release: langgraph-prebuilt 1.0.5 (#6473)

## overview

The main purpose of this is to respect tool signatures that request
injected args (like `ToolRuntime`) even when the explicitly specified
`args_schema` does not.

Ex in the following example, we should still inject `runtime` despite
its absence in `ArgsSchema`

```py
class ArgsSchema(BaseModel):
    some_arg: int = Field(...)

@tool(args_schema=ArgsSchema)
def my_tool(some_arg: int, runtime: ToolRuntime): ...
```

This is accompanied by
https://github.com/langchain-ai/langchain/pull/34051 which has tests
that pass w/ this change. This tests injection w/ `create_agent` (more
end to end than tests added in
https://github.com/langchain-ai/langchain/pull/33999.

This unblocks the injection of `ToolRuntime` into MCP tools which is
exciting bc that exposes tool call id and state, which we previously
were unable to do.

## other benefits

* Cleaner code structure w/ more helpful docs about injected args.
* Nice perf boost, we're no longer inspecting the annotations of a
tool's schema 3 different times to detect store, state, and runtime
injections.

## additional notes

1. I could see a world where we want more of this logic to reside on the
tools themselves, but tools don't now about LG specific injection types
(like `ToolRuntime`, hence having this logic here for now).
2. We could separately add validation for the case where something is
specified in `args_schema` and not in the function signature (probably
at the tool level though).

libs/langgraph/uv.lock

@@ -1710,7 +1710,7 @@ test = [
 
 [[package]]
 name = "langgraph-prebuilt"
-version = "1.0.4"
+version = "1.0.5"
 source = { editable = "../prebuilt" }
 dependencies = [
     { name = "langchain-core" },

libs/prebuilt/langgraph/prebuilt/tool_node.py

@@ -142,6 +142,25 @@ class ToolCallRequest:
     state: Any
     runtime: ToolRuntime
 
+    def __setattr__(self, name: str, value: Any) -> None:
+        """Raise deprecation warning when setting attributes directly.
+
+        Direct attribute assignment is deprecated. Use the `override()` method instead.
+        """
+        import warnings
+
+        # Allow setting attributes during initialization
+        if not hasattr(self, "__dataclass_fields__") or not hasattr(self, name):
+            object.__setattr__(self, name, value)
+        else:
+            warnings.warn(
+                f"Setting attribute '{name}' on ToolCallRequest is deprecated. "
+                "Use the override() method instead to create a new instance with modified values.",
+                DeprecationWarning,
+                stacklevel=2,
+            )
+            object.__setattr__(self, name, value)
+
     def override(
         self, **overrides: Unpack[_ToolCallRequestOverrides]
     ) -> ToolCallRequest:
@@ -202,8 +221,9 @@ Examples:
 
     ```python
     def handler(request, execute):
-        request.tool_call["args"]["value"] *= 2
-        return execute(request)
+        modified_call = {**request.tool_call, "args": {**request.tool_call["args"], "value": request.tool_call["args"]["value"] * 2}}
+        modified_request = request.override(tool_call=modified_call)
+        return execute(modified_request)
     ```
 
     Retry on error (execute multiple times):
@@ -479,9 +499,7 @@ def _infer_handled_types(handler: Callable[..., str]) -> tuple[type[Exception],
 
 def _filter_validation_errors(
     validation_error: ValidationError,
-    tool_to_state_args: dict[str, str | None],
-    tool_to_store_arg: str | None,
-    tool_to_runtime_arg: str | None,
+    injected_args: _InjectedArgs | None,
 ) -> list[ErrorDetails]:
     """Filter validation errors to only include LLM-controlled arguments.
 
@@ -496,25 +514,28 @@ def _filter_validation_errors(
 
     Args:
         validation_error: The Pydantic ValidationError raised during tool invocation.
-        tool_to_state_args: Mapping of state argument names to state field names.
-        tool_to_store_arg: Name of the store argument, if any.
-        tool_to_runtime_arg: Name of the runtime argument, if any.
+        injected_args: The _InjectedArgs structure containing all injected arguments,
+            or None if there are no injected arguments.
 
     Returns:
         List of ErrorDetails containing only errors for LLM-controlled arguments,
         with system-injected argument values removed from the input field.
     """
-    injected_args = set(tool_to_state_args.keys())
-    if tool_to_store_arg:
-        injected_args.add(tool_to_store_arg)
-    if tool_to_runtime_arg:
-        injected_args.add(tool_to_runtime_arg)
+    # Collect all injected argument names
+    injected_arg_names: set[str] = set()
+    if injected_args:
+        if injected_args.state:
+            injected_arg_names.update(injected_args.state.keys())
+        if injected_args.store:
+            injected_arg_names.add(injected_args.store)
+        if injected_args.runtime:
+            injected_arg_names.add(injected_args.runtime)
 
     filtered_errors: list[ErrorDetails] = []
     for error in validation_error.errors():
         # Check if error location contains any injected argument
         # error['loc'] is a tuple like ('field_name',) or ('field_name', 'nested_field')
-        if error["loc"] and error["loc"][0] not in injected_args:
+        if error["loc"] and error["loc"][0] not in injected_arg_names:
             # Create a copy of the error dict to avoid mutating the original
             error_copy: dict[str, Any] = {**error}
 
@@ -522,7 +543,7 @@ def _filter_validation_errors(
             if isinstance(error_copy.get("input"), dict):
                 input_dict = error_copy["input"]
                 input_copy = {
-                    k: v for k, v in input_dict.items() if k not in injected_args
+                    k: v for k, v in input_dict.items() if k not in injected_arg_names
                 }
                 error_copy["input"] = input_copy
 
@@ -532,6 +553,60 @@ def _filter_validation_errors(
     return filtered_errors
 
 
+@dataclass
+class _InjectedArgs:
+    """Internal structure for tracking injected arguments for a tool.
+
+    This data structure is built once during ToolNode initialization by analyzing
+    the tool's signature and args schema, then reused during execution for efficient
+    injection without repeated reflection.
+
+    The structure maps from tool parameter names to their injection sources, enabling
+    the ToolNode to know exactly which arguments need to be injected and where to
+    get their values from.
+
+    Attributes:
+        state: Mapping from tool parameter names to state field names for injection.
+            Keys are tool parameter names, values are either:
+            - str: Name of the state field to extract and inject
+            - None: Inject the entire state object
+            Empty dict if no state injection is needed.
+        store: Name of the tool parameter where the store should be injected,
+            or None if no store injection is needed.
+        runtime: Name of the tool parameter where the runtime should be injected,
+            or None if no runtime injection is needed.
+
+    Example:
+        For a tool with signature:
+        ```python
+        def my_tool(
+            x: int,
+            messages: Annotated[list, InjectedState("messages")],
+            full_state: Annotated[dict, InjectedState()],
+            store: Annotated[BaseStore, InjectedStore()],
+            runtime: ToolRuntime,
+        ) -> str:
+            ...
+        ```
+
+        The resulting `_InjectedArgs` would be:
+        ```python
+        _InjectedArgs(
+            state={
+                "messages": "messages",  # Extract state["messages"]
+                "full_state": None,      # Inject entire state
+            },
+            store="store",               # Inject into "store" parameter
+            runtime="runtime",           # Inject into "runtime" parameter
+        )
+        ```
+    """
+
+    state: dict[str, str | None]
+    store: str | None
+    runtime: str | None
+
+
 class ToolNode(RunnableCallable):
     """A node for executing tools in LangGraph workflows.
 
@@ -676,9 +751,7 @@ class ToolNode(RunnableCallable):
         """
         super().__init__(self._func, self._afunc, name=name, tags=tags, trace=False)
         self._tools_by_name: dict[str, BaseTool] = {}
-        self._tool_to_state_args: dict[str, dict[str, str | None]] = {}
-        self._tool_to_store_arg: dict[str, str | None] = {}
-        self._tool_to_runtime_arg: dict[str, str | None] = {}
+        self._injected_args: dict[str, _InjectedArgs] = {}
         self._handle_tool_errors = handle_tool_errors
         self._messages_key = messages_key
         self._wrap_tool_call = wrap_tool_call
@@ -689,9 +762,8 @@ class ToolNode(RunnableCallable):
             else:
                 tool_ = tool
             self._tools_by_name[tool_.name] = tool_
-            self._tool_to_state_args[tool_.name] = _get_state_args(tool_)
-            self._tool_to_store_arg[tool_.name] = _get_store_arg(tool_)
-            self._tool_to_runtime_arg[tool_.name] = _get_runtime_arg(tool_)
+            # Build injected args mapping once during initialization in a single pass
+            self._injected_args[tool_.name] = _get_all_injected_args(tool_)
 
     @property
     def tools_by_name(self) -> dict[str, BaseTool]:
@@ -844,12 +916,8 @@ class ToolNode(RunnableCallable):
                 response = tool.invoke(call_args, config)
             except ValidationError as exc:
                 # Filter out errors for injected arguments
-                filtered_errors = _filter_validation_errors(
-                    exc,
-                    self._tool_to_state_args.get(call["name"], {}),
-                    self._tool_to_store_arg.get(call["name"]),
-                    self._tool_to_runtime_arg.get(call["name"]),
-                )
+                injected = self._injected_args.get(call["name"])
+                filtered_errors = _filter_validation_errors(exc, injected)
                 # Use original call["args"] without injected values for error reporting
                 raise ToolInvocationError(
                     call["name"], exc, call["args"], filtered_errors
@@ -1001,12 +1069,8 @@ class ToolNode(RunnableCallable):
                 response = await tool.ainvoke(call_args, config)
             except ValidationError as exc:
                 # Filter out errors for injected arguments
-                filtered_errors = _filter_validation_errors(
-                    exc,
-                    self._tool_to_state_args.get(call["name"], {}),
-                    self._tool_to_store_arg.get(call["name"]),
-                    self._tool_to_runtime_arg.get(call["name"]),
-                )
+                injected = self._injected_args.get(call["name"])
+                filtered_errors = _filter_validation_errors(exc, injected)
                 # Use original call["args"] without injected values for error reporting
                 raise ToolInvocationError(
                     call["name"], exc, call["args"], filtered_errors
@@ -1199,86 +1263,6 @@ class ToolNode(RunnableCallable):
             return input["state"]
         return input
 
-    def _inject_state(
-        self,
-        tool_call: ToolCall,
-        state: list[AnyMessage] | dict[str, Any] | BaseModel,
-    ) -> ToolCall:
-        state_args = self._tool_to_state_args[tool_call["name"]]
-
-        if state_args and isinstance(state, list):
-            required_fields = list(state_args.values())
-            if (
-                len(required_fields) == 1 and required_fields[0] == self._messages_key
-            ) or required_fields[0] is None:
-                state = {self._messages_key: state}
-            else:
-                err_msg = (
-                    f"Invalid input to ToolNode. Tool {tool_call['name']} requires "
-                    f"graph state dict as input."
-                )
-                if any(state_field for state_field in state_args.values()):
-                    required_fields_str = ", ".join(f for f in required_fields if f)
-                    err_msg += f" State should contain fields {required_fields_str}."
-                raise ValueError(err_msg)
-
-        if isinstance(state, dict):
-            tool_state_args = {
-                tool_arg: state[state_field] if state_field else state
-                for tool_arg, state_field in state_args.items()
-            }
-        else:
-            tool_state_args = {
-                tool_arg: getattr(state, state_field) if state_field else state
-                for tool_arg, state_field in state_args.items()
-            }
-
-        tool_call["args"] = {
-            **tool_call["args"],
-            **tool_state_args,
-        }
-        return tool_call
-
-    def _inject_store(self, tool_call: ToolCall, store: BaseStore | None) -> ToolCall:
-        store_arg = self._tool_to_store_arg[tool_call["name"]]
-        if not store_arg:
-            return tool_call
-
-        if store is None:
-            msg = (
-                "Cannot inject store into tools with InjectedStore annotations - "
-                "please compile your graph with a store."
-            )
-            raise ValueError(msg)
-
-        tool_call["args"] = {
-            **tool_call["args"],
-            store_arg: store,
-        }
-        return tool_call
-
-    def _inject_runtime(
-        self, tool_call: ToolCall, tool_runtime: ToolRuntime
-    ) -> ToolCall:
-        """Inject ToolRuntime into tool call arguments.
-
-        Args:
-            tool_call: The tool call to inject runtime into.
-            tool_runtime: The ToolRuntime instance to inject.
-
-        Returns:
-            The tool call with runtime injected if needed.
-        """
-        runtime_arg = self._tool_to_runtime_arg.get(tool_call["name"])
-        if not runtime_arg:
-            return tool_call
-
-        tool_call["args"] = {
-            **tool_call["args"],
-            runtime_arg: tool_runtime,
-        }
-        return tool_call
-
     def _inject_tool_args(
         self,
         tool_call: ToolCall,
@@ -1317,12 +1301,64 @@ class ToolNode(RunnableCallable):
         if tool_call["name"] not in self.tools_by_name:
             return tool_call
 
+        injected = self._injected_args.get(tool_call["name"])
+        if not injected:
+            return tool_call
+
         tool_call_copy: ToolCall = copy(tool_call)
-        tool_call_with_state = self._inject_state(tool_call_copy, tool_runtime.state)
-        tool_call_with_store = self._inject_store(
-            tool_call_with_state, tool_runtime.store
-        )
-        return self._inject_runtime(tool_call_with_store, tool_runtime)
+        injected_args = {}
+
+        # Inject state
+        if injected.state:
+            state = tool_runtime.state
+            # Handle list state by converting to dict
+            if isinstance(state, list):
+                required_fields = list(injected.state.values())
+                if (
+                    len(required_fields) == 1
+                    and required_fields[0] == self._messages_key
+                ) or required_fields[0] is None:
+                    state = {self._messages_key: state}
+                else:
+                    err_msg = (
+                        f"Invalid input to ToolNode. Tool {tool_call['name']} requires "
+                        f"graph state dict as input."
+                    )
+                    if any(state_field for state_field in injected.state.values()):
+                        required_fields_str = ", ".join(f for f in required_fields if f)
+                        err_msg += (
+                            f" State should contain fields {required_fields_str}."
+                        )
+                    raise ValueError(err_msg)
+
+            # Extract state values
+            if isinstance(state, dict):
+                for tool_arg, state_field in injected.state.items():
+                    injected_args[tool_arg] = (
+                        state[state_field] if state_field else state
+                    )
+            else:
+                for tool_arg, state_field in injected.state.items():
+                    injected_args[tool_arg] = (
+                        getattr(state, state_field) if state_field else state
+                    )
+
+        # Inject store
+        if injected.store:
+            if tool_runtime.store is None:
+                msg = (
+                    "Cannot inject store into tools with InjectedStore annotations - "
+                    "please compile your graph with a store."
+                )
+                raise ValueError(msg)
+            injected_args[injected.store] = tool_runtime.store
+
+        # Inject runtime
+        if injected.runtime:
+            injected_args[injected.runtime] = tool_runtime
+
+        tool_call_copy["args"] = {**tool_call_copy["args"], **injected_args}
+        return tool_call_copy
 
     def _validate_tool_command(
         self,
@@ -1715,120 +1751,86 @@ def _is_injection(
     return False
 
 
-def _get_state_args(tool: BaseTool) -> dict[str, str | None]:
-    """Extract state injection mappings from tool annotations.
-
-    This function analyzes a tool's input schema to identify arguments that should
-    be injected with graph state. It processes InjectedState annotations to build
-    a mapping of tool argument names to state field names.
+def _get_injection_from_type(
+    type_: Any, injection_type: type[InjectedState | InjectedStore | ToolRuntime]
+) -> Any | None:
+    """Extract injection instance from a type annotation.
 
     Args:
-        tool: The tool to analyze for state injection requirements.
+        type_: The type annotation to check.
+        injection_type: The injection type to look for.
 
     Returns:
-        A dictionary mapping tool argument names to state field names. If a field
-        name is None, the entire state should be injected for that argument.
+        The injection instance if found, True if injection marker found without instance, None otherwise.
     """
-    full_schema = tool.get_input_schema()
-    tool_args_to_state_fields: dict = {}
+    type_args = get_args(type_)
+    matches = [arg for arg in type_args if _is_injection(arg, injection_type)]
 
-    for name, type_ in get_all_basemodel_annotations(full_schema).items():
-        injections = [
-            type_arg
-            for type_arg in get_args(type_)
-            if _is_injection(type_arg, InjectedState)
-        ]
-        if len(injections) > 1:
-            msg = (
-                "A tool argument should not be annotated with InjectedState more than "
-                f"once. Received arg {name} with annotations {injections}."
-            )
-            raise ValueError(msg)
-        if len(injections) == 1:
-            injection = injections[0]
-            if isinstance(injection, InjectedState) and injection.field:
-                tool_args_to_state_fields[name] = injection.field
-            else:
-                tool_args_to_state_fields[name] = None
-        else:
-            pass
-    return tool_args_to_state_fields
+    if len(matches) > 1:
+        msg = (
+            f"A tool argument should not be annotated with {injection_type.__name__} "
+            f"more than once. Found: {matches}"
+        )
+        raise ValueError(msg)
 
-
-def _get_store_arg(tool: BaseTool) -> str | None:
-    """Extract store injection argument from tool annotations.
-
-    This function analyzes a tool's input schema to identify the argument that
-    should be injected with the graph store. Only one store argument is supported
-    per tool.
-
-    Args:
-        tool: The tool to analyze for store injection requirements.
-
-    Returns:
-        The name of the argument that should receive the store injection, or None
-        if no store injection is required.
-
-    Raises:
-        ValueError: If a tool argument has multiple InjectedStore annotations.
-    """
-    full_schema = tool.get_input_schema()
-    for name, type_ in get_all_basemodel_annotations(full_schema).items():
-        injections = [
-            type_arg
-            for type_arg in get_args(type_)
-            if _is_injection(type_arg, InjectedStore)
-        ]
-        if len(injections) > 1:
-            msg = (
-                "A tool argument should not be annotated with InjectedStore more than "
-                f"once. Received arg {name} with annotations {injections}."
-            )
-            raise ValueError(msg)
-        if len(injections) == 1:
-            return name
+    if len(matches) == 1:
+        return matches[0]
+    elif _is_injection(type_, injection_type):
+        return True
 
     return None
 
 
-def _get_runtime_arg(tool: BaseTool) -> str | None:
-    """Extract runtime injection argument from tool annotations.
+def _get_all_injected_args(tool: BaseTool) -> _InjectedArgs:
+    """Extract all injected arguments from tool in a single pass.
 
-    This function analyzes a tool's input schema to identify the argument that
-    should be injected with the ToolRuntime instance. Only one runtime argument
-    is supported per tool.
+    This function analyzes both the tool's input schema and function signature
+    to identify all arguments that should be injected (state, store, runtime).
 
     Args:
-        tool: The tool to analyze for runtime injection requirements.
+        tool: The tool to analyze for injection requirements.
 
     Returns:
-        The name of the argument that should receive the runtime injection, or None
-        if no runtime injection is required.
-
-    Raises:
-        ValueError: If a tool argument has multiple ToolRuntime annotations.
+        _InjectedArgs structure containing all detected injections.
     """
+    # Get annotations from both schema and function signature
     full_schema = tool.get_input_schema()
-    for name, type_ in get_all_basemodel_annotations(full_schema).items():
-        # Check if the parameter name is "runtime" (regardless of type)
+    schema_annotations = get_all_basemodel_annotations(full_schema)
+
+    func = getattr(tool, "func", None) or getattr(tool, "coroutine", None)
+    func_annotations = get_type_hints(func, include_extras=True) if func else {}
+
+    # Combine both annotation sources, preferring schema annotations
+    # In the future, we might want to add more restrictions here...
+    all_annotations = {**func_annotations, **schema_annotations}
+
+    # Track injected args
+    state_args: dict[str, str | None] = {}
+    store_arg: str | None = None
+    runtime_arg: str | None = None
+
+    for name, type_ in all_annotations.items():
+        # Check for runtime (special case: parameter named "runtime")
         if name == "runtime":
-            return name
-        # Check if the type itself is ToolRuntime (direct usage)
-        if _is_injection(type_, ToolRuntime):
-            return name
-        # Check if ToolRuntime is in Annotated args
-        injections = [
-            type_arg
-            for type_arg in get_args(type_)
-            if _is_injection(type_arg, ToolRuntime)
-        ]
-        if len(injections) > 1:
-            msg = (
-                "A tool argument should not be annotated with ToolRuntime more than "
-                f"once. Received arg {name} with annotations {injections}."
-            )
-            raise ValueError(msg)
-        if len(injections) == 1:
-            return name
+            runtime_arg = name
 
-    return None
+        # Check for InjectedState
+        if state_inj := _get_injection_from_type(type_, InjectedState):
+            if isinstance(state_inj, InjectedState) and state_inj.field:
+                state_args[name] = state_inj.field
+            else:
+                state_args[name] = None
+
+        # Check for InjectedStore
+        if _get_injection_from_type(type_, InjectedStore):
+            store_arg = name
+
+        # Check for ToolRuntime
+        if _get_injection_from_type(type_, ToolRuntime):
+            runtime_arg = name
+
+    return _InjectedArgs(
+        state=state_args,
+        store=store_arg,
+        runtime=runtime_arg,
+    )

libs/prebuilt/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "langgraph-prebuilt"
-version = "1.0.4"
+version = "1.0.5"
 description = "Library with high-level APIs for creating and executing LangGraph agents and tools."
 authors = []
 requires-python = ">=3.10"

libs/prebuilt/tests/test_on_tool_call.py

@@ -130,11 +130,17 @@ def test_modify_arguments() -> None:
         execute: Callable[[ToolCallRequest], ToolMessage | Command],
     ) -> ToolMessage | Command:
         """Handler that doubles the input arguments."""
-        # Modify the arguments
-        request.tool_call["args"]["a"] *= 2
-        request.tool_call["args"]["b"] *= 2
-
-        return execute(request)
+        # Modify the arguments using override method
+        modified_call = {
+            **request.tool_call,
+            "args": {
+                **request.tool_call["args"],
+                "a": request.tool_call["args"]["a"] * 2,
+                "b": request.tool_call["args"]["b"] * 2,
+            },
+        }
+        modified_request = request.override(tool_call=modified_call)
+        return execute(modified_request)
 
     tool_node = ToolNode([add], wrap_tool_call=modify_args_handler)
 
@@ -362,10 +368,17 @@ async def test_handler_with_async_execution() -> None:
         execute: Callable[[ToolCallRequest], ToolMessage | Command],
     ) -> ToolMessage | Command:
         """Handler that modifies arguments."""
-        # Add 10 to both arguments
-        request.tool_call["args"]["a"] += 10
-        request.tool_call["args"]["b"] += 10
-        return execute(request)
+        # Add 10 to both arguments using override method
+        modified_call = {
+            **request.tool_call,
+            "args": {
+                **request.tool_call["args"],
+                "a": request.tool_call["args"]["a"] + 10,
+                "b": request.tool_call["args"]["b"] + 10,
+            },
+        }
+        modified_request = request.override(tool_call=modified_call)
+        return execute(modified_request)
 
     tool_node = ToolNode([async_add], wrap_tool_call=modifying_handler)
 
@@ -1305,3 +1318,64 @@ async def test_state_extraction_with_tool_call_with_context_async() -> None:
     assert state_seen[0] == actual_state
     assert "__type" not in state_seen[0]
     assert "tool_call" not in state_seen[0]
+
+
+def test_tool_call_request_is_frozen() -> None:
+    """Test that ToolCallRequest raises deprecation warnings on direct attribute reassignment."""
+    tool_call: ToolCall = {"name": "add", "args": {"a": 1, "b": 2}, "id": "call_1"}
+    state: dict = {"messages": []}
+    runtime = None
+
+    request = ToolCallRequest(
+        tool_call=tool_call, tool=add, state=state, runtime=runtime
+    )  # type: ignore[arg-type]
+
+    # Test that direct attribute reassignment raises DeprecationWarning
+    with pytest.warns(
+        DeprecationWarning,
+        match="Setting attribute 'tool_call' on ToolCallRequest is deprecated",
+    ):
+        request.tool_call = {"name": "other", "args": {}, "id": "call_2"}  # type: ignore[misc]
+
+    with pytest.warns(
+        DeprecationWarning,
+        match="Setting attribute 'tool' on ToolCallRequest is deprecated",
+    ):
+        request.tool = None  # type: ignore[misc]
+
+    with pytest.warns(
+        DeprecationWarning,
+        match="Setting attribute 'state' on ToolCallRequest is deprecated",
+    ):
+        request.state = {}  # type: ignore[misc]
+
+    with pytest.warns(
+        DeprecationWarning,
+        match="Setting attribute 'runtime' on ToolCallRequest is deprecated",
+    ):
+        request.runtime = None  # type: ignore[misc]
+
+    # Test that override method works correctly
+    new_tool_call: ToolCall = {
+        "name": "multiply",
+        "args": {"x": 5, "y": 10},
+        "id": "call_3",
+    }
+
+    # Original request should be unchanged (note: it was modified by the warnings tests above)
+    # So we create a fresh request to test override properly
+    fresh_request = ToolCallRequest(
+        tool_call=tool_call, tool=add, state=state, runtime=runtime
+    )  # type: ignore[arg-type]
+    fresh_new_request = fresh_request.override(tool_call=new_tool_call)
+
+    # Original request should be unchanged
+    assert fresh_request.tool_call == tool_call
+    assert fresh_request.tool_call["name"] == "add"
+
+    # New request should have the updated tool_call
+    assert fresh_new_request.tool_call == new_tool_call
+    assert fresh_new_request.tool_call["name"] == "multiply"
+    assert fresh_new_request.tool == add  # Other fields should remain the same
+    assert fresh_new_request.state == state
+    assert fresh_new_request.runtime is None

libs/prebuilt/tests/test_react_agent.py

@@ -53,7 +53,6 @@ from langgraph.prebuilt.chat_agent_executor import (
 from langgraph.prebuilt.tool_node import (
     InjectedState,
     InjectedStore,
-    _get_state_args,
     _infer_handled_types,
 )
 from tests.any_str import AnyStr
@@ -1084,21 +1083,6 @@ async def test_return_direct(version: str) -> None:
     ]
 
 
-def test__get_state_args() -> None:
-    class Schema1(BaseModel):
-        a: Annotated[str, InjectedState]
-
-    class Schema2(Schema1):
-        b: Annotated[int, InjectedState("bar")]
-
-    @dec_tool(args_schema=Schema2)
-    def foo(a: str, b: int) -> float:
-        """return"""
-        return 0.0
-
-    assert _get_state_args(foo) == {"a": None, "b": "bar"}
-
-
 def test_inspect_react() -> None:
     model = FakeToolCallingModel(tool_calls=[])
     agent = create_react_agent(model, [])

libs/prebuilt/tests/test_tool_node.py

@@ -42,6 +42,7 @@ from langgraph.prebuilt import (
 from langgraph.prebuilt.tool_node import (
     TOOL_CALL_ERROR_TEMPLATE,
     ToolInvocationError,
+    ToolRuntime,
     tools_condition,
 )
 
@@ -1610,3 +1611,252 @@ def test_tool_node_stream_writer() -> None:
             },
         ),
     ]
+
+
+def test_tool_call_request_setattr_deprecation_warning():
+    """Test that ToolCallRequest raises a deprecation warning on direct attribute modification."""
+    import warnings
+
+    from langgraph.prebuilt.tool_node import ToolCallRequest
+
+    # Create a mock ToolCall
+    tool_call = {"name": "test", "args": {"a": 1}, "id": "call_1", "type": "tool_call"}
+
+    # Create a ToolCallRequest
+    request = ToolCallRequest(
+        tool_call=tool_call,
+        tool=None,
+        state={"messages": []},
+        runtime=None,
+    )
+
+    # Test 1: Direct attribute assignment should raise deprecation warning but still work
+    with pytest.warns(DeprecationWarning, match="deprecated.*override"):
+        request.tool_call = {"name": "other", "args": {}, "id": "call_2"}
+
+    # Verify the attribute was actually modified
+    assert request.tool_call == {"name": "other", "args": {}, "id": "call_2"}
+
+    # Reset for further tests
+    with warnings.catch_warnings():
+        warnings.simplefilter("ignore")
+        request.tool_call = tool_call
+
+    # Test 2: override method should work without warnings
+    with warnings.catch_warnings(record=True) as w:
+        warnings.simplefilter("always")
+        new_tool_call = {
+            "name": "new_tool",
+            "args": {"b": 2},
+            "id": "call_3",
+            "type": "tool_call",
+        }
+        new_request = request.override(tool_call=new_tool_call)
+
+        # Verify no warning was raised
+        assert len(w) == 0
+
+    # Verify original is unchanged
+    assert request.tool_call == tool_call
+
+    # Verify new request has updated values
+    assert new_request.tool_call == new_tool_call
+
+    # Test 3: Initialization should not trigger warning
+    with warnings.catch_warnings(record=True) as w:
+        warnings.simplefilter("always")
+        ToolCallRequest(
+            tool_call=tool_call,
+            tool=None,
+            state={"messages": []},
+            runtime=None,
+        )
+        # Verify no warning was raised during initialization
+        assert len(w) == 0
+
+
+async def test_tool_node_inject_async_all_types_signature_only() -> None:
+    """Test all injection types without @tool decorator."""
+    store = InMemoryStore()
+    namespace = ("test",)
+    store.put(namespace, "test_key", {"store_data": "from_store"})
+
+    class TestState(TypedDict):
+        messages: list
+        foo: str
+        bar: int
+
+    async def comprehensive_async_tool(
+        x: int,
+        whole_state: Annotated[TestState, InjectedState],
+        foo_field: Annotated[str, InjectedState("foo")],
+        store: Annotated[BaseStore, InjectedStore()],
+        runtime: ToolRuntime,
+    ) -> str:
+        """Async tool that uses all injection types."""
+        bar_from_whole = whole_state["bar"]
+        foo_value = foo_field
+        store_val = store.get(namespace, "test_key").value["store_data"]
+        foo_from_runtime = runtime.state["foo"]
+        tool_call_id = runtime.tool_call_id
+
+        return (
+            f"x={x}, "
+            f"bar_from_whole={bar_from_whole}, "
+            f"foo_field={foo_value}, "
+            f"store={store_val}, "
+            f"foo_from_runtime={foo_from_runtime}, "
+            f"tool_call_id={tool_call_id}"
+        )
+
+    node = ToolNode([comprehensive_async_tool], handle_tool_errors=True)
+    tool_call = {
+        "name": "comprehensive_async_tool",
+        "args": {"x": 42},
+        "id": "test_call_123",
+        "type": "tool_call",
+    }
+    msg = AIMessage("hi?", tool_calls=[tool_call])
+
+    config = _create_config_with_runtime(store=store)
+    result = await node.ainvoke(
+        {"messages": [msg], "foo": "foo_value", "bar": 99}, config=config
+    )
+
+    tool_message = result["messages"][-1]
+    assert tool_message.content == (
+        "x=42, "
+        "bar_from_whole=99, "
+        "foo_field=foo_value, "
+        "store=from_store, "
+        "foo_from_runtime=foo_value, "
+        "tool_call_id=test_call_123"
+    )
+
+
+async def test_tool_node_inject_async_all_types_with_decorator() -> None:
+    """Test all injection types with @tool decorator."""
+    store = InMemoryStore()
+    namespace = ("test",)
+    store.put(namespace, "test_key", {"store_data": "from_store"})
+
+    class TestState(TypedDict):
+        messages: list
+        foo: str
+        bar: int
+
+    @dec_tool
+    async def comprehensive_async_tool(
+        x: int,
+        whole_state: Annotated[TestState, InjectedState],
+        foo_field: Annotated[str, InjectedState("foo")],
+        store: Annotated[BaseStore, InjectedStore()],
+        runtime: ToolRuntime,
+    ) -> str:
+        """Async tool that uses all injection types."""
+        bar_from_whole = whole_state["bar"]
+        foo_value = foo_field
+        store_val = store.get(namespace, "test_key").value["store_data"]
+        foo_from_runtime = runtime.state["foo"]
+        tool_call_id = runtime.tool_call_id
+
+        return (
+            f"x={x}, "
+            f"bar_from_whole={bar_from_whole}, "
+            f"foo_field={foo_value}, "
+            f"store={store_val}, "
+            f"foo_from_runtime={foo_from_runtime}, "
+            f"tool_call_id={tool_call_id}"
+        )
+
+    node = ToolNode([comprehensive_async_tool], handle_tool_errors=True)
+    tool_call = {
+        "name": "comprehensive_async_tool",
+        "args": {"x": 42},
+        "id": "test_call_456",
+        "type": "tool_call",
+    }
+    msg = AIMessage("hi?", tool_calls=[tool_call])
+
+    config = _create_config_with_runtime(store=store)
+    result = await node.ainvoke(
+        {"messages": [msg], "foo": "foo_value", "bar": 99}, config=config
+    )
+
+    tool_message = result["messages"][-1]
+    assert tool_message.content == (
+        "x=42, "
+        "bar_from_whole=99, "
+        "foo_field=foo_value, "
+        "store=from_store, "
+        "foo_from_runtime=foo_value, "
+        "tool_call_id=test_call_456"
+    )
+
+
+async def test_tool_node_inject_async_all_types_with_schema() -> None:
+    """Test all injection types with explicit schema."""
+    store = InMemoryStore()
+    namespace = ("test",)
+    store.put(namespace, "test_key", {"store_data": "from_store"})
+
+    class TestState(TypedDict):
+        messages: list
+        foo: str
+        bar: int
+
+    class ComprehensiveToolSchema(BaseModel):
+        model_config = {"arbitrary_types_allowed": True}
+        x: int
+        whole_state: Annotated[TestState, InjectedState]
+        foo_field: Annotated[str, InjectedState("foo")]
+        store: Annotated[BaseStore, InjectedStore()]
+        runtime: ToolRuntime
+
+    @dec_tool(args_schema=ComprehensiveToolSchema)
+    async def comprehensive_async_tool(
+        x: int,
+        whole_state: Annotated[TestState, InjectedState],
+        foo_field: Annotated[str, InjectedState("foo")],
+        store: Annotated[BaseStore, InjectedStore()],
+        runtime: ToolRuntime,
+    ) -> str:
+        """Async tool that uses all injection types."""
+        bar_from_whole = whole_state["bar"]
+        foo_value = foo_field
+        store_val = store.get(namespace, "test_key").value["store_data"]
+        foo_from_runtime = runtime.state["foo"]
+        tool_call_id = runtime.tool_call_id
+
+        return (
+            f"x={x}, "
+            f"bar_from_whole={bar_from_whole}, "
+            f"foo_field={foo_value}, "
+            f"store={store_val}, "
+            f"foo_from_runtime={foo_from_runtime}, "
+            f"tool_call_id={tool_call_id}"
+        )
+
+    node = ToolNode([comprehensive_async_tool], handle_tool_errors=True)
+    tool_call = {
+        "name": "comprehensive_async_tool",
+        "args": {"x": 42},
+        "id": "test_call_789",
+        "type": "tool_call",
+    }
+    msg = AIMessage("hi?", tool_calls=[tool_call])
+
+    config = _create_config_with_runtime(store=store)
+    result = await node.ainvoke(
+        {"messages": [msg], "foo": "foo_value", "bar": 99}, config=config
+    )
+
+    tool_message = result["messages"][-1]
+    assert tool_message.content == (
+        "x=42, "
+        "bar_from_whole=99, "
+        "foo_field=foo_value, "
+        "store=from_store, "
+        "foo_from_runtime=foo_value, "
+        "tool_call_id=test_call_789"
+    )

libs/prebuilt/uv.lock

@@ -467,7 +467,7 @@ test = [
 
 [[package]]
 name = "langgraph-prebuilt"
-version = "1.0.4"
+version = "1.0.5"
 source = { editable = "." }
 dependencies = [
     { name = "langchain-core" },