Overview

MyMonad is a protocol for privacy-preserving similarity-based discovery, with a reference implementation for human matchmaking. This page covers the architecture of both.

Components

mymonad-ingest

The data ingestion daemon. Watches local directories for document changes and builds your Monad vector.

Responsibilities:

• File system watching (fsnotify)
• Content extraction from supported formats
• Embedding generation via Ollama
• Running average vector updates
• Encrypted Monad storage

mymonad-agent

The P2P networking daemon. Handles peer discovery, connections, and the handshake protocol.

Responsibilities:

• libp2p host management
• Kademlia DHT participation
• mDNS local discovery
• LSH-based peer discovery (commit-reveal)
• Handshake protocol state machine
• Adaptive anti-spam difficulty
• ZK proof exchange (optional)
• Encrypted chat messaging

mymonad-cli

Command-line interface for interacting with the daemons.

Responsibilities:

• Status queries
• Peer management
• Identity display
• Manual bootstrapping

Data Flow

Documents → Ingest → Embeddings → Monad Vector
                                      ↓
                               LSH Signatures
                                      ↓
                            Agent (P2P Network)
                                      ↓
                           Discovery (DHT + mDNS)
                                      ↓
                            5-Stage Handshake
                                      ↓
                            Compatible Match

Design Principles

1. Privacy First — Raw data never leaves your device
2. Decentralized — No central servers or authorities
3. Progressive Trust — Reveal information only as trust is established
4. Cryptographic Verification — Math replaces blind trust
5. User Sovereignty — You own your keys, your data, your connections

Protocol vs Application

Key Packages