EverMem Story Memory Demo

Built on EverMemOS - Open-source AI memory infrastructure

A demonstration web application showcasing EverMem's AI memory infrastructure through an interactive Q&A experience with "A Game of Thrones" (Book 1).

Ask questions about the book and watch two AI responses stream side-by-side: one with memory using EverMem to retrieve relevant passages, and one without memory using only the LLM's training data. See the difference memory makes.

Features

• Side-by-Side Comparison: Watch two responses stream simultaneously - with and without memory context
• Memory-Grounded Responses: See exactly which book passages are used to answer questions
• Real-time Streaming: Token-by-token AI response streaming via SSE
• Interactive Memory Chips: Hover over memory chips to see full excerpt details and metadata
• Follow-up Suggestions: AI-generated follow-up questions after each response
• Dark Theme UI: Modern, clean interface inspired by EverMind's design

Tech Stack

• Frontend: React 18 + TypeScript + Vite
• Backend: Node.js + Express + Bun
• AI: Claude Haiku (via OpenRouter)
• Memory: EverMind Cloud API

Quick Start

Prerequisites

• Bun (latest version)
• OpenAI API key (or OpenRouter API key)
• EverMind Cloud API key (apply at EverMemOS Cloud)

Installation

# Clone the repository
git clone <repository-url>
cd evermem-story-demo

# Install dependencies
bun install

Configuration

Backend (backend/.env):

cp backend/.env.example backend/.env

Edit backend/.env:

OPENAI_API_KEY=your-openrouter-api-key
OPENAI_MODEL=anthropic/claude-3-haiku
PORT=3001
FRONTEND_URL=http://localhost:3000

# EverMind Cloud
USE_EVERMEMOS=true
EVERMEMOS_URL=https://api.evermind.ai
EVERMEMOS_API_KEY=your-evermind-api-key

Frontend (frontend/.env):

cp frontend/.env.example frontend/.env

The default VITE_API_URL=http://localhost:3001 should work for local development.

Running

# Start both frontend and backend
bun run dev

• Frontend: http://localhost:3000
• Backend: http://localhost:3001

Loading Novel Content

Before using the demo, you need to load novel content into EverMind Cloud.

Quick Test with Sample

A sample file with 5 chapters is included for testing:

bun run load-novel-cloud \
  --file sample/got-sample.txt \
  --book-title "A Game of Thrones" \
  --book-abbrev "got" \
  --api-key YOUR_EVERMIND_API_KEY

Full Book

For the complete experience, obtain the full novel text file and load it:

bun run load-novel-cloud \
  --file path/to/got.txt \
  --book-title "A Game of Thrones" \
  --book-abbrev "got" \
  --api-key YOUR_EVERMIND_API_KEY

The script:

• Detects chapter boundaries automatically (PROLOGUE, character names in caps)
• Splits text into paragraphs
• Uploads to EverMind Cloud with metadata
• Supports resumption if interrupted

Project Structure

├── frontend/                # React frontend
│   ├── src/
│   │   ├── components/     # UI components
│   │   ├── hooks/          # Custom React hooks
│   │   ├── services/       # API client
│   │   └── types/          # TypeScript types
│   └── public/             # Static assets
├── backend/                 # Express backend
│   ├── src/
│   │   ├── routes/         # API endpoints
│   │   ├── services/       # Business logic
│   │   └── utils/          # Utilities
├── scripts/                 # CLI tools
│   ├── load-novel-cloud.ts # Load novel to EverMind
│   └── clear-memories-cloud.ts
└── sample/                  # Sample data for testing
    └── got-sample.txt      # 5 chapters from Book 1

Development

# Start dev servers
bun run dev

# Frontend only
bun run dev:frontend

# Backend only
bun run dev:backend

# Type check
bun run type-check

# Lint
bun run lint

License

MIT

Acknowledgments

• EverMind for the memory infrastructure
• George R.R. Martin for "A Song of Ice and Fire"