Show HN: Memori – Dual-Mode Memory Layer for AI Agents

3 boburumurzokov 2 8/7/2025, 8:15:02 AM github.com ↗
Hi Everyone,

I’ve been working on an open-source project called Memori — it’s a memory layer for AI agents.

Memori is designed to give agents a persistent memory layer that supports both:

Short-term “conscious” memory (like working memory in humans)

Long-term retrieval using SQL full-text search (not vector stores)

It’s built around a multi-agent architecture (memory agent, conscious agent, retrieval agent), with structured storage and zero config to get started. It supports SQLite, Postgres, and MySQL.

Would love to hear what you think — all critique welcome. Happy to answer questions or hear if you’ve solved this differently.

Comments (2)

lokibug · 2d ago
Super cool. Just built an agent using memori. Much better experience.
boburumurzokov · 2d ago
Thanks for the feedback!

Memori engin uses multiple agents to automatically identify and promote your most important long-term memories into short-term memory — so your AI can respond faster and with better context.

Let me know if you tested with AI Agent or the assistant with function calls?