Episodic Memory (L1)

Status: Prototype (TRL-3)

The Episodic Memory stores snapshots of market conditions (volatility, DQ, capital) together with the best-performing strategy parameters at that moment. It is used to:

• Seed initial population when a node restarts – the most relevant past strategies are injected.
• Bias genome import – if a foreign genome matches a previously successful pattern, it receives a bonus in local_score, increasing its chance of being selected.

Implementation

• src/intelligence/episodic_memory.py – simple list-based store with similarity search.
• Integrated into SwarmNode and used every 50 steps.
• Memory consolidation every 500 steps removes duplicates and limits size.