the large model, significantly reducing token consumption.
@@ -79,6 +90,7 @@ Description-Content-Type: text/markdown
 - **Agent Assessment** 📊: Built-in agent assessment tool for conveniently evaluating and optimizing the agents you build, aligning with business scenarios, and continuously improving intelligence levels.  
 
 ## Built-in "Thought Flow"
+### ToT now supports DeepSeek-R1.
 The Thought Flow method effectively addresses challenges in complex scenarios through systematic, structured, and flexible thinking processes. Here are the specific implementation steps:
 ```text
 Problem Definition: Clarify the core problems and objectives.
@@ -346,6 +358,7 @@ def get_weather(
 # Define tool information inside the function
 get_weather.tool_info = {
     "tool_name": "get_weather",
+    "tool_title": "get weather",
     "tool_description": "Get current weather information for the specified city.",
     "tool_params": [
         {"name": "city_name", "description": "The name of the city to query", "type": "string", "required": True},
@@ -368,6 +381,7 @@ def search_news(
 # Define tool information inside the function
 search_news.tool_info = {
     "tool_name": "search_news",
+    "tool_title": "search news",
     "tool_description": "Search news based on keywords.",
     "tool_params": [
         {"name": "keyword", "description": "Search keyword", "type": "string", "required": True},
@@ -488,7 +502,7 @@ agent.create_tool(text, tools_directory=tools_directory)
 After execution, two files will be generated in the tools directory: get_stock_kline_data.py and get_stock_realtime_data.py.
 
 ### 4. Tree of Thought (ToT)
-The built-in Tree of Thought module supports complex task decomposition and multi-step reasoning. Through the Tree of Thought, the agent can better handle complex tasks.
+Currently, it is already supported to independently customize the use of the deepseek-r1 model for planning and thinking.The built-in Tree of Thought module supports complex task decomposition and multi-step reasoning. Through the Tree of Thought, the agent can better handle complex tasks.
 
 ```python
 # Enable Tree of Thought
@@ -497,6 +511,9 @@ agent = LightAgent(
     api_key="your_api_key", 
     base_url="your_base_url", 
     tree_of_thought=True,  # Enable Tree of Thought
+    tot_model="deepseek-r1", 
+    tot_api_key="sk-uXx0H0B***17778F1",  # your deepseek r1 API Key
+    tot_base_url="https://api.deepseek.com/v1",  # api url
 )
 ```
 
@@ -758,10 +775,10 @@ We look forward to your feedback and work together to make LightAgent even stron
   <h1>LightAgent🚀（下一代Agentic AI框架）</h1>
 </div>
 
-**LightAgent** 是一个极其轻量的带记忆（`mem0`）、工具（`Tools`）、思维树（`ToT`）的主动式 Agentic Framework（自主性框架）。它支持比Openai Swarm更简单的多智能体协同，构建具备自我学习能力的agent，并支持Agent测评，底层模型支持 OpenAI、智谱 ChatGLM、DeepSeek、阶跃星辰、Qwen千问大模型等。同时，LightAgent 支持 OpenAI 流格式 API 服务输出，无缝接入各大主流 Chat 框架。🌟
+**LightAgent** 是一个极其轻量的带记忆（`mem0`）、工具（`Tools`）、思维树（`ToT`）的主动式 Agentic Framework（自主性框架）。它支持比Openai Swarm更简单的多智能体协同，构建具备自我学习能力的agent，并支持Agent测评，底层模型支持 OpenAI、智谱 ChatGLM、DeepSeek、阶跃星辰、Qwen通义千问大模型等。同时，LightAgent 支持 OpenAI 流格式 API 服务输出，无缝接入各大主流 Chat 框架。🌟
 
 ---
-
+![lightswarm_demo_cn.png](docs%2Fimages%2Flightswarm_demo_cn.png)
 ## ✨ 特性
 
 - **轻量高效** 🚀：极简设计，快速部署，适合各种规模的应用场景。（No LangChain, No LlamaIndex）100% Python 实现，无需额外依赖，核心代码仅1000行，完全开源。 
@@ -776,6 +793,14 @@ We look forward to your feedback and work together to make LightAgent even stron
 - **Tools工具生成器** 🚀：只需将您的API文档交给[Tools工具生成器]，它将自动化地为您打造专属的tools，助您在短短1小时内快速构建数百个个性化的自定义工具，提升效率，释放您的创新潜能。
 - **agent自我学习** 🧠️：每个agent拥有自己的场景记忆能力，拥有从用户的对话中进行自我学习能力。
 
+---
+## 新闻
+- <img src="https://img.alicdn.com/imgextra/i3/O1CN01SFL0Gu26nrQBFKXFR_!!6000000007707-2-tps-500-500.png" alt="new" width="30" height="30"/>**[2025-02-19]** LightAgent v0.2.7 支持单独采用 deepseek-r1 作为的agent推理规划ToT引擎，大幅度提升复杂任务的多工具Plan能力.
+- **[2025-02-06]** LightAgent version 0.2.5 is released now.
+- **[2025-01-20]** LightAgent version 0.2.0 is released now.
+- **[2025-01-05]** LightAgent version 0.1.0 is released now.
+
+---
 
 ## 🚧 即将推出
 
@@ -784,7 +809,8 @@ We look forward to your feedback and work together to make LightAgent even stron
 - **Agent 测评** 📊：内置 Agent 测评工具，方便评估和优化你构建的Agent，对齐业务场景，持续提升智能水平。  
 
 
-## 内置 “思考流”
+## 🔥内置 “思考流”
+### ToT现已支持DeepSeek-R1
 （Thought Flow）方法通过系统性、结构化和灵活的思维过程，能够有效应对复杂场景中的挑战。
  以下是具体实施步骤：
 ```text
@@ -1055,6 +1081,7 @@ def get_weather(
 # 在函数内部定义工具信息
 get_weather.tool_info = {
     "tool_name": "get_weather",
+    "tool_name": "获取天气",
     "tool_description": "获取指定城市的当前天气信息",
     "tool_params": [
         {"name": "city_name", "description": "要查询的城市名称", "type": "string", "required": True},
@@ -1077,6 +1104,7 @@ def search_news(
 # 在函数内部定义工具信息
 search_news.tool_info = {
     "tool_name": "search_news",
+    "tool_name": "联网搜索",
     "tool_description": "根据关键词搜索新闻",
     "tool_params": [
         {"name": "keyword", "description": "搜索关键词", "type": "string", "required": True},
@@ -1197,7 +1225,7 @@ agent.create_tool(text, tools_directory=tools_directory)
 执行后将在tools目录中生成2个文件：get_stock_kline_data.py和get_stock_realtime_data.py
 
 ### 4. 思维树（ToT）
-内置思维树模块，支持复杂任务分解和多步推理。通过思维树，Agent 可以更好地处理复杂任务。
+当前已经支持单独自定义使用deepseek-r1模型来做规划思考。内置思维树模块，支持复杂任务分解和多步推理。通过思维树，Agent 可以更好地处理复杂任务。
 
 ```python
 # 启用思维树
@@ -1206,6 +1234,9 @@ agent = LightAgent(
     api_key="your_api_key", 
     base_url= "your_base_url", 
     tree_of_thought=True,  # 启用思维树
+    tot_model="deepseek-r1", 
+    tot_api_key="sk-uXx0H0B***17778F1",  # 替换为你的 deepseek r1 API Key
+    tot_base_url="https://api.deepseek.com/v1",  # api url
 )
 ```
 

{lightagent-0.2.6 → lightagent-0.2.9}/README.md

@@ -33,10 +33,12 @@
   <h1>LightAgent🚀 (Next Generation Agentic AI Framework)</h1>
 </div>
 
-**LightAgent** is an extremely lightweight active Agentic Framework with memory (`mem0`), tools (`Tools`), and a Tree of Thought (`ToT`). It supports swarm-like multi-agent collaboration, automated tool generation, and agent assessment, with underlying model support for OpenAI, Zhipu ChatGLM, Baichuan Large Model, DeepSeek, Qwen series large models, and more. At the same time, LightAgent supports OpenAI streaming format API service output, seamlessly integrating with major mainstream chat frameworks. 🌟
+**LightAgent** is an extremely lightweight active Agentic Framework with memory (`mem0`), tools (`Tools`), and a Tree of Thought (`ToT`). It supports swarm-like multi-agent collaboration, automated tool generation, and agent assessment, with underlying model support for OpenAI, Zhipu ChatGLM, Baichuan Large Model, DeepSeek, Qwen, stepfun and more. At the same time, LightAgent supports OpenAI streaming format API service output, seamlessly integrating with major mainstream chat frameworks. 🌟
 
 ---
 
+![lightswarm_demo_en.png](docs%2Fimages%2Flightswarm_demo_en.png)
+
 ## ✨ Features
 
 - **Lightweight and Efficient** 🚀: Minimalist design, quick deployment, suitable for various application scenarios. (No LangChain, No LlamaIndex) 100% Python implementation, no additional dependencies, core code is only 1000 lines, fully open source. 
@@ -51,6 +53,15 @@
 - **Tool Generator** 🚀: Just provide your API documentation to the [Tool Generator], which will automatically create exclusive tools for you, allowing you to quickly build hundreds of personalized custom tools in just 1 hour to improve efficiency and unleash your creative potential.
 - **Agent Self-Learning** 🧠️: Each agent has its own scene memory capabilities and the ability to self-learn from user conversations.
 
+---
+## News
+- <img src="https://img.alicdn.com/imgextra/i3/O1CN01SFL0Gu26nrQBFKXFR_!!6000000007707-2-tps-500-500.png" alt="new" width="30" height="30"/>**[2025-02-19]** LightAgent v0.2.7  supports deepseek-r1 model for tot now.Significantly enhances the multi-tool planning capability for complex tasks.
+- **[2025-02-06]** LightAgent version 0.2.5 is released now.
+- **[2025-01-20]** LightAgent version 0.2.0 is released now.
+- **[2025-01-05]** LightAgent version 0.1.0 is released now.
+
+---
+
 ## 🚧 Coming Soon
 
 - **Adaptive Tool Mechanism** 🛠️: Supports adding an unlimited number of tools, allowing the large model to first select a candidate tool set from thousands of tools, filtering irrelevant tools before submitting context to the large model, significantly reducing token consumption.
@@ -58,6 +69,7 @@
 - **Agent Assessment** 📊: Built-in agent assessment tool for conveniently evaluating and optimizing the agents you build, aligning with business scenarios, and continuously improving intelligence levels.  
 
 ## Built-in "Thought Flow"
+### ToT now supports DeepSeek-R1.
 The Thought Flow method effectively addresses challenges in complex scenarios through systematic, structured, and flexible thinking processes. Here are the specific implementation steps:
 ```text
 Problem Definition: Clarify the core problems and objectives.
@@ -325,6 +337,7 @@ def get_weather(
 # Define tool information inside the function
 get_weather.tool_info = {
     "tool_name": "get_weather",
+    "tool_title": "get weather",
     "tool_description": "Get current weather information for the specified city.",
     "tool_params": [
         {"name": "city_name", "description": "The name of the city to query", "type": "string", "required": True},
@@ -347,6 +360,7 @@ def search_news(
 # Define tool information inside the function
 search_news.tool_info = {
     "tool_name": "search_news",
+    "tool_title": "search news",
     "tool_description": "Search news based on keywords.",
     "tool_params": [
         {"name": "keyword", "description": "Search keyword", "type": "string", "required": True},
@@ -467,7 +481,7 @@ agent.create_tool(text, tools_directory=tools_directory)
 After execution, two files will be generated in the tools directory: get_stock_kline_data.py and get_stock_realtime_data.py.
 
 ### 4. Tree of Thought (ToT)
-The built-in Tree of Thought module supports complex task decomposition and multi-step reasoning. Through the Tree of Thought, the agent can better handle complex tasks.
+Currently, it is already supported to independently customize the use of the deepseek-r1 model for planning and thinking.The built-in Tree of Thought module supports complex task decomposition and multi-step reasoning. Through the Tree of Thought, the agent can better handle complex tasks.
 
 ```python
 # Enable Tree of Thought
@@ -476,6 +490,9 @@ agent = LightAgent(
     api_key="your_api_key", 
     base_url="your_base_url", 
     tree_of_thought=True,  # Enable Tree of Thought
+    tot_model="deepseek-r1", 
+    tot_api_key="sk-uXx0H0B***17778F1",  # your deepseek r1 API Key
+    tot_base_url="https://api.deepseek.com/v1",  # api url
 )
 ```
 

{lightagent-0.2.6 → lightagent-0.2.9}/README.zh-CN.md

@@ -33,10 +33,10 @@
   <h1>LightAgent🚀（下一代Agentic AI框架）</h1>
 </div>
 
-**LightAgent** 是一个极其轻量的带记忆（`mem0`）、工具（`Tools`）、思维树（`ToT`）的主动式 Agentic Framework（自主性框架）。它支持比Openai Swarm更简单的多智能体协同，构建具备自我学习能力的agent，并支持Agent测评，底层模型支持 OpenAI、智谱 ChatGLM、DeepSeek、阶跃星辰、Qwen千问大模型等。同时，LightAgent 支持 OpenAI 流格式 API 服务输出，无缝接入各大主流 Chat 框架。🌟
+**LightAgent** 是一个极其轻量的带记忆（`mem0`）、工具（`Tools`）、思维树（`ToT`）的主动式 Agentic Framework（自主性框架）。它支持比Openai Swarm更简单的多智能体协同，构建具备自我学习能力的agent，并支持Agent测评，底层模型支持 OpenAI、智谱 ChatGLM、DeepSeek、阶跃星辰、Qwen通义千问大模型等。同时，LightAgent 支持 OpenAI 流格式 API 服务输出，无缝接入各大主流 Chat 框架。🌟
 
 ---
-
+![lightswarm_demo_cn.png](docs%2Fimages%2Flightswarm_demo_cn.png)
 ## ✨ 特性
 
 - **轻量高效** 🚀：极简设计，快速部署，适合各种规模的应用场景。（No LangChain, No LlamaIndex）100% Python 实现，无需额外依赖，核心代码仅1000行，完全开源。 
@@ -51,6 +51,14 @@
 - **Tools工具生成器** 🚀：只需将您的API文档交给[Tools工具生成器]，它将自动化地为您打造专属的tools，助您在短短1小时内快速构建数百个个性化的自定义工具，提升效率，释放您的创新潜能。
 - **agent自我学习** 🧠️：每个agent拥有自己的场景记忆能力，拥有从用户的对话中进行自我学习能力。
 
+---
+## 新闻
+- <img src="https://img.alicdn.com/imgextra/i3/O1CN01SFL0Gu26nrQBFKXFR_!!6000000007707-2-tps-500-500.png" alt="new" width="30" height="30"/>**[2025-02-19]** LightAgent v0.2.7 支持单独采用 deepseek-r1 作为的agent推理规划ToT引擎，大幅度提升复杂任务的多工具Plan能力.
+- **[2025-02-06]** LightAgent version 0.2.5 is released now.
+- **[2025-01-20]** LightAgent version 0.2.0 is released now.
+- **[2025-01-05]** LightAgent version 0.1.0 is released now.
+
+---
 
 ## 🚧 即将推出
 
@@ -59,7 +67,8 @@
 - **Agent 测评** 📊：内置 Agent 测评工具，方便评估和优化你构建的Agent，对齐业务场景，持续提升智能水平。  
 
 
-## 内置 “思考流”
+## 🔥内置 “思考流”
+### ToT现已支持DeepSeek-R1
 （Thought Flow）方法通过系统性、结构化和灵活的思维过程，能够有效应对复杂场景中的挑战。
  以下是具体实施步骤：
 ```text
@@ -330,6 +339,7 @@ def get_weather(
 # 在函数内部定义工具信息
 get_weather.tool_info = {
     "tool_name": "get_weather",
+    "tool_name": "获取天气",
     "tool_description": "获取指定城市的当前天气信息",
     "tool_params": [
         {"name": "city_name", "description": "要查询的城市名称", "type": "string", "required": True},
@@ -352,6 +362,7 @@ def search_news(
 # 在函数内部定义工具信息
 search_news.tool_info = {
     "tool_name": "search_news",
+    "tool_name": "联网搜索",
     "tool_description": "根据关键词搜索新闻",
     "tool_params": [
         {"name": "keyword", "description": "搜索关键词", "type": "string", "required": True},
@@ -472,7 +483,7 @@ agent.create_tool(text, tools_directory=tools_directory)
 执行后将在tools目录中生成2个文件：get_stock_kline_data.py和get_stock_realtime_data.py
 
 ### 4. 思维树（ToT）
-内置思维树模块，支持复杂任务分解和多步推理。通过思维树，Agent 可以更好地处理复杂任务。
+当前已经支持单独自定义使用deepseek-r1模型来做规划思考。内置思维树模块，支持复杂任务分解和多步推理。通过思维树，Agent 可以更好地处理复杂任务。
 
 ```python
 # 启用思维树
@@ -481,6 +492,9 @@ agent = LightAgent(
     api_key="your_api_key", 
     base_url= "your_base_url", 
     tree_of_thought=True,  # 启用思维树
+    tot_model="deepseek-r1", 
+    tot_api_key="sk-uXx0H0B***17778F1",  # 替换为你的 deepseek r1 API Key
+    tot_base_url="https://api.deepseek.com/v1",  # api url
 )
 ```
 

{lightagent-0.2.6 → lightagent-0.2.9}/pyproject.toml

@@ -1,7 +1,7 @@
 [tool.poetry]
 name = "LightAgent"
-version = "0.2.6"
-description = "**LightAgent** is an extremely lightweight active Agentic Framework with memory (`mem0`), tools (`Tools`), and a Tree of Thought (`ToT`). It supports swarm-like multi-agent collaboration, automated tool generation, and agent assessment, with underlying model support for OpenAI, Zhipu ChatGLM, Baichuan Large Model, DeepSeek, Qwen series large models, and more. At the same time, LightAgent supports OpenAI streaming format API service output, seamlessly integrating with major mainstream chat frameworks. 🌟"
+version = "0.2.9"
+description = "**LightAgent** is an extremely lightweight active Agentic Framework with memory (`mem0`), tools (`Tools`), and a Tree of Thought (`ToT`). It supports swarm-like multi-agent collaboration, automated tool generation, and agent assessment, with underlying model support for OpenAI, Zhipu ChatGLM, Baichuan Large Model, DeepSeek R1, Qwen series large models, and more. At the same time, LightAgent supports OpenAI streaming format API service output, seamlessly integrating with major mainstream chat frameworks. 🌟"
 authors = ["caiweige <caiweige@qq.com>"]
 license = "Apache 2.0"
 readme = [