Merge branch 'main' into joaomdmoura/multimodal-crew

src/crewai/flow/flow.py

@@ -80,10 +80,27 @@ def listen(condition):
     return decorator
 
 
-def router(method):
+def router(condition):
     def decorator(func):
         func.__is_router__ = True
-        func.__router_for__ = method.__name__
+        # Handle conditions like listen/start
+        if isinstance(condition, str):
+            func.__trigger_methods__ = [condition]
+            func.__condition_type__ = "OR"
+        elif (
+            isinstance(condition, dict)
+            and "type" in condition
+            and "methods" in condition
+        ):
+            func.__trigger_methods__ = condition["methods"]
+            func.__condition_type__ = condition["type"]
+        elif callable(condition) and hasattr(condition, "__name__"):
+            func.__trigger_methods__ = [condition.__name__]
+            func.__condition_type__ = "OR"
+        else:
+            raise ValueError(
+                "Condition must be a method, string, or a result of or_() or and_()"
+            )
         return func
 
     return decorator
@@ -123,8 +140,8 @@ class FlowMeta(type):
 
         start_methods = []
         listeners = {}
-        routers = {}
         router_paths = {}
+        routers = set()
 
         for attr_name, attr_value in dct.items():
             if hasattr(attr_value, "__is_start_method__"):
@@ -137,18 +154,11 @@ class FlowMeta(type):
                 methods = attr_value.__trigger_methods__
                 condition_type = getattr(attr_value, "__condition_type__", "OR")
                 listeners[attr_name] = (condition_type, methods)
-
-            elif hasattr(attr_value, "__is_router__"):
-                routers[attr_value.__router_for__] = attr_name
-                possible_returns = get_possible_return_constants(attr_value)
-                if possible_returns:
-                    router_paths[attr_name] = possible_returns
-
-                # Register router as a listener to its triggering method
-                trigger_method_name = attr_value.__router_for__
-                methods = [trigger_method_name]
-                condition_type = "OR"
-                listeners[attr_name] = (condition_type, methods)
+                if hasattr(attr_value, "__is_router__") and attr_value.__is_router__:
+                    routers.add(attr_name)
+                    possible_returns = get_possible_return_constants(attr_value)
+                    if possible_returns:
+                        router_paths[attr_name] = possible_returns
 
         setattr(cls, "_start_methods", start_methods)
         setattr(cls, "_listeners", listeners)
@@ -163,7 +173,7 @@ class Flow(Generic[T], metaclass=FlowMeta):
 
     _start_methods: List[str] = []
     _listeners: Dict[str, tuple[str, List[str]]] = {}
-    _routers: Dict[str, str] = {}
+    _routers: Set[str] = set()
     _router_paths: Dict[str, List[str]] = {}
     initial_state: Union[Type[T], T, None] = None
     event_emitter = Signal("event_emitter")
@@ -210,20 +220,10 @@ class Flow(Generic[T], metaclass=FlowMeta):
         return self._method_outputs
 
     def _initialize_state(self, inputs: Dict[str, Any]) -> None:
-        """
-        Initializes or updates the state with the provided inputs.
-
-        Args:
-            inputs: Dictionary of inputs to initialize or update the state.
-
-        Raises:
-            ValueError: If inputs do not match the structured state model.
-            TypeError: If state is neither a BaseModel instance nor a dictionary.
-        """
         if isinstance(self._state, BaseModel):
-            # Structured state management
+            # Structured state
             try:
-                # Define a function to create the dynamic class
+
                 def create_model_with_extra_forbid(
                     base_model: Type[BaseModel],
                 ) -> Type[BaseModel]:
@@ -233,34 +233,20 @@ class Flow(Generic[T], metaclass=FlowMeta):
 
                     return ModelWithExtraForbid
 
-                # Create the dynamic class
                 ModelWithExtraForbid = create_model_with_extra_forbid(
                     self._state.__class__
                 )
-
-                # Create a new instance using the combined state and inputs
                 self._state = cast(
                     T, ModelWithExtraForbid(**{**self._state.model_dump(), **inputs})
                 )
-
             except ValidationError as e:
                 raise ValueError(f"Invalid inputs for structured state: {e}") from e
         elif isinstance(self._state, dict):
-            # Unstructured state management
             self._state.update(inputs)
         else:
             raise TypeError("State must be a BaseModel instance or a dictionary.")
 
     def kickoff(self, inputs: Optional[Dict[str, Any]] = None) -> Any:
-        """
-        Starts the execution of the flow synchronously.
-
-        Args:
-            inputs: Optional dictionary of inputs to initialize or update the state.
-
-        Returns:
-            The final output from the flow execution.
-        """
         self.event_emitter.send(
             self,
             event=FlowStartedEvent(
@@ -274,15 +260,6 @@ class Flow(Generic[T], metaclass=FlowMeta):
         return asyncio.run(self.kickoff_async())
 
     async def kickoff_async(self, inputs: Optional[Dict[str, Any]] = None) -> Any:
-        """
-        Starts the execution of the flow asynchronously.
-
-        Args:
-            inputs: Optional dictionary of inputs to initialize or update the state.
-
-        Returns:
-            The final output from the flow execution.
-        """
         if not self._start_methods:
             raise ValueError("No start method defined")
 
@@ -290,16 +267,12 @@ class Flow(Generic[T], metaclass=FlowMeta):
             self.__class__.__name__, list(self._methods.keys())
         )
 
-        # Create tasks for all start methods
         tasks = [
             self._execute_start_method(start_method)
             for start_method in self._start_methods
         ]
-
-        # Run all start methods concurrently
         await asyncio.gather(*tasks)
 
-        # Determine the final output (from the last executed method)
         final_output = self._method_outputs[-1] if self._method_outputs else None
 
         self.event_emitter.send(
@@ -310,7 +283,6 @@ class Flow(Generic[T], metaclass=FlowMeta):
                 result=final_output,
             ),
         )
-
         return final_output
 
     async def _execute_start_method(self, start_method_name: str) -> None:
@@ -327,49 +299,68 @@ class Flow(Generic[T], metaclass=FlowMeta):
             if asyncio.iscoroutinefunction(method)
             else method(*args, **kwargs)
         )
-        self._method_outputs.append(result)  # Store the output
-
-        # Track method execution counts
+        self._method_outputs.append(result)
         self._method_execution_counts[method_name] = (
             self._method_execution_counts.get(method_name, 0) + 1
         )
-
         return result
 
     async def _execute_listeners(self, trigger_method: str, result: Any) -> None:
-        listener_tasks = []
-
-        if trigger_method in self._routers:
-            router_method = self._methods[self._routers[trigger_method]]
-            path = await self._execute_method(
-                self._routers[trigger_method], router_method
+        # First, handle routers repeatedly until no router triggers anymore
+        while True:
+            routers_triggered = self._find_triggered_methods(
+                trigger_method, router_only=True
             )
-            trigger_method = path
+            if not routers_triggered:
+                break
+            for router_name in routers_triggered:
+                await self._execute_single_listener(router_name, result)
+                # After executing router, the router's result is the path
+                # The last router executed sets the trigger_method
+                # The router result is the last element in self._method_outputs
+                trigger_method = self._method_outputs[-1]
 
+        # Now that no more routers are triggered by current trigger_method,
+        # execute normal listeners
+        listeners_triggered = self._find_triggered_methods(
+            trigger_method, router_only=False
+        )
+        if listeners_triggered:
+            tasks = [
+                self._execute_single_listener(listener_name, result)
+                for listener_name in listeners_triggered
+            ]
+            await asyncio.gather(*tasks)
+
+    def _find_triggered_methods(
+        self, trigger_method: str, router_only: bool
+    ) -> List[str]:
+        triggered = []
         for listener_name, (condition_type, methods) in self._listeners.items():
+            is_router = listener_name in self._routers
+
+            if router_only != is_router:
+                continue
+
             if condition_type == "OR":
+                # If the trigger_method matches any in methods, run this
                 if trigger_method in methods:
-                    # Schedule the listener without preventing re-execution
-                    listener_tasks.append(
-                        self._execute_single_listener(listener_name, result)
-                    )
+                    triggered.append(listener_name)
             elif condition_type == "AND":
                 # Initialize pending methods for this listener if not already done
                 if listener_name not in self._pending_and_listeners:
                     self._pending_and_listeners[listener_name] = set(methods)
                 # Remove the trigger method from pending methods
-                self._pending_and_listeners[listener_name].discard(trigger_method)
+                if trigger_method in self._pending_and_listeners[listener_name]:
+                    self._pending_and_listeners[listener_name].discard(trigger_method)
+
                 if not self._pending_and_listeners[listener_name]:
                     # All required methods have been executed
-                    listener_tasks.append(
-                        self._execute_single_listener(listener_name, result)
-                    )
+                    triggered.append(listener_name)
                     # Reset pending methods for this listener
                     self._pending_and_listeners.pop(listener_name, None)
 
-        # Run all listener tasks concurrently and wait for them to complete
-        if listener_tasks:
-            await asyncio.gather(*listener_tasks)
+        return triggered
 
     async def _execute_single_listener(self, listener_name: str, result: Any) -> None:
         try:
@@ -386,17 +377,13 @@ class Flow(Generic[T], metaclass=FlowMeta):
 
             sig = inspect.signature(method)
             params = list(sig.parameters.values())
-
-            # Exclude 'self' parameter
             method_params = [p for p in params if p.name != "self"]
 
             if method_params:
-                # If listener expects parameters, pass the result
                 listener_result = await self._execute_method(
                     listener_name, method, result
                 )
             else:
-                # If listener does not expect parameters, call without arguments
                 listener_result = await self._execute_method(listener_name, method)
 
             self.event_emitter.send(
@@ -408,8 +395,9 @@ class Flow(Generic[T], metaclass=FlowMeta):
                 ),
             )
 
-            # Execute listeners of this listener
+            # Execute listeners (and possibly routers) of this listener
             await self._execute_listeners(listener_name, listener_result)
+
         except Exception as e:
             print(
                 f"[Flow._execute_single_listener] Error in method {listener_name}: {e}"
@@ -422,5 +410,4 @@ class Flow(Generic[T], metaclass=FlowMeta):
         self._telemetry.flow_plotting_span(
             self.__class__.__name__, list(self._methods.keys())
         )
-
         plot_flow(self, filename)

src/crewai/flow/utils.py

@@ -31,16 +31,50 @@ def get_possible_return_constants(function):
         print(f"Source code:\n{source}")
         return None
 
-    return_values = []
+    return_values = set()
+    dict_definitions = {}
+
+    class DictionaryAssignmentVisitor(ast.NodeVisitor):
+        def visit_Assign(self, node):
+            # Check if this assignment is assigning a dictionary literal to a variable
+            if isinstance(node.value, ast.Dict) and len(node.targets) == 1:
+                target = node.targets[0]
+                if isinstance(target, ast.Name):
+                    var_name = target.id
+                    dict_values = []
+                    # Extract string values from the dictionary
+                    for val in node.value.values:
+                        if isinstance(val, ast.Constant) and isinstance(val.value, str):
+                            dict_values.append(val.value)
+                        # If non-string, skip or just ignore
+                    if dict_values:
+                        dict_definitions[var_name] = dict_values
+            self.generic_visit(node)
 
     class ReturnVisitor(ast.NodeVisitor):
         def visit_Return(self, node):
-            # Check if the return value is a constant (Python 3.8+)
-            if isinstance(node.value, ast.Constant):
-                return_values.append(node.value.value)
+            # Direct string return
+            if isinstance(node.value, ast.Constant) and isinstance(
+                node.value.value, str
+            ):
+                return_values.add(node.value.value)
+            # Dictionary-based return, like return paths[result]
+            elif isinstance(node.value, ast.Subscript):
+                # Check if we're subscripting a known dictionary variable
+                if isinstance(node.value.value, ast.Name):
+                    var_name = node.value.value.id
+                    if var_name in dict_definitions:
+                        # Add all possible dictionary values
+                        for v in dict_definitions[var_name]:
+                            return_values.add(v)
+            self.generic_visit(node)
 
+    # First pass: identify dictionary assignments
+    DictionaryAssignmentVisitor().visit(code_ast)
+    # Second pass: identify returns
     ReturnVisitor().visit(code_ast)
-    return return_values
+
+    return list(return_values) if return_values else None
 
 
 def calculate_node_levels(flow):
@@ -61,10 +95,7 @@ def calculate_node_levels(flow):
         current_level = levels[current]
         visited.add(current)
 
-        for listener_name, (
-            condition_type,
-            trigger_methods,
-        ) in flow._listeners.items():
+        for listener_name, (condition_type, trigger_methods) in flow._listeners.items():
             if condition_type == "OR":
                 if current in trigger_methods:
                     if (
@@ -89,7 +120,7 @@ def calculate_node_levels(flow):
                             queue.append(listener_name)
 
         # Handle router connections
-        if current in flow._routers.values():
+        if current in flow._routers:
             router_method_name = current
             paths = flow._router_paths.get(router_method_name, [])
             for path in paths:
@@ -105,6 +136,7 @@ def calculate_node_levels(flow):
                             levels[listener_name] = current_level + 1
                             if listener_name not in visited:
                                 queue.append(listener_name)
+
     return levels
 
 
@@ -142,7 +174,7 @@ def dfs_ancestors(node, ancestors, visited, flow):
             dfs_ancestors(listener_name, ancestors, visited, flow)
 
     # Handle router methods separately
-    if node in flow._routers.values():
+    if node in flow._routers:
         router_method_name = node
         paths = flow._router_paths.get(router_method_name, [])
         for path in paths:

src/crewai/flow/visualization_utils.py

@@ -94,12 +94,14 @@ def add_edges(net, flow, node_positions, colors):
     ancestors = build_ancestor_dict(flow)
     parent_children = build_parent_children_dict(flow)
 
+    # Edges for normal listeners
     for method_name in flow._listeners:
         condition_type, trigger_methods = flow._listeners[method_name]
         is_and_condition = condition_type == "AND"
 
         for trigger in trigger_methods:
-            if trigger in flow._methods or trigger in flow._routers.values():
+            # Check if nodes exist before adding edges
+            if trigger in node_positions and method_name in node_positions:
                 is_router_edge = any(
                     trigger in paths for paths in flow._router_paths.values()
                 )
@@ -135,7 +137,22 @@ def add_edges(net, flow, node_positions, colors):
                 }
 
                 net.add_edge(trigger, method_name, **edge_style)
+            else:
+                # Nodes not found in node_positions. Check if it's a known router outcome and a known method.
+                is_router_edge = any(
+                    trigger in paths for paths in flow._router_paths.values()
+                )
+                # Check if method_name is a known method
+                method_known = method_name in flow._methods
 
+                # If it's a known router edge and the method is known, don't warn.
+                # This means the path is legitimate, just not reflected as nodes here.
+                if not (is_router_edge and method_known):
+                    print(
+                        f"Warning: No node found for '{trigger}' or '{method_name}'. Skipping edge."
+                    )
+
+    # Edges for router return paths
     for router_method_name, paths in flow._router_paths.items():
         for path in paths:
             for listener_name, (
@@ -143,36 +160,49 @@ def add_edges(net, flow, node_positions, colors):
                 trigger_methods,
             ) in flow._listeners.items():
                 if path in trigger_methods:
-                    is_cycle_edge = is_ancestor(trigger, method_name, ancestors)
-                    parent_has_multiple_children = (
-                        len(parent_children.get(router_method_name, [])) > 1
-                    )
-                    needs_curvature = is_cycle_edge or parent_has_multiple_children
+                    if (
+                        router_method_name in node_positions
+                        and listener_name in node_positions
+                    ):
+                        is_cycle_edge = is_ancestor(
+                            router_method_name, listener_name, ancestors
+                        )
+                        parent_has_multiple_children = (
+                            len(parent_children.get(router_method_name, [])) > 1
+                        )
+                        needs_curvature = is_cycle_edge or parent_has_multiple_children
 
-                    if needs_curvature:
-                        source_pos = node_positions.get(router_method_name)
-                        target_pos = node_positions.get(listener_name)
+                        if needs_curvature:
+                            source_pos = node_positions.get(router_method_name)
+                            target_pos = node_positions.get(listener_name)
 
-                        if source_pos and target_pos:
-                            dx = target_pos[0] - source_pos[0]
-                            smooth_type = "curvedCCW" if dx <= 0 else "curvedCW"
-                            index = get_child_index(
-                                router_method_name, listener_name, parent_children
-                            )
-                            edge_smooth = {
-                                "type": smooth_type,
-                                "roundness": 0.2 + (0.1 * index),
-                            }
+                            if source_pos and target_pos:
+                                dx = target_pos[0] - source_pos[0]
+                                smooth_type = "curvedCCW" if dx <= 0 else "curvedCW"
+                                index = get_child_index(
+                                    router_method_name, listener_name, parent_children
+                                )
+                                edge_smooth = {
+                                    "type": smooth_type,
+                                    "roundness": 0.2 + (0.1 * index),
+                                }
+                            else:
+                                edge_smooth = {"type": "cubicBezier"}
                         else:
-                            edge_smooth = {"type": "cubicBezier"}
-                    else:
-                        edge_smooth = False
+                            edge_smooth = False
 
-                    edge_style = {
-                        "color": colors["router_edge"],
-                        "width": 2,
-                        "arrows": "to",
-                        "dashes": True,
-                        "smooth": edge_smooth,
-                    }
-                    net.add_edge(router_method_name, listener_name, **edge_style)
+                        edge_style = {
+                            "color": colors["router_edge"],
+                            "width": 2,
+                            "arrows": "to",
+                            "dashes": True,
+                            "smooth": edge_smooth,
+                        }
+                        net.add_edge(router_method_name, listener_name, **edge_style)
+                    else:
+                        # Same check here: known router edge and known method?
+                        method_known = listener_name in flow._methods
+                        if not method_known:
+                            print(
+                                f"Warning: No node found for '{router_method_name}' or '{listener_name}'. Skipping edge."
+                            )

src/crewai/knowledge/source/base_file_knowledge_source.py

@@ -1,8 +1,8 @@
 from abc import ABC, abstractmethod
 from pathlib import Path
-from typing import Dict, List, Union
+from typing import Dict, List, Optional, Union
 
-from pydantic import Field
+from pydantic import Field, field_validator
 
 from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
 from crewai.knowledge.storage.knowledge_storage import KnowledgeStorage
@@ -14,17 +14,28 @@ class BaseFileKnowledgeSource(BaseKnowledgeSource, ABC):
     """Base class for knowledge sources that load content from files."""
 
     _logger: Logger = Logger(verbose=True)
-    file_path: Union[Path, List[Path], str, List[str]] = Field(
-        ..., description="The path to the file"
+    file_path: Optional[Union[Path, List[Path], str, List[str]]] = Field(
+        default=None,
+        description="[Deprecated] The path to the file. Use file_paths instead.",
+    )
+    file_paths: Optional[Union[Path, List[Path], str, List[str]]] = Field(
+        default_factory=list, description="The path to the file"
     )
     content: Dict[Path, str] = Field(init=False, default_factory=dict)
     storage: KnowledgeStorage = Field(default_factory=KnowledgeStorage)
     safe_file_paths: List[Path] = Field(default_factory=list)
 
+    @field_validator("file_path", "file_paths", mode="before")
+    def validate_file_path(cls, v, values):
+        """Validate that at least one of file_path or file_paths is provided."""
+        if v is None and ("file_path" not in values or values.get("file_path") is None):
+            raise ValueError("Either file_path or file_paths must be provided")
+        return v
+
     def model_post_init(self, _):
         """Post-initialization method to load content."""
         self.safe_file_paths = self._process_file_paths()
-        self.validate_paths()
+        self.validate_content()
         self.content = self.load_content()
 
     @abstractmethod
@@ -32,7 +43,7 @@ class BaseFileKnowledgeSource(BaseKnowledgeSource, ABC):
         """Load and preprocess file content. Should be overridden by subclasses. Assume that the file path is relative to the project root in the knowledge directory."""
         pass
 
-    def validate_paths(self):
+    def validate_content(self):
         """Validate the paths."""
         for path in self.safe_file_paths:
             if not path.exists():
@@ -59,13 +70,30 @@ class BaseFileKnowledgeSource(BaseKnowledgeSource, ABC):
 
     def _process_file_paths(self) -> List[Path]:
         """Convert file_path to a list of Path objects."""
-        paths = (
-            [self.file_path]
-            if isinstance(self.file_path, (str, Path))
-            else self.file_path
+
+        if hasattr(self, "file_path") and self.file_path is not None:
+            self._logger.log(
+                "warning",
+                "The 'file_path' attribute is deprecated and will be removed in a future version. Please use 'file_paths' instead.",
+                color="yellow",
+            )
+            self.file_paths = self.file_path
+
+        if self.file_paths is None:
+            raise ValueError("Your source must be provided with a file_paths: []")
+
+        # Convert single path to list
+        path_list: List[Union[Path, str]] = (
+            [self.file_paths]
+            if isinstance(self.file_paths, (str, Path))
+            else list(self.file_paths)
+            if isinstance(self.file_paths, list)
+            else []
         )
 
-        if not isinstance(paths, list):
-            raise ValueError("file_path must be a Path, str, or a list of these types")
+        if not path_list:
+            raise ValueError(
+                "file_path/file_paths must be a Path, str, or a list of these types"
+            )
 
-        return [self.convert_to_path(path) for path in paths]
+        return [self.convert_to_path(path) for path in path_list]

src/crewai/knowledge/source/base_knowledge_source.py

@@ -21,7 +21,7 @@ class BaseKnowledgeSource(BaseModel, ABC):
     collection_name: Optional[str] = Field(default=None)
 
     @abstractmethod
-    def load_content(self) -> Dict[Any, str]:
+    def validate_content(self) -> Any:
         """Load and preprocess content from the source."""
         pass
 

src/crewai/knowledge/source/crew_docling_source.py

@@ -0,0 +1,120 @@
+from pathlib import Path
+from typing import Iterator, List, Optional, Union
+from urllib.parse import urlparse
+
+from docling.datamodel.base_models import InputFormat
+from docling.document_converter import DocumentConverter
+from docling.exceptions import ConversionError
+from docling_core.transforms.chunker.hierarchical_chunker import HierarchicalChunker
+from docling_core.types.doc.document import DoclingDocument
+from pydantic import Field
+
+from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
+from crewai.utilities.constants import KNOWLEDGE_DIRECTORY
+from crewai.utilities.logger import Logger
+
+
+class CrewDoclingSource(BaseKnowledgeSource):
+    """Default Source class for converting documents to markdown or json
+    This will auto support PDF, DOCX, and TXT, XLSX, Images, and HTML files without any additional dependencies and follows the docling package as the source of truth.
+    """
+
+    _logger: Logger = Logger(verbose=True)
+
+    file_path: Optional[List[Union[Path, str]]] = Field(default=None)
+    file_paths: List[Union[Path, str]] = Field(default_factory=list)
+    chunks: List[str] = Field(default_factory=list)
+    safe_file_paths: List[Union[Path, str]] = Field(default_factory=list)
+    content: List[DoclingDocument] = Field(default_factory=list)
+    document_converter: DocumentConverter = Field(
+        default_factory=lambda: DocumentConverter(
+            allowed_formats=[
+                InputFormat.MD,
+                InputFormat.ASCIIDOC,
+                InputFormat.PDF,
+                InputFormat.DOCX,
+                InputFormat.HTML,
+                InputFormat.IMAGE,
+                InputFormat.XLSX,
+                InputFormat.PPTX,
+            ]
+        )
+    )
+
+    def model_post_init(self, _) -> None:
+        if self.file_path:
+            self._logger.log(
+                "warning",
+                "The 'file_path' attribute is deprecated and will be removed in a future version. Please use 'file_paths' instead.",
+                color="yellow",
+            )
+            self.file_paths = self.file_path
+        self.safe_file_paths = self.validate_content()
+        self.content = self._load_content()
+
+    def _load_content(self) -> List[DoclingDocument]:
+        try:
+            return self._convert_source_to_docling_documents()
+        except ConversionError as e:
+            self._logger.log(
+                "error",
+                f"Error loading content: {e}. Supported formats: {self.document_converter.allowed_formats}",
+                "red",
+            )
+            raise e
+        except Exception as e:
+            self._logger.log("error", f"Error loading content: {e}")
+            raise e
+
+    def add(self) -> None:
+        if self.content is None:
+            return
+        for doc in self.content:
+            new_chunks_iterable = self._chunk_doc(doc)
+            self.chunks.extend(list(new_chunks_iterable))
+        self._save_documents()
+
+    def _convert_source_to_docling_documents(self) -> List[DoclingDocument]:
+        conv_results_iter = self.document_converter.convert_all(self.safe_file_paths)
+        return [result.document for result in conv_results_iter]
+
+    def _chunk_doc(self, doc: DoclingDocument) -> Iterator[str]:
+        chunker = HierarchicalChunker()
+        for chunk in chunker.chunk(doc):
+            yield chunk.text
+
+    def validate_content(self) -> List[Union[Path, str]]:
+        processed_paths: List[Union[Path, str]] = []
+        for path in self.file_paths:
+            if isinstance(path, str):
+                if path.startswith(("http://", "https://")):
+                    try:
+                        if self._validate_url(path):
+                            processed_paths.append(path)
+                        else:
+                            raise ValueError(f"Invalid URL format: {path}")
+                    except Exception as e:
+                        raise ValueError(f"Invalid URL: {path}. Error: {str(e)}")
+                else:
+                    local_path = Path(KNOWLEDGE_DIRECTORY + "/" + path)
+                    if local_path.exists():
+                        processed_paths.append(local_path)
+                    else:
+                        raise FileNotFoundError(f"File not found: {local_path}")
+            else:
+                # this is an instance of Path
+                processed_paths.append(path)
+        return processed_paths
+
+    def _validate_url(self, url: str) -> bool:
+        try:
+            result = urlparse(url)
+            return all(
+                [
+                    result.scheme in ("http", "https"),
+                    result.netloc,
+                    len(result.netloc.split(".")) >= 2,  # Ensure domain has TLD
+                ]
+            )
+        except Exception:
+            return False

src/crewai/knowledge/source/string_knowledge_source.py

@@ -13,9 +13,9 @@ class StringKnowledgeSource(BaseKnowledgeSource):
 
     def model_post_init(self, _):
         """Post-initialization method to validate content."""
-        self.load_content()
+        self.validate_content()
 
-    def load_content(self):
+    def validate_content(self):
         """Validate string content."""
         if not isinstance(self.content, str):
             raise ValueError("StringKnowledgeSource only accepts string content")

src/crewai/task.py

@@ -1,12 +1,25 @@
 import datetime
+import inspect
 import json
+import logging
 import threading
 import uuid
 from concurrent.futures import Future
 from copy import copy
 from hashlib import md5
 from pathlib import Path
-from typing import Any, Dict, List, Optional, Set, Tuple, Type, Union
+from typing import (
+    Any,
+    Callable,
+    ClassVar,
+    Dict,
+    List,
+    Optional,
+    Set,
+    Tuple,
+    Type,
+    Union,
+)
 
 from opentelemetry.trace import Span
 from pydantic import (
@@ -20,6 +33,7 @@ from pydantic import (
 from pydantic_core import PydanticCustomError
 
 from crewai.agents.agent_builder.base_agent import BaseAgent
+from crewai.tasks.guardrail_result import GuardrailResult
 from crewai.tasks.output_format import OutputFormat
 from crewai.tasks.task_output import TaskOutput
 from crewai.telemetry.telemetry import Telemetry
@@ -49,6 +63,7 @@ class Task(BaseModel):
     """
 
     __hash__ = object.__hash__  # type: ignore
+    logger: ClassVar[logging.Logger] = logging.getLogger(__name__)
     used_tools: int = 0
     tools_errors: int = 0
     delegations: int = 0
@@ -110,6 +125,55 @@ class Task(BaseModel):
         default=None,
     )
     processed_by_agents: Set[str] = Field(default_factory=set)
+    guardrail: Optional[Callable[[TaskOutput], Tuple[bool, Any]]] = Field(
+        default=None,
+        description="Function to validate task output before proceeding to next task"
+    )
+    max_retries: int = Field(
+        default=3,
+        description="Maximum number of retries when guardrail fails"
+    )
+    retry_count: int = Field(
+        default=0,
+        description="Current number of retries"
+    )
+
+    @field_validator("guardrail")
+    @classmethod
+    def validate_guardrail_function(cls, v: Optional[Callable]) -> Optional[Callable]:
+        """Validate that the guardrail function has the correct signature and behavior.
+        
+        While type hints provide static checking, this validator ensures runtime safety by:
+        1. Verifying the function accepts exactly one parameter (the TaskOutput)
+        2. Checking return type annotations match Tuple[bool, Any] if present
+        3. Providing clear, immediate error messages for debugging
+        
+        This runtime validation is crucial because:
+        - Type hints are optional and can be ignored at runtime
+        - Function signatures need immediate validation before task execution
+        - Clear error messages help users debug guardrail implementation issues
+        
+        Args:
+            v: The guardrail function to validate
+            
+        Returns:
+            The validated guardrail function
+            
+        Raises:
+            ValueError: If the function signature is invalid or return annotation
+                       doesn't match Tuple[bool, Any]
+        """
+        if v is not None:
+            sig = inspect.signature(v)
+            if len(sig.parameters) != 1:
+                raise ValueError("Guardrail function must accept exactly one parameter")
+
+            # Check return annotation if present, but don't require it
+            return_annotation = sig.return_annotation
+            if return_annotation != inspect.Signature.empty:
+                if not (return_annotation == Tuple[bool, Any] or str(return_annotation) == 'Tuple[bool, Any]'):
+                    raise ValueError("If return type is annotated, it must be Tuple[bool, Any]")
+        return v
 
     _telemetry: Telemetry = PrivateAttr(default_factory=Telemetry)
     _execution_span: Optional[Span] = PrivateAttr(default=None)
@@ -254,7 +318,6 @@ class Task(BaseModel):
         )
 
         pydantic_output, json_output = self._export_output(result)
-
         task_output = TaskOutput(
             name=self.name,
             description=self.description,
@@ -265,6 +328,37 @@ class Task(BaseModel):
             agent=agent.role,
             output_format=self._get_output_format(),
         )
+
+        if self.guardrail:
+            guardrail_result = GuardrailResult.from_tuple(self.guardrail(task_output))
+            if not guardrail_result.success:
+                if self.retry_count >= self.max_retries:
+                    raise Exception(
+                        f"Task failed guardrail validation after {self.max_retries} retries. "
+                        f"Last error: {guardrail_result.error}"
+                    )
+
+                self.retry_count += 1
+                context = (
+                    f"### Previous attempt failed validation: {guardrail_result.error}\n\n\n"
+                    f"### Previous result:\n{task_output.raw}\n\n\n"
+                    "Try again, making sure to address the validation error."
+                )
+                return self._execute_core(agent, context, tools)
+
+            if guardrail_result.result is None:
+                raise Exception(
+                    "Task guardrail returned None as result. This is not allowed."
+                )
+
+            if isinstance(guardrail_result.result, str):
+                task_output.raw = guardrail_result.result
+                pydantic_output, json_output = self._export_output(guardrail_result.result)
+                task_output.pydantic = pydantic_output
+                task_output.json_dict = json_output
+            elif isinstance(guardrail_result.result, TaskOutput):
+                task_output = guardrail_result.result
+
         self.output = task_output
 
         self._set_end_execution_time(start_time)
@@ -308,7 +402,18 @@ class Task(BaseModel):
 
         if inputs:
             self.description = self._original_description.format(**inputs)
-            self.expected_output = self._original_expected_output.format(**inputs)
+            self.expected_output = self.interpolate_only(
+                input_string=self._original_expected_output, inputs=inputs
+            )
+
+    def interpolate_only(self, input_string: str, inputs: Dict[str, Any]) -> str:
+        """Interpolate placeholders (e.g., {key}) in a string while leaving JSON untouched."""
+        escaped_string = input_string.replace("{", "{{").replace("}", "}}")
+
+        for key in inputs.keys():
+            escaped_string = escaped_string.replace(f"{{{{{key}}}}}", f"{{{key}}}")
+
+        return escaped_string.format(**inputs)
 
     def increment_tools_errors(self) -> None:
         """Increment the tools errors counter."""
@@ -390,22 +495,33 @@ class Task(BaseModel):
         return OutputFormat.RAW
 
     def _save_file(self, result: Any) -> None:
+        """Save task output to a file.
+        
+        Args:
+            result: The result to save to the file. Can be a dict or any stringifiable object.
+            
+        Raises:
+            ValueError: If output_file is not set
+            RuntimeError: If there is an error writing to the file
+        """
         if self.output_file is None:
             raise ValueError("output_file is not set.")
 
-        resolved_path = Path(self.output_file).expanduser().resolve()
-        directory = resolved_path.parent
+        try:
+            resolved_path = Path(self.output_file).expanduser().resolve()
+            directory = resolved_path.parent
 
-        if not directory.exists():
-            directory.mkdir(parents=True, exist_ok=True)
+            if not directory.exists():
+                directory.mkdir(parents=True, exist_ok=True)
 
-        with resolved_path.open("w", encoding="utf-8") as file:
-            if isinstance(result, dict):
-                import json
-
-                json.dump(result, file, ensure_ascii=False, indent=2)
-            else:
-                file.write(str(result))
+            with resolved_path.open("w", encoding="utf-8") as file:
+                if isinstance(result, dict):
+                    import json
+                    json.dump(result, file, ensure_ascii=False, indent=2)
+                else:
+                    file.write(str(result))
+        except (OSError, IOError) as e:
+            raise RuntimeError(f"Failed to save output file: {e}")
         return None
 
     def __repr__(self):

src/crewai/tasks/guardrail_result.py

@@ -0,0 +1,56 @@
+"""
+Module for handling task guardrail validation results.
+
+This module provides the GuardrailResult class which standardizes
+the way task guardrails return their validation results.
+"""
+
+from typing import Any, Optional, Tuple, Union
+
+from pydantic import BaseModel, field_validator
+
+
+class GuardrailResult(BaseModel):
+    """Result from a task guardrail execution.
+
+    This class standardizes the return format of task guardrails,
+    converting tuple responses into a structured format that can
+    be easily handled by the task execution system.
+
+    Attributes:
+        success (bool): Whether the guardrail validation passed
+        result (Any, optional): The validated/transformed result if successful
+        error (str, optional): Error message if validation failed
+    """
+    success: bool
+    result: Optional[Any] = None
+    error: Optional[str] = None
+
+    @field_validator("result", "error")
+    @classmethod
+    def validate_result_error_exclusivity(cls, v: Any, info) -> Any:
+        values = info.data
+        if "success" in values:
+            if values["success"] and v and "error" in values and values["error"]:
+                raise ValueError("Cannot have both result and error when success is True")
+            if not values["success"] and v and "result" in values and values["result"]:
+                raise ValueError("Cannot have both result and error when success is False")
+        return v
+
+    @classmethod
+    def from_tuple(cls, result: Tuple[bool, Union[Any, str]]) -> "GuardrailResult":
+        """Create a GuardrailResult from a validation tuple.
+
+        Args:
+            result: A tuple of (success, data) where data is either
+                   the validated result or error message.
+
+        Returns:
+            GuardrailResult: A new instance with the tuple data.
+        """
+        success, data = result
+        return cls(
+            success=success,
+            result=data if success else None,
+            error=data if not success else None
+        )