haive.games.base.factory

Factory module for creating game agents.

from typing import Any, Dict This module provides a factory class for creating game agents with standardized workflows and configurations. It simplifies the process of creating new game agents by providing a flexible, composable pattern.

Examples

>>> # Create a new chess agent
>>> chess_agent = GameAgentFactory.create_game_agent(
...     name="ChessAgent",
...     state_schema=ChessState,
...     state_manager=ChessStateManager,
...     enable_analysis=True
... )
>>>
>>> # Create a standard workflow
>>> graph = StateGraph()
>>> graph = GameAgentFactory.create_standard_workflow(graph)
Typical usage:

  • Use create_game_agent to generate new game agent classes

  • Use create_standard_workflow to set up game workflows

  • Customize agents with analysis, custom nodes, and edges

Classes

GameAgentFactory

Factory for creating game agents using a flexible, composable pattern.

Module Contents

class haive.games.base.factory.GameAgentFactory[source]

Factory for creating game agents using a flexible, composable pattern.

This factory class provides methods for creating game agents and their workflows. It simplifies the process of creating new game agents by: 1. Generating workflow nodes from a standardized template 2. Using DynamicGraph for simplified graph building 3. Supporting consistent extension patterns for analysis and custom workflows

The factory supports: - Creation of game agent classes with standard methods - Configuration of analysis steps and custom nodes - Flexible workflow definition with conditional edges - Registration of agents with their configs

Examples

>>> # Create a new chess agent
>>> chess_agent = GameAgentFactory.create_game_agent(
...     name="ChessAgent",
...     state_schema=ChessState,
...     state_manager=ChessStateManager
... )
>>>
>>> # Create an instance with custom config
>>> agent = chess_agent(ChessConfig())
static create_game_agent(name, state_schema, state_manager, player1_name='player1', player2_name='player2', enable_analysis=True, aug_llm_configs=None, custom_nodes=None, custom_edges=None, conditional_edges=None)[source]

Create a new game agent class with a complete workflow.

This method generates a new game agent class with all necessary methods and workflow configuration. It creates both the agent class and its corresponding config class.

Parameters:

  • name (str) – Name of the agent class.

  • state_schema (Type[GameState]) – The game state schema class.

  • state_manager (Type[GameStateManager]) – The game state manager class.

  • player1_name (str, optional) – Name for player 1. Defaults to “player1”.

  • player2_name (str, optional) – Name for player 2. Defaults to “player2”.

  • enable_analysis (bool, optional) – Whether to enable analysis steps. Defaults to True.

  • aug_llm_configs (Optional[Dict[str, AugLLMConfig]], optional) – LLM configurations. Defaults to None.

  • custom_nodes (Optional[Dict[str, Callable]], optional) – Custom node functions. Defaults to None.

  • custom_edges (Optional[List[Dict[str, Any]]], optional) – Custom edges. Defaults to None.

  • conditional_edges (Optional[Dict[str, Dict[str, Any]]], optional) – Conditional edges. Defaults to None.

Returns:

A new agent class with all methods and workflow configured.

Return type:

Type[Agent]

Examples

>>> # Create a chess agent with analysis
>>> chess_agent = GameAgentFactory.create_game_agent(
...     name="ChessAgent",
...     state_schema=ChessState,
...     state_manager=ChessStateManager,
...     enable_analysis=True,
...     aug_llm_configs={
...         "player1": player1_config,
...         "player2": player2_config
...     }
... )
static create_standard_workflow(graph, enable_analysis=True)[source]

Add a standard game workflow to an existing graph.

This method creates the typical workflow for a turn-based game with two players, optionally including analysis steps. It modifies the provided graph by adding nodes and edges for the standard game flow.

Parameters:

  • graph (StateGraph) – The state graph to modify.

  • enable_analysis (bool, optional) – Whether to include analysis steps. Defaults to True.

Returns:

The modified graph with standard workflow added.

Return type:

StateGraph

Examples

>>> # Create a basic graph and add standard workflow
>>> graph = StateGraph()
>>> graph = GameAgentFactory.create_standard_workflow(
...     graph,
...     enable_analysis=True
... )