- Add LLMOrchestrator for LLM-first decision making
- Create autonomous_mode.py as standalone entry point
- Implement ExplorationSession with structured logging
- Add GameContext for complete game state awareness
- LLM decides strategy: explore|combat|quest|social|grinding|retreat
- Automatic discovery logging (rooms, NPCs, dangers, items)
- Learning memory for dangerous rooms and valuable NPCs
- Session logs saved to data/autonomous_sessions/ as JSON
- Real-time console output with emojis and reasoning
- Fallback to rule-based when LLM unavailable
- Safety overrides for low HP and high danger
Usage: python -m src.autonomous_mode --duration 60
The bot now explores autonomously, learns patterns, makes strategic
decisions, and saves detailed session logs for review.
