Repolode
Submit
ClaudioDrews

ClaudioDrews/memory-os

A seven-layer memory system for Hermes Agent that persists context across sessions using Qdrant, SQLite, and a self-curating wiki, running locally with any LLM provider.

Open-source alternative to:mem0

View on GitHub
Stars
1,393
Forks
132
Open issues
14
Last commit
2 months ago

What is it?

Memory OS is a memory layer for Hermes Agent that lets the agent keep projects, decisions, and preferences across sessions instead of starting fresh each time. It combines flat workspace files, a full-text searchable session history, a structured fact store with trust scoring, and a Qdrant vector database with a four-level search fallback.

At each LLM call, the system retrieves relevant context from several sources, gated by relevance thresholds and deduplicated per session, and injects it into the prompt. A weekly decay scanner and semantic deduplication keep the store from growing without bound. A separate Ground Truth layer instructs the agent to treat injected memory as authoritative, so it does not re-query the same information.

The infrastructure runs on your own machine using Docker services (Qdrant, Redis, and an ARQ worker) with Python 3.11+. It works with OpenRouter, OpenAI, Anthropic, Ollama, and other providers Hermes supports. A setup script installs the stack in one command.

Key features

  • 01Persistent memory across sessions with workspace files, session search, and a structured fact store.
  • 02Qdrant vector search with hybrid dense and BM25 retrieval and a four-level fallback.
  • 03Automatic context injection gated by relevance and deduplicated per session.
  • 04Weekly decay scanning and semantic deduplication of stored memories.
  • 05Self-curating LLM wiki that is continuously ingested into the vector database.
  • 06One-command installer that sets up Docker services, databases, and the plugin.

Use cases

  • Long-running projects where an agent needs to remember past decisions and architecture choices.
  • Keeping personal preferences and a tech stack profile available across many conversations.
  • Running a private memory backend locally with Ollama or another self-hosted model.

Quick start

  1. 1

    Install the prerequisites: Hermes Agent, Docker, and Python 3.11+

  2. 2

    Run the official one-command installer from the README

  3. 3

    If the installer fails, follow the manual 10-step guide in setup/install.md

Things to know

  • Requires Hermes Agent and Docker running Qdrant, Redis, and an ARQ worker.
  • The bundled Icarus plugin is a heavily modified fork and is not compatible with upstream.

In collections

Long-term memory for AI agentsSelf-hosted appsPrivacy & local-first tools

Topics

#ai-memory#vector-database#rag#qdrant#self-hosted#local-first#persistent-memory#context-injection#docker#ground-truth

FAQ

What is memory-os?
A seven-layer memory system for Hermes Agent that persists context across sessions using Qdrant, SQLite, and a self-curating wiki, running locally with any LLM provider. Memory OS is a memory layer for Hermes Agent that lets the agent keep projects, decisions, and preferences across sessions instead of starting fresh each time. It combines flat workspace files, a full-text searchable session history, a structured fact store with trust scoring, and a Qdrant vector database with a four-level search fallback.
Is memory-os free to use? What license does it have?
memory-os is open source under the MIT license. Check the license terms before using it in commercial products.
Is memory-os still maintained?
Updated within 6 months. The last commit to memory-os was on Sep 20, 2026.
What can I use memory-os for?
Long-running projects where an agent needs to remember past decisions and architecture choices. Keeping personal preferences and a tech stack profile available across many conversations. Running a private memory backend locally with Ollama or another self-hosted model.
How do I install memory-os?
Install the prerequisites: Hermes Agent, Docker, and Python 3.11+ Run the official one-command installer from the README If the installer fails, follow the manual 10-step guide in setup/install.md
Where can I get memory-os?
The source code and installation instructions are on GitHub: https://github.com/ClaudioDrews/memory-os

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects