agno 2.2.13__py3-none-any.whl

agno/tools/function.py

@@ -0,0 +1,1187 @@
+from dataclasses import dataclass
+from functools import partial
+from importlib.metadata import version
+from typing import Any, Callable, Dict, List, Literal, Optional, Sequence, Type, TypeVar, get_type_hints
+
+from docstring_parser import parse
+from packaging.version import Version
+from pydantic import BaseModel, Field, validate_call
+
+from agno.exceptions import AgentRunException
+from agno.media import Audio, File, Image, Video
+from agno.run import RunContext
+from agno.utils.log import log_debug, log_error, log_exception, log_warning
+
+T = TypeVar("T")
+
+
+def get_entrypoint_docstring(entrypoint: Callable) -> str:
+    from inspect import getdoc
+
+    if isinstance(entrypoint, partial):
+        return str(entrypoint)
+
+    docstring = getdoc(entrypoint)
+    if not docstring:
+        return ""
+
+    parsed_doc = parse(docstring)
+
+    # Combine short and long descriptions
+    lines = []
+    if parsed_doc.short_description:
+        lines.append(parsed_doc.short_description)
+    if parsed_doc.long_description:
+        lines.extend(parsed_doc.long_description.split("\n"))
+
+    return "\n".join(lines)
+
+
+@dataclass
+class UserInputField:
+    name: str
+    field_type: Type
+    description: Optional[str] = None
+    value: Optional[Any] = None
+
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "name": self.name,
+            "field_type": str(self.field_type.__name__),
+            "description": self.description,
+            "value": self.value,
+        }
+
+    @classmethod
+    def from_dict(cls, data: Dict[str, Any]) -> "UserInputField":
+        return cls(
+            name=data["name"],
+            field_type=eval(data["field_type"]),  # Convert string type name to actual type
+            description=data["description"],
+            value=data["value"],
+        )
+
+
+class Function(BaseModel):
+    """Model for storing functions that can be called by an agent."""
+
+    # The name of the function to be called.
+    # Must be a-z, A-Z, 0-9, or contain underscores and dashes, with a maximum length of 64.
+    name: str
+    # A description of what the function does, used by the model to choose when and how to call the function.
+    description: Optional[str] = None
+    # The parameters the functions accepts, described as a JSON Schema object.
+    # To describe a function that accepts no parameters, provide the value {"type": "object", "properties": {}}.
+    parameters: Dict[str, Any] = Field(
+        default_factory=lambda: {"type": "object", "properties": {}, "required": []},
+        description="JSON Schema object describing function parameters",
+    )
+    strict: Optional[bool] = None
+
+    instructions: Optional[str] = None
+    # If True, add instructions to the Agent's system message
+    add_instructions: bool = True
+
+    # The function to be called.
+    entrypoint: Optional[Callable] = None
+    # If True, the entrypoint processing is skipped and the Function is used as is.
+    skip_entrypoint_processing: bool = False
+    # If True, the function call will show the result along with sending it to the model.
+    show_result: bool = False
+    # If True, the agent will stop after the function call.
+    stop_after_tool_call: bool = False
+    # Hook that runs before the function is executed.
+    # If defined, can accept the FunctionCall instance as a parameter.
+    pre_hook: Optional[Callable] = None
+    # Hook that runs after the function is executed, regardless of success/failure.
+    # If defined, can accept the FunctionCall instance as a parameter.
+    post_hook: Optional[Callable] = None
+
+    # A list of hooks to run around tool calls.
+    tool_hooks: Optional[List[Callable]] = None
+
+    # If True, the function will require confirmation before execution
+    requires_confirmation: Optional[bool] = None
+
+    # If True, the function will require user input before execution
+    requires_user_input: Optional[bool] = None
+    # List of fields that the user will provide as input and that should be ignored by the agent (empty list means all fields are provided by the user)
+    user_input_fields: Optional[List[str]] = None
+    # This is set during parsing, not by the user
+    user_input_schema: Optional[List[UserInputField]] = None
+
+    # If True, the function will be executed outside the agent's control.
+    external_execution: Optional[bool] = None
+
+    # Caching configuration
+    cache_results: bool = False
+    cache_dir: Optional[str] = None
+    cache_ttl: int = 3600
+
+    # --*-- FOR INTERNAL USE ONLY --*--
+    # The agent that the function is associated with
+    _agent: Optional[Any] = None
+    # The team that the function is associated with
+    _team: Optional[Any] = None
+    # The run context that the function is associated with
+    _run_context: Optional[RunContext] = None
+    # The session state that the function is associated with
+    _session_state: Optional[Dict[str, Any]] = None
+    # The dependencies that the function is associated with
+    _dependencies: Optional[Dict[str, Any]] = None
+
+    # Media context that the function is associated with
+    _images: Optional[Sequence[Image]] = None
+    _videos: Optional[Sequence[Video]] = None
+    _audios: Optional[Sequence[Audio]] = None
+    _files: Optional[Sequence[File]] = None
+
+    def to_dict(self) -> Dict[str, Any]:
+        return self.model_dump(
+            exclude_none=True,
+            include={"name", "description", "parameters", "strict", "requires_confirmation", "external_execution"},
+        )
+
+    def model_copy(self, *, deep: bool = False) -> "Function":
+        """
+        Override model_copy to handle callable fields that can't be deep copied (pickled).
+        Callables should always be shallow copied (referenced), not deep copied.
+        """
+        # For deep copy, we need to handle callable fields specially
+        if deep:
+            # Fields that should NOT be deep copied (callables and complex objects)
+            shallow_fields = {
+                "entrypoint",
+                "pre_hook",
+                "post_hook",
+                "tool_hooks",
+                "_agent",
+                "_team",
+            }
+
+            # Create a copy with shallow references to callable fields
+            copied_data = {}
+            for field_name, field_value in self.__dict__.items():
+                if field_name in shallow_fields:
+                    # Shallow copy - just reference the same object
+                    copied_data[field_name] = field_value
+                elif field_name == "parameters":
+                    # Deep copy the parameters dict
+                    from copy import deepcopy
+
+                    copied_data[field_name] = deepcopy(field_value)
+                else:
+                    # For simple types, just copy the value
+                    copied_data[field_name] = field_value
+
+            # Create new instance with copied data
+            new_instance = self.__class__.model_construct(**copied_data)
+
+            return new_instance
+        else:
+            # For shallow copy, use the default Pydantic behavior
+            return super().model_copy(deep=False)
+
+    @classmethod
+    def from_callable(cls, c: Callable, name: Optional[str] = None, strict: bool = False) -> "Function":
+        from inspect import getdoc, signature
+
+        from agno.utils.json_schema import get_json_schema
+
+        function_name = name or c.__name__
+        parameters = {"type": "object", "properties": {}, "required": []}
+        try:
+            sig = signature(c)
+            type_hints = get_type_hints(c)
+
+            # If function has an the agent argument, remove the agent parameter from the type hints
+            if "agent" in sig.parameters and "agent" in type_hints:
+                del type_hints["agent"]
+            if "team" in sig.parameters and "team" in type_hints:
+                del type_hints["team"]
+            if "run_context" in sig.parameters and "run_context" in type_hints:
+                del type_hints["run_context"]
+            if "session_state" in sig.parameters and "session_state" in type_hints:
+                del type_hints["session_state"]
+            if "dependencies" in sig.parameters and "dependencies" in type_hints:
+                del type_hints["dependencies"]
+
+            # Remove media parameters from type hints as they are injected automatically
+            if "images" in sig.parameters and "images" in type_hints:
+                del type_hints["images"]
+            if "videos" in sig.parameters and "videos" in type_hints:
+                del type_hints["videos"]
+            if "audios" in sig.parameters and "audios" in type_hints:
+                del type_hints["audios"]
+            if "files" in sig.parameters and "files" in type_hints:
+                del type_hints["files"]
+            # log_info(f"Type hints for {function_name}: {type_hints}")
+
+            # Filter out return type and only process parameters
+            param_type_hints = {
+                name: type_hints.get(name)
+                for name in sig.parameters
+                if name != "return"
+                and name
+                not in [
+                    "agent",
+                    "team",
+                    "run_context",
+                    "session_state",
+                    "dependencies",
+                    "self",
+                    "images",
+                    "videos",
+                    "audios",
+                    "files",
+                ]
+            }
+
+            # Parse docstring for parameters
+            param_descriptions: Dict[str, Any] = {}
+            if docstring := getdoc(c):
+                parsed_doc = parse(docstring)
+                param_docs = parsed_doc.params
+
+                if param_docs is not None:
+                    for param in param_docs:
+                        param_name = param.arg_name
+                        param_type = param.type_name
+                        if param_type is None:
+                            param_descriptions[param_name] = param.description
+                        else:
+                            param_descriptions[param_name] = f"({param_type}) {param.description}"
+
+            # Get JSON schema for parameters only
+            parameters = get_json_schema(
+                type_hints=param_type_hints, param_descriptions=param_descriptions, strict=strict
+            )
+
+            # If strict=True mark all fields as required
+            # See: https://platform.openai.com/docs/guides/structured-outputs/supported-schemas#all-fields-must-be-required
+            if strict:
+                parameters["required"] = [
+                    name
+                    for name in parameters["properties"]
+                    if name
+                    not in [
+                        "agent",
+                        "team",
+                        "run_context",
+                        "session_state",
+                        "dependencies",
+                        "self",
+                        "images",
+                        "videos",
+                        "audios",
+                        "files",
+                    ]
+                ]
+            else:
+                # Mark a field as required if it has no default value (this would include optional fields)
+                parameters["required"] = [
+                    name
+                    for name, param in sig.parameters.items()
+                    if param.default == param.empty
+                    and name
+                    not in [
+                        "agent",
+                        "team",
+                        "run_context",
+                        "session_state",
+                        "dependencies",
+                        "self",
+                        "images",
+                        "videos",
+                        "audios",
+                        "files",
+                    ]
+                ]
+
+            # log_debug(f"JSON schema for {function_name}: {parameters}")
+        except Exception as e:
+            log_warning(f"Could not parse args for {function_name}: {e}", exc_info=True)
+
+        entrypoint = cls._wrap_callable(c)
+
+        return cls(
+            name=function_name,
+            description=get_entrypoint_docstring(entrypoint=c),
+            parameters=parameters,
+            entrypoint=entrypoint,
+        )
+
+    def process_entrypoint(self, strict: bool = False):
+        """Process the entrypoint and make it ready for use by an agent."""
+        from inspect import getdoc, signature
+
+        from agno.utils.json_schema import get_json_schema
+
+        if self.skip_entrypoint_processing:
+            if strict:
+                self.process_schema_for_strict()
+            return
+
+        if self.entrypoint is None:
+            return
+
+        parameters = {"type": "object", "properties": {}, "required": []}
+
+        params_set_by_user = False
+        # If the user set the parameters (i.e. they are different from the default), we should keep them
+        if self.parameters != parameters:
+            params_set_by_user = True
+
+        if self.requires_user_input:
+            self.user_input_schema = self.user_input_schema or []
+
+        try:
+            sig = signature(self.entrypoint)
+            type_hints = get_type_hints(self.entrypoint)
+
+            # If function has an the agent argument, remove the agent parameter from the type hints
+            if "agent" in sig.parameters and "agent" in type_hints:
+                del type_hints["agent"]
+            if "team" in sig.parameters and "team" in type_hints:
+                del type_hints["team"]
+            if "run_context" in sig.parameters and "run_context" in type_hints:
+                del type_hints["run_context"]
+            if "session_state" in sig.parameters and "session_state" in type_hints:
+                del type_hints["session_state"]
+            if "dependencies" in sig.parameters and "dependencies" in type_hints:
+                del type_hints["dependencies"]
+            if "images" in sig.parameters and "images" in type_hints:
+                del type_hints["images"]
+            if "videos" in sig.parameters and "videos" in type_hints:
+                del type_hints["videos"]
+            if "audios" in sig.parameters and "audios" in type_hints:
+                del type_hints["audios"]
+            if "files" in sig.parameters and "files" in type_hints:
+                del type_hints["files"]
+            # log_info(f"Type hints for {self.name}: {type_hints}")
+
+            # Filter out return type and only process parameters
+            excluded_params = [
+                "return",
+                "agent",
+                "team",
+                "run_context",
+                "session_state",
+                "dependencies",
+                "self",
+                "images",
+                "videos",
+                "audios",
+                "files",
+            ]
+            if self.requires_user_input and self.user_input_fields:
+                if len(self.user_input_fields) == 0:
+                    excluded_params.extend(list(type_hints.keys()))
+                else:
+                    excluded_params.extend(self.user_input_fields)
+
+            # Get filtered list of parameter types
+            param_type_hints = {name: type_hints.get(name) for name in sig.parameters if name not in excluded_params}
+
+            # Parse docstring for parameters
+            param_descriptions = {}
+            param_descriptions_clean = {}
+            if docstring := getdoc(self.entrypoint):
+                parsed_doc = parse(docstring)
+                param_docs = parsed_doc.params
+
+                if param_docs is not None:
+                    for param in param_docs:
+                        param_name = param.arg_name
+                        param_type = param.type_name
+
+                        # TODO: We should use type hints first, then map param types in docs to json schema types.
+                        # This is temporary to not lose information
+                        param_descriptions[param_name] = f"({param_type}) {param.description}"
+                        param_descriptions_clean[param_name] = param.description
+
+            # If the function requires user input, we should set the user_input_schema to all parameters. The arguments provided by the model are filled in later.
+            if self.requires_user_input:
+                self.user_input_schema = [
+                    UserInputField(
+                        name=name,
+                        description=param_descriptions_clean.get(name),
+                        field_type=type_hints.get(name, str),
+                    )
+                    for name in sig.parameters
+                ]
+
+            # Get JSON schema for parameters only
+            parameters = get_json_schema(
+                type_hints=param_type_hints, param_descriptions=param_descriptions, strict=strict
+            )
+
+            # If strict=True mark all fields as required
+            # See: https://platform.openai.com/docs/guides/structured-outputs/supported-schemas#all-fields-must-be-required
+            if strict:
+                parameters["required"] = [name for name in parameters["properties"] if name not in excluded_params]
+            else:
+                # Mark a field as required if it has no default value
+                parameters["required"] = [
+                    name
+                    for name, param in sig.parameters.items()
+                    if param.default == param.empty and name != "self" and name not in excluded_params
+                ]
+
+            if params_set_by_user:
+                self.parameters["additionalProperties"] = False
+                if strict:
+                    self.parameters["required"] = [
+                        name for name in self.parameters["properties"] if name not in excluded_params
+                    ]
+                else:
+                    # Mark a field as required if it has no default value
+                    self.parameters["required"] = [
+                        name
+                        for name, param in sig.parameters.items()
+                        if param.default == param.empty and name != "self" and name not in excluded_params
+                    ]
+
+            self.description = self.description or get_entrypoint_docstring(self.entrypoint)
+
+            # log_debug(f"JSON schema for {self.name}: {parameters}")
+        except Exception as e:
+            log_warning(f"Could not parse args for {self.name}: {e}", exc_info=True)
+
+        if not params_set_by_user:
+            self.parameters = parameters
+
+        if strict:
+            self.process_schema_for_strict()
+
+        try:
+            self.entrypoint = self._wrap_callable(self.entrypoint)
+        except Exception as e:
+            log_warning(f"Failed to add validate decorator to entrypoint: {e}")
+
+    @staticmethod
+    def _wrap_callable(func: Callable) -> Callable:
+        """Wrap a callable with Pydantic's validate_call decorator, if relevant"""
+        from inspect import isasyncgenfunction, iscoroutinefunction, signature
+
+        pydantic_version = Version(version("pydantic"))
+
+        # Don't wrap async generators validate_call
+        if isasyncgenfunction(func):
+            return func
+
+        # Don't wrap coroutines with validate_call if pydantic version is less than 2.10.0
+        if iscoroutinefunction(func) and pydantic_version < Version("2.10.0"):
+            log_debug(
+                f"Skipping validate_call for {func.__name__} because pydantic version is less than 2.10.0, please consider upgrading to pydantic 2.10.0 or higher"
+            )
+            return func
+
+        # Don't wrap callables that are already wrapped with validate_call
+        elif getattr(func, "_wrapped_for_validation", False):
+            return func
+        # Don't wrap functions with session_state parameter
+        # session_state needs to be passed by reference, not copied by pydantic's validation
+        elif "session_state" in signature(func).parameters:
+            return func
+        # Wrap the callable with validate_call
+        else:
+            wrapped = validate_call(func, config=dict(arbitrary_types_allowed=True))  # type: ignore
+            wrapped._wrapped_for_validation = True  # Mark as wrapped to avoid infinite recursion
+            return wrapped
+
+    def process_schema_for_strict(self):
+        """Process the schema to make it strict mode compliant."""
+
+        def make_nested_strict(schema):
+            """Recursively ensure all object schemas have additionalProperties: false"""
+            if not isinstance(schema, dict):
+                return schema
+
+            # Make a copy to avoid modifying the original
+            result = schema.copy()
+
+            # If this is an object schema, ensure additionalProperties: false
+            if result.get("type") == "object" or "properties" in result:
+                result["additionalProperties"] = False
+
+            # If schema has no type but has other schema properties, give it a type
+            if "type" not in result:
+                if "properties" in result:
+                    result["type"] = "object"
+                    result["additionalProperties"] = False
+                elif result.get("title") and not any(
+                    key in result for key in ["properties", "items", "anyOf", "oneOf", "allOf", "enum"]
+                ):
+                    result["type"] = "string"
+
+            # Recursively process nested schemas
+            for key, value in result.items():
+                if key == "properties" and isinstance(value, dict):
+                    result[key] = {k: make_nested_strict(v) for k, v in value.items()}
+                elif key == "items" and isinstance(value, dict):
+                    # This handles array items like List[KnowledgeFilter]
+                    result[key] = make_nested_strict(value)
+                elif isinstance(value, dict):
+                    result[key] = make_nested_strict(value)
+
+            return result
+
+        # Apply strict mode to the entire schema
+        self.parameters = make_nested_strict(self.parameters)
+
+        self.parameters["required"] = [
+            name
+            for name in self.parameters["properties"]
+            if name
+            not in [
+                "agent",
+                "team",
+                "run_context",
+                "session_state",
+                "dependencies",
+                "images",
+                "videos",
+                "audios",
+                "files",
+                "self",
+            ]
+        ]
+
+    def _get_cache_key(self, entrypoint_args: Dict[str, Any], call_args: Optional[Dict[str, Any]] = None) -> str:
+        """Generate a cache key based on function name and arguments."""
+        import json
+        from hashlib import md5
+
+        copy_entrypoint_args = entrypoint_args.copy()
+        # Remove agent from entrypoint_args
+        if "agent" in copy_entrypoint_args:
+            del copy_entrypoint_args["agent"]
+        if "team" in copy_entrypoint_args:
+            del copy_entrypoint_args["team"]
+        if "run_context" in copy_entrypoint_args:
+            del copy_entrypoint_args["run_context"]
+        if "session_state" in copy_entrypoint_args:
+            del copy_entrypoint_args["session_state"]
+        if "dependencies" in copy_entrypoint_args:
+            del copy_entrypoint_args["dependencies"]
+        if "images" in copy_entrypoint_args:
+            del copy_entrypoint_args["images"]
+        if "videos" in copy_entrypoint_args:
+            del copy_entrypoint_args["videos"]
+        if "audios" in copy_entrypoint_args:
+            del copy_entrypoint_args["audios"]
+        if "files" in copy_entrypoint_args:
+            del copy_entrypoint_args["files"]
+        # Use json.dumps with sort_keys=True to ensure consistent ordering regardless of dict key order
+        args_str = json.dumps(copy_entrypoint_args, sort_keys=True, default=str)
+
+        kwargs_str = str(sorted((call_args or {}).items()))
+        key_str = f"{self.name}:{args_str}:{kwargs_str}"
+        return md5(key_str.encode()).hexdigest()
+
+    def _get_cache_file_path(self, cache_key: str) -> str:
+        """Get the full path for the cache file."""
+        from pathlib import Path
+        from tempfile import gettempdir
+
+        base_cache_dir = self.cache_dir or Path(gettempdir()) / "agno_cache"
+        func_cache_dir = Path(base_cache_dir) / "functions" / self.name
+        func_cache_dir.mkdir(parents=True, exist_ok=True)
+        return str(func_cache_dir / f"{cache_key}.json")
+
+    def _get_cached_result(self, cache_file: str) -> Optional[Any]:
+        """Retrieve cached result if valid."""
+        import json
+        from pathlib import Path
+        from time import time
+
+        cache_path = Path(cache_file)
+        if not cache_path.exists():
+            return None
+
+        try:
+            with cache_path.open("r") as f:
+                cache_data = json.load(f)
+
+            timestamp = cache_data.get("timestamp", 0)
+            result = cache_data.get("result")
+
+            if time() - timestamp <= self.cache_ttl:
+                return result
+
+            # Remove expired entry
+            cache_path.unlink()
+        except Exception as e:
+            log_error(f"Error reading cache: {e}")
+
+        return None
+
+    def _save_to_cache(self, cache_file: str, result: Any):
+        """Save result to cache."""
+        import json
+        from time import time
+
+        try:
+            with open(cache_file, "w") as f:
+                json.dump({"timestamp": time(), "result": result}, f)
+        except Exception as e:
+            log_error(f"Error writing cache: {e}")
+
+
+class FunctionExecutionResult(BaseModel):
+    status: Literal["success", "failure"]
+    result: Optional[Any] = None
+    error: Optional[str] = None
+
+    updated_session_state: Optional[Dict[str, Any]] = None
+
+    # New fields for media artifacts
+    images: Optional[List[Image]] = None
+    videos: Optional[List[Video]] = None
+    audios: Optional[List[Audio]] = None
+    files: Optional[List[File]] = None
+
+
+class FunctionCall(BaseModel):
+    """Model for Function Calls"""
+
+    # The function to be called.
+    function: Function
+    # The arguments to call the function with.
+    arguments: Optional[Dict[str, Any]] = None
+    # The result of the function call.
+    result: Optional[Any] = None
+    # The ID of the function call.
+    call_id: Optional[str] = None
+
+    # Error while parsing arguments or running the function.
+    error: Optional[str] = None
+
+    def get_call_str(self) -> str:
+        """Returns a string representation of the function call."""
+        import shutil
+
+        # Get terminal width, default to 80 if it can't be determined
+        term_width = shutil.get_terminal_size().columns or 80
+        max_arg_len = max(20, (term_width - len(self.function.name) - 4) // 2)
+
+        if self.arguments is None:
+            return f"{self.function.name}()"
+
+        trimmed_arguments = {}
+        for k, v in self.arguments.items():
+            if isinstance(v, str) and len(str(v)) > max_arg_len:
+                trimmed_arguments[k] = "..."
+            else:
+                trimmed_arguments[k] = v
+
+        call_str = f"{self.function.name}({', '.join([f'{k}={v}' for k, v in trimmed_arguments.items()])})"
+
+        # If call string is too long, truncate arguments
+        if len(call_str) > term_width:
+            return f"{self.function.name}(...)"
+
+        return call_str
+
+    def _handle_pre_hook(self):
+        """Handles the pre-hook for the function call."""
+        if self.function.pre_hook is not None:
+            try:
+                from inspect import signature
+
+                pre_hook_args = {}
+                # Check if the pre-hook has and agent argument
+                if "agent" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["agent"] = self.function._agent
+                # Check if the pre-hook has an team argument
+                if "team" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["team"] = self.function._team
+                # Check if the pre-hook has an session_state argument
+                if "run_context" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["run_context"] = self.function._run_context
+                # Check if the pre-hook has an session_state argument
+                if "session_state" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["session_state"] = self.function._session_state
+                # Check if the pre-hook has an dependencies argument
+                if "dependencies" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["dependencies"] = self.function._dependencies
+                # Check if the pre-hook has an fc argument
+                if "fc" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["fc"] = self
+                self.function.pre_hook(**pre_hook_args)
+            except AgentRunException as e:
+                log_debug(f"{e.__class__.__name__}: {e}")
+                self.error = str(e)
+                raise
+            except Exception as e:
+                log_warning(f"Error in pre-hook callback: {e}")
+                log_exception(e)
+
+    def _handle_post_hook(self):
+        """Handles the post-hook for the function call."""
+        if self.function.post_hook is not None:
+            try:
+                from inspect import signature
+
+                post_hook_args = {}
+                # Check if the post-hook has and agent argument
+                if "agent" in signature(self.function.post_hook).parameters:
+                    post_hook_args["agent"] = self.function._agent
+                # Check if the post-hook has an team argument
+                if "team" in signature(self.function.post_hook).parameters:
+                    post_hook_args["team"] = self.function._team
+                # Check if the post-hook has an session_state argument
+                if "run_context" in signature(self.function.post_hook).parameters:
+                    post_hook_args["run_context"] = self.function._run_context
+                # Check if the post-hook has an session_state argument
+                if "session_state" in signature(self.function.post_hook).parameters:
+                    post_hook_args["session_state"] = self.function._session_state
+                # Check if the post-hook has an dependencies argument
+                if "dependencies" in signature(self.function.post_hook).parameters:
+                    post_hook_args["dependencies"] = self.function._dependencies
+                # Check if the post-hook has an fc argument
+                if "fc" in signature(self.function.post_hook).parameters:
+                    post_hook_args["fc"] = self
+                self.function.post_hook(**post_hook_args)
+            except AgentRunException as e:
+                log_debug(f"{e.__class__.__name__}: {e}")
+                self.error = str(e)
+                raise
+            except Exception as e:
+                log_warning(f"Error in post-hook callback: {e}")
+                log_exception(e)
+
+    def _build_entrypoint_args(self) -> Dict[str, Any]:
+        """Builds the arguments for the entrypoint."""
+        from inspect import signature
+
+        entrypoint_args = {}
+        # Check if the entrypoint has an agent argument
+        if "agent" in signature(self.function.entrypoint).parameters:  # type: ignore
+            entrypoint_args["agent"] = self.function._agent
+        # Check if the entrypoint has an team argument
+        if "team" in signature(self.function.entrypoint).parameters:  # type: ignore
+            entrypoint_args["team"] = self.function._team
+        # Check if the entrypoint has an run_context argument
+        if "run_context" in signature(self.function.entrypoint).parameters:  # type: ignore
+            entrypoint_args["run_context"] = self.function._run_context
+        # Check if the entrypoint has an session_state argument
+        if "session_state" in signature(self.function.entrypoint).parameters:  # type: ignore
+            entrypoint_args["session_state"] = self.function._session_state
+        # Check if the entrypoint has an dependencies argument
+        if "dependencies" in signature(self.function.entrypoint).parameters:  # type: ignore
+            entrypoint_args["dependencies"] = self.function._dependencies
+        # Check if the entrypoint has an fc argument
+        if "fc" in signature(self.function.entrypoint).parameters:  # type: ignore
+            entrypoint_args["fc"] = self
+
+        # Check if the entrypoint has media arguments
+        if "images" in signature(self.function.entrypoint).parameters:  # type: ignore
+            entrypoint_args["images"] = self.function._images
+        if "videos" in signature(self.function.entrypoint).parameters:  # type: ignore
+            entrypoint_args["videos"] = self.function._videos
+        if "audios" in signature(self.function.entrypoint).parameters:  # type: ignore
+            entrypoint_args["audios"] = self.function._audios
+        if "files" in signature(self.function.entrypoint).parameters:  # type: ignore
+            entrypoint_args["files"] = self.function._files
+        return entrypoint_args
+
+    def _build_hook_args(self, hook: Callable, name: str, func: Callable, args: Dict[str, Any]) -> Dict[str, Any]:
+        """Build the arguments for the hook."""
+        from inspect import signature
+
+        hook_args = {}
+        # Check if the hook has an agent argument
+        if "agent" in signature(hook).parameters:
+            hook_args["agent"] = self.function._agent
+        # Check if the hook has an team argument
+        if "team" in signature(hook).parameters:
+            hook_args["team"] = self.function._team
+        # Check if the hook has an run_context argument
+        if "run_context" in signature(hook).parameters:
+            hook_args["run_context"] = self.function._run_context
+        # Check if the hook has an session_state argument
+        if "session_state" in signature(hook).parameters:
+            hook_args["session_state"] = self.function._session_state
+        # Check if the hook has an dependencies argument
+        if "dependencies" in signature(hook).parameters:
+            hook_args["dependencies"] = self.function._dependencies
+        if "name" in signature(hook).parameters:
+            hook_args["name"] = name
+        if "function_name" in signature(hook).parameters:
+            hook_args["function_name"] = name
+        if "function" in signature(hook).parameters:
+            hook_args["function"] = func
+        if "func" in signature(hook).parameters:
+            hook_args["func"] = func
+        if "function_call" in signature(hook).parameters:
+            hook_args["function_call"] = func
+        if "args" in signature(hook).parameters:
+            hook_args["args"] = args
+        if "arguments" in signature(hook).parameters:
+            hook_args["arguments"] = args
+        return hook_args
+
+    def _build_nested_execution_chain(self, entrypoint_args: Dict[str, Any]):
+        """Build a nested chain of hook executions with the entrypoint at the center.
+
+        This creates a chain where each hook wraps the next one, with the function call
+        at the innermost level. Returns bubble back up through each hook.
+        """
+        from functools import reduce
+        from inspect import iscoroutinefunction
+
+        def execute_entrypoint(name, func, args):
+            """Execute the entrypoint function."""
+            arguments = entrypoint_args.copy()
+            if self.arguments is not None:
+                arguments.update(self.arguments)
+            return self.function.entrypoint(**arguments)  # type: ignore
+
+        # If no hooks, just return the entrypoint execution function
+        if not self.function.tool_hooks:
+            return execute_entrypoint
+
+        def create_hook_wrapper(inner_func, hook):
+            """Create a nested wrapper for the hook."""
+
+            def wrapper(name, func, args):
+                # Pass the inner function as next_func to the hook
+                # The hook will call next_func to continue the chain
+                def next_func(**kwargs):
+                    return inner_func(name, func, kwargs)
+
+                hook_args = self._build_hook_args(hook, name, next_func, args)
+
+                return hook(**hook_args)
+
+            return wrapper
+
+        # Remove coroutine hooks
+        final_hooks = []
+        for hook in self.function.tool_hooks:
+            if iscoroutinefunction(hook):
+                log_warning(f"Cannot use async hooks with sync function calls. Skipping hook: {hook.__name__}")
+            else:
+                final_hooks.append(hook)
+
+        # Build the chain from inside out - reverse the hooks to start from the innermost
+        hooks = list(reversed(final_hooks))
+        chain = reduce(create_hook_wrapper, hooks, execute_entrypoint)
+        return chain
+
+    def execute(self) -> FunctionExecutionResult:
+        """Runs the function call."""
+        from inspect import isgenerator, isgeneratorfunction
+
+        if self.function.entrypoint is None:
+            return FunctionExecutionResult(status="failure", error="Entrypoint is not set")
+
+        log_debug(f"Running: {self.get_call_str()}")
+
+        # Execute pre-hook if it exists
+        self._handle_pre_hook()
+
+        entrypoint_args = self._build_entrypoint_args()
+
+        # Check cache if enabled and not a generator function
+        if self.function.cache_results and not isgeneratorfunction(self.function.entrypoint):
+            cache_key = self.function._get_cache_key(entrypoint_args, self.arguments)
+            cache_file = self.function._get_cache_file_path(cache_key)
+            cached_result = self.function._get_cached_result(cache_file)
+
+            if cached_result is not None:
+                log_debug(f"Cache hit for: {self.get_call_str()}")
+                self.result = cached_result
+                return FunctionExecutionResult(status="success", result=cached_result)
+
+        # Execute function
+        execution_result = None
+        exception_to_raise = None
+
+        try:
+            # Build and execute the nested chain of hooks
+            if self.function.tool_hooks is not None:
+                execution_chain = self._build_nested_execution_chain(entrypoint_args=entrypoint_args)
+                result = execution_chain(self.function.name, self.function.entrypoint, self.arguments or {})
+            else:
+                result = self.function.entrypoint(**entrypoint_args, **self.arguments)  # type: ignore
+
+            updated_session_state = None
+            if entrypoint_args.get("run_context") is not None:
+                run_context = entrypoint_args.get("run_context")
+                updated_session_state = (
+                    run_context.session_state
+                    if run_context is not None and run_context.session_state is not None
+                    else None
+                )
+            else:
+                if self.function._session_state is not None:
+                    updated_session_state = self.function._session_state
+
+            # Handle generator case
+            if isgenerator(result):
+                self.result = result  # Store generator directly, can't cache
+            else:
+                self.result = result
+                # Only cache non-generator results
+                if self.function.cache_results:
+                    cache_key = self.function._get_cache_key(entrypoint_args, self.arguments)
+                    cache_file = self.function._get_cache_file_path(cache_key)
+                    self.function._save_to_cache(cache_file, self.result)
+
+            execution_result = FunctionExecutionResult(
+                status="success", result=self.result, updated_session_state=updated_session_state
+            )
+
+        except AgentRunException as e:
+            log_debug(f"{e.__class__.__name__}: {e}")
+            self.error = str(e)
+            exception_to_raise = e
+        except Exception as e:
+            log_warning(f"Could not run function {self.get_call_str()}")
+            log_exception(e)
+            self.error = str(e)
+            execution_result = FunctionExecutionResult(status="failure", error=str(e))
+
+        finally:
+            self._handle_post_hook()
+
+            if exception_to_raise is not None:
+                raise exception_to_raise
+
+            return execution_result  # type: ignore[return-value]
+
+    async def _handle_pre_hook_async(self):
+        """Handles the async pre-hook for the function call."""
+        if self.function.pre_hook is not None:
+            try:
+                from inspect import signature
+
+                pre_hook_args = {}
+                # Check if the pre-hook has an agent argument
+                if "agent" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["agent"] = self.function._agent
+                # Check if the pre-hook has an team argument
+                if "team" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["team"] = self.function._team
+                # Check if the pre-hook has an run_context argument
+                if "run_context" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["run_context"] = self.function._run_context
+                # Check if the pre-hook has an session_state argument
+                if "session_state" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["session_state"] = self.function._session_state
+                # Check if the pre-hook has an dependencies argument
+                if "dependencies" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["dependencies"] = self.function._dependencies
+                # Check if the pre-hook has an fc argument
+                if "fc" in signature(self.function.pre_hook).parameters:
+                    pre_hook_args["fc"] = self
+
+                await self.function.pre_hook(**pre_hook_args)
+            except AgentRunException as e:
+                log_debug(f"{e.__class__.__name__}: {e}")
+                self.error = str(e)
+                raise
+            except Exception as e:
+                log_warning(f"Error in pre-hook callback: {e}")
+                log_exception(e)
+
+    async def _handle_post_hook_async(self):
+        """Handles the async post-hook for the function call."""
+        if self.function.post_hook is not None:
+            try:
+                from inspect import signature
+
+                post_hook_args = {}
+                # Check if the post-hook has an agent argument
+                if "agent" in signature(self.function.post_hook).parameters:
+                    post_hook_args["agent"] = self.function._agent
+                # Check if the post-hook has an team argument
+                if "team" in signature(self.function.post_hook).parameters:
+                    post_hook_args["team"] = self.function._team
+                # Check if the post-hook has an run_context argument
+                if "run_context" in signature(self.function.post_hook).parameters:
+                    post_hook_args["run_context"] = self.function._run_context
+                # Check if the post-hook has an session_state argument
+                if "session_state" in signature(self.function.post_hook).parameters:
+                    post_hook_args["session_state"] = self.function._session_state
+                # Check if the post-hook has an dependencies argument
+                if "dependencies" in signature(self.function.post_hook).parameters:
+                    post_hook_args["dependencies"] = self.function._dependencies
+
+                # Check if the post-hook has an fc argument
+                if "fc" in signature(self.function.post_hook).parameters:
+                    post_hook_args["fc"] = self
+
+                await self.function.post_hook(**post_hook_args)
+            except AgentRunException as e:
+                log_debug(f"{e.__class__.__name__}: {e}")
+                self.error = str(e)
+                raise
+            except Exception as e:
+                log_warning(f"Error in post-hook callback: {e}")
+                log_exception(e)
+
+    async def _build_nested_execution_chain_async(self, entrypoint_args: Dict[str, Any]):
+        """Build a nested chain of async hook executions with the entrypoint at the center.
+
+        Similar to _build_nested_execution_chain but for async execution.
+        """
+        from functools import reduce
+        from inspect import isasyncgenfunction, iscoroutinefunction
+
+        async def execute_entrypoint_async(name, func, args):
+            """Execute the entrypoint function asynchronously."""
+            arguments = entrypoint_args.copy()
+            if self.arguments is not None:
+                arguments.update(self.arguments)
+
+            result = self.function.entrypoint(**arguments)  # type: ignore
+            if iscoroutinefunction(self.function.entrypoint) and not isasyncgenfunction(self.function.entrypoint):
+                result = await result
+            return result
+
+        def execute_entrypoint(name, func, args):
+            """Execute the entrypoint function synchronously."""
+            arguments = entrypoint_args.copy()
+            if self.arguments is not None:
+                arguments.update(self.arguments)
+            return self.function.entrypoint(**arguments)  # type: ignore
+
+        # If no hooks, just return the entrypoint execution function
+        if not self.function.tool_hooks:
+            return execute_entrypoint
+
+        def create_hook_wrapper(inner_func, hook):
+            """Create a nested wrapper for the hook."""
+
+            async def wrapper(name, func, args):
+                """Create a nested wrapper for the hook."""
+
+                # Pass the inner function as next_func to the hook
+                # The hook will call next_func to continue the chain
+                async def next_func(**kwargs):
+                    if iscoroutinefunction(inner_func):
+                        return await inner_func(name, func, kwargs)
+                    else:
+                        return inner_func(name, func, kwargs)
+
+                hook_args = self._build_hook_args(hook, name, next_func, args)
+
+                if iscoroutinefunction(hook):
+                    return await hook(**hook_args)
+                else:
+                    return hook(**hook_args)
+
+            return wrapper
+
+        # Build the chain from inside out - reverse the hooks to start from the innermost
+        hooks = list(reversed(self.function.tool_hooks))
+
+        # Handle async and sync entrypoints
+        if iscoroutinefunction(self.function.entrypoint):
+            chain = reduce(create_hook_wrapper, hooks, execute_entrypoint_async)
+        else:
+            chain = reduce(create_hook_wrapper, hooks, execute_entrypoint)
+        return chain
+
+    async def aexecute(self) -> FunctionExecutionResult:
+        """Runs the function call asynchronously."""
+        from inspect import isasyncgen, isasyncgenfunction, iscoroutinefunction, isgenerator, isgeneratorfunction
+
+        if self.function.entrypoint is None:
+            return FunctionExecutionResult(status="failure", error="Entrypoint is not set")
+
+        log_debug(f"Running: {self.get_call_str()}")
+
+        # Execute pre-hook if it exists
+        if iscoroutinefunction(self.function.pre_hook):
+            await self._handle_pre_hook_async()
+        else:
+            self._handle_pre_hook()
+
+        entrypoint_args = self._build_entrypoint_args()
+
+        # Check cache if enabled and not a generator function
+        if self.function.cache_results and not (
+            isasyncgenfunction(self.function.entrypoint) or isgeneratorfunction(self.function.entrypoint)
+        ):
+            cache_key = self.function._get_cache_key(entrypoint_args, self.arguments)
+            cache_file = self.function._get_cache_file_path(cache_key)
+            cached_result = self.function._get_cached_result(cache_file)
+            if cached_result is not None:
+                log_debug(f"Cache hit for: {self.get_call_str()}")
+                self.result = cached_result
+                return FunctionExecutionResult(status="success", result=cached_result)
+
+        # Execute function
+        execution_result = None
+        exception_to_raise = None
+
+        try:
+            # Build and execute the nested chain of hooks
+            if self.function.tool_hooks is not None:
+                execution_chain = await self._build_nested_execution_chain_async(entrypoint_args)
+                self.result = await execution_chain(self.function.name, self.function.entrypoint, self.arguments or {})
+            else:
+                if self.arguments is None or self.arguments == {}:
+                    result = self.function.entrypoint(**entrypoint_args)
+                else:
+                    result = self.function.entrypoint(**entrypoint_args, **self.arguments)
+
+                if isasyncgenfunction(self.function.entrypoint):
+                    self.result = result  # Store async generator directly
+                else:
+                    self.result = await result
+
+            # Only cache if not a generator
+            if self.function.cache_results and not (isgenerator(self.result) or isasyncgen(self.result)):
+                cache_key = self.function._get_cache_key(entrypoint_args, self.arguments)
+                cache_file = self.function._get_cache_file_path(cache_key)
+                self.function._save_to_cache(cache_file, self.result)
+
+            updated_session_state = None
+            if entrypoint_args.get("run_context") is not None:
+                run_context = entrypoint_args.get("run_context")
+                updated_session_state = (
+                    run_context.session_state
+                    if run_context is not None and run_context.session_state is not None
+                    else None
+                )
+
+            execution_result = FunctionExecutionResult(
+                status="success", result=self.result, updated_session_state=updated_session_state
+            )
+
+        except AgentRunException as e:
+            log_debug(f"{e.__class__.__name__}: {e}")
+            self.error = str(e)
+            exception_to_raise = e
+        except Exception as e:
+            log_warning(f"Could not run function {self.get_call_str()}")
+            log_exception(e)
+            self.error = str(e)
+            execution_result = FunctionExecutionResult(status="failure", error=str(e))
+
+        finally:
+            if iscoroutinefunction(self.function.post_hook):
+                await self._handle_post_hook_async()
+            else:
+                self._handle_post_hook()
+
+            if exception_to_raise is not None:
+                raise exception_to_raise
+
+            return execution_result  # type: ignore[return-value]
+
+
+class ToolResult(BaseModel):
+    """Result from a tool that can include media artifacts."""
+
+    content: str
+    images: Optional[List[Image]] = None
+    videos: Optional[List[Video]] = None
+    audios: Optional[List[Audio]] = None
+    files: Optional[List[File]] = None