9il dZddlmZddlZddlZddlZddlmZmZddl m Z m Z ddl m Z m Z ddlmZddlmZmZmZmZmZmZmZmZmZmZmZdd lmZmZmZm Z m!Z!m"Z"dd l#m$Z$m%Z%m&Z&dd l'm(Z(m)Z)dd l'm*Z+dd l,m-Z-m.Z.m/Z/m0Z0ddl1m2Z2ddl3m4Z4ddl5m6Z6ddl7m8Z8ddl9m:Z:m;Z;mZ>m?Z?ddl@mAZAmBZBerddlmCZCddlDmEZEddlFmGZGeAdeHZIeAddZJdZKdZLdZMdZNGdd ed!"ZOe Gd#d$ZPeePeePge!e:zfge!e:zfZQ eePeePgee!e:zfgee!e:zfZR Gd%d&eZSdd+ZWd?d,ZX d@d-ZYe Gd.d/ZZGd0d1e2Z[ dA dBd2Z\e Gd3d4e/eeJeIfZ]Gd5d6e)Z^Gd7d8e)Z_ dCd9Z` dDd:ZadEd;Zby)FaTool execution node for LangGraph workflows. This module provides prebuilt functionality for executing tools in LangGraph. Tools are functions that models can call to interact with external systems, APIs, databases, or perform computations. The module implements design patterns for: - Parallel execution of multiple tool calls for efficiency - Robust error handling with customizable error messages - State injection for tools that need access to graph state - Store injection for tools that need persistent storage - Command-based state updates for advanced control flow Key Components: - `ToolNode`: Main class for executing tools in LangGraph workflows - `InjectedState`: Annotation for injecting graph state into tools - `InjectedStore`: Annotation for injecting persistent store into tools - `ToolRuntime`: Runtime information for tools, bundling together `state`, `context`, `config`, `stream_writer`, `tool_call_id`, and `store` - `tools_condition`: Utility function for conditional routing based on tool calls Typical Usage: ```python from langchain_core.tools import tool from langchain.tools import ToolNode @tool def my_tool(x: int) -> str: return f"Result: {x}" tool_node = ToolNode([my_tool]) ``` ) annotationsN) AwaitableCallable)copydeepcopy) dataclassreplace) UnionType) TYPE_CHECKING AnnotatedAnyGenericLiteral TypedDictUnioncastget_args get_originget_type_hints) AIMessage AnyMessage RemoveMessageToolCall ToolMessageconvert_to_messages)RunnableConfigget_config_listget_executor_for_config)BaseToolInjectedToolArg)tool)TOOL_MESSAGE_BLOCK_TYPES ToolException_DirectlyInjectedToolArgget_all_basemodel_annotations)RunnableCallable) GraphBubbleUp)REMOVE_ALL_MESSAGES) BaseStore)CommandSend StreamWriter) BaseModelValidationError)TypeVarUnpack)Sequence)Runtime) ErrorDetailsStateT)defaultContextTzLError: {requested_tool} is not a valid tool, try one of [{available_tools}].z)Error: {error} Please fix your mistakes.zvError executing tool '{tool_name}' with kwargs {tool_kwargs} with error: {error} Please fix the error and try again.zuError invoking tool '{tool_name}' with kwargs {tool_kwargs} with error: {error} Please fix the error and try again.ceZdZUdZded<y)_ToolCallRequestOverridesz9Possible overrides for ToolCallRequest.override() method.r tool_callN__name__ __module__ __qualname____doc____annotations__f/var/www/html/backtest/airagagent/rag_env/lib/python3.12/site-packages/langgraph/prebuilt/tool_node.pyr8r8xs CrAr8F)totalcReZdZUdZded<ded<ded<ded <d d Z dd Zy )ToolCallRequestafTool execution request passed to tool call interceptors. Attributes: tool_call: Tool call dict with name, args, and id from model output. tool: BaseTool instance to be invoked, or None if tool is not registered with the `ToolNode`. When tool is `None`, interceptors can handle the request without validation. If the interceptor calls `execute()`, validation will occur and raise an error for unregistered tools. state: Agent state (`dict`, `list`, or `BaseModel`). runtime: LangGraph runtime context (optional, `None` if outside graph). rr9zBaseTool | Noner!r state ToolRuntimeruntimecddl}t|dr t||stj|||y|j d|dt dtj|||y)zRaise deprecation warning when setting attributes directly. Direct attribute assignment is deprecated. Use the `override()` method instead. rN__dataclass_fields__zSetting attribute 'zt' on ToolCallRequest is deprecated. Use the override() method instead to create a new instance with modified values.) stacklevel)warningshasattrobject __setattr__warnDeprecationWarning)selfnamevaluerMs rBrPzToolCallRequest.__setattr__sm t34GD$r?rPrXr@rArBrErE~s:   J 2&*<* *rArEc2eZdZUdZded<ded< ded<y) ToolCallWithContextaToolCall with additional context for graph state. This is an internal data structure meant to help the `ToolNode` accept tool calls with additional context (e.g. state) when dispatched using the Send API. The Send API is used in create_agent to distribute tool calls in parallel and support human-in-the-loop workflows where graph execution may be paused for an indefinite time. rr9z!Literal['tool_call_with_context']_ToolCallWithContext__typer rFNr:r@rArBr]r]s$  -- J6rAr]ct|ts"t|trtd|Dr|S t j |dS#t $rt|cYSwxYw)aXConvert tool output to `ToolMessage` content format. Handles `str`, `list[dict]` (content blocks), and arbitrary objects by attempting JSON serialization with fallback to str(). Args: output: Tool execution output of any type. Returns: String or list of content blocks suitable for `ToolMessage.content`. c3nK|]-}t|txr|jdtv/yw)typeN) isinstancedictgetr").0xs rB z%msg_content_output..9s5  q$  MAEE&M5M$M M s35F) ensure_ascii)rbrYlistalljsondumps Exception)outputs rBmsg_content_outputro+sc&#64     zz&u55 6{sA A$#A$c:eZdZdZ d dfd ZxZS)ToolInvocationErrorzAn error occurred while invoking a tool due to invalid arguments. This exception is only raised when invoking a tool using the `ToolNode`! c|kg}|D]R}djd|jddD}|jdd}|j|d|Td j|} n t|} tj ||| |_||_||_||_ ||_ t |1|j y) aiInitialize the ToolInvocationError. Args: tool_name: The name of the tool that failed. source: The exception that occurred. tool_kwargs: The keyword arguments that were passed to the tool. filtered_errors: Optional list of filtered validation errors excluding injected arguments. N.c32K|]}t|ywN)rY)relocs rBrgz/ToolInvocationError.__init__..ds"L3s8"Lsrvr@msgz Unknown errorz:  ) tool_name tool_kwargserror) joinrdappendrYTOOL_INVOCATION_ERROR_TEMPLATEformatmessageryrzsourcefiltered_errorssuper__init__) rSryrrzrerror_str_partsr{loc_strrwerror_display_str __class__s rBrzToolInvocationError.__init__Os"  & O( <(("Luyy7K"LLii7&&'"SE':; <!% / :  #F 5<<[@Q=  #& . &rAru) ryrYrr.rzdict[str, Any]rzlist[ErrorDetails] | NonerZr[)r;r<r=r>r __classcell__rs@rBrqrqIsG6: #'#' #'$ #' 3 #'  #'#'rArqc>t|tr |jS|)zDefault error handler for tool errors. If the tool is a tool invocation error, return its message. Otherwise, raise the error. )rbrqr)es rB_default_handle_tool_errorsrus !()yy GrAc"t|ttfs t|tr1t |t r!t jt|}|St|tr|}|St|r ||}|Sd|}t|)aGenerate error message content based on exception handling configuration. This function centralizes error message generation logic, supporting different error handling strategies configured via the `ToolNode`'s `handle_tool_errors` parameter. Args: e: The exception that occurred during tool execution. flag: Configuration for how to handle the error. Can be: - bool: If `True`, use default error template - str: Use this string as the error message - Callable: Call this function with the exception to get error message - tuple: Not used in this context (handled by caller) Returns: A string containing the error message to include in the `ToolMessage`. Raises: ValueError: If flag is not one of the supported types. !!! note The tuple case is handled by the caller through exception type checking, not by this function directly. )r{zVGot unexpected type of `handle_tool_error`. Expected bool, str or callable. Received: ) rbbooltuplera issubclassrmTOOL_CALL_ERROR_TEMPLATErreprrYcallable ValueError)rflagcontentrws rB_handle_tool_errorrsB$u &4:dI#>*11Q1@ N D#  N $q' N &&*V - orAcTtj|}t|jj }|r|dj dvrt |dk(r|d}n|d}t|}|j |vrt|j}|ttfvrMt|j}td|Dr t|Sd|jd}t|||j }t |j"vr|fSd|d}t|t fS) a(Infer exception types handled by a custom error handler function. This function analyzes the type annotations of a custom error handler to determine which exception types it's designed to handle. This enables type-safe error handling where only specific exceptions are caught and processed by the handler. Args: handler: A callable that takes an exception and returns an error message string. The first parameter (after self/cls if present) should be type-annotated with the exception type(s) to handle. Returns: A tuple of exception types that the handler can process. Returns (Exception,) if no specific type information is available for backward compatibility. Raises: ValueError: If the handler's annotation contains non-Exception types or if Union types contain non-Exception types. !!! note This function supports both single exception types and Union types for handlers that need to handle multiple exception types differently. r)rSclsrKc3<K|]}t|tywru)rrm)reargs rBrgz'_infer_handled_types..sBcz#y1BzAll types in the error handler error annotation must be Exception types. For example, `def custom_handler(e: Union[ValueError, TypeError])`. Got 'z ' instead.aArbitrary types are not supported in the error handler signature. Please annotate the error with either a specific Exception type or a union of Exception types. For example, `def custom_handler(e: ValueError)` or `def custom_handler(e: Union[ValueError, TypeError])`. Got ')inspect signatureri parametersvaluesrTlenrr annotationrr rrjrrrm__mro__) handlersigparams first_param type_hintsoriginargsrwexception_types rB_infer_handled_typesrs90   G $C #..'') *F !9>>_ ,V1A )K )K#G,   z ) 6 67F%++ 6 67BTBB ;&(223:? !o%' (8(89NN222&(( ''z 3 S/ ! <rAcLt}|r|jr)|j|jj|jr|j |j|j r|j |j g}|jD]y}|ds |dd|vsi|}t|jdtr2|d}|jDcic] \}}||vs ||} }}| |d<|j|{|Scc}}w)aFilter validation errors to only include LLM-controlled arguments. When a tool invocation fails validation, only errors for arguments that the LLM controls should be included in error messages. This ensures the LLM receives focused, actionable feedback about parameters it can actually fix. System-injected arguments (state, store, runtime) are filtered out since the LLM has no control over them. This function also removes injected argument values from the `input` field in error details, ensuring that only LLM-provided arguments appear in error messages. Args: validation_error: The Pydantic ValidationError raised during tool invocation. injected_args: The _InjectedArgs structure containing all injected arguments, or None if there are no injected arguments. Returns: List of ErrorDetails containing only errors for LLM-controlled arguments, with system-injected argument values removed from the input field. rvrinput) setrFupdatekeysstoreaddrHerrorsrbrdrcitemsr}) validation_error injected_argsinjected_arg_namesrr{ error_copy input_dictkv input_copys rB_filter_validation_errorsrs)2$'5     % %m&9&9&>&>&@ A     " "=#6#6 7   " "=#8#8 9*,O!((*/ D c0eZdZUdZded<ded<ded<y) _InjectedArgsaSInternal 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 ) ``` zdict[str, str | None]rF str | NonerrHNr:r@rArBrr,s-^ !  rArcleZdZUdZdZded<ddedddd dfdZedd Z dd Z dd Z dd Z dd Z ddZ ddZ ddZ ddZddZ ddZ ddZ d dZxZS)!ToolNodeu"A node for executing tools in LangGraph workflows. Handles tool execution patterns including function calls, state injection, persistent storage, and control flow. Manages parallel execution, error handling. Input Formats: 1. Graph state with `messages` key that has a list of messages: - Common representation for agentic workflows - Supports custom messages key via `messages_key` parameter 2. **Message List**: `[AIMessage(..., tool_calls=[...])]` - List of messages with tool calls in the last AIMessage 3. **Direct Tool Calls**: `[{"name": "tool", "args": {...}, "id": "1", "type": "tool_call"}]` - Bypasses message parsing for direct tool execution - For programmatic tool invocation and testing Output Formats: Output format depends on input type and tool behavior: **For Regular tools**: - Dict input → `{"messages": [ToolMessage(...)]}` - List input → `[ToolMessage(...)]` **For Command tools**: - Returns `[Command(...)]` or mixed list with regular tool outputs - `Command` can update state, trigger navigation, or send messages Args: tools: A sequence of tools that can be invoked by this node. Supports: - **BaseTool instances**: Tools with schemas and metadata - **Plain functions**: Automatically converted to tools with inferred schemas name: The name identifier for this node in the graph. Used for debugging and visualization. tags: Optional metadata tags to associate with the node for filtering and organization. handle_tool_errors: Configuration for error handling during tool execution. Supports multiple strategies: - `True`: Catch all errors and return a `ToolMessage` with the default error template containing the exception details. - `str`: Catch all errors and return a `ToolMessage` with this custom error message string. - `type[Exception]`: Only catch exceptions with the specified type and return the default error message for it. - `tuple[type[Exception], ...]`: Only catch exceptions with the specified types and return default error messages for them. - `Callable[..., str]`: Catch exceptions matching the callable's signature and return the string result of calling it with the exception. - `False`: Disable error handling entirely, allowing exceptions to propagate. Defaults to a callable that: - Catches tool invocation errors (due to invalid arguments provided by the model) and returns a descriptive error message - Ignores tool execution errors (they will be re-raised) messages_key: The key in the state dictionary that contains the message list. This same key will be used for the output `ToolMessage` objects. Allows custom state schemas with different message field names. Examples: Basic usage: ```python from langchain.tools import ToolNode from langchain_core.tools import tool @tool def calculator(a: int, b: int) -> int: """Add two numbers.""" return a + b tool_node = ToolNode([calculator]) ``` State injection: ```python from typing_extensions import Annotated from langchain.tools import InjectedState @tool def context_tool(query: str, state: Annotated[dict, InjectedState]) -> str: """Some tool that uses state.""" return f"Query: {query}, Messages: {len(state['messages'])}" tool_node = ToolNode([context_tool]) ``` Error handling: ```python def handle_errors(e: ValueError) -> str: return "Invalid input provided" tool_node = ToolNode([my_tool], handle_tool_errors=handle_errors) ``` toolsrYrTNmessages)rTtagshandle_tool_errors messages_keywrap_tool_callawrap_tool_callc~t ||j|j||di|_i|_||_||_||_||_ |D]e}t|tsttd|} n|} | |j| j<t| |j | j<gy)a Initialize `ToolNode` with tools and configuration. Args: tools: Sequence of tools to make available for execution. name: Node name for graph identification. tags: Optional metadata tags. handle_tool_errors: Error handling configuration. messages_key: State key containing messages. wrap_tool_call: Sync wrapper function to intercept tool execution. Receives ToolCallRequest and execute callable, returns ToolMessage or Command. Enables retries, caching, request modification, and control flow. awrap_tool_call: Async wrapper function to intercept tool execution. If not provided, falls back to wrap_tool_call for async execution. F)rTrtraceztype[BaseTool]N)rr_func_afunc_tools_by_name_injected_args_handle_tool_errors _messages_key_wrap_tool_call_awrap_tool_callrbr create_toolrrT_get_all_injected_args) rSrrTrrrrrr!tool_rs rBrzToolNode.__init__s: T[[t$eT358:#5 )- / LDdH-#D)94$@A.3D   +.DU.KD   + LrAc|jS)z,Mapping from tool name to BaseTool instance.)r)rSs rB tools_by_namezToolNode.tools_by_names"""rAc |j|\}}t|t|}g}t||dD]Y\}} |j |} t | |d| |j |j|j} |j| [|gt|z} t|5} t| j|j|| |}ddd|j|S#1swYxYwNF)strictid)rF tool_call_idconfigcontextr stream_writer) _parse_inputrrzip_extract_staterGrrrr}rrimap_run_one_combine_tool_outputs)rSrrrH tool_calls input_type config_list tool_runtimescallcfgrF tool_runtime input_typesexecutoroutputss rBrzToolNode._funcs "&!2!25!9 J%fc*o>  ZUC /ID#''.E&!$Zmm%33 L   . /"lS_4 $V ,  T]]J ]SG  ))':>>   s 0(C22C;c K|j|\}}t|t|}g}t||dD]Y\}} |j |} t | |d| |j |j|j} |j| [g} t||dD]'\}} | j|j||| )tj| d{} |j| |S7wr)rrrrrrGrrrr} _arun_oneasynciogatherr)rSrrrHrrrrrrrFrcorosrs rBrzToolNode._afunc%s "&!2!25!9 J%fc*o>  ZUC /ID#''.E&!$Zmm%33 L   . /"%j-"N I D, LLj,G H I..))':>>/sC%D'C>(Dcrtd|Ds|dk(r|S|j|iSg}d}|D]}t|tr|jtj urt|j trstd|j DrW|r/t|td|j |j z}ttj |j }|j||j|dk(r|gn|j|gi|r|j||S)Nc3<K|]}t|tywru)rbr*)rerns rBrgz1ToolNode._combine_tool_outputs..KsE6:fg.Erric3<K|]}t|tywru)rbr+)resends rBrgz1ToolNode._combine_tool_outputs..\sKtJtT2Krz list[Send])goto)graphr) anyrrbr*rPARENTrrirjr rr})rSrrcombined_outputsparent_commandrns rBrzToolNode._combine_tool_outputsDsEWEE(F27 U9K9KW8U U   *. F&'*LLGNN2"6;;5Kv{{KK%)0*!%lN4G4G!H6;;!V* *1w~~FKK)X$++F3 '' *f 4VH4;M;MPVx:X# *   # #N 3rAc|j}|j}|)|j|x}r|Sd|dd}t||j ||j }i|ddi} |j | |} t| t.r|j1| |j|St| t,r&t3d t5| j6| _| Sd|dd t!| }t|#t$rD} |jj|d} t| | } t|d| |d| | d} ~ wwxYw#t$rt$r}t|jt r(t#|jtr|jf}nst|jt$r |j}nLt'|jr0t|jt st)|j}ntf}|jr t||st+||j}t-||d|d d cYd}~Sd}~wwxYw)aUExecute tool call with configured error handling. Args: request: Tool execution request. input_type: Input format. config: Runnable configuration. Returns: ToolMessage or Command. Raises: Exception: If tool fails and handle_tool_errors is False. NTool rT is not registered with ToolNoderar9rrrr{rrTrstatus str | list returned unexpected type: )r9r!_validate_tool_call TypeError_inject_tool_argsrHinvoker.rrdrrqr'rmrbrrarrrrrrr*_validate_tool_commandrrorrSrequestrrrr!invalid_tool_messagerw injected_call call_argsresponseexcinjectedrr handled_typesrs rB_execute_tool_synczToolNode._execute_tool_syncps`&  || <'+'?'?'EE#E++$v,'GHCC. ..tW__E :}:fk: 3  ;;y&9h h (..x9J9JJW W h ,#L2DXEUEU2VWH Od6l^#>tH~>NOnw# ..224<@";C"J)L#tF|_  $   $22D9j(()?"&!9!9 ; D44e< $ 8 8 $223J(($=!5T5M5M N "+ ++:a3O)1I1IJG&\!$Z  / s7*C?? E ?EE  EI&"C9I!I&!I&c jj|d}t|||j|}|j j j | Sd fd } j ||S#t$rW}jst|j}t||jd|jddcYd}~Sd}~wwxYw) aExecute single tool call with wrap_tool_call wrapper if configured. Args: call: Tool call dict. input_type: Input format. tool_runtime: Tool runtime. Returns: ToolMessage or Command. rTr9r!rFrHNc*j|S)>Execute tool with given request. Can be called multiple times.rreqrrrSs rBexecutez"ToolNode._run_one..execute**3 FC CrArrr{rr!rErZToolMessage | Command) rrdrErFrrrrmrrrr9) rSrrrr! tool_requestr"rrrs ` ` @rBrzToolNode._run_ones$!!%%d6l3'$$  $$    '**<VL L D '' g> > ++(1I1IJG!++F3)33D9   s/B C! A CC!C!cK|j}|j}|)|j|x}r|Sd|dd}t||j ||j }i|ddi} |j | |d{} t| t.r|j1| |j|St| t,r&t3d t5| j6| _| Sd|dd t!| }t|7#t$rD} |jj|d} t| | } t|d| |d| | d} ~ wwxYw#t$rt$r}t|jt r(t#|jtr|jf}nst|jt$r |j}nLt'|jr0t|jt st)|j}ntf}|jr t||st+||j}t-||d|d d cYd}~Sd}~wwxYww)adExecute tool call asynchronously with configured error handling. Args: request: Tool execution request. input_type: Input format. config: Runnable configuration. Returns: ToolMessage or Command. Raises: Exception: If tool fails and handle_tool_errors is False. NrrTrrar9rrrr{rr r )r9r!r r rrHainvoker.rrdrrqr'rmrbrrarrrrrrr*rrrorrs rB_execute_tool_asynczToolNode._execute_tool_async sj&  || <'+'?'?'EE#E++$v,'GHCC. ..tW__E :}:fk: 3  !%i!@@h h (..x9J9JJW W h ,#L2DXEUEU2VWH Od6l^#>tH~>NOnyA" ..224<@";C"J)L#tF|_  $   $22D9j(()?"&!9!9 ; D44e< $ 8 8 $223J(($=!5T5M5M N "+ ++:a3O)1I1IJG&\!$Z  / sbA(I5,D D D BI5 D E?EEEI2.C9I-'I2(I5-I22I5c Kjj|d}t|||j|}|j j 'j j| d{Sd fd }d fd } j j ||d{Stdj _j ||S7h73#t$rW}jst|j} t| |jd|jdd cYd}~Sd}~wwxYww) aExecute single tool call asynchronously with awrap_tool_call wrapper if configured. Args: call: Tool call dict. input_type: Input format. tool_runtime: Tool runtime. Returns: ToolMessage or Command. rTrNcFKj|d{S7w)rN)r)r s rBr"z#ToolNode._arun_one..executes"11#z6JJ JJs !!c*j|S)z'Sync execute fallback for sync wrapper.rr s rB _sync_executez)ToolNode._arun_one.._sync_executer#rAToolCallWrapperrrr{rr$)rrdrErFrrrr)rrmrrrr9) rSrrrr!r&r"r-rrrs ` ` @rBrzToolNode._arun_oneksD$!!%%d6l3'$$  $$  (T-A-A-I11, FSS S K D  $$0!22<III#'(94;O;O#PD '' mD D#TJ ++(1I1IJG!++F3)33D9   saA2E7C 8E !C$.C"/C$2E3,C$E"C$$ E-A D?9E:E?EEcjt|trAt|dtr)|djddk(rd}t d|}||fSd}|}nt|tr*|jddk(rt d |}d}|dg|fSt|tr!|j|j gx}rd }n)t ||j gx}rd }n d }t| td t|D}t|j}||fS#t$rd }t|wxYw)Nrar9rzlist[ToolCall]ri__typetool_call_with_contextr]rczNo message found in inputc3BK|]}t|ts|ywru)rbr)rems rBrgz(ToolNode._parse_input..s%Ay1I%szNo AIMessage found in input) rbrircrdrrgetattrrnextreversed StopIterationr)rSrrrrinput_with_ctxrwlatest_ai_messages rBrzToolNode._parse_inputsR eT "%)T*uRy}}V/D /S) !"2E: !:--JH ud # ((;?W(W ""7?N%J";/0*< < t $ $"4"4b9 9H 9J (:(:B? ?X ?J-CS/ ! " $%#H-%! +667 :%%  "/CS/ ! "s 'DD2c|d}||jvr[t|jj}tj |dj |}t |||ddSy)NrT, )requested_toolavailable_toolsrr{)rTrr )rrir INVALID_TOOL_NAME_ERROR_TEMPLATErr|r)rSrr=all_tool_namesrs rBr zToolNode._validate_tool_callswf !3!3 3!$"4"4"9"9";Gn4:g rAcXt|tr|jddk(r|dS|S)zExtract state from input, handling ToolCallWithContext if present. Args: input: The input which may be raw state or ToolCallWithContext. Returns: The actual state to pass to wrap_tool_call wrappers. r1r2rF)rbrcrd)rSrs rBrzToolNode._extract_states. eT "uyy':>V'V> ! rAc|d|jvr|S|jj|d}|s|St|}i}|jrF|j}t |t rt |jj}t|dk(r|d|jk(s|d|j|i}n_d|dd}td|jjDr!djd|D} |d | d z }t|t |tr/|jjD]\} } | r|| n||| <n5|jjD]\} } | r t|| n||| <|j r2|j d } t| |j ||j <|j"r|||j"<i|d ||d <|S) a+Inject graph state, store, and runtime into tool call arguments. This is an internal method that enables tools to access graph context that should not be controlled by the model. Tools can declare dependencies on graph state, persistent storage, or runtime context using InjectedState, InjectedStore, and ToolRuntime annotations. This method automatically identifies these dependencies and injects the appropriate values. The injection process preserves the original tool call structure while adding the necessary context arguments. This allows tools to be both model-callable and context-aware without exposing internal state management to the model. Args: tool_call: The tool call dictionary to augment with injected arguments. Must contain 'name', 'args', 'id', and 'type' fields. tool_runtime: The ToolRuntime instance containing all runtime context (state, config, store, context, stream_writer) to inject into tools. Returns: A new ToolCall dictionary with the same structure as the input but with additional arguments injected based on the tool's annotation requirements. Raises: ValueError: If a tool requires store injection but no store is provided, or if state injection requirements cannot be satisfied. !!! note This method is called automatically during tool execution. It should not be called from outside the `ToolNode`. rTrrz Invalid input to ToolNode. Tool z$ requires graph state dict as input.c3 K|]}|ywrur@)re state_fields rBrgz-ToolNode._inject_tool_args../sR;;Rs r<c3&K|] }|s| ywrur@)refs rBrgz-ToolNode._inject_tool_args..0s7XaVW7Xsz State should contain fields rszgCannot inject store into tools with InjectedStore annotations - please compile your graph with a store.r)rrrdrrFrbrirrrrr|rrcrr5rrH) rSr9rrtool_call_copyrrFrequired_fieldserr_msgrequired_fields_strtool_argrDrws rBrzToolNode._inject_tool_argss.F V D$6$6 6 &&**9V+<= #' ?  >> &&E%&"&x~~'<'<'>"?(A-'*d.@.@@$Q'/!//7E;9V;L:MN56R(..:O:O:QRR.2ii7X?7X.X+;>!!)>!o%,8,>,>M(.. )   .:M(** +!LN6$:!Lm!LvrAct|jtrt|dvr.d|jd|jd|dd}t |t |}t d|jxsi}|j|jg}nZt|jtr>|dk7r!d |jd|dd}t |t |}|j}n|St|}|tt gk(r|Sd }|D]2} t| ts| j|d k(s'|d| _d }4|j%|s#|dk(rdnd} d|dd|d| d}t ||S)N)rcrzFTools can provide a dict in Command.update only when using dict with 'z' key as ToolNode input, got: z for tool 'rT'rrizpTools can provide a list of messages in Command.update only when using list of messages as ToolNode input, got: )rFrTrcz]`Command(update={"messages": [ToolMessage("Success", tool_call_id=tool_call_id), ...]}, ...)`zO`Command(update=[ToolMessage("Success", tool_call_id=tool_call_id), ...], ...)`zDExpected to have a matching ToolMessage in Command.update for tool 'z', got: z. Every tool call (LLM requesting to call a tool) in the message history MUST have a corresponding ToolMessage. You can fix it by modifying the tool to return rs)rbrrcrrrrrdrirrr(rrrTr) rScommandrrrwupdated_command state_updatemessages_updatehas_matching_tool_messagerexample_updates rBrzToolNode._validate_tool_commandSs gnnd +!77!//01#NN+;tF|nAG !o%&w/O 0/2H2HIORL*..t/A/A2FO  -V##NN+;tF|nAG !o%&w/O-44ON.o> }0CDE E" "$)!& 1Gg{3##tDz1#F| ,0)  1  (1J'SR !&\N(?2CDBCQAQQR T S/ !rA)rzSequence[BaseTool | Callable]rTrYrzlist[str] | NonerObool | str | Callable[..., str] | type[Exception] | tuple[type[Exception], ...]rrYrzToolCallWrapper | NonerzAsyncToolCallWrapper | NonerZr[)rZzdict[str, BaseTool])r-list[AnyMessage] | dict[str, Any] | BaseModelrrrHr2rZr )rzlist[ToolMessage | Command]r%Literal['list', 'dict', 'tool_calls']rZz@list[Command | list[ToolMessage] | dict[str, list[ToolMessage]]])rrErrVrrrZr%)rrrrVrrGrZr%)rrUrZzrTr?rrpropertyrrrrrrr)rrr rrrrrs@rBrrbs[l\D# !% )D&157;+L,+L +L  +L & +L+L/+L5+L +LZ##?<?? ? ?@?<?? ? ?>* ,* :*  J * X` `:` `  `D55:5" 5  5n` `:` `  `D==:=" =  =~*&<*& F*&X  B  6 __"_  _BFFF: F  FrArc t|tr|d}nIt|tr|j|gx}st ||gx}r|d}nd|}t |t |drt|jdkDryy)a= Conditional routing function for tool-calling workflows. This utility function implements the standard conditional logic for ReAct-style agents: if the last `AIMessage` contains tool calls, route to the tool execution node; otherwise, end the workflow. This pattern is fundamental to most tool-calling agent architectures. The function handles multiple state formats commonly used in LangGraph applications, making it flexible for different graph designs while maintaining consistent behavior. Args: state: The current graph state to examine for tool calls. Supported formats: - Dictionary containing a messages key (for `StateGraph`) - `BaseModel` instance with a messages attribute messages_key: The key or attribute name containing the message list in the state. This allows customization for graphs using different state schemas. Returns: Either `'tools'` if tool calls are present in the last `AIMessage`, or `'__end__'` to terminate the workflow. These are the standard routing destinations for tool-calling conditional edges. Raises: ValueError: If no messages can be found in the provided state format. Example: Basic usage in a ReAct agent: ```python from langgraph.graph import StateGraph from langchain.tools import ToolNode from langchain.tools.tool_node import tools_condition from typing_extensions import TypedDict class State(TypedDict): messages: list graph = StateGraph(State) graph.add_node("llm", call_model) graph.add_node("tools", ToolNode([my_tool])) graph.add_conditional_edges( "llm", tools_condition, # Routes to "tools" or "__end__" {"tools": "tools", "__end__": "__end__"}, ) ``` Custom messages key: ```python def custom_condition(state): return tools_condition(state, messages_key="chat_history") ``` !!! note This function is designed to work seamlessly with `ToolNode` and standard LangGraph patterns. It expects the last message to be an `AIMessage` when tool calls are present, which is the standard output format for tool-calling language models. r0z/No messages found in input state to tool_edge: rrr__end__) rbrircrdr5rrNrr)rFr ai_messagerrws rBtools_conditionr[sD%2Y UD !599\23N'Nx'NE<444b\ ?wGoz<(S1F1F-G!-K rAcNeZdZUdZded<ded<ded<ded <d ed <d ed <y)rGaRuntime context automatically injected into tools. When a tool function has a parameter named `tool_runtime` with type hint `ToolRuntime`, the tool execution system will automatically inject an instance containing: - `state`: The current graph state - `tool_call_id`: The ID of the current tool call - `config`: `RunnableConfig` for the current execution - `context`: Runtime context (from langgraph `Runtime`) - `store`: `BaseStore` instance for persistent storage (from langgraph `Runtime`) - `stream_writer`: `StreamWriter` for streaming output (from langgraph `Runtime`) No `Annotated` wrapper is needed - just use `runtime: ToolRuntime` as a parameter. Example: ```python from langchain_core.tools import tool from langchain.tools import ToolRuntime @tool def my_tool(x: int, runtime: ToolRuntime) -> str: """Tool that accesses runtime context.""" # Access state messages = tool_runtime.state["messages"] # Access tool_call_id print(f"Tool call ID: {tool_runtime.tool_call_id}") # Access config print(f"Run ID: {tool_runtime.config.get('run_id')}") # Access runtime context user_id = tool_runtime.context.get("user_id") # Access store tool_runtime.store.put(("metrics",), "count", 1) # Stream output tool_runtime.stream_writer.write("Processing...") return f"Processed {x}" ``` !!! note This is a marker class used for type checking and detection. The actual runtime object will be constructed during tool execution. r4rFr6rrrr,rrrzBaseStore | NonerNr:r@rArBrGrGs-0d M   rArGceZdZdZdddZy) InjectedStatea Annotation for injecting graph state into tool arguments. This annotation enables tools to access graph state without exposing state management details to the language model. Tools annotated with `InjectedState` receive state data automatically during execution while remaining invisible to the model's tool-calling interface. Args: field: Optional key to extract from the state dictionary. If `None`, the entire state is injected. If specified, only that field's value is injected. This allows tools to request specific state components rather than processing the full state structure. Example: ```python from typing import List from typing_extensions import Annotated, TypedDict from langchain_core.messages import BaseMessage, AIMessage from langchain.tools import InjectedState, ToolNode, tool class AgentState(TypedDict): messages: List[BaseMessage] foo: str @tool def state_tool(x: int, state: Annotated[dict, InjectedState]) -> str: '''Do something with state.''' if len(state["messages"]) > 2: return state["foo"] + str(x) else: return "not enough messages" @tool def foo_tool(x: int, foo: Annotated[str, InjectedState("foo")]) -> str: '''Do something else with state.''' return foo + str(x + 1) node = ToolNode([state_tool, foo_tool]) tool_call1 = {"name": "state_tool", "args": {"x": 1}, "id": "1", "type": "tool_call"} tool_call2 = {"name": "foo_tool", "args": {"x": 1}, "id": "2", "type": "tool_call"} state = { "messages": [AIMessage("", tool_calls=[tool_call1, tool_call2])], "foo": "bar", } node.invoke(state) ``` ```python [ ToolMessage(content="not enough messages", name="state_tool", tool_call_id="1"), ToolMessage(content="bar2", name="foo_tool", tool_call_id="2"), ] ``` !!! note - `InjectedState` arguments are automatically excluded from tool schemas presented to language models - `ToolNode` handles the injection process during execution - Tools can mix regular arguments (controlled by the model) with injected arguments (controlled by the system) - State injection occurs after the model generates tool calls but before tool execution Nc||_y)z*Initialize the `InjectedState` annotation.N)field)rSr`s rBrzInjectedState.__init__os  rAru)r`rrZr[)r;r<r=r>rr@rArBr^r^(sDLrAr^ceZdZdZy) InjectedStorea Annotation for injecting persistent store into tool arguments. This annotation enables tools to access LangGraph's persistent storage system without exposing storage details to the language model. Tools annotated with `InjectedStore` receive the store instance automatically during execution while remaining invisible to the model's tool-calling interface. The store provides persistent, cross-session data storage that tools can use for maintaining context, user preferences, or any other data that needs to persist beyond individual workflow executions. !!! warning `InjectedStore` annotation requires `langchain-core >= 0.3.8` Example: ```python from typing_extensions import Annotated from langgraph.store.memory import InMemoryStore from langchain.tools import InjectedStore, ToolNode, tool @tool def save_preference( key: str, value: str, store: Annotated[Any, InjectedStore()] ) -> str: """Save user preference to persistent storage.""" store.put(("preferences",), key, value) return f"Saved {key} = {value}" @tool def get_preference( key: str, store: Annotated[Any, InjectedStore()] ) -> str: """Retrieve user preference from persistent storage.""" result = store.get(("preferences",), key) return result.value if result else "Not found" ``` Usage with `ToolNode` and graph compilation: ```python from langgraph.graph import StateGraph from langgraph.store.memory import InMemoryStore store = InMemoryStore() tool_node = ToolNode([save_preference, get_preference]) graph = StateGraph(State) graph.add_node("tools", tool_node) compiled_graph = graph.compile(store=store) # Store is injected automatically ``` Cross-session persistence: ```python # First session result1 = graph.invoke({"messages": [HumanMessage("Save my favorite color as blue")]}) # Later session - data persists result2 = graph.invoke({"messages": [HumanMessage("What's my favorite color?")]}) ``` !!! note - `InjectedStore` arguments are automatically excluded from tool schemas presented to language models - The store instance is automatically injected by `ToolNode` during execution - Tools can access namespaced storage using the store's get/put methods - Store injection requires the graph to be compiled with a store instance - Multiple tools can share the same store instance for data consistency N)r;r<r=r>r@rArBrbrbtsGrArbct|st|tr t|ryt|}|tus|t urt fdt|DSy)a5Check if a type argument represents an injection annotation. This utility function determines whether a type annotation indicates that an argument should be injected with state or store data. It handles both direct annotations and nested annotations within Union or Annotated types. Args: type_arg: The type argument to check for injection annotations. injection_type: The injection type to look for (InjectedState or InjectedStore). Returns: True if the type argument contains the specified injection annotation. Tc36K|]}t|ywru) _is_injection)retainjection_types rBrgz _is_injection..sR=^4RsF)rbrarrrr rr)type_argrgorigin_s ` rBreresW"(N+8T"z(N'K"G%7i/Rx?QRRR rAct|}|Dcgc]}t||s|}}t|dkDrd|jd|}t |t|dk(r|dSt||ryycc}w)aExtract injection instance from a type annotation. Args: type_: The type annotation to check. injection_type: The injection type to look for. Returns: The injection instance if found, True if injection marker found without instance, None otherwise. rz-A tool argument should not be annotated with z more than once. Found: rTN)rrerr;r)type_rg type_argsrmatchesrws rB_get_injection_from_typernsI'Ns=n+MsNGN 7|a;N __future__rrrrkcollections.abcrrrr dataclassesrr typesr typingr r r rrrrrrrrlangchain_core.messagesrrrrrrlangchain_core.runnables.configrrrlangchain_core.toolsrr r!rlangchain_core.tools.baser"r#r$r%langgraph._internal._runnabler&langgraph.errorsr'langgraph.graph.messager(langgraph.store.baser)langgraph.typesr*r+r,pydanticr-r.typing_extensionsr/r0r1langgraph.runtimer2 pydantic_corer3rcr4r6r?rTOOL_EXECUTION_ERROR_TEMPLATEr~r8rEr.AsyncToolCallWrapperr]rorqrrrrrrr[rGr^rbrernrr@rArBrs%N# /*     ;4 ;*7*77/-()* 4 ( :t ,S!H+ +   >*>* >*Bh0+2GGHI'GR h0)K'