iv-2SrSSKJr SSKrSSKJr SSKJrJrJ r J r \(aSSK J r J r SSKJr SSKJr SS KJr SS KJrJr SS KJr SS KJrJrJrJr SS KJr \R@"\!5r"Sr#\"SS55r$SSjr%SSjr&"SS\5r'g)z#LLM-based tool selector middleware.) annotationsN) dataclass) TYPE_CHECKING AnnotatedLiteralUnion) AwaitableCallable)BaseTool) BaseChatModel) HumanMessage)Field TypeAdapter) TypedDict)AgentMiddlewareModelCallResult ModelRequest ModelResponse)init_chat_modelzNYour goal is to select the most relevant tools for answering the user's query.cL\rSrSr%SrS\S'S\S'S\S'S \S 'S \S 'S rg)_SelectionRequest"z#Prepared inputs for tool selection.list[BaseTool]available_toolsstrsystem_messager last_user_messager model list[str]valid_tool_namesN__name__ __module__ __qualname____firstlineno____doc____annotations____static_attributes__r!d/home/james-whalen/.local/lib/python3.13/site-packages/langchain/agents/middleware/tool_selection.pyrr"s#-#### r*rc  U(d Sn[U5eUVs/sH0n[[UR[ UR S94PM2 nn[ [U5nSn"SS[5n[U5$s snf)zCreate a structured output schema for tool selection. Args: tools: Available tools to include in the schema. Returns: TypeAdapter for a schema where each tool name is a Literal with its description. z&Invalid usage: tools must be non-empty) descriptionz2Tools to use. Place the most relevant tools first.c$\rSrSr%SrS\S'Srg)>_create_tool_selection_response..ToolSelectionResponseCzUse to select relevant tools.zCAnnotated[list[selected_tool_type], Field(description=description)]toolsr!Nr"r!r*r+ToolSelectionResponser/Cs +RRr*r2) AssertionErrorrrnamerr-rtuplerr)r1msgtoolliteralsselected_tool_typer-r2s r+_create_tool_selection_responser:-s 6S!! X]W\t '$))$e8H8H&IIJW\ uX/FKS S , --s7Bc2SRSU55$)z~Format tools as markdown list. Args: tools: Tools to format. Returns: Markdown string with each tool on a new line.  c3Z# UH!nSURSUR3v M# g7f)z- z: N)r4r-).0r7s r+ $_render_tool_list..Ts(Ledr$))Bt'7'7&89es)+)join)r1s r+_render_tool_listrBKs 99LeL LLr*c^\rSrSrSrS\SSS.S U4SjjjrS SjrS SjrSSjr SS jr S r U=r $)LLMToolSelectorMiddlewareWaUses an LLM to select relevant tools before calling the main model. When an agent has many tools available, this middleware filters them down to only the most relevant ones for the user's query. This reduces token usage and helps the main model focus on the right tools. Examples: Limit to 3 tools: ```python from langchain.agents.middleware import LLMToolSelectorMiddleware middleware = LLMToolSelectorMiddleware(max_tools=3) agent = create_agent( model="openai:gpt-4o", tools=[tool1, tool2, tool3, tool4, tool5], middleware=[middleware], ) ``` Use a smaller model for selection: ```python middleware = LLMToolSelectorMiddleware(model="openai:gpt-4o-mini", max_tools=2) ``` N)r system_prompt max_toolsalways_includec>[TU]5 X lX0lU=(d /Ul[ U[ [S545(aXlg[U5Ulg)aInitialize the tool selector. Args: model: Model to use for selection. If not provided, uses the agent's main model. Can be a model identifier string or BaseChatModel instance. system_prompt: Instructions for the selection model. max_tools: Maximum number of tools to select. If the model selects more, only the first max_tools will be used. No limit if not specified. always_include: Tool names to always include regardless of selection. These do not count against the max_tools limit. N) super__init__rFrGrH isinstancer typerr)selfrrFrGrH __class__s r+rK"LLMToolSelectorMiddleware.__init__rsO& *",2 emT$Z8 9 9/4J(/DJr*cUR(a[UR5S:XagURVs/sHn[U[5(aMUPM nnUR(aaUVs1sHo"R iM nnURVs/sH oUU;dM UPM nnU(aSUS[ U53n[U5eUVs/sH o"R UR;dMUPM" nnU(dgURn URbU SURS3- n [UR5Hn [U [5(dMU n O Sn[U5eUR=(d URn UVs/sHo"R PM n n[UU U U U S9$s snfs snfs snfs snfs snf) z{Prepare inputs for tool selection. Returns: SelectionRequest with prepared inputs, or None if no selection is needed. rNz.Tools in always_include not found in request: z. Available tools: z IMPORTANT: List the tool names in order of relevance, with the most relevant first. If you exceed the maximum number of tools, only the first z will be used.z)No user message found in request messages)rrrrr )r1lenrLdictrHr4sorted ValueErrorrFrGreversedmessagesr r3rr)rNrequestr7 base_toolsavailable_tool_namesr4 missing_toolsr6rrmessagerrr s r+_prepare_selection_request4LLMToolSelectorMiddleware._prepare_selection_requests}}GMM 2a 7(/}}S}tJtT> % "#'..!1A N  0 01G'<00$+!2 >C % % +gmm2AB/$II/B +)/-   [T$F`6Cs/F;F;4G G&GG 6G Gc/n/nUSH[nXs;aURU5 MXu;dM"URb[U5UR:dMJURU5 M] U(aSU3n[U5eUV s/sHoRU;dMU PM n n UR V s/sH8n [ U [5(aMU RUR;dM6U PM: n n U RU 5 UR V s/sHn [ U [5(dMU PM n n /U QU QUlU$s sn fs sn fs sn f)z@Process the selection response and return filtered ModelRequest.r1zModel selected invalid tools: ) appendrGrRrUr4r1rLrSrHextend) rNresponserr rXselected_tool_namesinvalid_tool_selections tool_namer6r7selected_toolsalways_included_toolsprovider_toolss r+_process_selection_response5LLMToolSelectorMiddleware._process_selection_responsesM*,"$!'*I0'..y93&#.A*BT^^*S#**95+ #23J2KLCS/ !-* ,T =P0PD_ *  1 %dD) .2ii4;N;N.N % 1 34,3==S=4JtT: * 1 Ts*EE1EE*EE .E cURU5nUcU"U5$[UR5nUR5nURR U5nUR SURS.UR/5n[U[5(dS[U53n[U5eURXsRURU5n U"U 5$)JFilter tools based on LLM selection before invoking the model via handler.systemrolecontentExpected dict response, got )r]r:r json_schemarwith_structured_outputinvokerrrLrSrMr3rir rNrXhandlerselection_request type_adapterschemastructured_modelrbr6modified_requests r+wrap_model_call)LLMToolSelectorMiddleware.wrap_model_calls !;;GD  $7# #77H7X7XY ))+,22II&Q#**!.?.N.NO!33  (D))0h0@AC % %;; 779J9[9[]d '((r*c # URU5nUcU"U5IShvN $[UR5nUR5nURR U5nUR SURS.UR/5IShvN n[U[5(dS[U53n[U5eURXsRURU5n U"U 5IShvN $NNkN7f)rlNrmrnrq)r]r:rrrrrsainvokerrrLrSrMr3rir rus r+awrap_model_call*LLMToolSelectorMiddleware.awrap_model_calls !;;GD  $ )) )77H7X7XY ))+,22II&Q)11!.?.N.NO!33   (D))0h0@AC % %;; 779J9[9[]d -...+* /s5 DC>A1DDA$D9D:DDD)rHrGrrF) rzstr | BaseChatModel | NonerFrrGz int | NonerHzlist[str] | NonereturnNone)rXrrz_SelectionRequest | None) rbrSrrr rrXrrr)rXrrvz'Callable[[ModelRequest], ModelResponse]rr)rXrrvz2Callable[[ModelRequest], Awaitable[ModelResponse]]rr) r#r$r%r&r'DEFAULT_SYSTEM_PROMPTrKr]rir|rr) __classcell__)rOs@r+rDrDWs:-12 $+/ 0*0 0  0 ) 0 00:> @))()$ )  )  )V))9)  )>//D/  //r*rD)r1rrr)r1rrr)(r' __future__rlogging dataclassesrtypingrrrrcollections.abcr r langchain.toolsr *langchain_core.language_models.chat_modelsr langchain_core.messagesr pydanticrrtyping_extensionsr!langchain.agents.middleware.typesrrrrlangchain.chat_models.baser getLoggerr#loggerrrr:rBrDr!r*r+rs)"!;;3(D0'' 7   8 $U      .< M_/_/r*