langchain-dev-utils 1.2.8__tar.gz → 1.2.9__tar.gz

{langchain_dev_utils-1.2.8 → langchain_dev_utils-1.2.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langchain-dev-utils
-Version: 1.2.8
+Version: 1.2.9
 Summary: A practical utility library for LangChain and LangGraph development
 Project-URL: Source Code, https://github.com/TBice123123/langchain-dev-utils
 Project-URL: repository, https://github.com/TBice123123/langchain-dev-utils
@@ -192,52 +192,66 @@ def get_current_time() -> str:
 
 ### 4. **Agent Development**
 
-Includes the following features:
+Includes the following capabilities:
 
-- Predefined agent factory functions
-- Common middleware components
+- Multi-agent construction  
+- Commonly used middleware components  
 
-#### 4.1 Agent Factory Functions
+#### 4.1 Multi-Agent Construction
 
-In LangChain v1, the official `create_agent` function can be used to create a single agent; its `model` parameter accepts either a BaseChatModel instance or a specific string (when a string is provided, only models supported by `init_chat_model` are allowed). To extend the flexibility of specifying models via string, this library provides an equivalent `create_agent` function that lets you designate any model supported by `load_chat_model` (registration required beforehand).
+Wrapping an agent as a tool is a common implementation pattern in multi-agent systems, as elaborated in the official LangChain documentation. To support this pattern, this library provides a pre-built utility function `wrap_agent_as_tool`, which encapsulates an agent instance into a tool that can be invoked by other agents.
 
-Usage example:
+**Usage Example**:
 
 ```python
-from langchain_dev_utils.agents import create_agent
+import datetime
+from langchain_dev_utils.agents import create_agent, wrap_agent_as_tool
 from langchain.agents import AgentState
 
+
+@tool
+def get_current_time() -> str:
+    """Get the current time."""
+    return datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+
+
 agent = create_agent("vllm:qwen3-4b", tools=[get_current_time], name="time-agent")
-response = agent.invoke({"messages": [{"role": "user", "content": "What time is it?"}]})
+call_time_agent_tool = wrap_agent_as_tool(agent)
+response = call_time_agent_tool.invoke(
+    {"messages": [{"role": "user", "content": "What time is it now?"}]}
+)
 print(response)
 ```
 
 #### 4.2 Middleware
 
-Provides some commonly used middleware components. Below, we illustrate with `ToolCallRepairMiddleware` and `PlanMiddleware`.
-
-`ToolCallRepairMiddleware` is used to repair `invalid_tool_calls` generated by large language models.
+Provides several commonly used middleware components. Below are examples using `ToolCallRepairMiddleware` and `PlanMiddleware`.
 
-`PlanMiddleware` is used for agent planning.
+- `ToolCallRepairMiddleware` automatically repairs malformed tool calls found in the model's `invalid_tool_calls` output.
+- `PlanMiddleware` enables task planning capabilities for agents.
 
 ```python
 from langchain_dev_utils.agents.middleware import (
-    ToolcallRepairMiddleware,
+    ToolCallRepairMiddleware,
     PlanMiddleware,
 )
 
 agent = create_agent(
     "vllm:qwen3-4b",
     name="plan-agent",
-    middleware=[ToolCallRepairMiddleware(), PlanMiddleware(
-        use_read_plan_tool=False
-    )]
+    middleware=[
+        ToolCallRepairMiddleware(),
+        PlanMiddleware(use_read_plan_tool=False)
+    ]
 )
-response = agent.invoke({"messages": [{"role": "user", "content": "Give me a travel plan to New York"}]})
+response = agent.invoke({"messages": [{"role": "user", "content": "Give me a travel plan for visiting New York."}]})
 print(response)
 ```
 
-**For more information about agent development and all built-in middleware, please refer to**: [Pre-built Agent Functions](https://tbice123123.github.io/langchain-dev-utils-docs/en/agent-development/prebuilt.html), [Middleware](https://tbice123123.github.io/langchain-dev-utils-docs/en/agent-development/middleware.html)
+**For more details on agent development and a complete list of built-in middleware, please refer to**:  
+[Multi-Agent Construction](https://tbice123123.github.io/langchain-dev-utils-docs/en/agent-development/multi-agent.html),  
+[Middleware](https://tbice123123.github.io/langchain-dev-utils-docs/en/agent-development/middleware.html)
+
 
 ### 5. **State Graph Orchestration**
 

{langchain_dev_utils-1.2.8 → langchain_dev_utils-1.2.9}/README.md

@@ -175,52 +175,66 @@ def get_current_time() -> str:
 
 ### 4. **Agent Development**
 
-Includes the following features:
+Includes the following capabilities:
 
-- Predefined agent factory functions
-- Common middleware components
+- Multi-agent construction  
+- Commonly used middleware components  
 
-#### 4.1 Agent Factory Functions
+#### 4.1 Multi-Agent Construction
 
-In LangChain v1, the official `create_agent` function can be used to create a single agent; its `model` parameter accepts either a BaseChatModel instance or a specific string (when a string is provided, only models supported by `init_chat_model` are allowed). To extend the flexibility of specifying models via string, this library provides an equivalent `create_agent` function that lets you designate any model supported by `load_chat_model` (registration required beforehand).
+Wrapping an agent as a tool is a common implementation pattern in multi-agent systems, as elaborated in the official LangChain documentation. To support this pattern, this library provides a pre-built utility function `wrap_agent_as_tool`, which encapsulates an agent instance into a tool that can be invoked by other agents.
 
-Usage example:
+**Usage Example**:
 
 ```python
-from langchain_dev_utils.agents import create_agent
+import datetime
+from langchain_dev_utils.agents import create_agent, wrap_agent_as_tool
 from langchain.agents import AgentState
 
+
+@tool
+def get_current_time() -> str:
+    """Get the current time."""
+    return datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+
+
 agent = create_agent("vllm:qwen3-4b", tools=[get_current_time], name="time-agent")
-response = agent.invoke({"messages": [{"role": "user", "content": "What time is it?"}]})
+call_time_agent_tool = wrap_agent_as_tool(agent)
+response = call_time_agent_tool.invoke(
+    {"messages": [{"role": "user", "content": "What time is it now?"}]}
+)
 print(response)
 ```
 
 #### 4.2 Middleware
 
-Provides some commonly used middleware components. Below, we illustrate with `ToolCallRepairMiddleware` and `PlanMiddleware`.
-
-`ToolCallRepairMiddleware` is used to repair `invalid_tool_calls` generated by large language models.
+Provides several commonly used middleware components. Below are examples using `ToolCallRepairMiddleware` and `PlanMiddleware`.
 
-`PlanMiddleware` is used for agent planning.
+- `ToolCallRepairMiddleware` automatically repairs malformed tool calls found in the model's `invalid_tool_calls` output.
+- `PlanMiddleware` enables task planning capabilities for agents.
 
 ```python
 from langchain_dev_utils.agents.middleware import (
-    ToolcallRepairMiddleware,
+    ToolCallRepairMiddleware,
     PlanMiddleware,
 )
 
 agent = create_agent(
     "vllm:qwen3-4b",
     name="plan-agent",
-    middleware=[ToolCallRepairMiddleware(), PlanMiddleware(
-        use_read_plan_tool=False
-    )]
+    middleware=[
+        ToolCallRepairMiddleware(),
+        PlanMiddleware(use_read_plan_tool=False)
+    ]
 )
-response = agent.invoke({"messages": [{"role": "user", "content": "Give me a travel plan to New York"}]})
+response = agent.invoke({"messages": [{"role": "user", "content": "Give me a travel plan for visiting New York."}]})
 print(response)
 ```
 
-**For more information about agent development and all built-in middleware, please refer to**: [Pre-built Agent Functions](https://tbice123123.github.io/langchain-dev-utils-docs/en/agent-development/prebuilt.html), [Middleware](https://tbice123123.github.io/langchain-dev-utils-docs/en/agent-development/middleware.html)
+**For more details on agent development and a complete list of built-in middleware, please refer to**:  
+[Multi-Agent Construction](https://tbice123123.github.io/langchain-dev-utils-docs/en/agent-development/multi-agent.html),  
+[Middleware](https://tbice123123.github.io/langchain-dev-utils-docs/en/agent-development/middleware.html)
+
 
 ### 5. **State Graph Orchestration**
 

{langchain_dev_utils-1.2.8 → langchain_dev_utils-1.2.9}/README_cn.md

@@ -176,21 +176,32 @@ def get_current_time() -> str:
 
 包含以下功能：
 
-- 预设的智能体工厂函数
+- 多智能体构建
 - 常用的中间件组件
 
-#### 4.1 智能体工厂函数
+#### 4.1 多智能体构建
 
-LangChain v1 版本中，官方提供的 `create_agent` 函数可以用于创建单智能体，其中 model 参数支持传入 BaseChatModel 实例或特定字符串（当传入字符串时，仅限于 `init_chat_model` 支持的模型）。为扩展字符串指定模型的灵活性，本库提供了功能相同的 `create_agent` 函数，使您能够通过字符串指定 `load_chat_model` 支持的模型（需要提前注册）。
+将智能体封装为工具是多智能体系统中的一种常见实现模式，LangChain 官方文档对此有详细阐述。为此，本库提供了预构建函数`wrap_agent_as_tool` 来实现此模式，该函数能够将一个智能体实例封装成一个可供其它智能体调用的工具。
 
 使用示例：
 
 ```python
-from langchain_dev_utils.agents import create_agent
+import datetime
+from langchain_dev_utils.agents import create_agent, wrap_agent_as_tool
 from langchain.agents import AgentState
 
+
+@tool
+def get_current_time() -> str:
+    """获取当前时间"""
+    return datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+
+
 agent = create_agent("vllm:qwen3-4b", tools=[get_current_time], name="time-agent")
-response = agent.invoke({"messages": [{"role": "user", "content": "现在几点了？"}]})
+call_time_agent_tool = wrap_agent_as_tool(agent)
+response = call_time_agent_tool.invoke(
+    {"messages": [{"role": "user", "content": "现在几点了？"}]}
+)
 print(response)
 ```
 
@@ -219,7 +230,7 @@ response = agent.invoke({"messages": [{"role": "user", "content": "给我一个
 print(response)
 ```
 
-**对于更多关于智能体开发以及所有的内置中间件的相关介绍，请参考**: [预构建智能体函数](https://tbice123123.github.io/langchain-dev-utils-docs/zh/agent-development/prebuilt.html),[中间件](https://tbice123123.github.io/langchain-dev-utils-docs/zh/agent-development/middleware.html)
+**对于更多关于智能体开发以及所有的内置中间件的相关介绍，请参考**: [多智能体构建](https://tbice123123.github.io/langchain-dev-utils-docs/zh/agent-development/multi-agent.html),[中间件](https://tbice123123.github.io/langchain-dev-utils-docs/zh/agent-development/middleware.html)
 
 ### 5. **状态图编排**
 

{langchain_dev_utils-1.2.8 → langchain_dev_utils-1.2.9}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "langchain-dev-utils"
-version = "1.2.8"
+version = "1.2.9"
 description = "A practical utility library for LangChain and LangGraph development"
 readme = "README.md"
 authors = [{ name = "tiebingice", email = "tiebingice123@outlook.com" }]

langchain_dev_utils-1.2.9/src/langchain_dev_utils/__init__.py

@@ -0,0 +1 @@
+__version__ = "1.2.9"

{langchain_dev_utils-1.2.8 → langchain_dev_utils-1.2.9}/src/langchain_dev_utils/chat_models/adapters/openai_compatible.py

@@ -41,7 +41,7 @@ from typing_extensions import Self
 
 from ..types import (
     CompatibilityOptions,
-    ReasoningContentKeepType,
+    ReasoningKeepPolicy,
     ResponseFormatType,
     ToolChoiceType,
 )
@@ -119,7 +119,7 @@ class _BaseChatOpenAICompatible(BaseChatOpenAI):
     """Supported tool choice"""
     supported_response_format: ResponseFormatType = Field(default_factory=list)
     """Supported response format"""
-    reasoning_content_keep_type: ReasoningContentKeepType = Field(default="discard")
+    reasoning_keep_policy: ReasoningKeepPolicy = Field(default="never")
     """How to keep reasoning content in the messages"""
     include_usage: bool = Field(default=True)
     """Whether to include usage information in the output"""
@@ -159,7 +159,7 @@ class _BaseChatOpenAICompatible(BaseChatOpenAI):
 
         payload_messages = []
         last_human_index = -1
-        if self.reasoning_content_keep_type == "temp":
+        if self.reasoning_keep_policy == "current":
             last_human_index = _get_last_human_message_index(messages)
 
         for index, m in enumerate(messages):
@@ -167,15 +167,14 @@ class _BaseChatOpenAICompatible(BaseChatOpenAI):
                 msg_dict = _convert_message_to_dict(
                     _convert_from_v1_to_chat_completions(m)
                 )
-                if (
-                    self.reasoning_content_keep_type == "retain"
-                    and m.additional_kwargs.get("reasoning_content")
+                if self.reasoning_keep_policy == "all" and m.additional_kwargs.get(
+                    "reasoning_content"
                 ):
                     msg_dict["reasoning_content"] = m.additional_kwargs.get(
                         "reasoning_content"
                     )
                 elif (
-                    self.reasoning_content_keep_type == "temp"
+                    self.reasoning_keep_policy == "current"
                     and index > last_human_index
                     and m.additional_kwargs.get("reasoning_content")
                 ):
@@ -558,13 +557,9 @@ def _create_openai_compatible_model(
             ToolChoiceType,
             Field(default=compatibility_options.get("supported_tool_choice", ["auto"])),
         ),
-        reasoning_content_keep_type=(
-            ReasoningContentKeepType,
-            Field(
-                default=compatibility_options.get(
-                    "reasoning_content_keep_type", "discard"
-                )
-            ),
+        reasoning_keep_policy=(
+            ReasoningKeepPolicy,
+            Field(default=compatibility_options.get("reasoning_keep_policy", "never")),
         ),
         supported_response_format=(
             ResponseFormatType,

{langchain_dev_utils-1.2.8 → langchain_dev_utils-1.2.9}/src/langchain_dev_utils/chat_models/types.py

@@ -9,11 +9,11 @@ ToolChoiceType = list[Literal["auto", "none", "required", "specific"]]
 
 ResponseFormatType = list[Literal["json_schema", "json_mode"]]
 
-ReasoningContentKeepType = Literal["discard", "temp", "retain"]
+ReasoningKeepPolicy = Literal["never", "current", "all"]
 
 
 class CompatibilityOptions(TypedDict):
     supported_tool_choice: NotRequired[ToolChoiceType]
     supported_response_format: NotRequired[ResponseFormatType]
-    reasoning_content_keep_type: NotRequired[ReasoningContentKeepType]
+    reasoning_keep_policy: NotRequired[ReasoningKeepPolicy]
     include_usage: NotRequired[bool]

{langchain_dev_utils-1.2.8 → langchain_dev_utils-1.2.9}/src/langchain_dev_utils/embeddings/base.py

@@ -218,30 +218,30 @@ def load_embeddings(
     ):
         raise ValueError(f"Provider {provider} not registered")
 
-    if provider in _SUPPORTED_PROVIDERS:
-        return init_embeddings(model, provider=provider, **kwargs)
-
-    embeddings = _EMBEDDINGS_PROVIDERS_DICT[provider]["embeddings_model"]
-    if isinstance(embeddings, str):
-        if not (api_key := kwargs.get("api_key")):
-            api_key = secret_from_env(f"{provider.upper()}_API_KEY", default=None)()
-            if not api_key:
-                raise ValueError(
-                    f"API key for {provider} not found. Please set it in the environment."
-                )
-            kwargs["api_key"] = api_key
-            if embeddings == "openai-compatible":
-                kwargs["check_embedding_ctx_length"] = False
-                embeddings = "openai"
-        return init_embeddings(
-            model=model,
-            provider=embeddings,
-            base_url=_EMBEDDINGS_PROVIDERS_DICT[provider]["base_url"],
-            **kwargs,
-        )
+    if provider in _EMBEDDINGS_PROVIDERS_DICT:
+        embeddings = _EMBEDDINGS_PROVIDERS_DICT[provider]["embeddings_model"]
+        if isinstance(embeddings, str):
+            if not (api_key := kwargs.get("api_key")):
+                api_key = secret_from_env(f"{provider.upper()}_API_KEY", default=None)()
+                if not api_key:
+                    raise ValueError(
+                        f"API key for {provider} not found. Please set it in the environment."
+                    )
+                kwargs["api_key"] = api_key
+                if embeddings == "openai-compatible":
+                    kwargs["check_embedding_ctx_length"] = False
+                    embeddings = "openai"
+            return init_embeddings(
+                model=model,
+                provider=embeddings,
+                base_url=_EMBEDDINGS_PROVIDERS_DICT[provider]["base_url"],
+                **kwargs,
+            )
+        else:
+            if base_url := _EMBEDDINGS_PROVIDERS_DICT[provider].get("base_url"):
+                url_key = _get_base_url_field_name(embeddings)
+                if url_key is not None:
+                    kwargs.update({url_key: base_url})
+            return embeddings(model=model, **kwargs)
     else:
-        if base_url := _EMBEDDINGS_PROVIDERS_DICT[provider].get("base_url"):
-            url_key = _get_base_url_field_name(embeddings)
-            if url_key is not None:
-                kwargs.update({url_key: base_url})
-        return embeddings(model=model, **kwargs)
+        return init_embeddings(model, provider=provider, **kwargs)

{langchain_dev_utils-1.2.8 → langchain_dev_utils-1.2.9}/src/langchain_dev_utils/message_convert/content.py

@@ -52,8 +52,10 @@ def convert_reasoning_content_for_ai_message(
     reasoning_content = _get_reasoning_content(model_response)
 
     if reasoning_content:
-        model_response.content = (
-            f"{think_tag[0]}{reasoning_content}{think_tag[1]}{model_response.content}"
+        return model_response.model_copy(
+            update={
+                "content": f"{think_tag[0]}{reasoning_content}{think_tag[1]}{model_response.content}"
+            }
         )
     return model_response
 
@@ -99,12 +101,16 @@ def convert_reasoning_content_for_chunk_iterator(
             reasoning_content = _get_reasoning_content(chunk)
             if reasoning_content:
                 if isfirst:
-                    chunk.content = f"{think_tag[0]}{reasoning_content}"
+                    chunk = chunk.model_copy(
+                        update={"content": f"{think_tag[0]}{reasoning_content}"}
+                    )
                     isfirst = False
                 else:
-                    chunk.content = reasoning_content
+                    chunk = chunk.model_copy(update={"content": reasoning_content})
             elif chunk.content and isend and not isfirst:
-                chunk.content = f"{think_tag[1]}{chunk.content}"
+                chunk = chunk.model_copy(
+                    update={"content": f"{think_tag[1]}{chunk.content}"}
+                )
                 isend = False
         yield chunk
 
@@ -149,12 +155,16 @@ async def aconvert_reasoning_content_for_chunk_iterator(
             reasoning_content = _get_reasoning_content(chunk)
             if reasoning_content:
                 if isfirst:
-                    chunk.content = f"{think_tag[0]}{reasoning_content}"
+                    chunk = chunk.model_copy(
+                        update={"content": f"{think_tag[0]}{reasoning_content}"}
+                    )
                     isfirst = False
                 else:
-                    chunk.content = reasoning_content
+                    chunk = chunk.model_copy(update={"content": reasoning_content})
             elif chunk.content and isend and not isfirst:
-                chunk.content = f"{think_tag[1]}{chunk.content}"
+                chunk = chunk.model_copy(
+                    update={"content": f"{think_tag[1]}{chunk.content}"}
+                )
                 isend = False
         yield chunk
 

{langchain_dev_utils-1.2.8 → langchain_dev_utils-1.2.9}/src/langchain_dev_utils/tool_calling/utils.py

@@ -62,20 +62,20 @@ def parse_tool_calling(
         ...     tool_calls = parse_tool_calling(response)
     """
 
-    tool_call = None
+    tool_calls = None
 
     tool_call_blocks = [
         block for block in message.content_blocks if block["type"] == "tool_call"
     ]
     if tool_call_blocks:
-        tool_call = tool_call_blocks
+        tool_calls = tool_call_blocks
 
-    if not tool_call:
-        tool_call = message.tool_calls
+    if not tool_calls:
+        tool_calls = message.tool_calls
 
-    if not tool_call:
+    if not tool_calls:
         raise ValueError("No tool call found in message")
 
     if first_tool_call_only:
-        return (tool_call[0]["name"], tool_call[0]["args"])
-    return [(tool_call["name"], tool_call["args"]) for tool_call in tool_call]
+        return (tool_calls[0]["name"], tool_calls[0]["args"])
+    return [(tool_call["name"], tool_call["args"]) for tool_call in tool_calls]

langchain_dev_utils-1.2.8/src/langchain_dev_utils/__init__.py

@@ -1 +0,0 @@
-__version__ = "1.2.8"