alita-sdk 0.3.351__py3-none-any.whl → 0.3.499__py3-none-any.whl

alita_sdk/runtime/toolkits/planning.py

@@ -0,0 +1,178 @@
+"""
+PlanningToolkit - Runtime toolkit for agent plan management.
+
+Provides tools for creating, tracking, and completing multi-step execution plans.
+Supports two storage backends:
+1. PostgreSQL - when pgvector_configuration with connection_string is provided
+2. Filesystem - when no connection string (local CLI usage)
+"""
+
+from typing import ClassVar, List, Any, Literal, Optional, Callable
+
+from langchain_community.agent_toolkits.base import BaseToolkit
+from langchain_core.tools import BaseTool
+from pydantic import create_model, BaseModel, ConfigDict, Field
+from pydantic.fields import FieldInfo
+
+from ..tools.planning import PlanningWrapper
+from ...tools.base.tool import BaseAction
+from ...tools.utils import clean_string, get_max_toolkit_length
+
+
+class PlanningToolkit(BaseToolkit):
+    """
+    Toolkit for agent plan management.
+    
+    Provides tools for creating, updating, and tracking execution plans.
+    Supports PostgreSQL (production) and filesystem (local) storage backends.
+    Plans are scoped by conversation_id.
+    """
+    tools: List[BaseTool] = []
+    _toolkit_max_length: ClassVar[int] = 50  # Use ClassVar to avoid Pydantic treating it as field
+    
+    @staticmethod
+    def toolkit_config_schema() -> BaseModel:
+        """
+        Returns the configuration schema for the Planning toolkit.
+        
+        Used by the UI to generate the toolkit configuration form.
+        """
+        # Define available tools
+        selected_tools = {
+            'update_plan': {
+                'title': 'UpdatePlanInput',
+                'type': 'object',
+                'properties': {
+                    'title': {'type': 'string', 'description': "Title for the plan"},
+                    'steps': {'type': 'array', 'items': {'type': 'string'}, 'description': "List of step descriptions"},
+                    'conversation_id': {'type': 'string', 'description': "Conversation ID (auto-injected)"}
+                },
+                'required': ['title', 'steps', 'conversation_id']
+            },
+            'complete_step': {
+                'title': 'CompleteStepInput', 
+                'type': 'object',
+                'properties': {
+                    'step_number': {'type': 'integer', 'description': "Step number to complete (1-indexed)"},
+                    'conversation_id': {'type': 'string', 'description': "Conversation ID (auto-injected)"}
+                },
+                'required': ['step_number', 'conversation_id']
+            },
+            'get_plan_status': {
+                'title': 'GetPlanStatusInput',
+                'type': 'object', 
+                'properties': {
+                    'conversation_id': {'type': 'string', 'description': "Conversation ID (auto-injected)"}
+                },
+                'required': ['conversation_id']
+            },
+            'delete_plan': {
+                'title': 'DeletePlanInput',
+                'type': 'object',
+                'properties': {
+                    'conversation_id': {'type': 'string', 'description': "Conversation ID (auto-injected)"}
+                },
+                'required': ['conversation_id']
+            }
+        }
+        
+        PlanningToolkit._toolkit_max_length = get_max_toolkit_length(selected_tools)
+        
+        return create_model(
+            "planning",
+            # Tool selection
+            selected_tools=(
+                List[Literal[tuple(selected_tools)]], 
+                Field(
+                    default=list(selected_tools.keys()),
+                    json_schema_extra={'args_schemas': selected_tools}
+                )
+            ),
+            __config__=ConfigDict(
+                json_schema_extra={
+                    'metadata': {
+                        "label": "Planning",
+                        "description": "Tools for managing multi-step execution plans with progress tracking. Uses PostgreSQL when configured, filesystem otherwise.",
+                        "icon_url": None,
+                        "max_length": PlanningToolkit._toolkit_max_length,
+                        "categories": ["planning", "internal_tool"],
+                        "extra_categories": ["task management", "todo", "progress tracking"]
+                    }
+                }
+            )
+        )
+    
+    @classmethod
+    def get_toolkit(
+        cls,
+        toolkit_name: Optional[str] = None,
+        selected_tools: Optional[List[str]] = None,
+        pgvector_configuration: Optional[dict] = None,
+        storage_dir: Optional[str] = None,
+        plan_callback: Optional[Any] = None,
+        conversation_id: Optional[str] = None,
+        **kwargs
+    ):
+        """
+        Create a PlanningToolkit instance with configured tools.
+        
+        Args:
+            toolkit_name: Optional name prefix for tools
+            selected_tools: List of tool names to include (default: all)
+            pgvector_configuration: PostgreSQL configuration dict with connection_string.
+                                   If not provided, uses filesystem storage.
+            storage_dir: Directory for filesystem storage (when no pgvector_configuration)
+            plan_callback: Optional callback function called when plan changes (for CLI UI)
+            conversation_id: Conversation ID for scoping plans.
+                            For server: from elitea_core payload. For CLI: session_id.
+            **kwargs: Additional configuration options
+            
+        Returns:
+            PlanningToolkit instance with configured tools
+        """
+        if selected_tools is None:
+            selected_tools = ['update_plan', 'complete_step', 'get_plan_status', 'delete_plan']
+        
+        tools = []
+        
+        # Extract connection string from pgvector configuration (if provided)
+        connection_string = None
+        if pgvector_configuration:
+            connection_string = pgvector_configuration.get('connection_string', '')
+            if hasattr(connection_string, 'get_secret_value'):
+                connection_string = connection_string.get_secret_value()
+        
+        # Create wrapper - it will auto-select storage backend
+        wrapper = PlanningWrapper(
+            connection_string=connection_string if connection_string else None,
+            conversation_id=conversation_id,
+            storage_dir=storage_dir,
+            plan_callback=plan_callback,
+        )
+        
+        # Use clean toolkit name for context (max 1000 chars in description)
+        toolkit_context = f" [Toolkit: {clean_string(toolkit_name, 0)}]" if toolkit_name else ''
+        
+        # Create tools from wrapper
+        available_tools = wrapper.get_available_tools()
+        for tool in available_tools:
+            if tool["name"] not in selected_tools:
+                continue
+            
+            # Add toolkit context to description with character limit
+            description = tool["description"]
+            if toolkit_context and len(description + toolkit_context) <= 1000:
+                description = description + toolkit_context
+            
+            tools.append(BaseAction(
+                api_wrapper=wrapper,
+                name=tool["name"],
+                description=description,
+                args_schema=tool["args_schema"]
+            ))
+        
+        return cls(tools=tools)
+    
+    def get_tools(self) -> List[BaseTool]:
+        """Return the list of configured tools."""
+        return self.tools

alita_sdk/runtime/toolkits/tools.py

@@ -1,4 +1,5 @@
 import logging
+from typing import Optional
 
 from langchain_core.tools import ToolException
 from langgraph.store.base import BaseStore
@@ -8,14 +9,20 @@ from alita_sdk.tools import get_tools as alita_tools
 from .application import ApplicationToolkit
 from .artifact import ArtifactToolkit
 from .datasource import DatasourcesToolkit
+from .planning import PlanningToolkit
 from .prompt import PromptToolkit
 from .subgraph import SubgraphToolkit
 from .vectorstore import VectorStoreToolkit
+from .mcp import McpToolkit
 from ..tools.mcp_server_tool import McpServerTool
+from ..tools.sandbox import SandboxToolkit
+from ..tools.image_generation import ImageGenerationToolkit
 # Import community tools
 from ...community import get_toolkits as community_toolkits, get_tools as community_tools
 from ...tools.memory import MemoryToolkit
-from ...tools.utils import TOOLKIT_SPLITTER
+from ..utils.mcp_oauth import canonical_resource, McpAuthorizationRequired
+from ...tools.utils import clean_string
+from alita_sdk.tools import _inject_toolkit_id
 
 logger = logging.getLogger(__name__)
 
@@ -24,64 +31,190 @@ def get_toolkits():
     core_toolkits = [
         ArtifactToolkit.toolkit_config_schema(),
         MemoryToolkit.toolkit_config_schema(),
-        VectorStoreToolkit.toolkit_config_schema()
+        PlanningToolkit.toolkit_config_schema(),
+        VectorStoreToolkit.toolkit_config_schema(),
+        SandboxToolkit.toolkit_config_schema(),
+        ImageGenerationToolkit.toolkit_config_schema(),
+        McpToolkit.toolkit_config_schema()
     ]
 
     return core_toolkits + community_toolkits() + alita_toolkits()
 
 
-def get_tools(tools_list: list, alita_client, llm, memory_store: BaseStore = None) -> list:
+def get_tools(tools_list: list, alita_client=None, llm=None, memory_store: BaseStore = None, debug_mode: Optional[bool] = False, mcp_tokens: Optional[dict] = None, conversation_id: Optional[str] = None) -> list:
     prompts = []
     tools = []
 
     for tool in tools_list:
-        if tool['type'] == 'datasource':
-            tools.extend(DatasourcesToolkit.get_toolkit(
-                alita_client,
-                datasource_ids=[int(tool['settings']['datasource_id'])],
-                selected_tools=tool['settings']['selected_tools'],
-                toolkit_name=tool.get('toolkit_name', '') or tool.get('name', '')
-            ).get_tools())
-        elif tool['type'] == 'application' and tool.get('agent_type', '') != 'pipeline' :
-            tools.extend(ApplicationToolkit.get_toolkit(
-                alita_client,
-                application_id=int(tool['settings']['application_id']),
-                application_version_id=int(tool['settings']['application_version_id']),
-                selected_tools=[]
-            ).get_tools())
-        elif tool['type'] == 'application' and tool.get('agent_type', '') == 'pipeline':
-            # static get_toolkit returns a list of CompiledStateGraph stubs
-            tools.extend(SubgraphToolkit.get_toolkit(
-                alita_client,
-                application_id=int(tool['settings']['application_id']),
-                application_version_id=int(tool['settings']['application_version_id']),
-                app_api_key=alita_client.auth_token,
-                selected_tools=[],
-                llm=llm
-            ))
-        elif tool['type'] == 'memory':
-            tools += MemoryToolkit.get_toolkit(
-                namespace=tool['settings'].get('namespace', str(tool['id'])),
-                pgvector_configuration=tool['settings'].get('pgvector_configuration', {}),
-                store=memory_store,
-            ).get_tools()
-        elif tool['type'] == 'artifact':
-            tools.extend(ArtifactToolkit.get_toolkit(
-                client=alita_client,
-                bucket=tool['settings']['bucket'],
-                toolkit_name=tool.get('toolkit_name', ''),
-                selected_tools=tool['settings'].get('selected_tools', []),
-                llm=llm,
-                # indexer settings
-                pgvector_configuration=tool['settings'].get('pgvector_configuration', {}),
-                embedding_model=tool['settings'].get('embedding_model'),
-                collection_name=f"{tool.get('toolkit_name')}",
-            ).get_tools())
-        elif tool['type'] == 'vectorstore':
-            tools.extend(VectorStoreToolkit.get_toolkit(
-                llm=llm,
-                toolkit_name=tool.get('toolkit_name', ''),
-                **tool['settings']).get_tools())
+        try:
+            if tool['type'] == 'datasource':
+                tools.extend(DatasourcesToolkit.get_toolkit(
+                    alita_client,
+                    datasource_ids=[int(tool['settings']['datasource_id'])],
+                    selected_tools=tool['settings']['selected_tools'],
+                    toolkit_name=tool.get('toolkit_name', '') or tool.get('name', '')
+                ).get_tools())
+            elif tool['type'] == 'application':
+                tools.extend(ApplicationToolkit.get_toolkit(
+                    alita_client,
+                    application_id=int(tool['settings']['application_id']),
+                    application_version_id=int(tool['settings']['application_version_id']),
+                    selected_tools=[]
+                ).get_tools())
+                # backward compatibility for pipeline application type as subgraph node
+                if tool.get('agent_type', '') == 'pipeline':
+                    # static get_toolkit returns a list of CompiledStateGraph stubs
+                    tools.extend(SubgraphToolkit.get_toolkit(
+                        alita_client,
+                        application_id=int(tool['settings']['application_id']),
+                        application_version_id=int(tool['settings']['application_version_id']),
+                        app_api_key=alita_client.auth_token,
+                        selected_tools=[],
+                        llm=llm
+                    ))
+            elif tool['type'] == 'memory':
+                tools += MemoryToolkit.get_toolkit(
+                    namespace=tool['settings'].get('namespace', str(tool['id'])),
+                    pgvector_configuration=tool['settings'].get('pgvector_configuration', {}),
+                    store=memory_store,
+                ).get_tools()
+            # TODO: update configuration of internal tools
+            elif tool['type'] == 'internal_tool':
+                if tool['name'] == 'pyodide':
+                    tools += SandboxToolkit.get_toolkit(
+                        stateful=False,
+                        allow_net=True,
+                        alita_client=alita_client,
+                    ).get_tools()
+                elif tool['name'] == 'image_generation':
+                    if alita_client and alita_client.model_image_generation:
+                        tools += ImageGenerationToolkit.get_toolkit(
+                            client=alita_client,
+                        ).get_tools()
+                    else:
+                        logger.warning("Image generation internal tool requested "
+                                       "but no image generation model configured")
+                elif tool['name'] == 'planner':
+                    tools += PlanningToolkit.get_toolkit(
+                        pgvector_configuration=tool.get('settings', {}).get('pgvector_configuration'),
+                        conversation_id=conversation_id,
+                    ).get_tools()
+            elif tool['type'] == 'artifact':
+                toolkit_tools = ArtifactToolkit.get_toolkit(
+                    client=alita_client,
+                    bucket=tool['settings']['bucket'],
+                    toolkit_name=tool.get('toolkit_name', ''),
+                    selected_tools=tool['settings'].get('selected_tools', []),
+                    llm=llm,
+                    # indexer settings
+                    pgvector_configuration=tool['settings'].get('pgvector_configuration', {}),
+                    embedding_model=tool['settings'].get('embedding_model'),
+                    collection_name=f"{tool.get('toolkit_name')}",
+                    collection_schema=str(tool['settings'].get('id', tool.get('id', ''))),
+                ).get_tools()
+                # Inject toolkit_id for artifact tools as well
+                # Pass settings as the tool config since that's where the id field is
+                _inject_toolkit_id(tool['settings'], toolkit_tools)
+                tools.extend(toolkit_tools)
+
+            elif tool['type'] == 'vectorstore':
+                tools.extend(VectorStoreToolkit.get_toolkit(
+                    llm=llm,
+                    toolkit_name=tool.get('toolkit_name', ''),
+                    **tool['settings']).get_tools())
+            elif tool['type'] == 'planning':
+                # Planning toolkit for multi-step task tracking
+                settings = tool.get('settings', {})
+                
+                # Check if local mode is enabled (uses filesystem storage, ignores pgvector)
+                use_local = settings.get('local', False)
+                
+                if use_local:
+                    # Local mode - use filesystem storage
+                    logger.info("Planning toolkit using local filesystem storage (local=true)")
+                    pgvector_config = {}
+                else:
+                    # Check if explicit connection_string is provided in pgvector_configuration
+                    explicit_pgvector_config = settings.get('pgvector_configuration', {})
+                    explicit_connstr = explicit_pgvector_config.get('connection_string') if explicit_pgvector_config else None
+                    
+                    if explicit_connstr:
+                        # Use explicitly provided connection string (overrides project secrets)
+                        logger.info("Using explicit connection_string for planning toolkit")
+                        pgvector_config = explicit_pgvector_config
+                    else:
+                        # Try to fetch pgvector_project_connstr from project secrets
+                        pgvector_connstr = None
+                        if alita_client:
+                            try:
+                                pgvector_connstr = alita_client.unsecret('pgvector_project_connstr')
+                                if pgvector_connstr:
+                                    logger.info("Using pgvector_project_connstr for planning toolkit")
+                            except Exception as e:
+                                logger.debug(f"pgvector_project_connstr not available: {e}")
+                        
+                        pgvector_config = {'connection_string': pgvector_connstr} if pgvector_connstr else {}
+                
+                tools.extend(PlanningToolkit.get_toolkit(
+                    toolkit_name=tool.get('toolkit_name', ''),
+                    selected_tools=settings.get('selected_tools', []),
+                    pgvector_configuration=pgvector_config,
+                    conversation_id=conversation_id or settings.get('conversation_id'),
+                ).get_tools())
+            elif tool['type'] == 'mcp':
+                # remote mcp tool initialization with token injection
+                settings = dict(tool['settings'])
+                url = settings.get('url')
+                headers = settings.get('headers')
+                token_data = None
+                session_id = None
+                if mcp_tokens and url:
+                    canonical_url = canonical_resource(url)
+                    logger.info(f"[MCP Auth] Looking for token for URL: {url}")
+                    logger.info(f"[MCP Auth] Canonical URL: {canonical_url}")
+                    logger.info(f"[MCP Auth] Available tokens: {list(mcp_tokens.keys())}")
+                    token_data = mcp_tokens.get(canonical_url)
+                    if token_data:
+                        logger.info(f"[MCP Auth] Found token data for {canonical_url}")
+                        # Handle both old format (string) and new format (dict with access_token and session_id)
+                        if isinstance(token_data, dict):
+                            access_token = token_data.get('access_token')
+                            session_id = token_data.get('session_id')
+                            logger.info(f"[MCP Auth] Token data: access_token={'present' if access_token else 'missing'}, session_id={session_id or 'none'}")
+                        else:
+                            # Backward compatibility: treat as plain token string
+                            access_token = token_data
+                            logger.info(f"[MCP Auth] Using legacy token format (string)")
+                    else:
+                        access_token = None
+                        logger.warning(f"[MCP Auth] No token found for {canonical_url}")
+                else:
+                    access_token = None
+                    
+                if access_token:
+                    merged_headers = dict(headers) if headers else {}
+                    merged_headers.setdefault('Authorization', f'Bearer {access_token}')
+                    settings['headers'] = merged_headers
+                    logger.info(f"[MCP Auth] Added Authorization header for {url}")
+                    
+                # Pass session_id to MCP toolkit if available
+                if session_id:
+                    settings['session_id'] = session_id
+                    logger.info(f"[MCP Auth] Passing session_id to toolkit: {session_id}")
+                tools.extend(McpToolkit.get_toolkit(
+                    toolkit_name=tool.get('toolkit_name', ''),
+                    client=alita_client,
+                    **settings).get_tools())
+        except McpAuthorizationRequired:
+            # Re-raise auth required exceptions directly
+            raise
+        except Exception as e:
+            logger.error(f"Error initializing toolkit for tool '{tool.get('name', 'unknown')}': {e}", exc_info=True)
+            if debug_mode:
+                logger.info("Skipping tool initialization error due to debug mode.")
+                continue
+            else:
+                raise ToolException(f"Error initializing toolkit for tool '{tool.get('name', 'unknown')}': {e}")
     
     if len(prompts) > 0:
         tools += PromptToolkit.get_toolkit(alita_client, prompts).get_tools()
@@ -90,7 +223,8 @@ def get_tools(tools_list: list, alita_client, llm, memory_store: BaseStore = Non
     tools += community_tools(tools_list, alita_client, llm)
     # Add alita tools
     tools += alita_tools(tools_list, alita_client, llm, memory_store)
-    # Add MCP tools
+    # Add MCP tools registered via alita-mcp CLI (static registry)
+    # Note: Tools with type='mcp' are already handled in main loop above
     tools += _mcp_tools(tools_list, alita_client)
     
     # Sanitize tool names to meet OpenAI's function naming requirements
@@ -145,6 +279,10 @@ def _sanitize_tool_names(tools: list) -> list:
 
 
 def _mcp_tools(tools_list, alita):
+    """
+    Handle MCP tools registered via alita-mcp CLI (static registry).
+    Skips tools with type='mcp' as those are handled by dynamic discovery.
+    """
     try:
         all_available_toolkits = alita.get_mcp_toolkits()
         toolkit_lookup = {tk["name"]: tk for tk in all_available_toolkits}
@@ -152,6 +290,11 @@ def _mcp_tools(tools_list, alita):
         #
         for selected_toolkit in tools_list:
             server_toolkit_name = selected_toolkit['type']
+            
+            # Skip tools with type='mcp' - they're handled by dynamic discovery
+            if server_toolkit_name == 'mcp':
+                continue
+            
             toolkit_conf = toolkit_lookup.get(server_toolkit_name)
             #
             if not toolkit_conf:
@@ -165,7 +308,7 @@ def _mcp_tools(tools_list, alita):
                 if not selected_tools or tool_name in selected_tools:
                     if server_tool := _init_single_mcp_tool(server_toolkit_name,
                                                             # selected_toolkit["name"] is None for toolkit_test
-                                                            selected_toolkit["name"] if selected_toolkit.get("name")
+                                                            selected_toolkit["toolkit_name"] if selected_toolkit.get("toolkit_name")
                                                             else server_toolkit_name,
                                                             available_tool, alita, selected_toolkit['settings']):
                         tools.append(server_tool)
@@ -177,11 +320,18 @@ def _mcp_tools(tools_list, alita):
 
 def _init_single_mcp_tool(server_toolkit_name, toolkit_name, available_tool, alita, toolkit_settings):
     try:
-
-        tool_name = f'{toolkit_name}{TOOLKIT_SPLITTER}{available_tool["name"]}'
+        # Use clean tool name without prefix
+        tool_name = available_tool["name"]
+        # Add toolkit context to description (max 1000 chars)
+        toolkit_context = f" [Toolkit: {clean_string(toolkit_name)}]" if toolkit_name else ''
+        base_description = f"MCP for a tool '{tool_name}': {available_tool.get('description', '')}"
+        description = base_description
+        if toolkit_context and len(base_description + toolkit_context) <= 1000:
+            description = base_description + toolkit_context
+        
         return McpServerTool(
             name=tool_name,
-            description=f"MCP for a tool '{tool_name}': {available_tool.get("description", "")}",
+            description=description,
             args_schema=McpServerTool.create_pydantic_model_from_schema(
                 available_tool.get("inputSchema", {})
             ),

alita_sdk/runtime/toolkits/vectorstore.py

@@ -1,7 +1,7 @@
 from logging import getLogger
 from typing import Any, List, Literal, Optional
 
-from alita_sdk.tools.utils import clean_string, TOOLKIT_SPLITTER
+from alita_sdk.tools.utils import clean_string
 from pydantic import BaseModel, create_model, Field, ConfigDict
 from langchain_core.tools import BaseToolkit, BaseTool
 from alita_sdk.tools.base.tool import BaseAction
@@ -31,7 +31,8 @@ class VectorStoreToolkit(BaseToolkit):
                     toolkit_name: Optional[str] = None,
                     selected_tools: list[str] = []):
         logger.info("Selected tools: %s", selected_tools)
-        prefix = clean_string(toolkit_name) + TOOLKIT_SPLITTER if toolkit_name else ''
+        # Use clean toolkit name for context (max 1000 chars in description)
+        toolkit_context = f" [Toolkit: {clean_string(toolkit_name)}]" if toolkit_name else ''
         if selected_tools is None:
             selected_tools = []
         tools = []
@@ -46,10 +47,14 @@ class VectorStoreToolkit(BaseToolkit):
             # if selected_tools:
             #     if tool["name"] not in selected_tools:
             #         continue
+            # Add toolkit context to description with character limit
+            description = tool["description"]
+            if toolkit_context and len(description + toolkit_context) <= 1000:
+                description = description + toolkit_context
             tools.append(BaseAction(
                 api_wrapper=vectorstore_wrapper,
-                name=f'{prefix}{tool["name"]}',
-                description=tool["description"],
+                name=tool["name"],
+                description=description,
                 args_schema=tool["args_schema"]
             ))
         return cls(tools=tools)

alita_sdk/runtime/tools/__init__.py

@@ -5,10 +5,18 @@ This module provides various tools that can be used within LangGraph agents.
 
 from .sandbox import PyodideSandboxTool, StatefulPyodideSandboxTool, create_sandbox_tool
 from .echo import EchoTool
+from .image_generation import (
+    ImageGenerationTool,
+    create_image_generation_tool,
+    ImageGenerationToolkit
+)
 
 __all__ = [
     "PyodideSandboxTool",
-    "StatefulPyodideSandboxTool", 
+    "StatefulPyodideSandboxTool",
     "create_sandbox_tool",
-    "EchoTool"
-]
+    "EchoTool",
+    "ImageGenerationTool",
+    "ImageGenerationToolkit",
+    "create_image_generation_tool"
+]

alita_sdk/runtime/tools/application.py

@@ -50,6 +50,8 @@ class Application(BaseTool):
     application: Any
     args_schema: Type[BaseModel] = applicationToolSchema
     return_type: str = "str"
+    client: Any
+    args_runnable: dict = {}
     
     @field_validator('name', mode='before')
     @classmethod
@@ -66,6 +68,11 @@ class Application(BaseTool):
         return self._run(*config, **all_kwargs)
 
     def _run(self, *args, **kwargs):
+        if self.client and self.args_runnable:
+            # Recreate new LanggraphAgentRunnable in order to reflect the current input_mapping (it can be dynamic for pipelines).
+            # Actually, for pipelines agent toolkits LanggraphAgentRunnable is created (for LLMNode) before pipeline's schema parsing.
+            application_variables = {k: {"name": k, "value": v} for k, v in kwargs.items()}
+            self.application = self.client.application(**self.args_runnable, application_variables=application_variables)
         response = self.application.invoke(formulate_query(kwargs))
         if self.return_type == "str":
             return response["output"]