[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-71923":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":23,"hasPages":23,"topics":25,"createdAt":10,"pushedAt":10,"updatedAt":38,"readmeContent":39,"aiSummary":40,"trendingCount":16,"starSnapshotCount":16,"syncStatus":41,"lastSyncTime":42,"discoverSource":43},71923,"memU","NevaMind-AI\u002FmemU","NevaMind-AI","The memory harness for proactive AI agents — structured storage, intent capture, 10x token reduction.","https:\u002F\u002Fmemu.pro",null,"Python",13820,1035,66,59,0,35,79,225,105,119.05,"Other",false,"main",[26,27,28,29,30,31,32,33,34,35,36,37],"agent-memory","agentic-workflow","claude","claude-skills","mcp","memory","openclaw","openclaw-skills","proactive","proactive-ai","sandbox","skills","2026-06-12 04:01:02","![MemU Banner](assets\u002Fbanner.png)\n\n\u003Cdiv align=\"center\">\n\n# memU\n\n### 24\u002F7 Always-On Proactive Memory for AI Agents\n\n[![PyPI version](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Fmemu-py.svg)](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Fmemu-py)\n[![License: Apache 2.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202.0-blue.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FApache-2.0)\n[![Python 3.13+](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.13+-blue.svg)](https:\u002F\u002Fwww.python.org\u002Fdownloads\u002F)\n[![Discord](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDiscord-Join%20Chat-5865F2?logo=discord&logoColor=white)](https:\u002F\u002Fdiscord.com\u002Finvite\u002FhQZntfGsbJ)\n[![Twitter](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTwitter-Follow-1DA1F2?logo=x&logoColor=white)](https:\u002F\u002Fx.com\u002FmemU_ai)\n\n\u003Ca href=\"https:\u002F\u002Ftrendshift.io\u002Frepositories\u002F17374\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Ftrendshift.io\u002Fapi\u002Fbadge\u002Frepositories\u002F17374\" alt=\"NevaMind-AI%2FmemU | Trendshift\" style=\"width: 250px; height: 55px;\" width=\"250\" height=\"55\"\u002F>\u003C\u002Fa>\n\n**[English](readme\u002FREADME_en.md) | [中文](readme\u002FREADME_zh.md) | [日本語](readme\u002FREADME_ja.md) | [한국어](readme\u002FREADME_ko.md) | [Español](readme\u002FREADME_es.md) | [Français](readme\u002FREADME_fr.md)**\n\n\u003C\u002Fdiv>\n\n---\n\nmemU is a memory framework built for **24\u002F7 proactive agents**.\nIt is designed for long-running use and greatly **reduces the LLM token cost** of keeping agents always online, making always-on, evolving agents practical in production systems.\nmemU **continuously captures and understands user intent**. Even without a command, the agent can tell what you are about to do and act on it by itself.\n\n---\n\n## 🤖 [OpenClaw Alternative](https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemUBot)\n\n\u003Cimg width=\"100%\" src=\"https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemU\u002Fblob\u002Fmain\u002Fassets\u002FmemUbot.png\" \u002F>\n\n**[memU Bot](https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemUBot)** — Now open source. The enterprise-ready OpenClaw. Your proactive AI assistant that remembers everything.\n\n- **Download-and-use and simple** to get started (one-click install, < 3 min).\n- Builds long-term memory to **understand user intent** and act proactively (24\u002F7).\n- **Cuts LLM token cost** with smaller context (~1\u002F10 of comparable usage).\n\nTry now: [memu.bot](https:\u002F\u002Fmemu.bot) · Source: [memUBot on GitHub](https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemUBot)\n\n---\n\n## 🗃️ Memory as File System, File System as Memory\n\nmemU treats **memory like a file system**—structured, hierarchical, and instantly accessible.\n\n| File System | memU Memory |\n|-------------|-------------|\n| 📁 Folders | 🏷️ Categories (auto-organized topics) |\n| 📄 Files | 🧠 Memory Items (extracted facts, preferences, skills) |\n| 🔗 Symlinks | 🔄 Cross-references (related memories linked) |\n| 📂 Mount points | 📥 Resources (conversations, documents, images) |\n\n**Why this matters:**\n- **Navigate memories** like browsing directories—drill down from broad categories to specific facts\n- **Mount new knowledge** instantly—conversations and documents become queryable memory\n- **Cross-link everything**—memories reference each other, building a connected knowledge graph\n- **Persistent & portable**—export, backup, and transfer memory like files\n\n```\nmemory\u002F\n├── preferences\u002F\n│   ├── communication_style.md\n│   └── topic_interests.md\n├── relationships\u002F\n│   ├── contacts\u002F\n│   └── interaction_history\u002F\n├── knowledge\u002F\n│   ├── domain_expertise\u002F\n│   └── learned_skills\u002F\n└── context\u002F\n    ├── recent_conversations\u002F\n    └── pending_tasks\u002F\n```\n\nJust as a file system turns raw bytes into organized data, memU transforms raw interactions into **structured, searchable, proactive intelligence**.\n\n---\n\n## ⭐️ Star the repository\n\n\u003Cimg width=\"100%\" src=\"https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemU\u002Fblob\u002Fmain\u002Fassets\u002Fstar.gif\" \u002F>\nIf you find memU useful or interesting, a GitHub Star ⭐️ would be greatly appreciated.\n\n---\n\n\n## ✨ Core Features\n\n| Capability | Description |\n|------------|-------------|\n| 🤖 **24\u002F7 Proactive Agent** | Always-on memory agent that works continuously in the background |\n| 🎯 **User Intention Capture** | Understands and remembers user goals, preferences, and context across sessions automatically |\n| 💰 **Cost Efficient** | Reduces long-running token costs by caching insights and avoiding redundant LLM calls |\n---\n\n## 🔄 How Proactive Memory Works\n\n```bash\n\ncd examples\u002Fproactive\npython proactive.py\n\n```\n\n---\n\n### Proactive Memory Lifecycle\n```\n┌──────────────────────────────────────────────────────────────────────────────────────────────────┐\n│                                         USER QUERY                                               │\n└──────────────────────────────────────────────────────────────────────────────────────────────────┘\n                 │                                                           │\n                 ▼                                                           ▼\n┌────────────────────────────────────────┐         ┌────────────────────────────────────────────────┐\n│         🤖 MAIN AGENT                  │         │              🧠 MEMU BOT                        │\n│                                        │         │                                                │\n│  Handle user queries & execute tasks   │  ◄───►  │  Monitor, memorize & proactive intelligence    │\n├────────────────────────────────────────┤         ├────────────────────────────────────────────────┤\n│                                        │         │                                                │\n│  ┌──────────────────────────────────┐  │         │  ┌──────────────────────────────────────────┐  │\n│  │  1. RECEIVE USER INPUT           │  │         │  │  1. MONITOR INPUT\u002FOUTPUT                 │  │\n│  │     Parse query, understand      │  │   ───►  │  │     Observe agent interactions           │  │\n│  │     context and intent           │  │         │  │     Track conversation flow              │  │\n│  └──────────────────────────────────┘  │         │  └──────────────────────────────────────────┘  │\n│                 │                      │         │                    │                           │\n│                 ▼                      │         │                    ▼                           │\n│  ┌──────────────────────────────────┐  │         │  ┌──────────────────────────────────────────┐  │\n│  │  2. PLAN & EXECUTE               │  │         │  │  2. MEMORIZE & EXTRACT                   │  │\n│  │     Break down tasks             │  │   ◄───  │  │     Store insights, facts, preferences   │  │\n│  │     Call tools, retrieve data    │  │  inject │  │     Extract skills & knowledge           │  │\n│  │     Generate responses           │  │  memory │  │     Update user profile                  │  │\n│  └──────────────────────────────────┘  │         │  └──────────────────────────────────────────┘  │\n│                 │                      │         │                    │                           │\n│                 ▼                      │         │                    ▼                           │\n│  ┌──────────────────────────────────┐  │         │  ┌──────────────────────────────────────────┐  │\n│  │  3. RESPOND TO USER              │  │         │  │  3. PREDICT USER INTENT                  │  │\n│  │     Deliver answer\u002Fresult        │  │   ───►  │  │     Anticipate next steps                │  │\n│  │     Continue conversation        │  │         │  │     Identify upcoming needs              │  │\n│  └──────────────────────────────────┘  │         │  └──────────────────────────────────────────┘  │\n│                 │                      │         │                    │                           │\n│                 ▼                      │         │                    ▼                           │\n│  ┌──────────────────────────────────┐  │         │  ┌──────────────────────────────────────────┐  │\n│  │  4. LOOP                         │  │         │  │  4. RUN PROACTIVE TASKS                  │  │\n│  │     Wait for next user input     │  │   ◄───  │  │     Pre-fetch relevant context           │  │\n│  │     or proactive suggestions     │  │  suggest│  │     Prepare recommendations              │  │\n│  └──────────────────────────────────┘  │         │  │     Update todolist autonomously         │  │\n│                                        │         │  └──────────────────────────────────────────┘  │\n└────────────────────────────────────────┘         └────────────────────────────────────────────────┘\n                 │                                                           │\n                 └───────────────────────────┬───────────────────────────────┘\n                                             ▼\n                              ┌──────────────────────────────┐\n                              │     CONTINUOUS SYNC LOOP     │\n                              │  Agent ◄──► MemU Bot ◄──► DB │\n                              └──────────────────────────────┘\n```\n\n---\n\n## 🎯 Proactive Use Cases\n\n### 1. **Information Recommendation**\n*Agent monitors interests and proactively surfaces relevant content*\n```python\n# User has been researching AI topics\nMemU tracks: reading history, saved articles, search queries\n\n# When new content arrives:\nAgent: \"I found 3 new papers on RAG optimization that align with\n        your recent research on retrieval systems. One author\n        (Dr. Chen) you've cited before published yesterday.\"\n\n# Proactive behaviors:\n- Learns topic preferences from browsing patterns\n- Tracks author\u002Fsource credibility preferences\n- Filters noise based on engagement history\n- Times recommendations for optimal attention\n```\n\n### 2. **Email Management**\n*Agent learns communication patterns and handles routine correspondence*\n```python\n# MemU observes email patterns over time:\n- Response templates for common scenarios\n- Priority contacts and urgent keywords\n- Scheduling preferences and availability\n- Writing style and tone variations\n\n# Proactive email assistance:\nAgent: \"You have 12 new emails. I've drafted responses for 3 routine\n        requests and flagged 2 urgent items from your priority contacts.\n        Should I also reschedule tomorrow's meeting based on the\n        conflict John mentioned?\"\n\n# Autonomous actions:\n✓ Draft context-aware replies\n✓ Categorize and prioritize inbox\n✓ Detect scheduling conflicts\n✓ Summarize long threads with key decisions\n```\n\n### 3. **Trading & Financial Monitoring**\n*Agent tracks market context and user investment behavior*\n```python\n# MemU learns trading preferences:\n- Risk tolerance from historical decisions\n- Preferred sectors and asset classes\n- Response patterns to market events\n- Portfolio rebalancing triggers\n\n# Proactive alerts:\nAgent: \"NVDA dropped 5% in after-hours trading. Based on your past\n        behavior, you typically buy tech dips above 3%. Your current\n        allocation allows for $2,000 additional exposure while\n        maintaining your 70\u002F30 equity-bond target.\"\n\n# Continuous monitoring:\n- Track price alerts tied to user-defined thresholds\n- Correlate news events with portfolio impact\n- Learn from executed vs. ignored recommendations\n- Anticipate tax-loss harvesting opportunities\n```\n\n\n...\n\n---\n\n## 🗂️ Hierarchical Memory Architecture\n\nMemU's three-layer system enables both **reactive queries** and **proactive context loading**:\n\n\u003Cimg width=\"100%\" alt=\"structure\" src=\"assets\u002Fstructure.png\" \u002F>\n\n| Layer | Reactive Use | Proactive Use |\n|-------|--------------|---------------|\n| **Resource** | Direct access to original data | Background monitoring for new patterns |\n| **Item** | Targeted fact retrieval | Real-time extraction from ongoing interactions |\n| **Category** | Summary-level overview | Automatic context assembly for anticipation |\n\n**Proactive Benefits:**\n- **Auto-categorization**: New memories self-organize into topics\n- **Pattern Detection**: System identifies recurring themes\n- **Context Prediction**: Anticipates what information will be needed next\n\n---\n\n## 🚀 Quick Start\n\n### Option 1: Cloud Version\n\nExperience proactive memory instantly:\n\n👉 **[memu.so](https:\u002F\u002Fmemu.so)** - Hosted service with 7×24 continuous learning\n\nFor enterprise deployment with custom proactive workflows, contact **info@nevamind.ai**\n\n#### Cloud API (v3)\n\n| Base URL | `https:\u002F\u002Fapi.memu.so` |\n|----------|----------------------|\n| Auth | `Authorization: Bearer YOUR_API_KEY` |\n\n| Method | Endpoint | Description |\n|--------|----------|-------------|\n| `POST` | `\u002Fapi\u002Fv3\u002Fmemory\u002Fmemorize` | Register continuous learning task |\n| `GET` | `\u002Fapi\u002Fv3\u002Fmemory\u002Fmemorize\u002Fstatus\u002F{task_id}` | Check real-time processing status |\n| `POST` | `\u002Fapi\u002Fv3\u002Fmemory\u002Fcategories` | List auto-generated categories |\n| `POST` | `\u002Fapi\u002Fv3\u002Fmemory\u002Fretrieve` | Query memory (supports proactive context loading) |\n\n📚 **[Full API Documentation](https:\u002F\u002Fmemu.pro\u002Fdocs#cloud-version)**\n\n---\n\n### Option 2: Self-Hosted\n\n#### Installation\n```bash\npip install -e .\n```\n\n#### Basic Example\n\n> **Requirements**: Python 3.13+ and an OpenAI API key\n\n**Test Continuous Learning** (in-memory):\n```bash\nexport OPENAI_API_KEY=your_api_key\ncd tests\npython test_inmemory.py\n```\n\n**Test with Persistent Storage** (PostgreSQL):\n```bash\n# Start PostgreSQL with pgvector\ndocker run -d \\\n  --name memu-postgres \\\n  -e POSTGRES_USER=postgres \\\n  -e POSTGRES_PASSWORD=postgres \\\n  -e POSTGRES_DB=memu \\\n  -p 5432:5432 \\\n  pgvector\u002Fpgvector:pg16\n\n# Run continuous learning test\nexport OPENAI_API_KEY=your_api_key\ncd tests\npython test_postgres.py\n```\n\nBoth examples demonstrate **proactive memory workflows**:\n1. **Continuous Ingestion**: Process multiple files sequentially\n2. **Auto-Extraction**: Immediate memory creation\n3. **Proactive Retrieval**: Context-aware memory surfacing\n\nSee [`tests\u002Ftest_inmemory.py`](tests\u002Ftest_inmemory.py) and [`tests\u002Ftest_postgres.py`](tests\u002Ftest_postgres.py) for implementation details.\n\n---\n\n### Custom LLM and Embedding Providers\n\nMemU supports custom LLM and embedding providers beyond OpenAI. Configure them via `llm_profiles`:\n```python\nfrom memu import MemUService\n\nservice = MemUService(\n    llm_profiles={\n        # Default profile for LLM operations\n        \"default\": {\n            \"base_url\": \"https:\u002F\u002Fdashscope.aliyuncs.com\u002Fcompatible-mode\u002Fv1\",\n            \"api_key\": \"your_api_key\",\n            \"chat_model\": \"qwen3-max\",\n            \"client_backend\": \"sdk\"  # \"sdk\" or \"http\"\n        },\n        # Separate profile for embeddings\n        \"embedding\": {\n            \"base_url\": \"https:\u002F\u002Fapi.voyageai.com\u002Fv1\",\n            \"api_key\": \"your_voyage_api_key\",\n            \"embed_model\": \"voyage-3.5-lite\"\n        }\n    },\n    # ... other configuration\n)\n```\n\n---\n\n### OpenRouter Integration\n\nMemU supports [OpenRouter](https:\u002F\u002Fopenrouter.ai) as a model provider, giving you access to multiple LLM providers through a single API.\n\n#### Configuration\n```python\nfrom memu import MemoryService\n\nservice = MemoryService(\n    llm_profiles={\n        \"default\": {\n            \"provider\": \"openrouter\",\n            \"client_backend\": \"httpx\",\n            \"base_url\": \"https:\u002F\u002Fopenrouter.ai\",\n            \"api_key\": \"your_openrouter_api_key\",\n            \"chat_model\": \"anthropic\u002Fclaude-3.5-sonnet\",  # Any OpenRouter model\n            \"embed_model\": \"openai\u002Ftext-embedding-3-small\",  # Embedding model\n        },\n    },\n    database_config={\n        \"metadata_store\": {\"provider\": \"inmemory\"},\n    },\n)\n```\n\n#### Environment Variables\n\n| Variable | Description |\n|----------|-------------|\n| `OPENROUTER_API_KEY` | Your OpenRouter API key from [openrouter.ai\u002Fkeys](https:\u002F\u002Fopenrouter.ai\u002Fkeys) |\n\n#### Supported Features\n\n| Feature | Status | Notes |\n|---------|--------|-------|\n| Chat Completions | Supported | Works with any OpenRouter chat model |\n| Embeddings | Supported | Use OpenAI embedding models via OpenRouter |\n| Vision | Supported | Use vision-capable models (e.g., `openai\u002Fgpt-4o`) |\n\n#### Running OpenRouter Tests\n```bash\nexport OPENROUTER_API_KEY=your_api_key\n\n# Full workflow test (memorize + retrieve)\npython tests\u002Ftest_openrouter.py\n\n# Embedding-specific tests\npython tests\u002Ftest_openrouter_embedding.py\n\n# Vision-specific tests\npython tests\u002Ftest_openrouter_vision.py\n```\n\nSee [`examples\u002Fexample_4_openrouter_memory.py`](examples\u002Fexample_4_openrouter_memory.py) for a complete working example.\n\n---\n\n## 📖 Core APIs\n\n### `memorize()` - Continuous Learning Pipeline\n\nProcesses inputs in real-time and immediately updates memory:\n\n\u003Cimg width=\"100%\" alt=\"memorize\" src=\"assets\u002Fmemorize.png\" \u002F>\n\n```python\nresult = await service.memorize(\n    resource_url=\"path\u002Fto\u002Ffile.json\",  # File path or URL\n    modality=\"conversation\",            # conversation | document | image | video | audio\n    user={\"user_id\": \"123\"}             # Optional: scope to a user\n)\n\n# Returns immediately with extracted memory:\n{\n    \"resource\": {...},      # Stored resource metadata\n    \"items\": [...],         # Extracted memory items (available instantly)\n    \"categories\": [...]     # Auto-updated category structure\n}\n```\n\n**Proactive Features:**\n- Zero-delay processing—memories available immediately\n- Automatic categorization without manual tagging\n- Cross-reference with existing memories for pattern detection\n\n### `retrieve()` - Dual-Mode Intelligence\n\nMemU supports both **proactive context loading** and **reactive querying**:\n\n\u003Cimg width=\"100%\" alt=\"retrieve\" src=\"assets\u002Fretrieve.png\" \u002F>\n\n#### RAG-based Retrieval (`method=\"rag\"`)\n\nFast **proactive context assembly** using embeddings:\n\n- ✅ **Instant context**: Sub-second memory surfacing\n- ✅ **Background monitoring**: Can run continuously without LLM costs\n- ✅ **Similarity scoring**: Identifies most relevant memories automatically\n\n#### LLM-based Retrieval (`method=\"llm\"`)\n\nDeep **anticipatory reasoning** for complex contexts:\n\n- ✅ **Intent prediction**: LLM infers what user needs before they ask\n- ✅ **Query evolution**: Automatically refines search as context develops\n- ✅ **Early termination**: Stops when sufficient context is gathered\n\n#### Comparison\n\n| Aspect | RAG (Fast Context) | LLM (Deep Reasoning) |\n|--------|-------------------|---------------------|\n| **Speed** | ⚡ Milliseconds | 🐢 Seconds |\n| **Cost** | 💰 Embedding only | 💰💰 LLM inference |\n| **Proactive use** | Continuous monitoring | Triggered context loading |\n| **Best for** | Real-time suggestions | Complex anticipation |\n\n#### Usage\n```python\n# Proactive retrieval with context history\nresult = await service.retrieve(\n    queries=[\n        {\"role\": \"user\", \"content\": {\"text\": \"What are their preferences?\"}},\n        {\"role\": \"user\", \"content\": {\"text\": \"Tell me about work habits\"}}\n    ],\n    where={\"user_id\": \"123\"},  # Optional: scope filter\n    method=\"rag\"  # or \"llm\" for deeper reasoning\n)\n\n# Returns context-aware results:\n{\n    \"categories\": [...],     # Relevant topic areas (auto-prioritized)\n    \"items\": [...],          # Specific memory facts\n    \"resources\": [...],      # Original sources for traceability\n    \"next_step_query\": \"...\" # Predicted follow-up context\n}\n```\n\n**Proactive Filtering**: Use `where` to scope continuous monitoring:\n- `where={\"user_id\": \"123\"}` - User-specific context\n- `where={\"agent_id__in\": [\"1\", \"2\"]}` - Multi-agent coordination\n- Omit `where` for global context awareness\n\n---\n\n## 💡 Proactive Scenarios\n\n### Example 1: Always-Learning Assistant\n\nContinuously learns from every interaction without explicit memory commands:\n```bash\nexport OPENAI_API_KEY=your_api_key\npython examples\u002Fexample_1_conversation_memory.py\n```\n\n**Proactive Behavior:**\n- Automatically extracts preferences from casual mentions\n- Builds relationship models from interaction patterns\n- Surfaces relevant context in future conversations\n- Adapts communication style based on learned preferences\n\n**Best for:** Personal AI assistants, customer support that remembers, social chatbots\n\n---\n\n### Example 2: Self-Improving Agent\n\nLearns from execution logs and proactively suggests optimizations:\n```bash\nexport OPENAI_API_KEY=your_api_key\npython examples\u002Fexample_2_skill_extraction.py\n```\n\n**Proactive Behavior:**\n- Monitors agent actions and outcomes continuously\n- Identifies patterns in successes and failures\n- Auto-generates skill guides from experience\n- Proactively suggests strategies for similar future tasks\n\n**Best for:** DevOps automation, agent self-improvement, knowledge capture\n\n---\n\n### Example 3: Multimodal Context Builder\n\nUnifies memory across different input types for comprehensive context:\n```bash\nexport OPENAI_API_KEY=your_api_key\npython examples\u002Fexample_3_multimodal_memory.py\n```\n\n**Proactive Behavior:**\n- Cross-references text, images, and documents automatically\n- Builds unified understanding across modalities\n- Surfaces visual context when discussing related topics\n- Anticipates information needs by combining multiple sources\n\n**Best for:** Documentation systems, learning platforms, research assistants\n\n---\n\n## 📊 Performance\n\nMemU achieves **92.09% average accuracy** on the Locomo benchmark across all reasoning tasks, demonstrating reliable proactive memory operations.\n\n\u003Cimg width=\"100%\" alt=\"benchmark\" src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F6fec4884-94e5-4058-ad5c-baac3d7e76d9\" \u002F>\n\nView detailed experimental data: [memU-experiment](https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemU-experiment)\n\n---\n\n## 🧩 Ecosystem\n\n| Repository | Description | Proactive Features |\n|------------|-------------|-------------------|\n| **[memU](https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemU)** | Core proactive memory engine | 7×24 learning pipeline, auto-categorization |\n| **[memU-server](https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemU-server)** | Backend with continuous sync | Real-time memory updates, webhook triggers |\n| **[memU-ui](https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemU-ui)** | Visual memory dashboard | Live memory evolution monitoring |\n\n**Quick Links:**\n- 🚀 [Try MemU Cloud](https:\u002F\u002Fapp.memu.so\u002Fquick-start)\n- 📚 [API Documentation](https:\u002F\u002Fmemu.pro\u002Fdocs)\n- 💬 [Discord Community](https:\u002F\u002Fdiscord.com\u002Finvite\u002FhQZntfGsbJ)\n\n---\n\n## 🤝 Partners\n\n\u003Cdiv align=\"center\">\n\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FTEN-framework\u002Ften-framework\">\u003Cimg src=\"https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F113095513?s=200&v=4\" alt=\"Ten\" height=\"40\" style=\"margin: 10px;\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fopenagents.org\">\u003Cimg src=\"assets\u002Fpartners\u002Fopenagents.png\" alt=\"OpenAgents\" height=\"40\" style=\"margin: 10px;\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fmilvus-io\u002Fmilvus\">\u003Cimg src=\"https:\u002F\u002Fmiro.medium.com\u002Fv2\u002Fresize:fit:2400\u002F1*-VEGyAgcIBD62XtZWavy8w.png\" alt=\"Milvus\" height=\"40\" style=\"margin: 10px;\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fxroute.ai\u002F\">\u003Cimg src=\"assets\u002Fpartners\u002Fxroute.png\" alt=\"xRoute\" height=\"40\" style=\"margin: 10px;\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fjaaz.app\u002F\">\u003Cimg src=\"assets\u002Fpartners\u002Fjazz.png\" alt=\"Jazz\" height=\"40\" style=\"margin: 10px;\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FBuddie-AI\u002FBuddie\">\u003Cimg src=\"assets\u002Fpartners\u002Fbuddie.png\" alt=\"Buddie\" height=\"40\" style=\"margin: 10px;\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fbytebase\u002Fbytebase\">\u003Cimg src=\"assets\u002Fpartners\u002Fbytebase.png\" alt=\"Bytebase\" height=\"40\" style=\"margin: 10px;\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FLazyAGI\u002FLazyLLM\">\u003Cimg src=\"assets\u002Fpartners\u002FLazyLLM.png\" alt=\"LazyLLM\" height=\"40\" style=\"margin: 10px;\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fclawdchat.ai\u002F\">\u003Cimg src=\"assets\u002Fpartners\u002FClawdchat.png\" alt=\"Clawdchat\" height=\"40\" style=\"margin: 10px;\">\u003C\u002Fa>\n\n\u003C\u002Fdiv>\n\n---\n\n## 🤝 How to Contribute\n\nWe welcome contributions from the community! Whether you're fixing bugs, adding features, or improving documentation, your help is appreciated.\n\n### Getting Started\n\nTo start contributing to MemU, you'll need to set up your development environment:\n\n#### Prerequisites\n- Python 3.13+\n- [uv](https:\u002F\u002Fgithub.com\u002Fastral-sh\u002Fuv) (Python package manager)\n- Git\n\n#### Setup Development Environment\n```bash\n# 1. Fork and clone the repository\ngit clone https:\u002F\u002Fgithub.com\u002FYOUR_USERNAME\u002FmemU.git\ncd memU\n\n# 2. Install development dependencies\nmake install\n```\n\nThe `make install` command will:\n- Create a virtual environment using `uv`\n- Install all project dependencies\n- Set up pre-commit hooks for code quality checks\n\n#### Running Quality Checks\n\nBefore submitting your contribution, ensure your code passes all quality checks:\n```bash\nmake check\n```\n\nThe `make check` command runs:\n- **Lock file verification**: Ensures `pyproject.toml` consistency\n- **Pre-commit hooks**: Lints code with Ruff, formats with Black\n- **Type checking**: Runs `mypy` for static type analysis\n- **Dependency analysis**: Uses `deptry` to find obsolete dependencies\n\n### Contributing Guidelines\n\nFor detailed contribution guidelines, code standards, and development practices, please see [CONTRIBUTING.md](CONTRIBUTING.md).\n\n**Quick tips:**\n- Create a new branch for each feature or bug fix\n- Write clear commit messages\n- Add tests for new functionality\n- Update documentation as needed\n- Run `make check` before pushing\n\n---\n\n## 📄 License\n\n[Apache License 2.0](LICENSE.txt)\n\n---\n\n## 🌍 Community\n\n- **GitHub Issues**: [Report bugs & request features](https:\u002F\u002Fgithub.com\u002FNevaMind-AI\u002FmemU\u002Fissues)\n- **Discord**: [Join the community](https:\u002F\u002Fdiscord.com\u002Finvite\u002FhQZntfGsbJ)\n- **X (Twitter)**: [Follow @memU_ai](https:\u002F\u002Fx.com\u002FmemU_ai)\n- **Contact**: info@nevamind.ai\n\n---\n\n\u003Cdiv align=\"center\">\n\n⭐ **Star us on GitHub** to get notified about new releases!\n\n\u003C\u002Fdiv>\n","memU 是一个专为24\u002F7主动型AI代理设计的记忆框架。其核心功能包括显著降低保持代理在线所需的LLM令牌成本，持续捕捉并理解用户意图，即使在没有明确指令的情况下也能预测并执行用户可能的操作。技术特点上，memU将记忆视为文件系统，结构化、层次化且即时可访问，支持类别自动组织、记忆项提取及交叉引用等。适用于需要长期运行、能够自主学习和适应的智能代理场景，如客户服务、个人助理等。",2,"2026-06-11 03:39:28","high_star"]