games.checkers.engines¶

Checkers game engines using AugLLMConfig.

This module provides LLM engine configurations for checkers game agents, including:

  • Player engines for red and black

  • Analyzer engines for position evaluation

  • Prompt templates with checkers-specific instructions

  • Structured output models for moves and analysis

The engines use LLM configurations optimized for checkers gameplay, with prompt templates designed to generate high-quality moves and analysis.

Functions¶

build_checkers_aug_llms()

Build LLM configs for checkers players and analyzers.

generate_analysis_prompt(color)

Generate analysis prompt for checkers.

generate_move_prompt(color)

Generate move selection prompt for checkers.

Module Contents¶

games.checkers.engines.build_checkers_aug_llms()¶

Build LLM configs for checkers players and analyzers.

Creates a complete set of AugLLMConfig objects for checkers gameplay, including player and analyzer engines for both red and black.

All engines use the same LLM model (GPT-4o) with appropriate temperature for move generation and analysis, with the corresponding prompt templates and structured output models.

Returns:

Dictionary with configurations for all checkers roles:

  • ”red_player”: Engine for red’s moves

  • ”black_player”: Engine for black’s moves

  • ”red_analyzer”: Engine for analyzing positions from red’s perspective

  • ”black_analyzer”: Engine for analyzing positions from black’s perspective

Return type:

dict[str, AugLLMConfig]

Examples

>>> engines = build_checkers_aug_llms()
>>> len(engines)
4
>>> sorted(list(engines.keys()))
['black_analyzer', 'black_player', 'red_analyzer', 'red_player']
>>> engines["red_player"].structured_output_model
<class 'haive.games.checkers.models.CheckersPlayerDecision'>
games.checkers.engines.generate_analysis_prompt(color)¶

Generate analysis prompt for checkers.

Creates a ChatPromptTemplate with system and human messages designed to elicit detailed position analysis from an LLM. The prompt includes: - Instructions on what aspects to analyze - Current board state and game context - Output format specifications

Parameters:

color (str) – Player color to analyze for (“red” or “black”)

Returns:

A prompt template for generating position analysis

Return type:

ChatPromptTemplate

Examples

>>> black_prompt = generate_analysis_prompt("black")
>>> "BLACK" in black_prompt.messages[0][1]  # First message content
True
>>> "material_advantage" in black_prompt.messages[0][1]  # Output format specification
True
games.checkers.engines.generate_move_prompt(color)¶

Generate move selection prompt for checkers.

Creates a ChatPromptTemplate with system and human messages designed to elicit high-quality checkers moves from an LLM. The prompt includes: - Clear instructions on move format and rules - Game context like board state and move history - Error feedback for retries - Output format specifications

Parameters:

color (str) – Player color (“red” or “black”)

Returns:

A prompt template for generating checkers moves

Return type:

ChatPromptTemplate

Examples

>>> red_prompt = generate_move_prompt("red")
>>> red_prompt.messages[0][0]  # First message role
'system'
>>> "You are playing checkers as RED" in red_prompt.messages[0][1]  # First message content
True