MemoryMesh

A context and memory management system for business AI agents that stores business memories, assigns dynamic trust scores that decay over time, and surfaces contextually relevant memories for business decisions.

Features

• Entity Management: Create and manage business entities (suppliers, customers, warehouses, people)
• Memory Storage: Store business memories with financial impact, volatility classification, and tags
• Dynamic Trust Scoring: Trust scores decay over time based on volatility class and recency
• Trajectory Detection: Analyze entity performance trends (IMPROVING, DECLINING, VOLATILE, STABLE)
• Contextual Decision Support: Get relevant memories based on decision type and context budget

Tech Stack

• Backend: Node.js, Express, MongoDB Atlas, Mongoose
• Frontend: React, Tailwind CSS, Axios
• Trust Engine: Custom mathematical models for trust scoring and trajectory analysis

Getting Started

Prerequisites

• Node.js (v14 or higher)
• MongoDB Atlas account
• Git

Installation

1. Clone the repository
2. Navigate to the project directory
3. Install dependencies for both server and client

# Install root dependencies
npm install

# Install server dependencies
cd server && npm install

# Install client dependencies
cd ../client && npm install

Environment Configuration

1. Create a .env file in the root directory
2. Add your MongoDB Atlas connection string:

MONGO_URI=mongodb+srv://username:password@cluster.mongodb.net/memorymesh?retryWrites=true&w=majority
PORT=5001

Database Seeding

Run the seed script to populate the database with sample data:

npm run seed

This will create:

• XYZ Corporation (supplier) with quality incidents, delivery issues, and recent improvements
• TechCorp Inc. (customer) with contract renewals, technical issues, and leadership preferences

Running the Application

Start both the server and client concurrently:

npm run dev

Or run them separately:

# Start backend server (port 5001)
npm run server

# Start frontend client (port 3000)
npm run client

API Endpoints

Entities

• POST /api/entities - Create a new entity
• GET /api/entities - List all entities

Memories

• POST /api/memories - Add memory to an entity
• GET /api/memories/:entityId - Get all memories for an entity

Context Engine

• POST /api/context - Get contextual memories for decision making
• GET /api/context/dashboard - Get dashboard statistics

Trust Score Calculation

The trust score is calculated using:

trust_score = base_weight × recency_decay × consistency_factor × source_reliability

• Base Weight: Normalized financial impact (0-1)
• Recency Decay: Exponential decay based on volatility class
  • STRUCTURAL: λ = 0.0019 (half-life ~365 days)
  • BEHAVIORAL: λ = 0.0077 (half-life ~90 days)
  • ENVIRONMENTAL: λ = 0.0231 (half-life ~30 days)
• Consistency Factor: Sentiment analysis against recent memories
• Source Reliability: Reliability score of the information source

Trajectory Detection

Entity trajectories are determined by comparing average trust scores:

• IMPROVING: Recent trust > Previous trust + 0.1
• DECLINING: Recent trust < Previous trust - 0.1
• VOLATILE: High variance in trust scores
• STABLE: Neither improving nor declining significantly

Project Structure

memorymesh/
├── server/
│   ├── models/          # Mongoose schemas
│   ├── services/        # Business logic (trust engine)
│   ├── routes/          # API routes
│   ├── seed.js          # Database seeding
│   └── index.js         # Express server
├── client/
│   ├── src/
│   │   ├── components/  # React components
│   │   ├── pages/       # Page components
│   │   └── api/         # API client
│   └── package.json
├── .env                 # Environment variables
└── package.json         # Root package.json

Usage

1. View Entities: Browse all entities in the left panel
2. Explore Memories: Click an entity to view its memory timeline
3. Get Context: Use the decision briefing panel to get relevant memories for specific decisions
4. Trust Analysis: View trust scores and trajectory indicators for each entity

License

MIT License