langchain-dev-utils 1.3.4__tar.gz → 1.3.6__tar.gz

{langchain_dev_utils-1.3.4 → langchain_dev_utils-1.3.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langchain-dev-utils
-Version: 1.3.4
+Version: 1.3.6
 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

{langchain_dev_utils-1.3.4 → langchain_dev_utils-1.3.6}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "langchain-dev-utils"
-version = "1.3.4"
+version = "1.3.6"
 description = "A practical utility library for LangChain and LangGraph development"
 readme = "README.md"
 authors = [{ name = "tiebingice", email = "tiebingice123@outlook.com" }]

langchain_dev_utils-1.3.6/src/langchain_dev_utils/__init__.py

@@ -0,0 +1 @@
+__version__ = "1.3.6"

langchain_dev_utils-1.3.6/src/langchain_dev_utils/_utils.py

@@ -0,0 +1,131 @@
+from importlib import util
+from typing import Literal, Optional
+
+from pydantic import BaseModel
+
+
+def _check_pkg_install(
+    pkg: Literal["langchain_openai", "json_repair"],
+) -> None:
+    if not util.find_spec(pkg):
+        if pkg == "langchain_openai":
+            msg = "Please install langchain_dev_utils[standard],when use 'openai-compatible'"
+        else:
+            msg = "Please install langchain_dev_utils[standard] to use ToolCallRepairMiddleware."
+        raise ImportError(msg)
+
+
+def _get_base_url_field_name(model_cls: type[BaseModel]) -> str | None:
+    """
+    Return 'base_url' if the model has a field named or aliased as 'base_url',
+    else return 'api_base' if it has a field named or aliased as 'api_base',
+    else return None.
+    The return value is always either 'base_url', 'api_base', or None.
+    """
+    model_fields = model_cls.model_fields
+
+    # try model_fields first
+    if "base_url" in model_fields:
+        return "base_url"
+
+    if "api_base" in model_fields:
+        return "api_base"
+
+    # then try aliases
+    for field_info in model_fields.values():
+        if field_info.alias == "base_url":
+            return "base_url"
+
+    for field_info in model_fields.values():
+        if field_info.alias == "api_base":
+            return "api_base"
+
+    return None
+
+
+def _validate_base_url(base_url: Optional[str] = None) -> None:
+    """Validate base URL format.
+
+    Args:
+        base_url: Base URL to validate
+
+    Raises:
+        ValueError: If base URL is not a valid HTTP or HTTPS URL
+    """
+    if base_url is None:
+        return
+
+    from urllib.parse import urlparse
+
+    parsed = urlparse(base_url.strip())
+
+    if not parsed.scheme or not parsed.netloc:
+        raise ValueError(
+            f"base_url must be a valid HTTP or HTTPS URL. Received: {base_url}"
+        )
+
+    if parsed.scheme not in ("http", "https"):
+        raise ValueError(
+            f"base_url must use HTTP or HTTPS protocol. Received: {parsed.scheme}"
+        )
+
+
+def _validate_model_cls_name(model_cls_name: str) -> None:
+    """Validate model class name follows Python naming conventions.
+
+    Args:
+        model_cls_name: Class name to validate
+
+    Raises:
+        ValueError: If class name is invalid
+    """
+    if not model_cls_name:
+        raise ValueError("model_cls_name cannot be empty")
+
+    if not model_cls_name[0].isalpha():
+        raise ValueError(
+            f"model_cls_name must start with a letter. Received: {model_cls_name}"
+        )
+
+    if not all(c.isalnum() or c == "_" for c in model_cls_name):
+        raise ValueError(
+            f"model_cls_name can only contain letters, numbers, and underscores. Received: {model_cls_name}"
+        )
+
+    if model_cls_name[0].islower():
+        raise ValueError(
+            f"model_cls_name should start with an uppercase letter (PEP 8). Received: {model_cls_name}"
+        )
+
+    if len(model_cls_name) > 30:
+        raise ValueError(
+            f"model_cls_name must be 30 characters or fewer. Received: {model_cls_name}"
+        )
+
+
+def _validate_provider_name(provider_name: str) -> None:
+    """Validate provider name follows Python naming conventions.
+
+    Args:
+        provider_name: Provider name to validate
+
+    Raises:
+        ValueError: If provider name is invalid
+    """
+    if not provider_name:
+        raise ValueError("provider_name cannot be empty")
+
+    if not provider_name[0].isalnum():
+        raise ValueError(
+            f"provider_name must start with a letter or number. Received: {provider_name}"
+        )
+
+    if not all(c.isalnum() or c == "_" for c in provider_name):
+        raise ValueError(
+            f"provider_name can only contain letters, numbers, underscores. Received: {provider_name}"
+        )
+
+    if len(provider_name) > 20:
+        raise ValueError(
+            f"provider_name must be 20 characters or fewer. Received: {provider_name}"
+        )

{langchain_dev_utils-1.3.4 → langchain_dev_utils-1.3.6}/src/langchain_dev_utils/agents/middleware/format_prompt.py

@@ -11,7 +11,7 @@ def format_prompt(request: ModelRequest) -> str:
     Variables are first resolved from the state, then from the context if not found.
 
     Example:
-        >>> from langchain_dev_utils.agents.middleware.format_prompt import format_prompt
+        >>> from langchain_dev_utils.agents.middleware import format_prompt
         >>> from langchain.agents import create_agent
         >>> from langchain_core.messages import HumanMessage
         >>> from dataclasses import dataclass

{langchain_dev_utils-1.3.4 → langchain_dev_utils-1.3.6}/src/langchain_dev_utils/agents/wrap.py

@@ -1,8 +1,8 @@
 import asyncio
-from typing import Any, Awaitable, Callable, Optional, cast
+from typing import Any, Awaitable, Callable, Optional
 
 from langchain.tools import ToolRuntime
-from langchain_core.messages import AnyMessage, HumanMessage
+from langchain_core.messages import HumanMessage
 from langchain_core.tools import BaseTool, StructuredTool
 from langgraph.graph.state import CompiledStateGraph
 
@@ -14,9 +14,9 @@ def _process_input(request: str, runtime: ToolRuntime) -> str:
 
 
 def _process_output(
-    request: str, response: list[AnyMessage], runtime: ToolRuntime
+    request: str, response: dict[str, Any], runtime: ToolRuntime
 ) -> Any:
-    return response[-1].content
+    return response["messages"][-1].content
 
 
 def wrap_agent_as_tool(
@@ -25,17 +25,17 @@ def wrap_agent_as_tool(
     tool_description: Optional[str] = None,
     pre_input_hooks: Optional[
         tuple[
-            Callable[[str, ToolRuntime], str],
-            Callable[[str, ToolRuntime], Awaitable[str]],
+            Callable[[str, ToolRuntime], str | dict[str, Any]],
+            Callable[[str, ToolRuntime], Awaitable[str | dict[str, Any]]],
         ]
-        | Callable[[str, ToolRuntime], str]
+        | Callable[[str, ToolRuntime], str | dict[str, Any]]
     ] = None,
     post_output_hooks: Optional[
         tuple[
-            Callable[[str, list[AnyMessage], ToolRuntime], Any],
-            Callable[[str, list[AnyMessage], ToolRuntime], Awaitable[Any]],
+            Callable[[str, dict[str, Any], ToolRuntime], Any],
+            Callable[[str, dict[str, Any], ToolRuntime], Awaitable[Any]],
         ]
-        | Callable[[str, list[AnyMessage], ToolRuntime], Any]
+        | Callable[[str, dict[str, Any], ToolRuntime], Any]
     ] = None,
 ) -> BaseTool:
     """Wraps an agent as a tool
@@ -91,33 +91,58 @@ def wrap_agent_as_tool(
     def call_agent(
         request: str,
         runtime: ToolRuntime,
-    ) -> str:
-        request = process_input(request, runtime) if process_input else request
+    ):
+        _processed_input = process_input(request, runtime) if process_input else request
+        if isinstance(_processed_input, str):
+            agent_input = {"messages": [HumanMessage(content=_processed_input)]}
+        elif isinstance(_processed_input, dict):
+            if "messages" not in _processed_input:
+                raise ValueError("Agent input must contain 'messages' key")
+            agent_input = _processed_input
+        else:
+            raise ValueError("Pre Hooks must return a string or a dict")
 
-        messages = [HumanMessage(content=request)]
-        response = agent.invoke({"messages": messages})
+        response = agent.invoke(agent_input)
 
-        response = process_output(request, response["messages"], runtime)
+        response = (
+            process_output(request, response, runtime)
+            if process_output
+            else response["messages"][-1].content
+        )
         return response
 
     async def acall_agent(
         request: str,
         runtime: ToolRuntime,
-    ) -> str:
+    ):
         if asyncio.iscoroutinefunction(process_input_async):
-            request = await process_input_async(request, runtime)
+            _processed_input = await process_input_async(request, runtime)
+        else:
+            _processed_input = (
+                process_input_async(request, runtime)
+                if process_input_async
+                else request
+            )
+
+        if isinstance(_processed_input, str):
+            agent_input = {"messages": [HumanMessage(content=_processed_input)]}
+        elif isinstance(_processed_input, dict):
+            if "messages" not in _processed_input:
+                raise ValueError("Agent input must contain 'messages' key")
+            agent_input = _processed_input
         else:
-            request = cast(str, process_input_async(request, runtime))
+            raise ValueError("Pre Hooks must return a string or a dict")
 
-        messages = [HumanMessage(content=request)]
-        response = await agent.ainvoke({"messages": messages})
+        response = await agent.ainvoke(agent_input)
 
         if asyncio.iscoroutinefunction(process_output_async):
-            response = await process_output_async(
-                request, response["messages"], runtime
-            )
+            response = await process_output_async(request, response, runtime)
         else:
-            response = process_output(request, response["messages"], runtime)
+            response = (
+                process_output(request, response, runtime)
+                if process_output
+                else response["messages"][-1].content
+            )
 
         return response
 
@@ -143,17 +168,17 @@ def wrap_all_agents_as_tool(
     tool_description: Optional[str] = None,
     pre_input_hooks: Optional[
         tuple[
-            Callable[[str, ToolRuntime], str],
-            Callable[[str, ToolRuntime], Awaitable[str]],
+            Callable[[str, ToolRuntime], str | dict[str, Any]],
+            Callable[[str, ToolRuntime], Awaitable[str | dict[str, Any]]],
         ]
-        | Callable[[str, ToolRuntime], str]
+        | Callable[[str, ToolRuntime], str | dict[str, Any]]
     ] = None,
     post_output_hooks: Optional[
         tuple[
-            Callable[[str, list[AnyMessage], ToolRuntime], Any],
-            Callable[[str, list[AnyMessage], ToolRuntime], Awaitable[Any]],
+            Callable[[str, dict[str, Any], ToolRuntime], Any],
+            Callable[[str, dict[str, Any], ToolRuntime], Awaitable[Any]],
         ]
-        | Callable[[str, list[AnyMessage], ToolRuntime], Any]
+        | Callable[[str, dict[str, Any], ToolRuntime], Any]
     ] = None,
 ) -> BaseTool:
     """Wraps all agents as single tool
@@ -219,42 +244,67 @@ def wrap_all_agents_as_tool(
         agent_name: str,
         description: str,
         runtime: ToolRuntime,
-    ) -> str:
-        task_description = (
-            process_input(description, runtime) if process_input else description
-        )
-
+    ):
         if agent_name not in agents_map:
             raise ValueError(f"Agent {agent_name} not found")
 
-        messages = [HumanMessage(content=task_description)]
-        response = agents_map[agent_name].invoke({"messages": messages})
+        _processed_input = (
+            process_input(description, runtime) if process_input else description
+        )
+        if isinstance(_processed_input, str):
+            agent_input = {"messages": [HumanMessage(content=_processed_input)]}
+        elif isinstance(_processed_input, dict):
+            if "messages" not in _processed_input:
+                raise ValueError("Agent input must contain 'messages' key")
+            agent_input = _processed_input
+        else:
+            raise ValueError("Pre Hooks must return str or dict")
+
+        response = agent.invoke(agent_input)
 
-        response = process_output(task_description, response["messages"], runtime)
+        response = (
+            process_output(description, response, runtime)
+            if process_output
+            else response["messages"][-1].content
+        )
         return response
 
     async def acall_agent(
         agent_name: str,
         description: str,
         runtime: ToolRuntime,
-    ) -> str:
+    ):
+        if agent_name not in agents_map:
+            raise ValueError(f"Agent {agent_name} not found")
+
         if asyncio.iscoroutinefunction(process_input_async):
-            task_description = await process_input_async(description, runtime)
+            _processed_input = await process_input_async(description, runtime)
         else:
-            task_description = cast(str, process_input_async(description, runtime))
+            _processed_input = (
+                process_input_async(description, runtime)
+                if process_input_async
+                else description
+            )
 
-        if agent_name not in agents_map:
-            raise ValueError(f"Agent {agent_name} not found")
+        if isinstance(_processed_input, str):
+            agent_input = {"messages": [HumanMessage(content=_processed_input)]}
+        elif isinstance(_processed_input, dict):
+            if "messages" not in _processed_input:
+                raise ValueError("Agent input must contain 'messages' key")
+            agent_input = _processed_input
+        else:
+            raise ValueError("Pre Hooks must return str or dict")
 
-        messages = [HumanMessage(content=task_description)]
-        response = await agents_map[agent_name].ainvoke({"messages": messages})
+        response = await agents_map[agent_name].ainvoke(agent_input)
 
         if asyncio.iscoroutinefunction(process_output_async):
-            response = await process_output_async(
-                task_description, response["messages"], runtime
-            )
+            response = await process_output_async(description, response, runtime)
         else:
-            response = process_output(task_description, response["messages"], runtime)
+            response = (
+                process_output(description, response, runtime)
+                if process_output
+                else response["messages"][-1].content
+            )
 
         return response
 

{langchain_dev_utils-1.3.4 → langchain_dev_utils-1.3.6}/src/langchain_dev_utils/chat_models/adapters/openai_compatible.py

@@ -50,6 +50,11 @@ from pydantic import (
 )
 from typing_extensions import Self
 
+from ..._utils import (
+    _validate_base_url,
+    _validate_model_cls_name,
+    _validate_provider_name,
+)
 from ..types import (
     CompatibilityOptions,
     ReasoningKeepPolicy,
@@ -586,6 +591,51 @@ class _BaseChatOpenAICompatible(BaseChatOpenAI):
         )
 
 
+def _validate_compatibility_options(
+    compatibility_options: Optional[CompatibilityOptions] = None,
+) -> None:
+    """Validate provider configuration against supported features.
+
+    Args:
+        compatibility_options: Optional configuration for the provider
+
+    Raises:
+        ValueError: If provider configuration is invalid
+    """
+    if compatibility_options is None:
+        compatibility_options = {}
+
+    if "supported_tool_choice" in compatibility_options:
+        _supported_tool_choice = compatibility_options["supported_tool_choice"]
+        for tool_choice in _supported_tool_choice:
+            if tool_choice not in ["auto", "none", "required", "specific"]:
+                raise ValueError(
+                    f"Unsupported tool_choice: {tool_choice}. Please choose from 'auto', 'none', 'required','specific'."
+                )
+
+    if "supported_response_format" in compatibility_options:
+        _supported_response_format = compatibility_options["supported_response_format"]
+        for response_format in _supported_response_format:
+            if response_format not in ["json_schema", "json_mode"]:
+                raise ValueError(
+                    f"Unsupported response_format: {response_format}. Please choose from 'json_schema', 'json_mode'."
+                )
+
+    if "reasoning_keep_policy" in compatibility_options:
+        _reasoning_keep_policy = compatibility_options["reasoning_keep_policy"]
+        if _reasoning_keep_policy not in ["never", "current", "all"]:
+            raise ValueError(
+                f"Unsupported reasoning_keep_policy: {_reasoning_keep_policy}. Please choose from 'never', 'current', 'all'."
+            )
+
+    if "include_usage" in compatibility_options:
+        _include_usage = compatibility_options["include_usage"]
+        if not isinstance(_include_usage, bool):
+            raise ValueError(
+                f"include_usage must be a boolean value. Received: {_include_usage}"
+            )
+
+
 def _create_openai_compatible_model(
     provider: str,
     base_url: str,
@@ -615,6 +665,14 @@ def _create_openai_compatible_model(
     if profiles is not None:
         _register_profile_with_provider(provider, profiles)
 
+    _validate_compatibility_options(compatibility_options)
+
+    _validate_provider_name(provider)
+
+    _validate_model_cls_name(chat_model_cls_name)
+
+    _validate_base_url(base_url)
+
     return create_model(
         chat_model_cls_name,
         __base__=_BaseChatOpenAICompatible,

{langchain_dev_utils-1.3.4 → langchain_dev_utils-1.3.6}/src/langchain_dev_utils/chat_models/base.py

@@ -7,6 +7,7 @@ from langchain_core.utils import from_env
 from langchain_dev_utils._utils import (
     _check_pkg_install,
     _get_base_url_field_name,
+    _validate_provider_name,
 )
 
 from .types import ChatModelProvider, ChatModelType, CompatibilityOptions
@@ -126,6 +127,7 @@ def register_model_provider(
         >>> model = load_chat_model(model="vllm:qwen3-4b")
         >>> model.invoke("Hello")
     """
+    _validate_provider_name(provider_name)
     base_url = base_url or from_env(f"{provider_name.upper()}_API_BASE", default=None)()
     if isinstance(chat_model, str):
         _check_pkg_install("langchain_openai")

{langchain_dev_utils-1.3.4 → langchain_dev_utils-1.3.6}/src/langchain_dev_utils/embeddings/adapters/openai_compatible.py

@@ -4,6 +4,12 @@ from langchain_core.utils import from_env, secret_from_env
 from langchain_openai.embeddings import OpenAIEmbeddings
 from pydantic import Field, SecretStr, create_model
 
+from ..._utils import (
+    _validate_base_url,
+    _validate_model_cls_name,
+    _validate_provider_name,
+)
+
 
 class _BaseEmbeddingOpenAICompatible(OpenAIEmbeddings):
     """Base class for OpenAI-Compatible embeddings.
@@ -53,6 +59,16 @@ def _create_openai_compatible_embedding(
     """
     embeddings_cls_name = embeddings_cls_name or f"{provider.title()}Embeddings"
 
+    if len(provider) >= 20:
+        raise ValueError(
+            f"provider must be less than 50 characters. Received: {provider}"
+        )
+
+    _validate_model_cls_name(embeddings_cls_name)
+    _validate_provider_name(provider)
+
+    _validate_base_url(base_url)
+
     return create_model(
         embeddings_cls_name,
         __base__=_BaseEmbeddingOpenAICompatible,

{langchain_dev_utils-1.3.4 → langchain_dev_utils-1.3.6}/src/langchain_dev_utils/embeddings/base.py

@@ -6,6 +6,7 @@ from langchain_core.utils import from_env
 from langchain_dev_utils._utils import (
     _check_pkg_install,
     _get_base_url_field_name,
+    _validate_provider_name,
 )
 
 _EMBEDDINGS_PROVIDERS_DICT = {}
@@ -87,7 +88,7 @@ def register_embeddings_provider(
         >>> embeddings = load_embeddings("vllm:qwen3-embedding-4b")
         >>> embeddings.embed_query("hello world")
     """
-
+    _validate_provider_name(provider_name)
     base_url = base_url or from_env(f"{provider_name.upper()}_API_BASE", default=None)()
     if isinstance(embeddings_model, str):
         if base_url is None:

{langchain_dev_utils-1.3.4 → langchain_dev_utils-1.3.6}/tests/test_wrap_agent.py

@@ -31,18 +31,22 @@ async def process_input_async(request: str, runtime: ToolRuntime) -> str:
     return "<task_description>" + request + "</task_description>"
 
 
-def process_output(request: str, messages: list, runtime: ToolRuntime) -> str:
-    assert request.startswith("<task_description>")
-    assert request.endswith("</task_description>")
-    return "<task_response>" + messages[-1].content + "</task_response>"
+def process_output(request: str, response: dict[str, Any], runtime: ToolRuntime) -> str:
+    human_message = response["messages"][0]
+    assert human_message.content.startswith(
+        "<task_description>"
+    ) and human_message.content.endswith("</task_description>")
+    return "<task_response>" + response["messages"][-1].content + "</task_response>"
 
 
 async def process_output_async(
-    request: str, messages: list, runtime: ToolRuntime
+    request: str, response: dict[str, Any], runtime: ToolRuntime
 ) -> str:
-    assert request.startswith("<task_description>")
-    assert request.endswith("</task_description>")
-    return "<task_response>" + messages[-1].content + "</task_response>"
+    human_message = response["messages"][0]
+    assert human_message.content.startswith(
+        "<task_description>"
+    ) and human_message.content.endswith("</task_description>")
+    return "<task_response>" + response["messages"][-1].content + "</task_response>"
 
 
 def test_wrap_agent():

langchain_dev_utils-1.3.4/src/langchain_dev_utils/__init__.py

@@ -1 +0,0 @@
-__version__ = "1.3.4"

langchain_dev_utils-1.3.4/src/langchain_dev_utils/_utils.py

@@ -1,43 +0,0 @@
-from importlib import util
-from typing import Literal
-
-from pydantic import BaseModel
-
-
-def _check_pkg_install(
-    pkg: Literal["langchain_openai", "json_repair"],
-) -> None:
-    if not util.find_spec(pkg):
-        if pkg == "langchain_openai":
-            msg = "Please install langchain_dev_utils[standard],when use 'openai-compatible'"
-        else:
-            msg = "Please install langchain_dev_utils[standard] to use ToolCallRepairMiddleware."
-        raise ImportError(msg)
-
-
-def _get_base_url_field_name(model_cls: type[BaseModel]) -> str | None:
-    """
-    Return 'base_url' if the model has a field named or aliased as 'base_url',
-    else return 'api_base' if it has a field named or aliased as 'api_base',
-    else return None.
-    The return value is always either 'base_url', 'api_base', or None.
-    """
-    model_fields = model_cls.model_fields
-
-    # try model_fields first
-    if "base_url" in model_fields:
-        return "base_url"
-
-    if "api_base" in model_fields:
-        return "api_base"
-
-    # then try aliases
-    for field_info in model_fields.values():
-        if field_info.alias == "base_url":
-            return "base_url"
-
-    for field_info in model_fields.values():
-        if field_info.alias == "api_base":
-            return "api_base"
-
-    return None