Skill: memory-consolidation (Dream Memory)

Purpose

Inspired by hippocampal replay: long-running AI systems accumulate session
artifacts, and periodically consolidating them into structured memories
improves future recall and context reuse.

Consolidation Lifecycle

COLLECTING → DREAMING → CONSOLIDATED

Memory Kinds

• SESSION_NOTE — General session observations
• LEARNED_RULE — Patterns/corrections to persist
• BUG_PATTERN — Recurring bug signatures
• ARCHITECTURAL_DECISION — ADR-like records
• CODE_SNIPPET — Reusable code fragments

Quick Start

from agents.memory_consolidation import DreamMemory, MemoryKind, PatternConsolidation

pc = PatternConsolidation()
pc.add_memory(DreamMemory("m1", MemoryKind.BUG_PATTERN, "null deref in auth", tags=["bug", "auth"]))
pc.add_memory(DreamMemory("m2", MemoryKind.BUG_PATTERN, "timeout on login", tags=["bug", "auth"]))
result = pc.consolidate()
print(result)  # clusters found, memories consolidated

Testing

pytest tests/test_memory_consolidation.py -v

Branch

fix/quick-note-259-memory-dreams