haive.games.chess.aug_llms¶

Chess game augmented LLM configurations module.

This module provides augmented LLM configurations for the chess game, including:

Move generation prompts for white and black players
Position analysis prompts for both sides
Structured output models for moves and analysis
Pre-configured LLM configurations for easy agent setup

The module provides an alternative to the engines.py approach, with a focus on customizability and different prompt styles for chess gameplay.

Example

>>> from haive.games.chess.aug_llms import aug_llm_configs
>>>
>>> # Access white player's move generation configuration
>>> white_config = aug_llm_configs["white_player"]
>>> white_prompt = white_config.prompt_template

Functions¶

`build_chess_aug_llms_per_color`(*[, white_llm, black_llm])	Build LLM configs for both players and analyzers using per-color LLMs.
`generate_analysis_prompt`(color)	Generate an analysis prompt for a given player color.
`generate_move_prompt`(color)	Generate a move prompt for a given player color.

Module Contents¶

haive.games.chess.aug_llms.build_chess_aug_llms_per_color(*, white_llm=AzureLLMConfig(), black_llm=DeepSeekLLMConfig())¶

Build LLM configs for both players and analyzers using per-color LLMs.

Creates a comprehensive set of AugLLMConfig objects for chess gameplay, allowing different LLM providers for white and black players.

Parameters:

white_llm (LLMConfig | None) – LLM configuration for white player and analyzer
black_llm (LLMConfig | None) – LLM configuration for black player and analyzer

Returns:

Dictionary with configurations for all chess roles

Return type:

dict[str, AugLLMConfig]

Examples

>>> # Using default LLMs
>>> configs = build_chess_aug_llms_per_color()
>>> len(configs)
4
>>> sorted(list(configs.keys()))
['black_analyzer', 'black_player', 'white_analyzer', 'white_player']

>>> # Using custom LLMs
>>> from haive.core.models.llm.base import AnthropicLLMConfig
>>> configs = build_chess_aug_llms_per_color(
...     white_llm=AnthropicLLMConfig(model="claude-3-opus-20240229"),
...     black_llm=AnthropicLLMConfig(model="claude-3-sonnet-20240229"),
... )
>>> configs["white_player"].llm_config.model
'claude-3-opus-20240229'

haive.games.chess.aug_llms.generate_analysis_prompt(color)¶

Generate an analysis prompt for a given player color.

Creates a ChatPromptTemplate with system and human messages designed to elicit detailed position analysis from an LLM.

Parameters:: color (str) – Player color (“white” or “black”)
Returns:: A prompt template for generating chess position analysis
Return type:: ChatPromptTemplate

Examples

>>> black_prompt = generate_analysis_prompt("black")
>>> isinstance(black_prompt, ChatPromptTemplate)
True
>>> "strategic themes" in black_prompt.messages[1][1]
True

haive.games.chess.aug_llms.generate_move_prompt(color)¶

Generate a move prompt for a given player color.

Creates a ChatPromptTemplate with system and human messages designed to elicit high-quality chess moves from an LLM.

Parameters:: color (str) – Player color (“white” or “black”)
Returns:: A prompt template for generating chess moves
Return type:: ChatPromptTemplate

Examples

>>> white_prompt = generate_move_prompt("white")
>>> isinstance(white_prompt, ChatPromptTemplate)
True
>>> "UCI format" in white_prompt.messages[0][1]
True