| """Adversarial self-play module for RecallTrace. | |
| Two agents co-evolve in a shared environment: | |
| - InvestigatorAgent: finds and quarantines contaminated nodes. | |
| - AdversaryAgent: chooses where and how to hide contamination. | |
| """ | |
| from selfplay.adversary import AdversaryAgent | |
| from selfplay.investigator import InvestigatorAgent | |
| from selfplay.trainer import SelfPlayTrainer | |
| from selfplay.visualization import show_training_curves, show_episode_comparison | |
| from selfplay.demo_replay import render_demo | |
| from selfplay.belief_tracker import BeliefStateTracker | |
| __all__ = [ | |
| "AdversaryAgent", | |
| "InvestigatorAgent", | |
| "SelfPlayTrainer", | |
| "show_training_curves", | |
| "show_episode_comparison", | |
| "render_demo", | |
| "BeliefStateTracker", | |
| ] | |