APEX-MEM: property graph + append-only events for LLM long-term memory
burkov · x · 2026-10-18
The article tackles LLMs' core challenge in long-term conversations: maintaining coherent memory across sessions. Bigger context windows or RAG introduce noise, lose critical details, and fail to track evolving facts or resolve contradictions; existing entity-centric graph systems limit entity classes or overwrite prior information, losing temporal context.
APEX-MEM combines:
- Property graph + domain-agnostic ontology (no restricted entity classes)
- Append-only event storage preserving temporal evolution
- Multi-tool retrieval agent that resolves conflicting/evolving information at query time
It evaluates whether this structured approach improves long-term conversational reasoning versus existing structured memory systems.
More from coding & agent
- DeepSeek V4.1 Flash 1017 matches Opus 5.5 on long-horizon browser use at 19x lower cost — zizhpan · 2026-10-18
- Writer has Claude redesign her Obsidian plugin UI with zero backend changes — evielync · 2026-10-18
- AI agents hunting a $465 train refund invented excuses, leaked medical data, and spammed an MP — The Decoder · 2026-10-18
- Open-source Grok Bot alternative cuts coding agent costs from $110/day to $4 — Arindam_1729 · 2026-10-18
- DeepSeek harness lifts V4.1 Flash by 2-3 points at 80% higher cost; model caught cheating — PawelHuryn · 2026-10-18
- Agent engineering debate: when is a retry defensible after a timed-out non-idempotent write? — InformationLive493 · 2026-10-18