CoACT initialize (#292)

mm_agents/coact/autogen/agentchat/group/patterns/__init__.py

@@ -0,0 +1,18 @@
+# Copyright (c) 2023 - 2025, AG2ai, Inc., AG2ai open-source projects maintainers and core contributors
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+
+from .auto import AutoPattern
+from .manual import ManualPattern
+from .pattern import DefaultPattern
+from .random import RandomPattern
+from .round_robin import RoundRobinPattern
+
+__all__ = [
+    "AutoPattern",
+    "DefaultPattern",
+    "ManualPattern",
+    "RandomPattern",
+    "RoundRobinPattern",
+]

mm_agents/coact/autogen/agentchat/group/patterns/auto.py

@@ -0,0 +1,159 @@
+# Copyright (c) 2023 - 2025, AG2ai, Inc., AG2ai open-source projects maintainers and core contributors
+#
+# SPDX-License-Identifier: Apache-2.0
+
+from typing import TYPE_CHECKING, Any, Callable, Optional, Tuple, Union
+
+from ..context_variables import ContextVariables
+from ..targets.group_manager_target import GroupManagerSelectionMessage, GroupManagerTarget
+from ..targets.transition_target import TransitionTarget
+from .pattern import Pattern
+
+if TYPE_CHECKING:
+    from ...conversable_agent import ConversableAgent
+    from ...groupchat import GroupChat, GroupChatManager
+    from ..group_tool_executor import GroupToolExecutor
+
+
+class AutoPattern(Pattern):
+    """AutoPattern implements a flexible pattern where agents are selected based on their expertise.
+
+    In this pattern, a group manager automatically selects the next agent to speak based on the context
+    of the conversation and agent descriptions. The after_work is always set to "group_manager" as
+    this is the defining characteristic of this pattern.
+    """
+
+    def __init__(
+        self,
+        initial_agent: "ConversableAgent",
+        agents: list["ConversableAgent"],
+        user_agent: Optional["ConversableAgent"] = None,
+        group_manager_args: Optional[dict[str, Any]] = None,
+        context_variables: Optional[ContextVariables] = None,
+        selection_message: Optional[GroupManagerSelectionMessage] = None,
+        exclude_transit_message: bool = True,
+        summary_method: Optional[Union[str, Callable[..., Any]]] = "last_msg",
+    ):
+        """Initialize the AutoPattern.
+
+        The after_work is always set to group_manager selection, which is the defining
+        characteristic of this pattern. You can customize the selection message used
+        by the group manager when selecting the next agent.
+
+        Args:
+            initial_agent: The first agent to speak in the group chat.
+            agents: List of all agents participating in the chat.
+            user_agent: Optional user proxy agent.
+            group_manager_args: Optional arguments for the GroupChatManager.
+            context_variables: Initial context variables for the chat.
+            selection_message: Custom message to use when the group manager is selecting agents.
+            exclude_transit_message: Whether to exclude transit messages from the conversation.
+            summary_method: Method for summarizing the conversation.
+        """
+        # Create the group_manager after_work with the provided selection message
+        group_manager_after_work = GroupManagerTarget(selection_message=selection_message)
+
+        super().__init__(
+            initial_agent=initial_agent,
+            agents=agents,
+            user_agent=user_agent,
+            group_manager_args=group_manager_args,
+            context_variables=context_variables,
+            group_after_work=group_manager_after_work,
+            exclude_transit_message=exclude_transit_message,
+            summary_method=summary_method,
+        )
+
+        # Store the selection message for potential use
+        self.selection_message = selection_message
+
+    def prepare_group_chat(
+        self,
+        max_rounds: int,
+        messages: Union[list[dict[str, Any]], str],
+    ) -> Tuple[
+        list["ConversableAgent"],
+        list["ConversableAgent"],
+        Optional["ConversableAgent"],
+        ContextVariables,
+        "ConversableAgent",
+        TransitionTarget,
+        "GroupToolExecutor",
+        "GroupChat",
+        "GroupChatManager",
+        list[dict[str, Any]],
+        Any,
+        list[str],
+        list[Any],
+    ]:
+        """Prepare the group chat for organic agent selection.
+
+        Ensures that:
+        1. The group manager has a valid LLM config
+        2. All agents have appropriate descriptions for the group manager to use
+
+        Args:
+            max_rounds: Maximum number of conversation rounds.
+            messages: Initial message(s) to start the conversation.
+
+        Returns:
+            Tuple containing all necessary components for the group chat.
+        """
+        # Validate that group_manager_args has an LLM config which is required for this pattern
+        if not self.group_manager_args.get("llm_config", False):
+            # Check if any agent has an LLM config we can use
+            has_llm_config = any(getattr(agent, "llm_config", False) for agent in self.agents)
+
+            if not has_llm_config:
+                raise ValueError(
+                    "AutoPattern requires the group_manager_args to include an llm_config, "
+                    "or at least one agent to have an llm_config"
+                )
+
+        # Check that all agents have descriptions for effective group manager selection
+        for agent in self.agents:
+            if not hasattr(agent, "description") or not agent.description:
+                agent.description = f"Agent {agent.name}"
+
+        # Use the parent class's implementation to prepare the agents and group chat
+        components = super().prepare_group_chat(
+            max_rounds=max_rounds,
+            messages=messages,
+        )
+
+        # Extract the group_after_work and the rest of the components
+        (
+            agents,
+            wrapped_agents,
+            user_agent,
+            context_variables,
+            initial_agent,
+            _,
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        ) = components
+
+        # Ensure we're using the group_manager after_work
+        group_after_work = self.group_after_work
+
+        # Return all components with our group_after_work
+        return (
+            agents,
+            wrapped_agents,
+            user_agent,
+            context_variables,
+            initial_agent,
+            group_after_work,
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        )

mm_agents/coact/autogen/agentchat/group/patterns/manual.py

@@ -0,0 +1,176 @@
+# Copyright (c) 2023 - 2025, AG2ai, Inc., AG2ai open-source projects maintainers and core contributors
+#
+# SPDX-License-Identifier: Apache-2.0
+
+from typing import TYPE_CHECKING, Any, Callable, Optional, Tuple, Union
+
+from ..context_variables import ContextVariables
+from ..group_tool_executor import GroupToolExecutor
+from ..targets.transition_target import AskUserTarget, TransitionTarget
+from .pattern import Pattern
+
+if TYPE_CHECKING:
+    from ...conversable_agent import ConversableAgent
+    from ...groupchat import GroupChat, GroupChatManager
+
+
+class ManualPattern(Pattern):
+    """ManualPattern will ask the user to nominate the next agent to speak at each turn."""
+
+    def __init__(
+        self,
+        initial_agent: "ConversableAgent",
+        agents: list["ConversableAgent"],
+        user_agent: Optional["ConversableAgent"] = None,
+        group_manager_args: Optional[dict[str, Any]] = None,
+        context_variables: Optional[ContextVariables] = None,
+        exclude_transit_message: bool = True,
+        summary_method: Optional[Union[str, Callable[..., Any]]] = "last_msg",
+    ):
+        """Initialize the ManualPattern.
+
+        The after_work is always set to ask_user, which will prompt the user for the next agent
+
+        Args:
+            initial_agent: The first agent to speak in the group chat.
+            agents: List of all agents participating in the chat.
+            user_agent: Optional user proxy agent.
+            group_manager_args: Optional arguments for the GroupChatManager.
+            context_variables: Initial context variables for the chat.
+            exclude_transit_message: Whether to exclude transit messages from the conversation.
+            summary_method: Method for summarizing the conversation.
+        """
+        # The group after work will be to ask the user
+        group_after_work = AskUserTarget()
+
+        super().__init__(
+            initial_agent=initial_agent,
+            agents=agents,
+            user_agent=user_agent,
+            group_manager_args=group_manager_args,
+            context_variables=context_variables,
+            group_after_work=group_after_work,
+            exclude_transit_message=exclude_transit_message,
+            summary_method=summary_method,
+        )
+
+    def prepare_group_chat(
+        self,
+        max_rounds: int,
+        messages: Union[list[dict[str, Any]], str],
+    ) -> Tuple[
+        list["ConversableAgent"],
+        list["ConversableAgent"],
+        Optional["ConversableAgent"],
+        ContextVariables,
+        "ConversableAgent",
+        TransitionTarget,
+        "GroupToolExecutor",
+        "GroupChat",
+        "GroupChatManager",
+        list[dict[str, Any]],
+        Any,
+        list[str],
+        list[Any],
+    ]:
+        """Prepare the group chat for organic agent selection.
+
+        Ensures that:
+        1. The group manager has a valid LLM config
+        2. All agents have appropriate descriptions for the group manager to use
+
+        Args:
+            max_rounds: Maximum number of conversation rounds.
+            messages: Initial message(s) to start the conversation.
+
+        Returns:
+            Tuple containing all necessary components for the group chat.
+        """
+        # Use the parent class's implementation to prepare the agents and group chat
+        components = super().prepare_group_chat(
+            max_rounds=max_rounds,
+            messages=messages,
+        )
+
+        # Extract the group_after_work and the rest of the components
+        (
+            agents,
+            wrapped_agents,
+            user_agent,
+            context_variables,
+            initial_agent,
+            _,
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        ) = components
+
+        # Ensure we're using the group_manager after_work
+        group_after_work = self.group_after_work
+
+        # Set up the allowed speaker transitions to exclude user_agent and GroupToolExecutor
+        self._setup_allowed_transitions(groupchat, user_agent, tool_executor)
+
+        # Return all components with our group_after_work
+        return (
+            agents,
+            wrapped_agents,
+            user_agent,
+            context_variables,
+            initial_agent,
+            group_after_work,
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        )
+
+    def _setup_allowed_transitions(
+        self, groupchat: "GroupChat", user_agent: Optional["ConversableAgent"], tool_executor: "GroupToolExecutor"
+    ) -> None:
+        """Set up the allowed speaker transitions for the group chat so that when a user selects the next agent the tool executor and user agent don't appear as options.
+
+        Creates transitions where:
+        1. Any agent can speak after any other agent, including themselves
+        2. The user_agent and GroupToolExecutor are excluded from transitions
+
+        Args:
+            groupchat: The GroupChat instance to configure
+            user_agent: The user agent to exclude from transitions
+            tool_executor: The GroupToolExecutor to exclude from transitions
+        """
+        # NOTE: THIS IS NOT WORKING - THE TRANSITIONS ARE NOT BEING KEPT?!
+        """
+        # Get all agents in the group chat
+        all_agents = groupchat.agents
+
+        # Filter out user_agent and group tool executor
+        eligible_agents = []
+        for agent in all_agents:
+            # Skip user_agent
+            if agent == user_agent:
+                continue
+
+            # Skip GroupToolExecutor
+            if isinstance(agent, GroupToolExecutor):
+                continue
+
+            eligible_agents.append(agent)
+
+        # Create a fully connected graph among eligible agents
+        # Each agent can be followed by any other eligible agent
+        allowed_transitions = {}
+        for agent in eligible_agents:
+            # For each agent, every other eligible agent can follow
+            allowed_transitions[agent] = eligible_agents
+
+        # Set the transitions in the group chat
+        groupchat.allowed_speaker_transitions_dict = allowed_transitions
+        """

mm_agents/coact/autogen/agentchat/group/patterns/pattern.py

@@ -0,0 +1,294 @@
+# Copyright (c) 2023 - 2025, AG2ai, Inc., AG2ai open-source projects maintainers and core contributors
+#
+# SPDX-License-Identifier: Apache-2.0
+
+# Patterns of agent orchestrations
+# Uses the group chat or the agents' handoffs to create a pattern
+
+from abc import ABC, abstractmethod
+from typing import TYPE_CHECKING, Any, Callable, Optional, Tuple, Union
+
+from ..context_variables import ContextVariables
+from ..group_utils import (
+    create_group_manager,
+    create_group_transition,
+    link_agents_to_group_manager,
+    prepare_group_agents,
+    process_initial_messages,
+    setup_context_variables,
+)
+from ..targets.transition_target import TerminateTarget, TransitionTarget
+
+if TYPE_CHECKING:
+    from ...agent import Agent
+    from ...conversable_agent import ConversableAgent
+    from ...groupchat import GroupChat, GroupChatManager
+    from ..group_tool_executor import GroupToolExecutor
+
+
+class Pattern(ABC):
+    """Base abstract class for all orchestration patterns.
+
+    Patterns provide a reusable way to define how agents interact within a group chat.
+    Each pattern encapsulates the logic for setting up agents, configuring handoffs,
+    and determining the flow of conversation.
+
+    This is an abstract base class and should not be instantiated directly.
+    Use one of the concrete pattern implementations like AutoPattern,
+    RoundRobinPattern, RandomPattern, or ManualPattern.
+    """
+
+    def __init__(
+        self,
+        initial_agent: "ConversableAgent",
+        agents: list["ConversableAgent"],
+        user_agent: Optional["ConversableAgent"] = None,
+        group_manager_args: Optional[dict[str, Any]] = None,
+        context_variables: Optional[ContextVariables] = None,
+        group_after_work: Optional[TransitionTarget] = None,
+        exclude_transit_message: bool = True,
+        summary_method: Optional[Union[str, Callable[..., Any]]] = "last_msg",
+    ):
+        """Initialize the pattern with the required components.
+
+        Args:
+            initial_agent: The first agent to speak in the group chat.
+            agents: List of all agents participating in the chat.
+            user_agent: Optional user proxy agent.
+            group_manager_args: Optional arguments for the GroupChatManager.
+            context_variables: Initial context variables for the chat.
+            group_after_work: Default after work transition behavior when no specific next agent is determined.
+            exclude_transit_message: Whether to exclude transit messages from the conversation.
+            summary_method: Method for summarizing the conversation.
+        """
+        self.initial_agent = initial_agent
+        self.agents = agents
+        self.user_agent = user_agent
+        self.group_manager_args = group_manager_args or {}
+        self.context_variables = context_variables or ContextVariables()
+        self.group_after_work = group_after_work if group_after_work is not None else TerminateTarget()
+        self.exclude_transit_message = exclude_transit_message
+        self.summary_method = summary_method
+
+    @abstractmethod
+    def prepare_group_chat(
+        self,
+        max_rounds: int,
+        messages: Union[list[dict[str, Any]], str],
+    ) -> Tuple[
+        list["ConversableAgent"],
+        list["ConversableAgent"],
+        Optional["ConversableAgent"],
+        ContextVariables,
+        "ConversableAgent",
+        TransitionTarget,
+        "GroupToolExecutor",
+        "GroupChat",
+        "GroupChatManager",
+        list[dict[str, Any]],
+        "ConversableAgent",
+        list[str],
+        list["Agent"],
+    ]:
+        """Prepare the group chat for orchestration.
+
+        This is the main method called by initiate_group_chat to set up the pattern.
+        Subclasses must implement or extend this method to define pattern-specific behavior.
+
+        Args:
+            max_rounds: Maximum number of conversation rounds.
+            messages: Initial message(s) to start the conversation.
+
+        Returns:
+            Tuple containing:
+            - List of agents involved in the group chat
+            - List of wrapped agents
+            - User agent, if applicable
+            - Context variables for the group chat
+            - Initial agent for the group chat
+            - Group-level after work transition for the group chat
+            - Tool executor for the group chat
+            - GroupChat instance
+            - GroupChatManager instance
+            - Processed messages
+            - Last agent to speak
+            - List of group agent names
+            - List of temporary user agents
+        """
+        from ...groupchat import GroupChat
+
+        # Prepare the agents using the existing helper function
+        tool_executor, wrapped_agents = prepare_group_agents(
+            self.agents, self.context_variables, self.exclude_transit_message
+        )
+
+        # Process the initial messages BEFORE creating the GroupChat
+        # This will create a temporary user agent if needed
+        processed_messages, last_agent, group_agent_names, temp_user_list = process_initial_messages(
+            messages, self.user_agent, self.agents, wrapped_agents
+        )
+
+        # Create transition function (has enclosed state for initial agent)
+        group_transition = create_group_transition(
+            initial_agent=self.initial_agent,
+            tool_execution=tool_executor,
+            group_agent_names=group_agent_names,
+            user_agent=self.user_agent,
+            group_after_work=self.group_after_work,
+        )
+
+        # Create the group chat - now we use temp_user_list if no user_agent
+        groupchat = GroupChat(
+            agents=[tool_executor]
+            + self.agents
+            + wrapped_agents
+            + ([self.user_agent] if self.user_agent else temp_user_list),
+            messages=[],
+            max_round=max_rounds,
+            speaker_selection_method=group_transition,
+        )
+
+        # Create the group manager
+        manager = create_group_manager(groupchat, self.group_manager_args, self.agents, self.group_after_work)
+
+        # Point all agent's context variables to this function's context_variables
+        setup_context_variables(
+            tool_execution=tool_executor,
+            agents=self.agents,
+            manager=manager,
+            user_agent=self.user_agent,
+            context_variables=self.context_variables,
+        )
+
+        # Link all agents with the GroupChatManager to allow access to the group chat
+        link_agents_to_group_manager(groupchat.agents, manager)
+
+        return (
+            self.agents,
+            wrapped_agents,
+            self.user_agent,
+            self.context_variables,
+            self.initial_agent,
+            self.group_after_work,
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        )  # type: ignore[return-value]
+
+    @classmethod
+    def create_default(
+        cls,
+        initial_agent: "ConversableAgent",
+        agents: list["ConversableAgent"],
+        user_agent: Optional["ConversableAgent"] = None,
+        group_manager_args: Optional[dict[str, Any]] = None,
+        context_variables: Optional[ContextVariables] = None,
+        exclude_transit_message: bool = True,
+        summary_method: Optional[Union[str, Callable[..., Any]]] = "last_msg",
+    ) -> "DefaultPattern":
+        """Create a default pattern with minimal configuration.
+
+        This replaces the need for a separate BasePattern class by providing
+        a factory method that creates a simple DefaultPattern instance.
+
+        Args:
+            initial_agent: The first agent to speak in the group chat.
+            agents: List of all agents participating in the chat.
+            user_agent: Optional user proxy agent.
+            group_manager_args: Optional arguments for the GroupChatManager.
+            context_variables: Initial context variables for the chat.
+            exclude_transit_message: Whether to exclude transit messages from the conversation.
+            summary_method: Method for summarizing the conversation.
+
+        Returns:
+            A DefaultPattern instance with basic configuration.
+        """
+        return DefaultPattern(
+            initial_agent=initial_agent,
+            agents=agents,
+            user_agent=user_agent,
+            group_manager_args=group_manager_args,
+            context_variables=context_variables,
+            exclude_transit_message=exclude_transit_message,
+            summary_method=summary_method,
+        )
+
+
+class DefaultPattern(Pattern):
+    """DefaultPattern implements a minimal pattern for simple agent interactions.
+
+    This replaces the previous BasePattern and provides a concrete implementation
+    of the Pattern abstract base class.
+    """
+
+    def prepare_group_chat(
+        self,
+        max_rounds: int,
+        messages: Union[list[dict[str, Any]], str],
+    ) -> Tuple[
+        list["ConversableAgent"],
+        list["ConversableAgent"],
+        Optional["ConversableAgent"],
+        ContextVariables,
+        "ConversableAgent",
+        TransitionTarget,
+        "GroupToolExecutor",
+        "GroupChat",
+        "GroupChatManager",
+        list[dict[str, Any]],
+        Any,
+        list[str],
+        list[Any],
+    ]:
+        """Prepare the group chat with default configuration.
+
+        This implementation calls the parent class method but ensures that
+        the group_after_work in the returned tuple is the pattern's own.
+
+        Args:
+            max_rounds: Maximum number of conversation rounds.
+            messages: Initial message(s) to start the conversation.
+
+        Returns:
+            Tuple containing all necessary components for the group chat.
+        """
+        # Use the parent class's implementation to prepare the agents and group chat
+        (
+            agents,
+            wrapped_agents,
+            user_agent,
+            context_variables,
+            initial_agent,
+            _,  # Ignore the group_after_work from parent
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        ) = super().prepare_group_chat(
+            max_rounds=max_rounds,
+            messages=messages,
+        )
+
+        # Return all components with our group_after_work
+        return (
+            agents,
+            wrapped_agents,
+            user_agent,
+            context_variables,
+            initial_agent,
+            self.group_after_work,  # Use our own group_after_work
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        )

mm_agents/coact/autogen/agentchat/group/patterns/random.py

@@ -0,0 +1,106 @@
+# Copyright (c) 2023 - 2025, AG2ai, Inc., AG2ai open-source projects maintainers and core contributors
+#
+# SPDX-License-Identifier: Apache-2.0
+
+from typing import TYPE_CHECKING, Any, Optional, Tuple, Union
+
+from ..context_variables import ContextVariables
+from ..targets.transition_target import RandomAgentTarget, TransitionTarget
+from .pattern import Pattern
+
+if TYPE_CHECKING:
+    from ...conversable_agent import ConversableAgent
+    from ...groupchat import GroupChat, GroupChatManager
+    from ..group_tool_executor import GroupToolExecutor
+
+
+class RandomPattern(Pattern):
+    """RandomPattern implements a random agent selection process."""
+
+    def _generate_handoffs(
+        self,
+        initial_agent: "ConversableAgent",
+        agents: list["ConversableAgent"],
+        user_agent: Optional["ConversableAgent"],
+    ) -> None:
+        """Generate handoffs between agents in a random fashion."""
+        agent_list = agents + ([user_agent] if user_agent is not None else [])
+
+        for agent in agent_list:
+            # Get the list of agents except itself
+            other_agents = [a for a in agent_list if a != agent]
+
+            # Create a random after work
+            agent.handoffs.set_after_work(target=RandomAgentTarget(agents=other_agents))
+
+    def prepare_group_chat(
+        self,
+        max_rounds: int,
+        messages: Union[list[dict[str, Any]], str],
+    ) -> Tuple[
+        list["ConversableAgent"],
+        list["ConversableAgent"],
+        Optional["ConversableAgent"],
+        ContextVariables,
+        "ConversableAgent",
+        TransitionTarget,
+        "GroupToolExecutor",
+        "GroupChat",
+        "GroupChatManager",
+        list[dict[str, Any]],
+        Any,
+        list[str],
+        list[Any],
+    ]:
+        """Prepare the group chat for organic agent selection.
+
+        Ensures that:
+        1. The group manager has a valid LLM config
+        2. All agents have appropriate descriptions for the group manager to use
+
+        Args:
+            max_rounds: Maximum number of conversation rounds.
+            messages: Initial message(s) to start the conversation.
+
+        Returns:
+            Tuple containing all necessary components for the group chat.
+        """
+        # Use the parent class's implementation to prepare the agents and group chat
+        (
+            agents,
+            wrapped_agents,
+            user_agent,
+            context_variables,
+            initial_agent,
+            group_after_work,
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        ) = super().prepare_group_chat(
+            max_rounds=max_rounds,
+            messages=messages,
+        )
+
+        # Create the random handoffs between agents
+        self._generate_handoffs(initial_agent=initial_agent, agents=agents, user_agent=user_agent)
+
+        # Return all components with our group_after_work
+        return (
+            agents,
+            wrapped_agents,
+            user_agent,
+            context_variables,
+            initial_agent,
+            group_after_work,
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        )

mm_agents/coact/autogen/agentchat/group/patterns/round_robin.py

@@ -0,0 +1,117 @@
+# Copyright (c) 2023 - 2025, AG2ai, Inc., AG2ai open-source projects maintainers and core contributors
+#
+# SPDX-License-Identifier: Apache-2.0
+
+from typing import TYPE_CHECKING, Any, Optional, Tuple, Union
+
+from ..context_variables import ContextVariables
+from ..targets.transition_target import AgentTarget, TransitionTarget
+from .pattern import Pattern
+
+if TYPE_CHECKING:
+    from ...conversable_agent import ConversableAgent
+    from ...groupchat import GroupChat, GroupChatManager
+    from ..group_tool_executor import GroupToolExecutor
+
+
+class RoundRobinPattern(Pattern):
+    """RoundRobinPattern implements a round robin with handoffs between agents."""
+
+    def _generate_handoffs(
+        self,
+        initial_agent: "ConversableAgent",
+        agents: list["ConversableAgent"],
+        user_agent: Optional["ConversableAgent"],
+    ) -> None:
+        """Generate handoffs between agents in a round-robin fashion."""
+        # Create a list of the agents and the user_agent but put the initial_agent first
+        agent_list = [initial_agent]
+
+        # Add the rest of the agents, excluding the initial_agent and user_agent
+        for agent in agents:
+            if agent != initial_agent and (user_agent is None or agent != user_agent):
+                agent_list.append(agent)
+
+        # Add the user_agent last if it exists
+        if user_agent is not None:
+            agent_list.append(user_agent)
+
+        # Create handoffs in a round-robin fashion
+        for i, agent in enumerate(agent_list):
+            # Last agent hands off to the first agent
+            # Otherwise agent hands off to the next one
+            handoff_target = agent_list[0] if i == len(agent_list) - 1 else agent_list[i + 1]
+
+            agent.handoffs.set_after_work(target=AgentTarget(agent=handoff_target))
+
+    def prepare_group_chat(
+        self,
+        max_rounds: int,
+        messages: Union[list[dict[str, Any]], str],
+    ) -> Tuple[
+        list["ConversableAgent"],
+        list["ConversableAgent"],
+        Optional["ConversableAgent"],
+        ContextVariables,
+        "ConversableAgent",
+        TransitionTarget,
+        "GroupToolExecutor",
+        "GroupChat",
+        "GroupChatManager",
+        list[dict[str, Any]],
+        Any,
+        list[str],
+        list[Any],
+    ]:
+        """Prepare the group chat for organic agent selection.
+
+        Ensures that:
+        1. The group manager has a valid LLM config
+        2. All agents have appropriate descriptions for the group manager to use
+
+        Args:
+            max_rounds: Maximum number of conversation rounds.
+            messages: Initial message(s) to start the conversation.
+
+        Returns:
+            Tuple containing all necessary components for the group chat.
+        """
+        # Use the parent class's implementation to prepare the agents and group chat
+        (
+            agents,
+            wrapped_agents,
+            user_agent,
+            context_variables,
+            initial_agent,
+            group_after_work,
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        ) = super().prepare_group_chat(
+            max_rounds=max_rounds,
+            messages=messages,
+        )
+
+        # Create the handoffs between agents
+        self._generate_handoffs(initial_agent=initial_agent, agents=agents, user_agent=user_agent)
+
+        # Return all components with our group_after_work
+        return (
+            agents,
+            wrapped_agents,
+            user_agent,
+            context_variables,
+            initial_agent,
+            group_after_work,
+            tool_executor,
+            groupchat,
+            manager,
+            processed_messages,
+            last_agent,
+            group_agent_names,
+            temp_user_list,
+        )