memory-os
A 7-layer memory operating system for Hermes Agent — persistent memory with Qdrant, structured facts, fabric recall, auto-curated wiki, and surgical context inj
What is memory-os?
A 7-layer memory operating system for Hermes Agent — persistent memory with Qdrant, structured facts, fabric recall, auto-curated wiki, and surgical context injection. Runs locally, any LLM provider.
Overview of memory-os
What it is
Memory OS is a 7-layer memory operating system for Hermes Agent. It provides persistent memory using Qdrant, structured facts, fabric recall, an auto-curated wiki, and surgical context injection. The system runs locally and works with any LLM provider, including OpenRouter, OpenAI, Anthropic, Ollama, and local models. It is designed to turn Hermes Agent into a long-term collaborator that remembers projects, decisions, and reasoning across sessions.
Key features
- Seven memory layers: workspace, sessions, structured facts, fabric cross-session recall, vector database (Qdrant), LLM wiki, and a Ground Truth hierarchy.
- Structured facts with trust scoring and an automatic feedback loop that trains trust scores over time.
- Qdrant vector database with 4-level fallback (hybrid → dense → lexical → SQLite), weekly decay scanner, and semantic dedup (cosine >0.92 → merge).
- Auto-curated LLM wiki vault with concepts, entities, and comparisons, continuously ingested into Qdrant.
- Ground Truth hierarchy (SOUL.md, rulebook.md) that instructs the agent to treat injected memory as authoritative.
- Surgical context injection with relevance thresholds, per-session deduplication, and a social-closer filter to skip trivial messages.
- One-command install via
curl -sSL https://raw.githubusercontent.com/ClaudioDrews/memory-os/main/setup.sh | bash.
Use cases
Memory OS is intended for users who need persistent, local memory for Hermes Agent. It addresses problems such as the agent forgetting context between sessions, losing track of important decisions, and lacking a place for structured facts like stack, projects, and patterns. It is suitable for anyone who wants a private memory infrastructure without cloud dependency or vendor lock-in.
Getting started
bashcurl -sSL https://raw.githubusercontent.com/ClaudioDrews/memory-os/main/setup.sh | bashRequirements: Hermes Agent, Docker (Qdrant, Redis, ARQ Worker), Python 3.11+. For detailed manual setup, see the repository README.
License and pricing
License: MIT. Pricing: open source.
Based on the project's public README and repository metadata.
