greeum 3.1.1rc6

Context-Dependent Neural Memory System for LLMs with GraphIndex beam search, STM/LTM consolidation, and semantic tagging

Greeum

AI conversations that remember everything. No more repeating context every time.

⚡ Quick Start

# Install
pip install greeum

# Add your first memory
greeum memory add "Started working on the new dashboard project"

# Set memory anchors for quick access
greeum anchors set A 123  # Pin important memory to slot A

# Search with anchor-based localization
greeum memory search "dashboard project" --slot A --radius 2

That's it. Your AI now remembers.

✨ What It Does

🧠 Remembers context - AI recalls previous conversations and decisions ⚡ 280x faster search - Checkpoint-based memory retrieval 🔄 Works with any AI - GPT, Claude, or your custom model 🛡️ Your data stays yours - Local storage, no cloud required

🔧 Installation

Basic Setup

pip install greeum

With All Features

pip install greeum[all]  # includes vector search, embeddings

For Claude Code Users

# Install and start MCP server
pip install greeum
greeum mcp serve

For Codex/OpenAI MCP Clients

pip install greeum
greeum mcp serve -t http --host 0.0.0.0 --port 8800
# Register http://127.0.0.1:8800/mcp as the MCP endpoint

📝 Usage

Adding Memories

# Add important context
greeum memory add "Client prefers minimal UI design"

# Add with expiration
greeum stm add "Working on login page today" --ttl 24h

Memory Anchors (v2.2.5+)

# View current anchor status
greeum anchors status

# Set anchors for quick access
greeum anchors set A 123     # Pin memory #123 to slot A
greeum anchors set B 456     # Pin memory #456 to slot B

# Search near anchored memories
greeum memory search "UI design" --slot A --radius 3

# Pin/unpin anchors
greeum anchors pin A         # Prevent auto-movement
greeum anchors unpin A       # Allow auto-movement

# Clear all anchors
greeum anchors clear

Searching

# Find relevant memories
greeum memory search "UI design preferences" --count 5

# Anchor-based localized search (faster)
greeum memory search "login" --slot B --radius 2 --fallback

# Global search (traditional)
greeum memory search "login" --count 10

Python API

from greeum import BlockManager, DatabaseManager

# Initialize
db_manager = DatabaseManager()
memory = BlockManager(db_manager)

# Add block to long-term memory
block = memory.add_block(
    context="User wants dark mode toggle",
    keywords=["dark", "mode", "toggle"],
    tags=["ui", "preference"],
    embedding=[],  # Auto-generated if empty
    importance=0.7
)

# Search memories
results = memory.search_memories("dark mode", limit=3)

🤖 Claude Integration

Setup MCP Server

Add to your Claude Desktop config:

{
  "mcpServers": {
    "greeum": {
      "command": "greeum",
      "args": ["mcp", "serve"],
      "env": {
        "GREEUM_DATA_DIR": "/path/to/your/data"
      }
    }
  }
}

Available Tools

• add_memory - Store important context
• search_memory - Find relevant memories
• get_memory_stats - View memory statistics

📚 Documentation

• Getting Started - Installation and first steps
• API Reference - Complete API documentation
• MCP Integration - Claude Code setup

🏗️ Architecture

Your Input → Working Memory → Cache → Checkpoints → Long-term Storage
             0.04ms          0.08ms   0.7ms        Permanent

Four-layer memory system optimized for speed and relevance.

📋 Version Updates

v2.1.1 (2025-08)

• Enhanced search with temporal boost for recent information prioritization
• Optimized codebase with 955 lines of code reduction and improved test architecture
• Resolved import dependencies and improved memory management
• Added intelligent date keyword detection for search result ranking
• Improved test stability with BaseGreeumTestCase standardization
• Performance optimizations with minimal overhead (+1.0%)

🤝 Contributing

1. Fork the repository
2. Create your feature branch
3. Submit a pull request

See CONTRIBUTING.md for details.

📄 License

MIT License - see LICENSE file.

---

Greeum - Memory for AI that actually works. Made with ❤️ by the open source community.