Skip to content

A programming framework for agentic AI

Fix chess sample (#4932)

Eric Zhu52c2a70e95df
diff --git a/python/samples/core_chess_game/main.py b/python/samples/core_chess_game/main.py
index 78b478775..ccc77deba 100644
--- a/python/samples/core_chess_game/main.py
+++ b/python/samples/core_chess_game/main.py
@@ -1,6 +1,7 @@
"""This is an example of simulating a chess game with two agents
that play against each other, using tools to reason about the game state
-and make moves, and using a group chat manager to orchestrate the conversation."""
+and make moves. The agents subscribe to the default topic and publish their
+moves to the default topic."""
import argparse
import asyncio
@@ -19,7 +20,12 @@ from autogen_core import (
message_handler,
)
from autogen_core.model_context import BufferedChatCompletionContext, ChatCompletionContext
-from autogen_core.models import AssistantMessage, ChatCompletionClient, LLMMessage, SystemMessage, UserMessage
+from autogen_core.models import (
+ ChatCompletionClient,
+ LLMMessage,
+ SystemMessage,
+ UserMessage,
+)
from autogen_core.tool_agent import ToolAgent, tool_agent_caller_loop
from autogen_core.tools import FunctionTool, Tool, ToolSchema
from chess import BLACK, SQUARE_NAMES, WHITE, Board, Move
@@ -33,7 +39,7 @@ class TextMessage(BaseModel):
@default_subscription
-class ToolUseAgent(RoutedAgent):
+class PlayerAgent(RoutedAgent):
def __init__(
self,
description: str,
@@ -59,14 +65,15 @@ class ToolUseAgent(RoutedAgent):
self,
tool_agent_id=self._tool_agent_id,
model_client=self._model_client,
- input_messages=(await self._model_context.get_messages()),
+ input_messages=self._system_messages + (await self._model_context.get_messages()),
tool_schema=self._tool_schema,
cancellation_token=ctx.cancellation_token,
)
- assert isinstance(messages[-1].content, str)
# Add the assistant message to the model context.
- await self._model_context.add_message(AssistantMessage(content=messages[-1].content, source=self.id.type))
+ for msg in messages:
+ await self._model_context.add_message(msg)
# Publish the final response.
+ assert isinstance(messages[-1].content, str)
await self.publish_message(TextMessage(content=messages[-1].content, source=self.id.type), DefaultTopicId())
@@ -203,39 +210,39 @@ async def chess_game(runtime: AgentRuntime, model_config: Dict[str, Any]) -> Non
# Register the agents.
await ToolAgent.register(
runtime,
- "ToolAgent",
- lambda: ToolAgent(description="Tool agent for chess game.", tools=black_tools + white_tools),
+ "PlayerBlackToolAgent",
+ lambda: ToolAgent(description="Tool agent for chess game.", tools=black_tools),
+ )
+
+ await ToolAgent.register(
+ runtime,
+ "PlayerWhiteToolAgent",
+ lambda: ToolAgent(description="Tool agent for chess game.", tools=white_tools),
)
- await ToolUseAgent.register(
+ await PlayerAgent.register(
runtime,
"PlayerBlack",
- lambda: ToolUseAgent(
+ lambda: PlayerAgent(
description="Player playing black.",
- instructions="You are a chess player and you play as black. "
- "Use get_legal_moves() to get list of legal moves. "
- "Use get_board() to get the current board state. "
- "Think about your strategy and call make_move(thinking, move) to make a move.",
+ instructions="You are a chess player and you play as black. Use the tool 'get_board' and 'get_legal_moves' to get the legal moves and 'make_move' to make a move.",
model_client=model_client,
model_context=BufferedChatCompletionContext(buffer_size=10),
tool_schema=[tool.schema for tool in black_tools],
- tool_agent_type="ToolAgent",
+ tool_agent_type="PlayerBlackToolAgent",
),
)
- await ToolUseAgent.register(
+ await PlayerAgent.register(
runtime,
"PlayerWhite",
- lambda: ToolUseAgent(
+ lambda: PlayerAgent(
description="Player playing white.",
- instructions="You are a chess player and you play as white. "
- "Use get_legal_moves() to get list of legal moves. "
- "Use get_board() to get the current board state. "
- "Think about your strategy and call make_move(thinking, move) to make a move.",
+ instructions="You are a chess player and you play as white. Use the tool 'get_board' and 'get_legal_moves' to get the legal moves and 'make_move' to make a move.",
model_client=model_client,
model_context=BufferedChatCompletionContext(buffer_size=10),
tool_schema=[tool.schema for tool in white_tools],
- tool_agent_type="ToolAgent",
+ tool_agent_type="PlayerWhiteToolAgent",
),
)
@@ -249,7 +256,7 @@ async def main(model_config: Dict[str, Any]) -> None:
# orchestration.
# Send an initial message to player white to start the game.
await runtime.send_message(
- TextMessage(content="Game started.", source="System"),
+ TextMessage(content="Game started, white player your move.", source="System"),
AgentId("PlayerWhite", "default"),
)
await runtime.stop_when_idle()
diff --git a/python/samples/core_chess_game/utils.py b/python/samples/core_chess_game/utils.py
deleted file mode 100644
index 5fc21dc20..000000000
--- a/python/samples/core_chess_game/utils.py
+++ /dev/null
@@ -1,98 +0,0 @@
-from typing import List, Optional, Union
-
-from autogen_core.models import (
- AssistantMessage,
- FunctionExecutionResult,
- FunctionExecutionResultMessage,
- LLMMessage,
- UserMessage,
-)
-from typing_extensions import Literal
-
-from .messages import (
- FunctionCallMessage,
- Message,
- MultiModalMessage,
- TextMessage,
-)
-
-
-def convert_content_message_to_assistant_message(
- message: Union[TextMessage, MultiModalMessage, FunctionCallMessage],
- handle_unrepresentable: Literal["error", "ignore", "try_slice"] = "error",
-) -> Optional[AssistantMessage]:
- match message:
- case TextMessage() | FunctionCallMessage():
- return AssistantMessage(content=message.content, source=message.source)
- case MultiModalMessage():
- if handle_unrepresentable == "error":
- raise ValueError("Cannot represent multimodal message as AssistantMessage")
- elif handle_unrepresentable == "ignore":
- return None
- elif handle_unrepresentable == "try_slice":
- return AssistantMessage(
- content="".join([x for x in message.content if isinstance(x, str)]),
- source=message.source,
- )
-
-
-def convert_content_message_to_user_message(
- message: Union[TextMessage, MultiModalMessage, FunctionCallMessage],
- handle_unrepresentable: Literal["error", "ignore", "try_slice"] = "error",
-) -> Optional[UserMessage]:
- match message:
- case TextMessage() | MultiModalMessage():
- return UserMessage(content=message.content, source=message.source)
- case FunctionCallMessage():
- if handle_unrepresentable == "error":
- raise ValueError("Cannot represent multimodal message as UserMessage")
- elif handle_unrepresentable == "ignore":
- return None
- elif handle_unrepresentable == "try_slice":
- # TODO: what is a sliced function call?
- raise NotImplementedError("Sliced function calls not yet implemented")
-
-
-def convert_tool_call_response_message(
- message: FunctionExecutionResultMessage,
- handle_unrepresentable: Literal["error", "ignore", "try_slice"] = "error",
-) -> Optional[FunctionExecutionResultMessage]:
- match message:
- case FunctionExecutionResultMessage():
- return FunctionExecutionResultMessage(
- content=[FunctionExecutionResult(content=x.content, call_id=x.call_id) for x in message.content]
- )
-
-
-def convert_messages_to_llm_messages(
- messages: List[Message],
- self_name: str,
- handle_unrepresentable: Literal["error", "ignore", "try_slice"] = "error",
-) -> List[LLMMessage]:
- result: List[LLMMessage] = []
- for message in messages:
- match message:
- case (
- TextMessage(content=_, source=source)
- | MultiModalMessage(content=_, source=source)
- | FunctionCallMessage(content=_, source=source)
- ) if source == self_name:
- converted_message_1 = convert_content_message_to_assistant_message(message, handle_unrepresentable)
- if converted_message_1 is not None:
- result.append(converted_message_1)
- case (
- TextMessage(content=_, source=source)
- | MultiModalMessage(content=_, source=source)
- | FunctionCallMessage(content=_, source=source)
- ) if source != self_name:
- converted_message_2 = convert_content_message_to_user_message(message, handle_unrepresentable)
- if converted_message_2 is not None:
- result.append(converted_message_2)
- case FunctionExecutionResultMessage(content=_):
- converted_message_3 = convert_tool_call_response_message(message, handle_unrepresentable)
- if converted_message_3 is not None:
- result.append(converted_message_3)
- case _:
- raise AssertionError("unreachable")
-
- return result