langchain-dev-utils 1.2.11__tar.gz → 1.2.13__tar.gz

{langchain_dev_utils-1.2.11 → langchain_dev_utils-1.2.13}/PKG-INFO

@@ -1,10 +1,10 @@
 Metadata-Version: 2.4
 Name: langchain-dev-utils
-Version: 1.2.11
+Version: 1.2.13
 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
-Project-URL: documentation, https://tbice123123.github.io/langchain-dev-utils-docs/en/
+Project-URL: documentation, https://tbice123123.github.io/langchain-dev-utils
 Author-email: tiebingice <tiebingice123@outlook.com>
 License-File: LICENSE
 Requires-Python: >=3.11

{langchain_dev_utils-1.2.11 → langchain_dev_utils-1.2.13}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "langchain-dev-utils"
-version = "1.2.11"
+version = "1.2.13"
 description = "A practical utility library for LangChain and LangGraph development"
 readme = "README.md"
 authors = [{ name = "tiebingice", email = "tiebingice123@outlook.com" }]
@@ -10,7 +10,7 @@ dependencies = ["langchain>=1.1.0", "langgraph>=1.0.0"]
 [project.urls]
 "Source Code" = "https://github.com/TBice123123/langchain-dev-utils"
 repository = "https://github.com/TBice123123/langchain-dev-utils"
-documentation = "https://tbice123123.github.io/langchain-dev-utils-docs/en/"
+documentation = "https://tbice123123.github.io/langchain-dev-utils"
 
 
 [project.optional-dependencies]
@@ -20,9 +20,6 @@ standard = ["json-repair>=0.53.1", "langchain-openai"]
 requires = ["hatchling"]
 build-backend = "hatchling.build"
 
-[tool.uv.build-backend]
-source-exclude = ["/data", "/docs", "mkdocs.yml"]
-
 [tool.hatch.build]
 exclude = ["/data", "/docs", "mkdocs.yml"]
 

langchain_dev_utils-1.2.13/src/langchain_dev_utils/__init__.py

@@ -0,0 +1 @@
+__version__ = "1.2.13"

{langchain_dev_utils-1.2.11 → langchain_dev_utils-1.2.13}/src/langchain_dev_utils/agents/middleware/__init__.py

@@ -1,3 +1,4 @@
+from .format_prompt import format_prompt
 from .model_fallback import ModelFallbackMiddleware
 from .model_router import ModelRouterMiddleware
 from .plan import (
@@ -22,4 +23,5 @@ __all__ = [
     "LLMToolEmulator",
     "ModelRouterMiddleware",
     "ToolCallRepairMiddleware",
+    "format_prompt",
 ]

langchain_dev_utils-1.2.13/src/langchain_dev_utils/agents/middleware/format_prompt.py

@@ -0,0 +1,66 @@
+from langchain.agents.middleware import ModelRequest, dynamic_prompt
+from langchain_core.prompts.string import get_template_variables
+
+
+@dynamic_prompt
+def format_prompt(request: ModelRequest) -> str:
+    """Format the system prompt with variables from state and context.
+
+    This middleware function extracts template variables from the system prompt
+    and populates them with values from the agent's state and runtime context.
+    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.agents import create_agent
+        >>> from langchain_core.messages import HumanMessage
+        >>> from dataclasses import dataclass
+        >>>
+        >>> @dataclass
+        ... class Context:
+        ...       name: str
+        ...       user: str
+        >>>
+        >>> agent=create_agent(
+        ...     model=model,
+        ...     tools=tools,
+        ...     system_prompt="You are a helpful assistant. Your name is {name}. Your user is {user}.",
+        ...     middleware=[format_prompt],
+        ...     context_schema=Context,
+        ... )
+        >>> agent.invoke(
+        ...     {
+        ...         "messages": [HumanMessage(content="Hello")],
+        ...     },
+        ...     context=Context(name="assistant", user="tbice"),
+        ... )
+
+    """
+    system_msg = request.system_message
+    if system_msg is None:
+        raise ValueError(
+            "system_message must be provided,while use format_prompt in middleware."
+        )
+
+    system_prompt = "\n".join(
+        [content.get("text", "") for content in system_msg.content_blocks]
+    )
+    variables = get_template_variables(system_prompt, "f-string")
+
+    format_params = {}
+
+    state = request.state
+    for key in variables:
+        if var := state.get(key, None):
+            format_params[key] = var
+
+    other_var_keys = set(variables) - set(format_params.keys())
+
+    if other_var_keys:
+        context = request.runtime.context
+        if context is not None:
+            for key in other_var_keys:
+                if var := getattr(context, key, None):
+                    format_params[key] = var
+
+    return system_prompt.format(**format_params)

{langchain_dev_utils-1.2.11 → langchain_dev_utils-1.2.13}/src/langchain_dev_utils/chat_models/adapters/openai_compatible.py

@@ -20,7 +20,13 @@ from langchain_core.callbacks import (
     CallbackManagerForLLMRun,
 )
 from langchain_core.language_models import LangSmithParams, LanguageModelInput
-from langchain_core.messages import AIMessage, AIMessageChunk, BaseMessage, HumanMessage
+from langchain_core.messages import (
+    AIMessage,
+    AIMessageChunk,
+    BaseMessage,
+    HumanMessage,
+    ToolMessage,
+)
 from langchain_core.outputs import ChatGenerationChunk, ChatResult
 from langchain_core.runnables import Runnable
 from langchain_core.tools import BaseTool
@@ -63,6 +69,58 @@ def _get_last_human_message_index(messages: list[BaseMessage]) -> int:
     )
 
 
+def _transform_video_block(block: dict[str, Any]) -> dict:
+    """Transform video block to video_url block."""
+    if "url" in block:
+        return {
+            "type": "video_url",
+            "video_url": {
+                "url": block["url"],
+            },
+        }
+    if "base64" in block or block.get("source_type") == "base64":
+        if "mime_type" not in block:
+            error_message = "mime_type key is required for base64 data."
+            raise ValueError(error_message)
+        mime_type = block["mime_type"]
+        base64_data = block["data"] if "data" in block else block["base64"]
+        return {
+            "type": "video_url",
+            "video_url": {
+                "url": f"data:{mime_type};base64,{base64_data}",
+            },
+        }
+    error_message = "Unsupported source type. Only 'url' and 'base64' are supported."
+    raise ValueError(error_message)
+
+
+def _process_video_input(message: BaseMessage):
+    """
+    Process BaseMessage with video input.
+
+    Args:
+        message (BaseMessage): The HumanMessage instance to process.
+
+    Returns:
+        None: The method modifies the message in-place.
+    """
+    if not message.content:
+        return message
+    content = message.content
+
+    if not isinstance(content, list):
+        return message
+
+    formatted_content = []
+    for block in content:
+        if isinstance(block, dict) and block.get("type") == "video":
+            formatted_content.append(_transform_video_block(block))
+        else:
+            formatted_content.append(block)
+    message = message.model_copy(update={"content": formatted_content})
+    return message
+
+
 class _BaseChatOpenAICompatible(BaseChatOpenAI):
     """
     Base template class for OpenAI-compatible chat model implementations.
@@ -183,6 +241,10 @@ class _BaseChatOpenAICompatible(BaseChatOpenAI):
                     )
                 payload_messages.append(msg_dict)
             else:
+                if (
+                    isinstance(m, HumanMessage) or isinstance(m, ToolMessage)
+                ) and isinstance(m.content, list):
+                    m = _process_video_input(m)
                 payload_messages.append(_convert_message_to_dict(m))
 
         payload["messages"] = payload_messages

{langchain_dev_utils-1.2.11 → langchain_dev_utils-1.2.13}/src/langchain_dev_utils/chat_models/base.py

@@ -11,7 +11,6 @@ from langchain_dev_utils._utils import (
 
 from .types import ChatModelProvider, ChatModelType, CompatibilityOptions
 
-
 _MODEL_PROVIDERS_DICT = {}
 
 

langchain_dev_utils-1.2.13/tests/test_chat_models.py

@@ -0,0 +1,185 @@
+import base64
+from typing import Any, cast
+
+import httpx
+import pytest
+from langchain_core.language_models import BaseChatModel
+from langchain_core.messages import AIMessage, HumanMessage, ToolMessage
+from langchain_tests.integration_tests.chat_models import ChatModelIntegrationTests
+
+from langchain_dev_utils.chat_models.base import load_chat_model
+
+
+class TestStandard(ChatModelIntegrationTests):
+    @pytest.fixture
+    def model(self, request: Any) -> BaseChatModel:
+        """Model fixture."""
+        extra_init_params = getattr(request, "param", None) or {}
+        if extra_init_params.get("output_version") == "v1":
+            pytest.skip("Output version v1 is not supported")
+        return self.chat_model_class(
+            **{
+                **self.standard_chat_model_params,
+                **self.chat_model_params,
+                **extra_init_params,
+            },
+        )
+
+    @property
+    def chat_model_class(self) -> type[BaseChatModel]:
+        return cast("type[BaseChatModel]", load_chat_model)
+
+    @property
+    def chat_model_params(self) -> dict:
+        return {
+            "model": "zai:glm-4.5",
+            "extra_body": {
+                "thinking": {
+                    "type": "disabled",
+                }
+            },
+        }
+
+    @property
+    def has_tool_calling(self) -> bool:
+        return True
+
+    @property
+    def has_structured_output(self) -> bool:
+        return True
+
+    @property
+    def has_tool_choice(self) -> bool:
+        return False
+
+    @property
+    def supports_image_tool_message(self) -> bool:
+        return False
+
+    @property
+    def supports_json_mode(self) -> bool:
+        """(bool) whether the chat model supports JSON mode."""
+        return False
+
+
+class TestImageProcessing:
+    @pytest.fixture(autouse=True)
+    def setup(self) -> None:
+        self.model = load_chat_model("zai:glm-4.6v")
+        self.image_url = "https://cloudcovert-1305175928.cos.ap-guangzhou.myqcloud.com/%E5%9B%BE%E7%89%87grounding.PNG"
+        self.image_data = base64.b64encode(httpx.get(self.image_url).content).decode(
+            "utf-8"
+        )
+
+    def test_image_url_format(self) -> None:
+        message = HumanMessage(
+            content=[
+                {"type": "text", "text": "Give a concise description of this image."},
+                {
+                    "type": "image_url",
+                    "image_url": {"url": f"data:image/png;base64,{self.image_data}"},
+                },
+            ],
+        )
+        _ = self.model.invoke([message])
+
+    def test_image_base64_format(self) -> None:
+        message = HumanMessage(
+            content=[
+                {"type": "text", "text": "Give a concise description of this image."},
+                {
+                    "type": "image",
+                    "base64": self.image_data,
+                    "mime_type": "image/png",
+                },
+            ],
+        )
+        _ = self.model.invoke([message])
+
+    def test_image_url_direct_format(self) -> None:
+        message = HumanMessage(
+            content=[
+                {
+                    "type": "text",
+                    "text": "Give a concise description of this image.  answer in Chinese",
+                },
+                {
+                    "type": "image",
+                    "url": self.image_url,
+                },
+            ],
+        )
+        _ = self.model.invoke([message])
+
+    def test_oai_format_tool_message(self) -> None:
+        oai_format_message = ToolMessage(
+            content=[
+                {
+                    "type": "image_url",
+                    "image_url": {"url": f"data:image/png;base64,{self.image_data}"},
+                },
+            ],
+            tool_call_id="1",
+            name="random_image",
+        )
+
+        messages = [
+            HumanMessage(
+                "get a random diagram using the tool and give it a concise description，answer in Chinese"
+            ),
+            AIMessage(
+                [],
+                tool_calls=[
+                    {
+                        "type": "tool_call",
+                        "id": "1",
+                        "name": "random_image",
+                        "args": {},
+                    }
+                ],
+            ),
+            oai_format_message,
+        ]
+
+        def random_image() -> str:
+            """Return a random image."""
+            return ""
+
+        _ = self.model.bind_tools([random_image]).invoke(messages)
+
+    def test_standard_format_tool_message(self) -> None:
+        standard_format_message = ToolMessage(
+            content=[
+                {
+                    "type": "image",
+                    "base64": self.image_data,
+                    "mime_type": "image/png",
+                },
+            ],
+            tool_call_id="1",
+            name="random_image",
+        )
+
+        messages = [
+            HumanMessage(
+                "get a random diagram using the tool and give it a concise description，answer in Chinese"
+            ),
+            AIMessage(
+                [],
+                tool_calls=[
+                    {
+                        "type": "tool_call",
+                        "id": "1",
+                        "name": "random_image",
+                        "args": {},
+                    }
+                ],
+            ),
+            standard_format_message,
+        ]
+
+        def random_image() -> str:
+            """Return a random image."""
+            return ""
+
+        _ = self.model.bind_tools([random_image]).invoke(messages)

langchain_dev_utils-1.2.11/src/langchain_dev_utils/__init__.py

@@ -1 +0,0 @@
-__version__ = "1.2.11"

langchain_dev_utils-1.2.11/tests/test_chat_models.py

@@ -1,59 +0,0 @@
-from typing import Any, cast
-
-import pytest
-from langchain_core.language_models import BaseChatModel
-from langchain_tests.integration_tests.chat_models import ChatModelIntegrationTests
-
-from langchain_dev_utils.chat_models.base import load_chat_model
-
-
-class TestStandard(ChatModelIntegrationTests):
-    @pytest.fixture
-    def model(self, request: Any) -> BaseChatModel:
-        """Model fixture."""
-        extra_init_params = getattr(request, "param", None) or {}
-        if extra_init_params.get("output_version") == "v1":
-            pytest.skip("Output version v1 is not supported")
-        return self.chat_model_class(
-            **{
-                **self.standard_chat_model_params,
-                **self.chat_model_params,
-                **extra_init_params,
-            },
-        )
-
-    @property
-    def chat_model_class(self) -> type[BaseChatModel]:
-        return cast("type[BaseChatModel]", load_chat_model)
-
-    @property
-    def chat_model_params(self) -> dict:
-        return {
-            "model": "zai:glm-4.5",
-            "extra_body": {
-                "thinking": {
-                    "type": "disabled",
-                }
-            },
-        }
-
-    @property
-    def has_tool_calling(self) -> bool:
-        return True
-
-    @property
-    def has_structured_output(self) -> bool:
-        return True
-
-    @property
-    def has_tool_choice(self) -> bool:
-        return False
-
-    @property
-    def supports_image_tool_message(self) -> bool:
-        return False
-
-    @property
-    def supports_json_mode(self) -> bool:
-        """(bool) whether the chat model supports JSON mode."""
-        return False