# SPDX-FileCopyrightText: 2022-present deepset GmbH # # SPDX-License-Identifier: Apache-2.0 import itertools import json from collections import defaultdict from collections.abc import Iterator, Mapping, Sequence from contextlib import AbstractContextManager as ContextManager from datetime import datetime from enum import IntEnum from pathlib import Path from typing import Any, TextIO, TypeVar, Union, get_args import networkx from haystack import logging, tracing from haystack.core.component import Component, InputSocket, OutputSocket, component from haystack.core.errors import ( DeserializationError, PipelineComponentsBlockedError, PipelineConnectError, PipelineDrawingError, PipelineError, PipelineMaxComponentRuns, PipelineRuntimeError, PipelineValidationError, ) from haystack.core.pipeline.component_checks import ( _NO_OUTPUT_PRODUCED, all_predecessors_executed, are_all_sockets_ready, can_component_run, is_any_greedy_socket_ready, ) from haystack.core.pipeline.utils import FIFOPriorityQueue, _deepcopy_with_exceptions, parse_connect_string from haystack.core.serialization import ( DeserializationCallbacks, component_from_dict, component_to_dict, generate_qualified_class_name, ) from haystack.core.serialization_security import _check_module_allowed, _deserialization_context from haystack.core.type_utils import ( ConversionStrategyType, _convert_value, _safe_get_origin, _type_name, _types_are_compatible, ) from haystack.marshal import Marshaller, YamlMarshaller from haystack.utils import is_in_jupyter, type_serialization from .descriptions import find_pipeline_inputs, find_pipeline_outputs from .draw import _to_mermaid_image DEFAULT_MARSHALLER = YamlMarshaller() # We use a generic type to annotate the return value of class methods, # so that static analyzers won't be confused when derived classes # use those methods. T = TypeVar("T", bound="PipelineBase") logger = logging.getLogger(__name__) # Constants for tracing tags _COMPONENT_INPUT = "haystack.component.input" _COMPONENT_OUTPUT = "haystack.component.output" _COMPONENT_VISITS = "haystack.component.visits" InputsType = dict[str, dict[str, list[dict[str, Any]]]] class ComponentPriority(IntEnum): HIGHEST = 1 READY = 2 DEFER = 3 BLOCKED = 4 class PipelineBase: # noqa: PLW1641 """ Components orchestration engine. Builds a graph of components and orchestrates their execution according to the execution graph. """ def __init__( self, metadata: dict[str, Any] | None = None, max_runs_per_component: int = 100, connection_type_validation: bool = True, ) -> None: """ Creates the Pipeline. :param metadata: Arbitrary dictionary to store metadata about this `Pipeline`. Make sure all the values contained in this dictionary can be serialized and deserialized if you wish to save this `Pipeline` to file. :param max_runs_per_component: How many times the `Pipeline` can run the same Component. If this limit is reached a `PipelineMaxComponentRuns` exception is raised. If not set defaults to 100 runs per Component. :param connection_type_validation: Whether the pipeline will validate the types of the connections. Defaults to True. """ self._telemetry_runs = 0 self._last_telemetry_sent: datetime | None = None self.metadata = metadata or {} self.graph = networkx.MultiDiGraph() self._max_runs_per_component = max_runs_per_component self._connection_type_validation = connection_type_validation def __eq__(self, other: object) -> bool: """ Pipeline equality is defined by their type and the equality of their serialized form. Pipelines of the same type share every metadata, node and edge, but they're not required to use the same node instances: this allows pipeline saved and then loaded back to be equal to themselves. """ if not isinstance(self, type(other)): return False assert isinstance(other, PipelineBase) return self.to_dict() == other.to_dict() def __repr__(self) -> str: """ Returns a text representation of the Pipeline. """ res = f"{object.__repr__(self)}\n" if self.metadata: res += "🧱 Metadata\n" for k, v in self.metadata.items(): res += f" - {k}: {v}\n" res += "🚅 Components\n" for name, instance in self.graph.nodes(data="instance"): res += f" - {name}: {instance.__class__.__name__}\n" res += "🛤️ Connections\n" for sender, receiver, edge_data in self.graph.edges(data=True): sender_socket = edge_data["from_socket"].name receiver_socket = edge_data["to_socket"].name res += f" - {sender}.{sender_socket} -> {receiver}.{receiver_socket} ({edge_data['conn_type']})\n" return res def to_dict(self) -> dict[str, Any]: """ Serializes the pipeline to a dictionary. This is meant to be an intermediate representation but it can be also used to save a pipeline to file. :returns: Dictionary with serialized data. """ components = {} for name, instance in self.graph.nodes(data="instance"): components[name] = component_to_dict(instance, name) connections = [] for sender, receiver, edge_data in self.graph.edges.data(): sender_socket = edge_data["from_socket"].name receiver_socket = edge_data["to_socket"].name connections.append({"sender": f"{sender}.{sender_socket}", "receiver": f"{receiver}.{receiver_socket}"}) return { "metadata": self.metadata, "max_runs_per_component": self._max_runs_per_component, "components": components, "connections": connections, "connection_type_validation": self._connection_type_validation, } @classmethod def from_dict( cls: type[T], data: dict[str, Any], callbacks: DeserializationCallbacks | None = None, *, allowed_modules: list[str] | None = None, unsafe: bool = False, **kwargs: Any, ) -> T: """ Deserializes the pipeline from a dictionary. :param data: Dictionary to deserialize from. :param callbacks: Callbacks to invoke during deserialization. :param allowed_modules: Additional module patterns whose classes may be imported during deserialization. By default, only modules under `haystack`, `haystack_integrations`, `haystack_experimental`, `builtins`, `typing`, and `collections` are trusted. See `haystack.core.serialization.allow_deserialization_module` for the matching semantics. :param unsafe: If `True`, bypass the deserialization allowlist entirely. Only use this when you fully trust the source of the serialized data — any class in any importable module can be instantiated. :param kwargs: `components`: a dictionary of `{name: instance}` to reuse instances of components instead of creating new ones. :returns: Deserialized component. """ with _deserialization_context(allowed_modules=allowed_modules, unsafe=unsafe): return cls._from_dict_impl(data, callbacks, **kwargs) @classmethod def _from_dict_impl( cls: type[T], data: dict[str, Any], callbacks: DeserializationCallbacks | None = None, **kwargs: Any ) -> T: data_copy = _deepcopy_with_exceptions(data) # to prevent modification of original data metadata = data_copy.get("metadata", {}) max_runs_per_component = data_copy.get("max_runs_per_component", 100) connection_type_validation = data_copy.get("connection_type_validation", True) pipe = cls( metadata=metadata, max_runs_per_component=max_runs_per_component, connection_type_validation=connection_type_validation, ) components_to_reuse = kwargs.get("components", {}) for name, component_data in data_copy.get("components", {}).items(): if name in components_to_reuse: # Reuse an instance instance = components_to_reuse[name] else: if "type" not in component_data: raise PipelineError(f"Missing 'type' in component '{name}'") component_type = component_data["type"] if isinstance(component_type, str) and "." in component_type: _check_module_allowed(component_type.rsplit(".", 1)[0]) if component_type not in component.registry: try: # Import the module first... module, _ = component_type.rsplit(".", 1) logger.debug("Trying to import module {module_name}", module_name=module) type_serialization.thread_safe_import(module) # ...then try again if component_type not in component.registry: raise PipelineError( # noqa: TRY301 f"Successfully imported module '{module}' but couldn't find " f"'{component_type}' in the component registry.\n" f"The component might be registered under a different path. " f"Here are the registered components:\n {list(component.registry.keys())}\n" ) except (ImportError, PipelineError, ValueError) as e: raise PipelineError( f"Component '{component_type}' (name: '{name}') not imported. Please " f"check that the package is installed and the component path is correct." ) from e # Create a new one component_class = component.registry[component_type] try: instance = component_from_dict(component_class, component_data, name, callbacks) except Exception as e: # Convert to JSON with indentation, truncate if too long try: data_str = json.dumps(component_data, default=str, indent=2) except Exception: data_str = str(component_data) max_len = 1000 if len(data_str) > max_len: data_str = data_str[:max_len] + "\n... (truncated)" msg = ( f"Couldn't deserialize component '{name}' of class '{component_class.__name__}' " f"with the following data:\n{data_str}\n\n" f"Original error: {e}" ) raise DeserializationError(msg) from e pipe.add_component(name=name, instance=instance) for connection in data.get("connections", []): if "sender" not in connection: raise PipelineError(f"Missing sender in connection: {connection}") if "receiver" not in connection: raise PipelineError(f"Missing receiver in connection: {connection}") pipe.connect(sender=connection["sender"], receiver=connection["receiver"]) return pipe def dumps(self, marshaller: Marshaller = DEFAULT_MARSHALLER) -> str: """ Returns the string representation of this pipeline according to the format dictated by the `Marshaller` in use. :param marshaller: The Marshaller used to create the string representation. Defaults to `YamlMarshaller`. :returns: A string representing the pipeline. """ return marshaller.marshal(self.to_dict()) def dump(self, fp: TextIO, marshaller: Marshaller = DEFAULT_MARSHALLER) -> None: """ Writes the string representation of this pipeline to the file-like object passed in the `fp` argument. :param fp: A file-like object ready to be written to. :param marshaller: The Marshaller used to create the string representation. Defaults to `YamlMarshaller`. """ fp.write(marshaller.marshal(self.to_dict())) @classmethod def loads( cls: type[T], data: str | bytes | bytearray, marshaller: Marshaller = DEFAULT_MARSHALLER, callbacks: DeserializationCallbacks | None = None, *, allowed_modules: list[str] | None = None, unsafe: bool = False, ) -> T: """ Creates a `Pipeline` object from the string representation passed in the `data` argument. :param data: The string representation of the pipeline, can be `str`, `bytes` or `bytearray`. :param marshaller: The Marshaller used to create the string representation. Defaults to `YamlMarshaller`. :param callbacks: Callbacks to invoke during deserialization. :param allowed_modules: Additional module patterns whose classes may be imported during deserialization. By default, only modules under `haystack`, `haystack_integrations`, `haystack_experimental`, `builtins`, `typing`, and `collections` are trusted. :param unsafe: If `True`, bypass the deserialization allowlist entirely. Only use this when you fully trust the source of the serialized data — any class in any importable module can be instantiated. :raises DeserializationError: If an error occurs during deserialization. :returns: A `Pipeline` object. """ try: deserialized_data = marshaller.unmarshal(data) except Exception as e: raise DeserializationError( "Error while unmarshalling serialized pipeline data. This is usually " "caused by malformed or invalid syntax in the serialized representation." ) from e return cls.from_dict(deserialized_data, callbacks, allowed_modules=allowed_modules, unsafe=unsafe) @classmethod def load( cls: type[T], fp: TextIO, marshaller: Marshaller = DEFAULT_MARSHALLER, callbacks: DeserializationCallbacks | None = None, *, allowed_modules: list[str] | None = None, unsafe: bool = False, ) -> T: """ Creates a `Pipeline` object from a string representation. The string representation is read from the file-like object passed in the `fp` argument. :param fp: A file-like object ready to be read from. :param marshaller: The Marshaller used to create the string representation. Defaults to `YamlMarshaller`. :param callbacks: Callbacks to invoke during deserialization. :param allowed_modules: Additional module patterns whose classes may be imported during deserialization. By default, only modules under `haystack`, `haystack_integrations`, `haystack_experimental`, `builtins`, `typing`, and `collections` are trusted. :param unsafe: If `True`, bypass the deserialization allowlist entirely. Only use this when you fully trust the source of the serialized data — any class in any importable module can be instantiated. :raises DeserializationError: If an error occurs during deserialization. :returns: A `Pipeline` object. """ return cls.loads(fp.read(), marshaller, callbacks, allowed_modules=allowed_modules, unsafe=unsafe) def add_component(self, name: str, instance: Component) -> None: """ Add the given component to the pipeline. Components are not connected to anything by default: use `Pipeline.connect()` to connect components together. Component names must be unique, but component instances can be reused if needed. :param name: The name of the component to add. :param instance: The component instance to add. :raises ValueError: If a component with the same name already exists. :raises PipelineValidationError: If the given instance is not a component. """ # Component names are unique if name in self.graph.nodes: raise ValueError(f"A component named '{name}' already exists in this pipeline: choose another name.") # Components can't be named `_debug` if name == "_debug": raise ValueError("'_debug' is a reserved name for debug output. Choose another name.") # Component names can't have "." if "." in name: raise ValueError(f"{name} is an invalid component name, cannot contain '.' (dot) characters.") # Component instances must be components if not isinstance(instance, Component): raise PipelineValidationError( f"'{type(instance)}' doesn't seem to be a component. Is this class decorated with @component?" ) if getattr(instance, "__haystack_added_to_pipeline__", None): msg = ( "Component has already been added in another Pipeline. Components can't be shared between Pipelines. " "Create a new instance instead." ) raise PipelineError(msg) setattr(instance, "__haystack_added_to_pipeline__", self) # noqa: B010 setattr(instance, "__component_name__", name) # noqa: B010 # Add component to the graph, disconnected logger.debug("Adding component '{component_name}' ({component})", component_name=name, component=instance) # We're completely sure the fields exist so we ignore the type error self.graph.add_node( name, instance=instance, input_sockets=instance.__haystack_input__._sockets_dict, # type: ignore[attr-defined] output_sockets=instance.__haystack_output__._sockets_dict, # type: ignore[attr-defined] visits=0, ) def remove_component(self, name: str) -> Component: """ Remove and returns component from the pipeline. Remove an existing component from the pipeline by providing its name. All edges that connect to the component will also be deleted. :param name: The name of the component to remove. :returns: The removed Component instance. :raises ValueError: If there is no component with that name already in the Pipeline. """ # Check that a component with that name is in the Pipeline try: instance = self.get_component(name) except ValueError as exc: raise ValueError( f"There is no component named '{name}' in the pipeline. The valid component names are: ", ", ".join(n for n in self.graph.nodes), ) from exc # Delete component from the graph, deleting all its connections self.graph.remove_node(name) # Reset the Component sockets' senders and receivers input_sockets = instance.__haystack_input__._sockets_dict # type: ignore[attr-defined] for socket in input_sockets.values(): socket.senders = [] output_sockets = instance.__haystack_output__._sockets_dict # type: ignore[attr-defined] for socket in output_sockets.values(): socket.receivers = [] # Reset the Component's pipeline reference setattr(instance, "__haystack_added_to_pipeline__", None) # noqa: B010 return instance def connect(self, sender: str, receiver: str) -> "PipelineBase": # noqa: PLR0915 PLR0912 C901 """ Connects two components together. All components to connect must exist in the pipeline. If connecting to a component that has several output connections, specify the inputs and output names as 'component_name.connections_name'. If multiple senders are connected to the same list-typed receiver socket, the socket is promoted to a lazy variadic socket so it can accept all incoming values. With the synchronous `run`, the resulting list is ordered alphabetically by sender component name, not by the order in which `connect()` was called. With the asynchronous run path (`run_async`), no ordering is guaranteed, since components in different branches may run in parallel. :param sender: The component that delivers the value. This can be either just a component name or can be in the format `component_name.connection_name` if the component has multiple outputs. :param receiver: The component that receives the value. This can be either just a component name or can be in the format `component_name.connection_name` if the component has multiple inputs. :returns: The Pipeline instance. :raises PipelineConnectError: If the two components cannot be connected (for example if one of the components is not present in the pipeline, or the connections don't match by type, and so on). """ # Edges may be named explicitly by passing 'node_name.edge_name' to connect(). sender_component_name, sender_socket_name = parse_connect_string(sender) receiver_component_name, receiver_socket_name = parse_connect_string(receiver) if sender_component_name == receiver_component_name: raise PipelineConnectError("Connecting a Component to itself is not supported.") # Get the nodes data. try: sender_sockets = self.graph.nodes[sender_component_name]["output_sockets"] except KeyError as exc: raise ValueError(f"Component named {sender_component_name} not found in the pipeline.") from exc try: receiver_sockets = self.graph.nodes[receiver_component_name]["input_sockets"] except KeyError as exc: raise ValueError(f"Component named {receiver_component_name} not found in the pipeline.") from exc if not sender_sockets: raise PipelineConnectError( f"'{sender_component_name}' does not have any output connections. " f"Please check that the output types of '{sender_component_name}.run' are set, " f"for example by using the '@component.output_types' decorator." ) # If the name of either socket is given, get the socket sender_socket: OutputSocket | None = None if sender_socket_name: sender_socket = sender_sockets.get(sender_socket_name) if not sender_socket: raise PipelineConnectError( f"'{sender}' does not exist. " f"Output connections of {sender_component_name} are: " + ", ".join([f"{name} (type {_type_name(socket.type)})" for name, socket in sender_sockets.items()]) ) receiver_socket: InputSocket | None = None if receiver_socket_name: receiver_socket = receiver_sockets.get(receiver_socket_name) if not receiver_socket: raise PipelineConnectError( f"'{receiver} does not exist. " f"Input connections of {receiver_component_name} are: " + ", ".join( [f"{name} (type {_type_name(socket.type)})" for name, socket in receiver_sockets.items()] ) ) # Look for a matching connection among the possible ones. # Note that if there is more than one possible connection but two sockets match by name, they're paired. sender_socket_candidates: list[OutputSocket] = ( [sender_socket] if sender_socket else list(sender_sockets.values()) ) receiver_socket_candidates: list[InputSocket] = ( [receiver_socket] if receiver_socket else list(receiver_sockets.values()) ) conversion_strategy = None # Find all possible connections between these two components possible_connections: list[tuple[OutputSocket, InputSocket, ConversionStrategyType]] = [] for sender_sock, receiver_sock in itertools.product(sender_socket_candidates, receiver_socket_candidates): is_compat, conversion_strategy = _types_are_compatible( sender_sock.type, receiver_sock.type, self._connection_type_validation ) if is_compat: possible_connections.append((sender_sock, receiver_sock, conversion_strategy)) # If there are multiple possibilities, prioritize strict matches over convertible ones. # This ensures backward compatibility: previously, pipelines did not allow type conversion. if len(possible_connections) > 1 and self._connection_type_validation: strict_matches = [ (out_sock, in_sock, None) for out_sock, in_sock, conversion_strategy in possible_connections if conversion_strategy is None ] if strict_matches: possible_connections[:] = strict_matches # We need this status for error messages, since we might need it in multiple places we calculate it here status = _connections_status( sender_node=sender_component_name, sender_sockets=sender_socket_candidates, receiver_node=receiver_component_name, receiver_sockets=receiver_socket_candidates, ) if not possible_connections: # There's no possible connection between these two components if len(sender_socket_candidates) == len(receiver_socket_candidates) == 1: msg = ( f"Cannot connect '{sender_component_name}.{sender_socket_candidates[0].name}' with " f"'{receiver_component_name}.{receiver_socket_candidates[0].name}': " f"their declared input and output types do not match.\n{status}" ) else: msg = ( f"Cannot connect '{sender_component_name}' with '{receiver_component_name}': " f"no matching connections available.\n{status}" ) raise PipelineConnectError(msg) if len(possible_connections) == 1: # There's only one possible connection, use it sender_socket = possible_connections[0][0] receiver_socket = possible_connections[0][1] conversion_strategy = possible_connections[0][2] if len(possible_connections) > 1: # There are multiple possible connection, let's try to match them by name name_matches = [ (out_sock, in_sock, conversion_strategy_) for out_sock, in_sock, conversion_strategy_ in possible_connections if in_sock.name == out_sock.name ] if len(name_matches) != 1: # There's are either no matches or more than one, we can't pick one reliably msg = ( f"Cannot connect '{sender_component_name}' with " f"'{receiver_component_name}': more than one connection is possible " "between these components. Please specify the connection name, like: " f"pipeline.connect('{sender_component_name}.{possible_connections[0][0].name}', " f"'{receiver_component_name}.{possible_connections[0][1].name}').\n{status}" ) raise PipelineConnectError(msg) # Get the only possible match sender_socket = name_matches[0][0] receiver_socket = name_matches[0][1] conversion_strategy = name_matches[0][2] # Connection must be valid on both sender/receiver sides if not sender_socket or not receiver_socket or not sender_component_name or not receiver_component_name: if sender_component_name and sender_socket: sender_repr = f"{sender_component_name}.{sender_socket.name} ({_type_name(sender_socket.type)})" else: sender_repr = "input needed" if receiver_component_name and receiver_socket: receiver_repr = f"({_type_name(receiver_socket.type)}) {receiver_component_name}.{receiver_socket.name}" else: receiver_repr = "output" msg = f"Connection must have both sender and receiver: {sender_repr} -> {receiver_repr}" raise PipelineConnectError(msg) logger.debug( "Connecting '{sender_component}.{sender_socket_name}' to '{receiver_component}.{receiver_socket_name}'", sender_component=sender_component_name, sender_socket_name=sender_socket.name, receiver_component=receiver_component_name, receiver_socket_name=receiver_socket.name, ) if receiver_component_name in sender_socket.receivers and sender_component_name in receiver_socket.senders: # This is already connected, nothing to do return self if receiver_socket.senders: receiver_socket = self._make_socket_auto_variadic( component_name=receiver_component_name, receiver_socket=receiver_socket, error_type=PipelineConnectError ) # Update the sockets with the new connection sender_socket.receivers.append(receiver_component_name) receiver_socket.senders.append(sender_component_name) # Create the new connection self.graph.add_edge( sender_component_name, receiver_component_name, key=f"{sender_socket.name}/{receiver_socket.name}", conn_type=_type_name(sender_socket.type), from_socket=sender_socket, to_socket=receiver_socket, mandatory=receiver_socket.is_mandatory, conversion_strategy=conversion_strategy, ) return self def get_component(self, name: str) -> Component: """ Get the component with the specified name from the pipeline. :param name: The name of the component. :returns: The instance of that component. :raises ValueError: If a component with that name is not present in the pipeline. """ try: return self.graph.nodes[name]["instance"] except KeyError as exc: raise ValueError(f"Component named {name} not found in the pipeline.") from exc def get_component_name(self, instance: Component) -> str: """ Returns the name of the Component instance if it has been added to this Pipeline or an empty string otherwise. :param instance: The Component instance to look for. :returns: The name of the Component instance. """ for name, inst in self.graph.nodes(data="instance"): if inst == instance: return name return "" def inputs(self, include_components_with_connected_inputs: bool = False) -> dict[str, dict[str, Any]]: """ Returns a dictionary containing the inputs of a pipeline. Each key in the dictionary corresponds to a component name, and its value is another dictionary that describes the input sockets of that component, including their types and whether they are optional. :param include_components_with_connected_inputs: If `False`, only components that have disconnected input edges are included in the output. :returns: A dictionary where each key is a pipeline component name and each value is a dictionary of inputs sockets of that component. """ inputs: dict[str, dict[str, Any]] = {} for component_name, data in find_pipeline_inputs(self.graph, include_components_with_connected_inputs).items(): sockets_description = {} for socket in data: # Variadic mandatory sockets with existing connections don't require user input, so treat them as # optional. is_mandatory = socket.is_mandatory and not socket.senders sockets_description[socket.name] = {"type": socket.type, "is_mandatory": is_mandatory} if not socket.is_mandatory: sockets_description[socket.name]["default_value"] = socket.default_value if sockets_description: inputs[component_name] = sockets_description return inputs def outputs(self, include_components_with_connected_outputs: bool = False) -> dict[str, dict[str, Any]]: """ Returns a dictionary containing the outputs of a pipeline. Each key in the dictionary corresponds to a component name, and its value is another dictionary that describes the output sockets of that component. :param include_components_with_connected_outputs: If `False`, only components that have disconnected output edges are included in the output. :returns: A dictionary where each key is a pipeline component name and each value is a dictionary of output sockets of that component. """ return { comp: {socket.name: {"type": socket.type} for socket in data} for comp, data in find_pipeline_outputs(self.graph, include_components_with_connected_outputs).items() if data } def show( self, *, server_url: str = "https://mermaid.ink", params: dict | None = None, timeout: int = 30, super_component_expansion: bool = False, ) -> None: """ Display an image representing this `Pipeline` in a Jupyter notebook. This function generates a diagram of the `Pipeline` using a Mermaid server and displays it directly in the notebook. :param server_url: The base URL of the Mermaid server used for rendering (default: 'https://mermaid.ink'). See https://github.com/jihchi/mermaid.ink and https://github.com/mermaid-js/mermaid-live-editor for more info on how to set up your own Mermaid server. :param params: Dictionary of customization parameters to modify the output. Refer to Mermaid documentation for more details Supported keys: - format: Output format ('img', 'svg', or 'pdf'). Default: 'img'. - type: Image type for /img endpoint ('jpeg', 'png', 'webp'). Default: 'png'. - theme: Mermaid theme ('default', 'neutral', 'dark', 'forest'). Default: 'neutral'. - bgColor: Background color in hexadecimal (e.g., 'FFFFFF') or named format (e.g., '!white'). - width: Width of the output image (integer). - height: Height of the output image (integer). - scale: Scaling factor (1–3). Only applicable if 'width' or 'height' is specified. - fit: Whether to fit the diagram size to the page (PDF only, boolean). - paper: Paper size for PDFs (e.g., 'a4', 'a3'). Ignored if 'fit' is true. - landscape: Landscape orientation for PDFs (boolean). Ignored if 'fit' is true. :param timeout: Timeout in seconds for the request to the Mermaid server. :param super_component_expansion: If set to True and the pipeline contains SuperComponents the diagram will show the internal structure of super-components as if they were components part of the pipeline instead of a "black-box". Otherwise, only the super-component itself will be displayed. :raises PipelineDrawingError: If the function is called outside of a Jupyter notebook or if there is an issue with rendering. """ if is_in_jupyter(): from IPython.display import Image, display if super_component_expansion: graph, super_component_mapping = self._merge_super_component_pipelines() else: graph = self.graph super_component_mapping = None image_data = _to_mermaid_image( graph, server_url=server_url, params=params, timeout=timeout, super_component_mapping=super_component_mapping, ) display(Image(image_data)) else: msg = "This method is only supported in Jupyter notebooks. Use Pipeline.draw() to save an image locally." raise PipelineDrawingError(msg) def draw( self, *, path: Path, server_url: str = "https://mermaid.ink", params: dict | None = None, timeout: int = 30, super_component_expansion: bool = False, ) -> None: """ Save an image representing this `Pipeline` to the specified file path. This function generates a diagram of the `Pipeline` using the Mermaid server and saves it to the provided path. :param path: The file path where the generated image will be saved. :param server_url: The base URL of the Mermaid server used for rendering (default: 'https://mermaid.ink'). See https://github.com/jihchi/mermaid.ink and https://github.com/mermaid-js/mermaid-live-editor for more info on how to set up your own Mermaid server. :param params: Dictionary of customization parameters to modify the output. Refer to Mermaid documentation for more details Supported keys: - format: Output format ('img', 'svg', or 'pdf'). Default: 'img'. - type: Image type for /img endpoint ('jpeg', 'png', 'webp'). Default: 'png'. - theme: Mermaid theme ('default', 'neutral', 'dark', 'forest'). Default: 'neutral'. - bgColor: Background color in hexadecimal (e.g., 'FFFFFF') or named format (e.g., '!white'). - width: Width of the output image (integer). - height: Height of the output image (integer). - scale: Scaling factor (1–3). Only applicable if 'width' or 'height' is specified. - fit: Whether to fit the diagram size to the page (PDF only, boolean). - paper: Paper size for PDFs (e.g., 'a4', 'a3'). Ignored if 'fit' is true. - landscape: Landscape orientation for PDFs (boolean). Ignored if 'fit' is true. :param timeout: Timeout in seconds for the request to the Mermaid server. :param super_component_expansion: If set to True and the pipeline contains SuperComponents the diagram will show the internal structure of super-components as if they were components part of the pipeline instead of a "black-box". Otherwise, only the super-component itself will be displayed. :raises PipelineDrawingError: If there is an issue with rendering or saving the image. """ # Before drawing we edit a bit the graph, to avoid modifying the original that is # used for running the pipeline we copy it. if super_component_expansion: graph, super_component_mapping = self._merge_super_component_pipelines() else: graph = self.graph super_component_mapping = None image_data = _to_mermaid_image( graph, server_url=server_url, params=params, timeout=timeout, super_component_mapping=super_component_mapping, ) Path(path).write_bytes(image_data) def walk(self) -> Iterator[tuple[str, Component]]: """ Visits each component in the pipeline exactly once and yields its name and instance. No guarantees are provided on the visiting order. :returns: An iterator of tuples of component name and component instance. """ for component_name, instance in self.graph.nodes(data="instance"): # noqa: UP028 yield component_name, instance def warm_up(self) -> None: """ Make sure all components are warm. It's the component's responsibility to make sure this method can be called at every `Pipeline.run()` without re-initializing everything. """ for component_name in self.graph.nodes: if hasattr(self.graph.nodes[component_name]["instance"], "warm_up"): logger.info("Warming up component {component_name}...", component_name=component_name) self.graph.nodes[component_name]["instance"].warm_up() async def warm_up_async(self) -> None: """ Make sure all components are warm, using the async warm-up path where available. Each component is warmed up with `warm_up_async` if it has one, otherwise with its sync `warm_up`. Both run on the event loop, never offloaded to a worker thread. This ensures that if an async client is created during `warm-up` (residual scenario), it binds to the loop that `run_async` will use. """ for component_name in self.graph.nodes: instance = self.graph.nodes[component_name]["instance"] if hasattr(instance, "warm_up_async"): logger.info("Warming up component {component_name}...", component_name=component_name) await instance.warm_up_async() elif hasattr(instance, "warm_up"): logger.info("Warming up component {component_name}...", component_name=component_name) instance.warm_up() def close(self) -> None: """ Release resources held by the pipeline's components by calling each component's `close` method. Only the synchronous side of each component is released here; use `close_async` to release async clients. """ for component_name in self.graph.nodes: instance = self.graph.nodes[component_name]["instance"] if hasattr(instance, "close"): logger.info("Closing component {component_name}...", component_name=component_name) instance.close() async def close_async(self) -> None: """ Release resources held by the pipeline's components, using the async close path where available. Each component is closed with `close_async` if it has one, otherwise with its sync `close`. """ for component_name in self.graph.nodes: instance = self.graph.nodes[component_name]["instance"] if hasattr(instance, "close_async"): logger.info("Closing component {component_name}...", component_name=component_name) await instance.close_async() elif hasattr(instance, "close"): logger.info("Closing component {component_name}...", component_name=component_name) instance.close() @staticmethod def _create_component_span( component_name: str, instance: Component, inputs: dict[str, Any], parent_span: tracing.Span | None = None ) -> ContextManager[tracing.Span]: return tracing.tracer.trace( "haystack.component.run", tags={ "haystack.component.name": component_name, "haystack.component.type": instance.__class__.__name__, "haystack.component.fully_qualified_type": generate_qualified_class_name(type(instance)), "haystack.component.input_types": {k: type(v).__name__ for k, v in inputs.items()}, "haystack.component.input_spec": { key: { "type": value.type.__name__ if type(value.type) is type else str(value.type), "senders": value.senders, } for key, value in instance.__haystack_input__._sockets_dict.items() # type: ignore }, "haystack.component.output_spec": { key: { "type": value.type.__name__ if type(value.type) is type else str(value.type), "receivers": value.receivers, } for key, value in instance.__haystack_output__._sockets_dict.items() # type: ignore }, }, parent_span=parent_span, ) def validate_input(self, data: dict[str, Any]) -> None: """ Validates pipeline input data. Validates that data: * Each Component name actually exists in the Pipeline * Each Component is not missing any input * Each Component has only one input per input socket, if not variadic * Each Component doesn't receive inputs that are already sent by another Component :param data: A dictionary of inputs for the pipeline's components. Each key is a component name. :raises ValueError: If inputs are invalid according to the above. """ for component_name, component_inputs in data.items(): # Check that the component exists if component_name not in self.graph.nodes: raise ValueError(f"Component named '{component_name}' not found in the pipeline.") # Check that no input is provided that the component can't accept instance = self.graph.nodes[component_name]["instance"] for input_name in component_inputs.keys(): if input_name not in instance.__haystack_input__._sockets_dict: raise ValueError(f"Input '{input_name}' not found in component '{component_name}'.") for component_name in self.graph.nodes: instance = self.graph.nodes[component_name]["instance"] for socket_name, socket in instance.__haystack_input__._sockets_dict.items(): component_inputs = data.get(component_name, {}) # Check that no mandatory input is missing for any component in the pipeline if socket.senders == [] and socket.is_mandatory and socket_name not in component_inputs: raise ValueError(f"Missing mandatory input '{socket_name}' for component '{component_name}'.") # Check if an input is provided more than once for non-variadic sockets if socket.senders and socket_name in component_inputs: self._make_socket_auto_variadic( component_name=component_name, receiver_socket=socket, error_type=ValueError ) def _make_socket_auto_variadic( self, component_name: str, receiver_socket: InputSocket, error_type: type[Exception] ) -> InputSocket: """ Attempts to make the receiver socket lazy variadic in-place to accommodate a new sender. A socket is automatically made lazy variadic when: - It already has at least one connected sender - It is not already variadic - Its type is list, Optional[list], a union of list types When auto-variadicity is applied, `wrap_input_in_list` is also set to False so that sender output types match the receiver socket's declared list type directly. :param component_name: Name of the component owning the receiver socket, used in error messages. :param receiver_socket: The receiver socket to inspect and potentially modify in-place. :param error_type: Exception class to raise on failure (e.g. ValueError or PipelineConnectError). :returns: The (possibly modified) receiver socket. :raises error_type: If the socket cannot accept multiple senders given its type constraints. """ # If it's already variadic, we return as-is if receiver_socket.is_variadic: return receiver_socket # Get receiver origin receiver_origin = _safe_get_origin(receiver_socket.type) # Handle Union types if receiver_origin == Union: # Unwrap Optional types non_none_args = [a for a in get_args(receiver_socket.type) if a is not type(None)] if len(non_none_args) == 1: receiver_origin = _safe_get_origin(non_none_args[0]) # Handle Union of list types (e.g. list[int] | list[str]) elif all(_safe_get_origin(arg) == list for arg in non_none_args): receiver_origin = list # If the receiver origin is a list, we can make the socket lazy variadic if receiver_origin == list: receiver_socket.is_lazy_variadic = True receiver_socket.wrap_input_in_list = False return receiver_socket raise error_type( f"Component '{component_name}' cannot accept multiple inputs to '{receiver_socket.name}'. " f"It is already connected to component '{receiver_socket.senders[0]}', and it can only accept " f"inputs from multiple senders if its type is list, Optional[list], or union of list types." ) def _prepare_component_input_data(self, data: dict[str, Any]) -> dict[str, dict[str, Any]]: """ Prepares input data for pipeline components. Organizes input data for pipeline components and identifies any inputs that are not matched to any component's input slots. Deep-copies data items to avoid sharing mutables across multiple components. This method processes a flat dictionary of input data, where each key-value pair represents an input name and its corresponding value. It distributes these inputs to the appropriate pipeline components based on their input requirements. Inputs that don't match any component's input slots are classified as unresolved. :param data: A dictionary potentially having input names as keys and input values as values. :returns: A dictionary mapping component names to their respective matched inputs. """ # check whether the data is a nested dictionary of component inputs where each key is a component name # and each value is a dictionary of input parameters for that component is_nested_component_input = all(isinstance(value, dict) for value in data.values()) if not is_nested_component_input: # flat input, a dict where keys are input names and values are the corresponding values # we need to convert it to a nested dictionary of component inputs and then run the pipeline # just like in the previous case pipeline_input_data: dict[str, dict[str, Any]] = defaultdict(dict) unresolved_kwargs = {} # Retrieve the input slots for each component in the pipeline available_inputs: dict[str, dict[str, Any]] = self.inputs() # Go through all provided to distribute them to the appropriate component inputs for input_name, input_value in data.items(): resolved_at_least_once = False # Check each component to see if it has a slot for the current kwarg for component_name, component_inputs in available_inputs.items(): if input_name in component_inputs: # If a match is found, add the kwarg to the component's input data pipeline_input_data[component_name][input_name] = input_value resolved_at_least_once = True if not resolved_at_least_once: unresolved_kwargs[input_name] = input_value if unresolved_kwargs: logger.warning( "Inputs {input_keys} were not matched to any component inputs, please check your run parameters.", input_keys=list(unresolved_kwargs.keys()), ) data = dict(pipeline_input_data) # deepcopying the inputs prevents the Pipeline run logic from being altered unexpectedly # when the same input reference is passed to multiple components. for component_name, component_inputs in data.items(): data[component_name] = {k: _deepcopy_with_exceptions(v) for k, v in component_inputs.items()} return data def _find_receivers_from( self, component_name: str ) -> list[tuple[str, OutputSocket, InputSocket, ConversionStrategyType]]: """ Utility function to find all Components that receive input from `component_name`. :param component_name: Name of the sender Component :returns: A list of tuples containing: - receiver component name - sender OutputSocket - receiver InputSocket - ConversionStrategy if conversion is required, otherwise None. """ res = [] for _, receiver_name, connection in self.graph.edges(nbunch=component_name, data=True): sender_socket: OutputSocket = connection["from_socket"] receiver_socket: InputSocket = connection["to_socket"] res.append((receiver_name, sender_socket, receiver_socket, connection.get("conversion_strategy"))) return res @staticmethod def _convert_to_internal_format(pipeline_inputs: dict[str, Any]) -> InputsType: """ Converts the inputs to the pipeline to the format that is needed for the internal `Pipeline.run` logic. Example Input: {'prompt_builder': {'question': 'Who lives in Paris?'}, 'retriever': {'query': 'Who lives in Paris?'}} Example Output: {'prompt_builder': {'question': [{'sender': None, 'value': 'Who lives in Paris?'}]}, 'retriever': {'query': [{'sender': None, 'value': 'Who lives in Paris?'}]}} :param pipeline_inputs: Inputs to the pipeline. :returns: Converted inputs that can be used by the internal `Pipeline.run` logic. """ inputs: InputsType = {} for component_name, socket_dict in pipeline_inputs.items(): inputs[component_name] = {} for socket_name, value in socket_dict.items(): inputs[component_name][socket_name] = [{"sender": None, "value": value}] return inputs @staticmethod def _consume_component_inputs( component_name: str, component: dict, inputs: InputsType, is_resume: bool = False ) -> dict[str, Any]: """ Extracts the inputs needed to run for the component and removes them from the global inputs state. :param component_name: The name of a component. :param component: Component with component metadata. :param inputs: Global inputs state. :returns: The inputs for the component. """ component_inputs = inputs.get(component_name, {}) consumed_inputs = {} greedy_inputs_to_remove = set() for socket_name, socket in component["input_sockets"].items(): socket_inputs = component_inputs.get(socket_name, []) socket_inputs_values = [sock["value"] for sock in socket_inputs if sock["value"] is not _NO_OUTPUT_PRODUCED] # if we are resuming a component, the inputs are already consumed, so we just return the first input if is_resume: consumed_inputs[socket_name] = socket_inputs_values[0] continue if socket_inputs_values: if socket.is_greedy: # We need to keep track of greedy inputs because we always remove them, even if they come from # outside the pipeline. Otherwise, a greedy input from the user would trigger a pipeline to run # indefinitely. greedy_inputs_to_remove.add(socket_name) consumed_inputs[socket_name] = [socket_inputs_values[0]] elif socket.is_lazy_variadic: if socket.wrap_input_in_list: # We use all inputs provided to the socket on a lazy variadic socket. # So keep it wrapped in a list. consumed_inputs[socket_name] = socket_inputs_values else: # We flatten one-level of lists for lazy variadic sockets that don't wrap inputs in lists. # This way the incoming inputs match the expected type of the socket. consumed_inputs[socket_name] = list(itertools.chain.from_iterable(socket_inputs_values)) else: # For a normal socket we only care about the first input provided to the socket. consumed_inputs[socket_name] = socket_inputs_values[0] # We prune all inputs except for those that were provided from outside the pipeline (e.g. user inputs). pruned_inputs = { socket_name: [ sock for sock in socket if sock["sender"] is None and socket_name not in greedy_inputs_to_remove ] for socket_name, socket in component_inputs.items() } pruned_inputs = {socket_name: socket for socket_name, socket in pruned_inputs.items() if len(socket) > 0} inputs[component_name] = pruned_inputs return consumed_inputs def _fill_queue( self, component_names: list[str], inputs: InputsType, component_visits: dict[str, int] ) -> FIFOPriorityQueue: """ Calculates the execution priority for each component and inserts it into the priority queue. :param component_names: Names of the components to put into the queue. :param inputs: Inputs to the components. :param component_visits: Current state of component visits. :returns: A prioritized queue of component names. """ priority_queue = FIFOPriorityQueue() for component_name in component_names: comp = self._get_component_with_graph_metadata_and_visits(component_name, component_visits[component_name]) priority = self._calculate_priority(comp, inputs.get(component_name, {})) priority_queue.push(component_name, priority) return priority_queue @staticmethod def _calculate_priority(comp: dict, comp_inputs: dict[str, list[dict[str, Any]]]) -> ComponentPriority: """ Calculates the execution priority for a component depending on the component's inputs. :param comp: Component metadata and component instance. :param comp_inputs: Inputs to the component. :returns: Priority value for the component. """ # NOTE: Even if a component can run, it doesn't mean it's ready to run since it could be waiting for optional # inputs. This is why it's only used to determine if a component is BLOCKED or not. if not can_component_run(comp, comp_inputs): return ComponentPriority.BLOCKED if is_any_greedy_socket_ready(comp, comp_inputs) and are_all_sockets_ready(comp, comp_inputs): # This priority is explicitly used in the async run path + implicitly in _is_queue_stale # Implicit b/c it checks via ">" operator if there is a component with HIGHEST priority return ComponentPriority.HIGHEST if all_predecessors_executed(comp, comp_inputs): # This priority is explicitly used in the async run path + in _is_queue_stale return ComponentPriority.READY # If we make it here it means the component can run but is waiting for more inputs, so we give it the lowest # priority. This way, components that are ready to run will be prioritized over ones assigned with this prio. return ComponentPriority.DEFER def _get_component_with_graph_metadata_and_visits(self, component_name: str, visits: int) -> dict[str, Any]: """ Returns the component instance alongside input/output-socket metadata from the graph and adds current visits. We can't store visits in the pipeline graph because this would prevent reentrance / thread-safe execution. :param component_name: The name of the component. :param visits: Number of visits for the component. :returns: dict with keys: - instance: the component instance - input_sockets: the component input sockets metadata from the graph - output_sockets: the component output sockets metadata from the graph """ comp_dict = self.graph.nodes[component_name] # We inject the visits into the component dict here to avoid storing it in the graph, which would prevent # thread-safe execution return {**comp_dict, "visits": visits} def _get_next_runnable_component( self, priority_queue: FIFOPriorityQueue, component_visits: dict[str, int] ) -> tuple[ComponentPriority, str, dict[str, Any]] | None: """ Returns the next runnable component alongside its metadata from the priority queue. :param priority_queue: Priority queue of component names. :param component_visits: Current state of component visits. :returns: The next runnable component, the component name, and its priority or None if no component in the queue can run. :raises: PipelineMaxComponentRuns if the next runnable component has exceeded the maximum number of runs. """ item = priority_queue.get() # If no component is runnable, return None if item is None: return None component_name = item[1] comp = self._get_component_with_graph_metadata_and_visits(component_name, component_visits[component_name]) # Only raise the max run count error if the component is not blocked, since if it's blocked it means it # can't run anyway. if item[0] < ComponentPriority.BLOCKED and comp["visits"] >= self._max_runs_per_component: msg = f"Maximum run count {self._max_runs_per_component} reached for component '{component_name}'" raise PipelineMaxComponentRuns(msg) return ComponentPriority(item[0]), component_name, comp @staticmethod def _add_missing_input_defaults( component_inputs: dict[str, list[dict[str, Any]]], component_input_sockets: dict[str, InputSocket] ) -> dict[str, Any]: """ Updates the inputs with the default values for the inputs that are missing :param component_inputs: Inputs for the component. :param component_input_sockets: Input sockets of the component. """ for name, socket in component_input_sockets.items(): if not socket.is_mandatory and name not in component_inputs: # NOTE: Variadic inputs expect a single default value in the function signature that matches the inner # type, for example Variadic[str] = "default". When executed inside a pipeline, we wrap this # default into a list, resulting in ["default"], which is the intended behavior. # # However, when the component is executed directly by calling run(), the default is not wrapped and # is treated as an iterable. # For strings, this would produce ["d", "e", "f", ...] instead of ["default"]. if socket.is_variadic: component_inputs[name] = [socket.default_value] else: component_inputs[name] = socket.default_value return component_inputs def _topological_sort(self) -> dict[str, int]: """ Returns a topological sort of the components in the pipeline. If the graph is a DAG, we use lexicographical topological sort to get a deterministic order of the components. If the graph is not a DAG, we use the condensation of the graph to get a topological sort of the strongly connected components. This way, components that are part of the same cycle will have the same priority and will be tie-broken by their name in lexicographical order, while components that are not part of the same cycle will be tie-broken by their topological order. :returns: A dictionary mapping component names to their position in the topological sort. """ if networkx.is_directed_acyclic_graph(self.graph): topological_sort = networkx.lexicographical_topological_sort(self.graph) return {node: idx for idx, node in enumerate(topological_sort)} else: condensed = networkx.condensation(self.graph) condensed_sorted = {node: idx for idx, node in enumerate(networkx.topological_sort(condensed))} return { component_name: condensed_sorted[node] for component_name, node in condensed.graph["mapping"].items() } def _tiebreak_waiting_components( self, component_name: str, priority: ComponentPriority, priority_queue: FIFOPriorityQueue, topological_sort: dict[str, int] | None, ) -> tuple[str, dict[str, int] | None]: """ Decides which component to run when multiple components are waiting for inputs with the same priority. NOTE: This was designed to only tie-break for components with the priority DEFER. Since this function also removes these components from the priority queue we rely on _is_queue_stale to then refill the priority queue. And _is_queue_stale only triggers when all remaining components have BLOCKED priority. :param component_name: The name of the component. :param priority: Priority of the component. :param priority_queue: Priority queue of component names. :param topological_sort: Cached topological sort of all components in the pipeline. :returns: The name of the component to run and the cached topological sort of all components in the pipeline. """ # Create a list of all components that have the same priority as the current component, including the # current component itself and remove them from the priority queue. components_with_same_priority = [component_name] while len(priority_queue) > 0: next_priority, next_component_name = priority_queue.peek() if priority == ComponentPriority.DEFER and next_priority == ComponentPriority.DEFER: priority_queue.pop() # actually remove the component components_with_same_priority.append(next_component_name) else: break # If there are multiple components with the same priority, we tiebreak them to decide which one to run first. if len(components_with_same_priority) > 1: topological_sort = topological_sort or self._topological_sort() components_with_same_priority = sorted( components_with_same_priority, key=lambda comp_name: (topological_sort[comp_name], comp_name.lower()) ) component_name = components_with_same_priority[0] return component_name, topological_sort def _find_components_blocking_pipeline( self, priority_queue: FIFOPriorityQueue, component_visits: dict[str, int], inputs: InputsType ) -> tuple[list[str], list[str]]: """ Finds the components that are most likely blocking the pipeline execution. :returns: The list of component names that are most likely blocking the pipeline execution and their corresponding component types. """ # 1. Go through all components in priority queue (should all be blocked at this point) comps_in_queue: list[str] = [comp_name for _, _, comp_name in priority_queue._queue] # 2. Check which components have entries in inputs. comps_with_inputs = [] for comp_name in comps_in_queue: # If component has non-empty inputs and is blocked it means that the component is waiting for more inputs # to run, so it could be blocking the pipeline. if inputs.get(comp_name): comps_with_inputs.append(comp_name) # If there are no components with any inputs we fallback to checking all components in the queue. # This isn't always ideal since already executed components can also be in the queue at this point also # with blocked priority. if not comps_with_inputs: comps_with_inputs = comps_in_queue # 3. Only keep components with the lowest number of visits. Mostly needed to handle the fallback case if no # components with inputs are found. ordered_comps_with_inputs = sorted(comps_with_inputs, key=lambda x: component_visits[x]) lowest_component_visit = component_visits[ordered_comps_with_inputs[0]] possible_blocking_comps = [ comp for comp in ordered_comps_with_inputs if component_visits[comp] == lowest_component_visit ] # If there is only one component with the lowest visits, return it as the most likely blocking component. if len(possible_blocking_comps) == 1: blocking_comp_types = [self.graph.nodes[possible_blocking_comps[0]]["instance"].__class__.__name__] self._log_warning_for_blocking_components( blocking_comp_names=possible_blocking_comps, blocking_comp_types=blocking_comp_types ) return possible_blocking_comps, blocking_comp_types # 4. Then for all components with the same lowest component visits we sort topologically before returning. topological_sort = self._topological_sort() possible_blocking_comps = sorted( possible_blocking_comps, key=lambda comp_name: (topological_sort[comp_name], comp_name.lower()) ) possible_blocking_comp_types = [ self.graph.nodes[comp_name]["instance"].__class__.__name__ for comp_name in possible_blocking_comps ] self._log_warning_for_blocking_components( blocking_comp_names=possible_blocking_comps, blocking_comp_types=possible_blocking_comp_types ) return possible_blocking_comps, possible_blocking_comp_types def _log_warning_for_blocking_components( self, blocking_comp_names: list[str], blocking_comp_types: list[dict] ) -> None: """ Logs a warning about the components that are most likely blocking the pipeline execution. :param blocking_comp_names: The list of component names that are most likely blocking the pipeline execution. :param blocking_comp_types: The list of component types that are most likely blocking the pipeline execution. """ comp_details = "\n".join( f" - '{name}' ({comp_type})" for name, comp_type in zip(blocking_comp_names, blocking_comp_types, strict=True) ) logger.warning( "Cannot run pipeline - the pipeline appears to be blocked.\n" "The following components could not be run and may be waiting on inputs that will " "never arrive:\n" + comp_details + "\n" "Note that some of these components may be intentionally inactive due to conditional " "branching. If this is unexpected, check the connections to these components and " "ensure all required inputs are provided.", component_names=blocking_comp_names, component_types=blocking_comp_types, ) def _write_component_outputs( self, *, component_name: str, component_outputs: Mapping[str, Any], inputs: InputsType, receivers: Sequence[tuple[str, OutputSocket, InputSocket, ConversionStrategyType]], include_outputs_from: set[str], ) -> Mapping[str, Any]: """ Distributes the outputs of a component to the input sockets that it is connected to. :param component_name: The name of the component. :param component_outputs: The outputs of the component. :param inputs: The current global input state. :param receivers: A sequence of tuples containing: - receiver component name, - output socket of the sender, - input socket of the receiver, - ConversionStrategy to be used to convert the value if required, otherwise None. :param include_outputs_from: Set of component names that should always return an output from the pipeline. """ for receiver_name, sender_socket, receiver_socket, conversion_strategy in receivers: # We either get the value that was produced by the actor or we use the _NO_OUTPUT_PRODUCED class to indicate # that the sender did not produce an output for this socket. # This allows us to track if a predecessor already ran but did not produce an output. value = component_outputs.get(sender_socket.name, _NO_OUTPUT_PRODUCED) if value is not _NO_OUTPUT_PRODUCED and conversion_strategy: try: value = _convert_value(value=value, conversion_strategy=conversion_strategy) except Exception as e: sender_node = self.graph.nodes.get(component_name) sender_instance = sender_node.get("instance") if sender_node else None sender_type_name = type(sender_instance).__name__ if sender_instance else "unknown" receiver_node = self.graph.nodes.get(receiver_name) receiver_instance = receiver_node.get("instance") if receiver_node else None receiver_type_name = type(receiver_instance).__name__ if receiver_instance else "unknown" msg = ( f"Failed to perform conversion between components:\n" f"Sender component: '{component_name}' (type: '{sender_type_name}')\n" f"Sender socket: '{sender_socket.name}'\n" f"Receiver component: '{receiver_name}' (type: '{receiver_type_name}')\n" f"Receiver socket: '{receiver_socket.name}'\n" f"Error: {e}" ) raise PipelineRuntimeError(component_name=None, component_type=None, message=msg) from e if receiver_name not in inputs: inputs[receiver_name] = {} if receiver_socket.is_lazy_variadic: # If the receiver socket is lazy variadic, we append the new input. # Lazy variadic sockets can collect multiple inputs. _write_to_lazy_variadic_socket( inputs=inputs, receiver_name=receiver_name, receiver_socket_name=receiver_socket.name, component_name=component_name, value=value, ) else: # If the receiver socket is not lazy variadic, it is greedy variadic or non-variadic. # We overwrite with the new input if it's not _NO_OUTPUT_PRODUCED or if the current value is None. _write_to_standard_socket( inputs=inputs, receiver_name=receiver_name, receiver_socket_name=receiver_socket.name, component_name=component_name, value=value, ) # If we want to include all outputs from this actor in the final outputs, we don't need to prune any consumed # outputs if component_name in include_outputs_from: return component_outputs # We prune outputs that were consumed by any receiving sockets. # All remaining outputs will be added to the final outputs of the pipeline. consumed_outputs = {sender_socket.name for _, sender_socket, __, ___ in receivers} return {key: value for key, value in component_outputs.items() if key not in consumed_outputs} @staticmethod def _is_queue_stale(priority_queue: FIFOPriorityQueue) -> bool: """ Checks if the priority queue needs to be recomputed because the priorities might have changed. The queue is considered stale if it is empty or if the highest priority component is not READY or HIGHEST. For example, if the next component in a queue has the priority READY then the equality becomes ComponentPriority.READY > ComponentPriority.READY which is false. However, if the next component has priority DEFER (or BLOCKED) then the equality becomes ComponentPriority.DEFER > ComponentPriority.READY which is true, indicating that the queue is stale. :param priority_queue: Priority queue of component names. """ return len(priority_queue) == 0 or priority_queue.peek()[0] > ComponentPriority.READY @staticmethod def validate_pipeline(priority_queue: FIFOPriorityQueue) -> None: """ Validate the pipeline to check if it is blocked or has no valid entry point. :param priority_queue: Priority queue of component names. :raises PipelineRuntimeError: If the pipeline is blocked or has no valid entry point. """ if len(priority_queue) == 0: return candidate = priority_queue.peek() if candidate is not None and candidate[0] == ComponentPriority.BLOCKED: raise PipelineComponentsBlockedError() def _find_super_components(self) -> list[tuple[str, Component]]: """ Find all SuperComponents in the pipeline. :returns: List of tuples containing (component_name, component_instance) representing a SuperComponent. """ super_components = [] for comp_name, comp in self.walk(): # a SuperComponent has a "pipeline" attribute which itself a Pipeline instance # we don't test against SuperComponent because doing so always lead to circular imports if hasattr(comp, "pipeline") and isinstance(comp.pipeline, self.__class__): super_components.append((comp_name, comp)) return super_components def _merge_super_component_pipelines(self) -> tuple[networkx.MultiDiGraph, dict[str, str]]: """ Merge the internal pipelines of SuperComponents into the main pipeline graph structure. This creates a new networkx.MultiDiGraph containing all the components from both the main pipeline and all the internal SuperComponents' pipelines. The SuperComponents are removed and their internal components are connected to corresponding input and output sockets of the main pipeline. :returns: A tuple containing: - A networkx.MultiDiGraph with the expanded structure of the main pipeline and all it's SuperComponents - A dictionary mapping component names to boolean indicating that this component was part of a SuperComponent - A dictionary mapping component names to their SuperComponent name """ merged_graph = self.graph.copy() super_component_mapping: dict[str, str] = {} for super_name, super_component in self._find_super_components(): internal_pipeline = super_component.pipeline # type: ignore internal_graph = internal_pipeline.graph.copy() # Mark all components in the internal pipeline as being part of a SuperComponent for node in internal_graph.nodes(): super_component_mapping[node] = super_name # edges connected to the super component incoming_edges = list(merged_graph.in_edges(super_name, data=True)) outgoing_edges = list(merged_graph.out_edges(super_name, data=True)) # merge the SuperComponent graph into the main graph and remove the super component node # since its components are now part of the main graph merged_graph = networkx.compose(merged_graph, internal_graph) merged_graph.remove_node(super_name) # get the entry and exit points of the SuperComponent internal pipeline entry_points = [n for n in internal_graph.nodes() if internal_graph.in_degree(n) == 0] exit_points = [n for n in internal_graph.nodes() if internal_graph.out_degree(n) == 0] # connect the incoming edges to entry points for sender, _, edge_data in incoming_edges: sender_socket = edge_data["from_socket"] for entry_point in entry_points: # find a matching input socket in the entry point entry_point_sockets = internal_graph.nodes[entry_point]["input_sockets"] for socket_name, socket in entry_point_sockets.items(): if _types_are_compatible(sender_socket.type, socket.type, self._connection_type_validation)[0]: merged_graph.add_edge( sender, entry_point, key=f"{sender_socket.name}/{socket_name}", conn_type=_type_name(sender_socket.type), from_socket=sender_socket, to_socket=socket, mandatory=socket.is_mandatory, ) # connect outgoing edges from exit points for _, receiver, edge_data in outgoing_edges: receiver_socket = edge_data["to_socket"] for exit_point in exit_points: # find a matching output socket in the exit point exit_point_sockets = internal_graph.nodes[exit_point]["output_sockets"] for socket_name, socket in exit_point_sockets.items(): if _types_are_compatible(socket.type, receiver_socket.type, self._connection_type_validation)[ 0 ]: merged_graph.add_edge( exit_point, receiver, key=f"{socket_name}/{receiver_socket.name}", conn_type=_type_name(socket.type), from_socket=socket, to_socket=receiver_socket, mandatory=receiver_socket.is_mandatory, ) return merged_graph, super_component_mapping def _is_pipeline_possibly_blocked(self, current_pipeline_outputs: dict[str, Any]) -> bool: """ Heuristically determines whether the pipeline is possibly blocked based on its current outputs. This method checks if the pipeline has produced any of the expected outputs. - If no outputs are expected (i.e., `self.outputs()` returns an empty list), the method assumes the pipeline is not blocked. - If at least one expected output is present in `current_pipeline_outputs`, the pipeline is also assumed to not be blocked. - If none of the expected outputs are present, the pipeline is considered to be possibly blocked. Note: This check is not definitive—it is intended as a best-effort guess to detect a stalled or misconfigured pipeline when there are no more runnable components. :param current_pipeline_outputs: A dictionary of outputs currently produced by the pipeline. :returns: bool: True if the pipeline is possibly blocked (i.e., expected outputs are missing), False otherwise. """ expected_outputs = self.outputs() return bool(expected_outputs) and not any(k in current_pipeline_outputs for k in expected_outputs) def _connections_status( sender_node: str, receiver_node: str, sender_sockets: list[OutputSocket], receiver_sockets: list[InputSocket] ) -> str: """ Lists the status of the sockets, for error messages. """ sender_sockets_entries = [] for sender_socket in sender_sockets: sender_sockets_entries.append(f" - {sender_socket.name}: {_type_name(sender_socket.type)}") sender_sockets_list = "\n".join(sender_sockets_entries) receiver_sockets_entries = [] for receiver_socket in receiver_sockets: if receiver_socket.senders: sender_status = f"sent by {','.join(receiver_socket.senders)}" else: sender_status = "available" receiver_sockets_entries.append( f" - {receiver_socket.name}: {_type_name(receiver_socket.type)} ({sender_status})" ) receiver_sockets_list = "\n".join(receiver_sockets_entries) return f"'{sender_node}':\n{sender_sockets_list}\n'{receiver_node}':\n{receiver_sockets_list}" # Utility functions def _validate_component_output_keys( component_name: str, comp: dict[str, Any], component_output: Mapping[str, Any] ) -> None: """ Validate that the output keys returned by a component match its declared output types. Logs a warning for any actually returned output key(s) that was not declared as an output socket(s). This helps catch bugs where a component returns wrong keys, which would otherwise cause downstream components to wait forever for expected data, resulting in a confusing "Pipeline Blocked" error that points to an unexpected component. :param component_name: Name of the Component as registered in the Pipeline. :param comp: The component metadata dictionary containing the component instance and its input/output socket metadata. :param component_output: The actual output dictionary returned by the component's run method. """ output_sockets = comp.get("output_sockets", {}) if not output_sockets: return declared_keys = set(output_sockets.keys()) actual_keys = set(component_output.keys()) extra_keys = actual_keys - declared_keys instance = comp["instance"] component_type = instance.__class__.__name__ if extra_keys: logger.warning( "Component '{component_name}' (type: {component_type}) returned output keys {extra_keys} " "that are not declared in its output types. " "These keys will be ignored and not passed to downstream components. " "Make sure the component's output keys match its declared @component.output_types.", component_name=component_name, component_type=component_type, extra_keys=extra_keys, ) def _write_to_lazy_variadic_socket( inputs: InputsType, receiver_name: str, receiver_socket_name: str, component_name: str, value: Any ) -> None: """ Write to a lazy variadic socket. Mutates inputs in place. :param inputs: The global inputs state to be mutated. :param receiver_name: The name of the component receiving the input. :param receiver_socket_name: The name of the socket receiving the input. :param component_name: The name of the component sending the input. :param value: The value to be sent to the socket. """ if not inputs[receiver_name].get(receiver_socket_name): inputs[receiver_name][receiver_socket_name] = [] inputs[receiver_name][receiver_socket_name].append({"sender": component_name, "value": value}) def _write_to_standard_socket( inputs: InputsType, receiver_name: str, receiver_socket_name: str, component_name: str, value: Any ) -> None: """ Write to a greedy variadic or non-variadic socket. Mutates inputs in place. :param inputs: The global inputs state to be mutated. :param receiver_name: The name of the component receiving the input. :param receiver_socket_name: The name of the socket receiving the input. :param component_name: The name of the component sending the input. :param value: The value to be sent to the socket. """ current_value = inputs[receiver_name].get(receiver_socket_name) # Only overwrite if there's no existing value, or we have a new value to provide if current_value is None or value is not _NO_OUTPUT_PRODUCED: inputs[receiver_name][receiver_socket_name] = [{"sender": component_name, "value": value}]