[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-82790":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":15,"forks30d":15,"starsTrendScore":19,"compositeScore":20,"rankGlobal":9,"rankLanguage":9,"license":21,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":24,"hasPages":22,"topics":25,"createdAt":9,"pushedAt":9,"updatedAt":26,"readmeContent":27,"aiSummary":28,"trendingCount":15,"starSnapshotCount":15,"syncStatus":29,"lastSyncTime":30,"discoverSource":31},82790,"memory-os","ClaudioDrews\u002Fmemory-os","ClaudioDrews","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.",null,"Python",1083,100,10,3,0,16,269,1008,84,104,"MIT License",false,"main",true,[],"2026-06-12 04:01:39","# Memory OS — Hermes Agent Memory Operating System\n\n![Memory OS Banner](assets\u002Fbanner.jpg)\n\n> **Your agent finally stops forgetting.**  \\\n> Permanent memory. Local memory infrastructure. API-provider agnostic. Surgically token-efficient.\n\nSeven memory layers. Automatic, intelligent context injection. Structured facts with trust scoring. A self-curating wiki pipeline. Semantic search across **every conversation you've ever had**.\n\nMemory OS turns Hermes Agent into a real long-term collaborator — one that remembers your projects, your decisions, your reasoning, and brings exactly the right context back at exactly the right moment. Like talking to a colleague who was there for every session.\n\n**Memory infrastructure runs entirely on your machine. Works with any LLM provider — OpenRouter, OpenAI, Anthropic, Ollama, or local models. No memory subscription. No vendor lock-in.**\n\n---\n\n## The problem every serious Hermes user knows\n\nYou spend hours configuring the agent, teaching it your preferences, solving hard problems together — and in the next session it acts like it's meeting you for the first time.\n\n- Repeating context at the start of every conversation\n- Losing the thread of important decisions made weeks ago\n- Structured facts — your stack, your projects, your patterns — with nowhere to live\n- Every memory solution you've tried is either cloud-locked or too shallow to matter\n\nAfter months of hitting these walls in production, I built something that actually works.\n\n---\n\n## What Memory OS is\n\nNot just another plugin. A complete **memory operating system** — 7 layers working in concert, from flat files to a vector database, with surgical context injection, a knowledge pipeline that organizes itself, **and an explicit Ground Truth hierarchy that tells the agent to actually use the injected memory**.\n\nDesigned and refined by someone who ran headfirst into every limitation of stock Hermes and every existing memory solution.\n\n**Requirements:** Hermes Agent + Docker (Qdrant + Redis + ARQ Worker) + Python 3.11+.  \nCompatible with any LLM provider Hermes supports — OpenRouter, OpenAI, Anthropic, Ollama, and more.\n\n---\n\n## Architecture: 7 memory layers\n\n```\n┌──────────────────────────────────────────────────────────────────┐\n│  LAYER 1 · WORKSPACE                                              │\n│  MEMORY.md · USER.md · CREATIVE.md                               │\n│  → Injected into the system prompt every single turn             │\n├──────────────────────────────────────────────────────────────────┤\n│  LAYER 2 · SESSIONS                                               │\n│  state.db (SQLite + FTS5)                                         │\n│  → Full-text search across your entire conversation history       │\n├──────────────────────────────────────────────────────────────────┤\n│  LAYER 3 · STRUCTURED FACTS                                       │\n│  memory_store.db (SQLite + HRR + FTS5 + trust scoring)            │\n│  → Durable facts with entity resolution and an automatic          │\n│    feedback loop that trains trust scores over time               │\n├──────────────────────────────────────────────────────────────────┤\n│  LAYER 4 · FABRIC (CROSS-SESSION)                                 │\n│  Icarus Plugin (heavily forked)                                   │\n│  → LLM-powered session extraction + multi-source injection        │\n│  → 16 tools: fabric_recall, fabric_write, fabric_brief, etc.      │\n├──────────────────────────────────────────────────────────────────┤\n│  LAYER 5 · VECTOR DATABASE                                        │\n│  Qdrant (4096d Cosine + BM25 sparse)                              │\n│  → 4-level fallback: hybrid → dense → lexical → SQLite            │\n│  → Weekly decay scanner + semantic dedup (cosine >0.92 → merge)  │\n├──────────────────────────────────────────────────────────────────┤\n│  LAYER 6 · LLM WIKI                                               │\n│  Auto-curated vault: concepts\u002F · entities\u002F · comparisons\u002F         │\n│  → Continuously ingested into Qdrant via wiki-continuous-ingest   │\n├──────────────────────────────────────────────────────────────────┤\n│  ⚡ LAYER 7 · GROUND TRUTH HIERARCHY (identity layer)              │\n│  SOUL.md · rulebook.md                                             │\n│  → Tells the agent that injected memory is authoritative           │\n│  → Without this, layers 2-6 deliver context the agent ignores     │\n└──────────────────────────────────────────────────────────────────┘\n```\n\n**How it flows:**\n\n`pre_llm_call` → surgical recall from all four sources (Fabric + Qdrant + Sessions + Facts)\n\n**But recall is not enough.** The agent must be explicitly instructed to treat this injected context as authoritative. That's what [Layer 7](layers\u002F07-ground-truth.md) provides — without it, the agent rediscovers knowledge that's already in the prompt.\n\n`post_llm_call` + `on_session_end` → automatic learning extraction and capture\n\nEach source is gated by relevance thresholds. Per-session deduplication prevents the same context from appearing twice. A social-closer filter skips trivial messages entirely. No padding. No firehose. The LLM gets exactly what it needs — nothing more.\n\n---\n\n## Why Layer 7 is the most important layer\n\nLayers 1-6 ensure memory is **captured, stored, and injected**. Layer 7 ensures the injected memory is **used**.\n\nWithout the Ground Truth hierarchy:\n- Qdrant points are injected but the agent calls the Qdrant API to verify them\n- Fabric entries are injected but the agent runs `fabric_recall` to re-find them  \n- Session history is injected but the agent runs `session_search` to re-discover it\n- Facts are injected but the agent probes `fact_store` to confirm them\n\nThe result: **memory-zero behavior** despite perfect injection. Every rediscovery burns tokens, context, and time.\n\n→ **[Read Layer 7: Ground Truth Hierarchy](layers\u002F07-ground-truth.md)** — the critical fix.\n\n---\n\n## Memory OS vs. stock Hermes\n\n| Aspect | Stock Hermes | Memory OS |\n|---|---|---|\n| Workspace memory | MEMORY.md + USER.md | + CREATIVE.md + intelligent injection |\n| Session memory | Basic state.db | + FTS5 full-text search + session injection |\n| Structured facts | Not present | Fact store + trust scoring + feedback loop |\n| Cross-session recall | Limited | Fabric fork + multi-source injection |\n| Vector search | Not present | Qdrant hybrid + 4-level fallback cascade |\n| Cleanup and deduplication | Not present | Decay scanner + semantic dedup + archival |\n| Knowledge pipeline | Not present | Self-curating LLM Wiki |\n| **Ground Truth hierarchy** | **Not present** | **Injected memory ranked as authoritative; agent must use context provided** |\n| Token efficiency | — | Surgical: gated retrieval + per-session dedup + no wasted rediscovery |\n| Infrastructure | — | Local memory stack (Qdrant + Redis + ARQ) + any LLM provider |\n\n---\n\n## Why not mem0, Zep, Letta, or other providers?\n\nBecause almost every modern memory solution is **cloud-first**. If you want real, private memory infrastructure running on your own machine — with no cloud memory subscription, full provider flexibility, and no data leaving your local stack — none of them deliver what Memory OS delivers.\n\n| | Memory OS | mem0 | Zep | Letta |\n|---|---|---|---|---|\n| Local memory infrastructure | ✓ | ✗ | ✗ | ✗ |\n| No memory subscription | ✓ | ✗ | ✗ | ✗ |\n| Provider agnostic (OpenRouter, Ollama…) | ✓ | Partial | Partial | Partial |\n| Hermes-native | ✓ | ✗ | ✗ | ✗ |\n| Structured facts + trust scores | ✓ | Partial | ✗ | ✗ |\n| Self-curating wiki | ✓ | ✗ | ✗ | ✗ |\n| Intelligent decay + archival | ✓ | ✗ | ✗ | ✗ |\n| **Ground Truth hierarchy** | **✓** | **✗** | **✗** | **✗** |\n\n---\n\n## Included components\n\n- **Icarus Plugin (heavily modified fork)** — bundled in `icarus\u002F`  \n  The upstream [esaradev\u002Ficarus-plugin](https:\u002F\u002Fgithub.com\u002Fesaradev\u002Ficarus-plugin) is the base, but this fork is not upstream-compatible. Key additions: LLM-powered session extraction (replaces `text[:500]` truncation), multi-source injection (Qdrant + sessions + facts — upstream is fabric only), CREATIVE.md isolation (fixes `§` delimiter corruption from dual-writer conflict), backtick sanitization, system injection filter, and social closer detection.\n\n- **Vault Curator v3** — [ClaudioDrews\u002Fvault-curator](https:\u002F\u002Fgithub.com\u002FClaudioDrews\u002Fvault-curator)  \n  Frontmatter enrichment, semantic linking, and MOC index generation for the wiki layer.\n\n---\n\n## Who this is for\n\nFor people who take Hermes Agent seriously.  \nFor people who want an agent that **actually evolves** over time — one that doesn't need the world re-explained every session.  \nFor people who value clean engineering, extreme efficiency, and solutions that hold up in real local production.\n\nIf you're like me — tired of amnesiac agents — Memory OS was built for you.\n\n---\n\n**Want to see the agent remember for real?**  \nClone it, run it, feel the difference.\n\n→ [Setup guide](setup\u002Finstall.md) · [Layer deep-dives](layers\u002F) · [Infrastructure docs](infrastructure\u002Farchitecture.md) · [Operational skills](skills\u002F) · [License](LICENSE)\n\nMIT License · Built with obsession by someone who runs Hermes every single day.","Memory OS 是一个为Hermes Agent设计的7层记忆操作系统，旨在提供持久化记忆功能。它利用Qdrant作为持久化存储，支持结构化事实记录、织物回忆、自动整理wiki以及精准上下文注入等功能。项目采用Python编写，并且能够在本地运行，兼容包括OpenAI、Anthropic等在内的多种语言模型提供商。Memory OS非常适合需要长期记忆能力的应用场景，如持续性对话系统或复杂任务协作工具，能够帮助用户保持项目、决策和推理过程的记忆，从而在恰当的时间点准确地回溯相关信息。此外，其完全本地化的记忆基础设施确保了数据的安全性和隐私性，避免了对特定云服务提供商的依赖。",2,"2026-06-11 04:09:16","CREATED_QUERY"]