Skip to content

Bryan-cmf/memory-hub

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🧠 MemoryHub — Persistent Memory Enhancement System

Vector semantic memory + Multi-platform auto-capture + Decade-scale lifecycle Version: 2.0.0 | License: MIT

One-Liner

Cure Agent amnesia. A fourth-layer enhancement to existing memory systems — vector search + session scanning + decade-scale lifecycle.

Quick Start

# One-command install
pip install memory-hub

# Interactive setup
memory-hub setup

# Start daemon
memory-hub start

# Open dashboard
open http://localhost:3872

Why MemoryHub?

Problem Current MemoryHub Solution
Agent forgets daily log High task density Auto-capture from 4 platforms
Cross-session amnesia Blank slate each session Vector search across time
Non-main sessions ignored WhatsApp etc. no rules Daemon scans all platforms
Memories fade over time Old buried by new Consolidate, never delete

Architecture

AI Platforms (OpenClaw / Hermes / DeepSeek / Claude)
        │                          │
        │ MODE B: File scan        │ MODE A: MCP real-time
        ▼                          ▼
   Capture Daemon ──────── Qdrant Vector DB
        │                    (localhost:6333)
        │
   Dashboard (:3872)    MCP Server (6 tools)

Commands

memory-hub setup       # Interactive installation wizard
memory-hub start       # Start the capture daemon
memory-hub status      # Show system health
memory-hub stop        # Stop the daemon
memory-hub verify      # Run verification checks
memory-hub backup      # Run backup

Supported Platforms

Platform Capture Mode Format
🦙 OpenClaw Auto-scan + MCP JSONL
🪽 Hermes Auto-scan + MCP JSONL
🐋 DeepSeek TUI Auto-scan + MCP JSON
🦫 Claude Code Auto-scan + MCP JSONL

Backends

Backend Required Install Purpose
💾 File System Built-in Source of Truth
🗄️ SQLite Built-in Metadata + full-text index
🧠 Qdrant Docker Vector search (primary)
📦 Chroma Optional pip Lightweight vector (no Docker)
🪶 LanceDB Optional pip Embedded vector (lightest)

Key Features

  • 🔍 Semantic search — "last month's DD report" not grep
  • 📡 4-Platform capture — OpenClaw / Hermes / DeepSeek / Claude
  • 🧠 Vector memory — Qdrant + BGE-m3, local inference
  • 🛡️ 3-tier backup — hourly / daily / weekly
  • 📊 Live Dashboard:3872 real-time feed + charts + search
  • 🔌 MCP Integration — 6 tools, Mode A real-time capture
  • 🌐 Decade memory — daily → weekly → monthly → yearly consolidation

Documentation

Guide File
Quick Start QUICKSTART.md
Installation INSTALL.md
User Guide docs/USER_GUIDE.md
Architecture docs/ARCHITECTURE.md
MCP Setup docs/MCP_SETUP.md
API Reference API.md
Troubleshooting docs/TROUBLESHOOTING.md

License

MIT © Bryan Chan

About

No description, website, or topics provided.

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages