john/McRogueFace
1
0
Fork 0
Code Issues 32 Pull requests Projects Releases Wiki Activity

Update "LLM-Agent-Testbed-Architecture.-"

John McCardle 2026-02-07 23:02:33 +00:00
commit ffc51fd392
1 changed files with 660 additions and 657 deletions
Download patch file Download diff file Expand all files Collapse all files

15
LLM-Agent-Testbed-Architecture.md → LLM-Agent-Testbed-Architecture.-.md

@ -1,8 +1,8 @@
# LLM Agent Testbed Architecture # LLM Agent Testbed Architecture
**Status**: Planning **Status**: Design Complete, Implementation Pending
**Parent Issue**: #154 - Grounded Multi-Agent Testbed **Parent Issue**: #154 - Grounded Multi-Agent Testbed
**Last Updated**: 2025-12-14 **Last Updated**: 2026-02-07
## Overview ## Overview
@ -13,6 +13,8 @@ This document describes the architecture for running LLM agents in McRogueFace e
The key insight is that both modes share the same physics layer, but differ in their perception layer. An **ActionLog** bridges this gap, providing structured event data that humans can view as a combat log and agents receive as text context. The key insight is that both modes share the same physics layer, but differ in their perception layer. An **ActionLog** bridges this gap, providing structured event data that humans can view as a combat log and agents receive as text context.
> **Headless Mode Available**: McRogueFace now supports true headless execution (`--headless --exec`) with deterministic frame stepping via `mcrfpy.step()` (#157 completed). This enables agent testing without an X11 display server or GPU, making it suitable for CI pipelines, remote servers, and automated evaluation harnesses.
--- ---
## Architecture Layers ## Architecture Layers
@ -644,10 +646,11 @@ src/scripts/
## Related Issues ## Related Issues
- #154 - Grounded Multi-Agent Testbed (parent) - #153 - Separate render loop from game state loop (**complete**)
- #155 - Deterministic Text Descriptions (closed) - #154 - Grounded Multi-Agent Testbed (parent, **open**, 2 comments)
- #156 - Turn-based LLM Agent Orchestration (in progress) - #155 - Deterministic Text Descriptions (**closed**)
- #153 - Separate render loop from game state loop (complete) - #156 - Turn-based LLM Agent Orchestration (**open**, 4 comments, in progress)
- #157 - Headless mode (**complete** - enables display-free agent testing)
--- ---