Vector store vs memory “vivante”
Summary
L'article clarifie la différence entre le RAG classique (stateless) et la mémoire d'agent (stateful), et présente deux approches émergentes : Mem0 avec une mémoire à trois niveaux et auto-édition, et Letta avec une mémoire façon OS (core/archive). Il souligne que le vrai défi est la logique de mise à jour et d'oubli, pas le stockage vectoriel.
Similar Articles
@PrajwalTomar_: YOUR VECTOR DB ISN'T MEMORY. IT'S A GRAVEYARD. london stored cancel ignored gym still london embedding worked database …
The article argues that vector databases alone act as data graveyards without true memory capabilities, and describes building a local AI memory layer with contradiction detection and forgetting cycles on top of Actian VectorAI DB.
Unpopular opinion: AI agents don't always need a Vector DB for project memory
The article argues that for small to medium coding projects, using Vector DBs for AI agent memory may be unnecessary, and proposes using Markdown and Git as a simpler, auditable alternative for project context management.
Feedback on V1 memory architecture for multi-agent setup (supervisor/sub-agents) – targeted retrieval vs unified store?
A developer shares a prototype memory architecture for multi-agent systems with separate storage for episodic, semantic, and procedural memories and intent-based retrieval, seeking feedback on scalability and design choices.
Understanding Agent Memory (38 minute read)
This article compares three common shapes of agent memory systems—file-based, structured store, and experience-based—and evaluates their effectiveness through a benchmark using a common agent loop and open-weight model.
AIPass Memory: Why agents that own small JSON files beat giant vector stores
AIPass introduces a memory system where AI agents manage their identity and session history in small JSON files, with hooks for context injection and archival to vector stores only when necessary, enhancing efficiency in multi-agent setups.