Memoire-cli

An AI-powered terminal chat assistant with persistent memory - it actually remembers you.

Memoire (French: memory) is a lightweight CLI chat app that connects Google's Gemini models with Mem0's long-term memory layer. Every conversation is stored and retrieved semantically, so the AI can reference what you told it last week without you having to repeat yourself.

---

Demo

---

Features

• Streaming responses - text prints word-by-word as the model generates it, no waiting for a full response
• Persistent memory - powered by Mem0, past conversations are stored and injected as context automatically
• Graceful memory fallback - skip the MEM0_API_KEY and the assistant works fine, just without cross-session memory
• Minimal commands - /help, /clear, /exit - stays out of the way
• Environment-first config - everything lives in a .env file, no flags to memorize

---

How it works

You type a message
       │
       ▼
  Mem0 searches past memories for relevant context
       │
       ▼
  System prompt assembled: base persona + retrieved memories
       │
       ▼
  Gemini streams a response token-by-token → printed to terminal
       │
       ▼
  Full conversation turn stored back into Mem0

The USER_ID is set to "myself" by default in - change it to your name or a unique identifier in .env if you plan to run multiple users against the same Mem0 account.

---

Tech stack

Layer	Library	Purpose
LLM inference	@ai-sdk/google + ai (Vercel AI SDK)	Stream responses from Gemini models
Memory	mem0ai	Store and semantically search conversation history
Config	dotenv	Load environment variables from .env
UX	ora	Terminal spinner while waiting for responses
Runtime	Node.js + TypeScript	-

---

Prerequisites

• Node.js 18+
• A Google AI Studio API key
• (Optional) A Mem0 API key for persistent memory

---

Setup

1. Clone the repo

git clone https://github.com/rrahul-1/Memoire-cli.git
cd Memoire-cli

2. Install dependencies

npm install

3. Create a .env file

cp .env.example .env   # or create it manually

.env contents:

# Required - the Gemini model to use
LLM_MODEL=gemini-2.5-flash-lite

# Required - your Google Generative AI key
GOOGLE_GENERATIVE_AI_API_KEY=your_google_api_key_here

# Optional - enables persistent memory across sessions
MEM0_API_KEY=your_mem0_api_key_here

# Optional - scopes your memories in Mem0 (defaults to "myself")
USER_ID=your_name

---

Environment variables

Variable	Required	Description
LLM_MODEL	Yes	Gemini model identifier, e.g. gemini-2.5-flash-lite, gemini-2.0-flash
GOOGLE_GENERATIVE_AI_API_KEY	Yes	API key from Google AI Studio
MEM0_API_KEY	No	API key from Mem0. If omitted, memory is disabled but the assistant still works.
USER_ID	No	Identifier used to scope memories in Mem0. Defaults to "myself". Set this if you share a Mem0 account across multiple people or projects.

Missing LLM_MODEL causes the app to exit immediately with a clear error message listing which variables are absent.

---

Running

Development (TypeScript, no build step)

npm run dev

Production (build first, then run)

npm run build
npm start

---

Usage

Once started, you'll see:

    Welcome to memoire-cli ✨
    Your AI-powered memory + chat assistant.

    Memory: ✓ enabled

    What's on your mind today?

    Use /help to see available commands.

 You  >  

Type anything to start chatting. The AI's response streams directly into the terminal.

Commands

Command	Description
/help	Show the command guide
/clear	Clear the terminal screen
/memory	Print all memories stored for your USER_ID
/clear-memory	Permanently wipe all stored memories for your USER_ID
/exit	Quit the application (also works with Ctrl+C)

Unrecognized /commands display an error and remind you to use /help.

Note: /memory and /clear-memory do nothing when MEM0_API_KEY is not set.

---

Screenshots

Startup screen

Chat with memory enabled

---

Project structure

memoire-cli/
├── src/
│   └── index.ts        # Entire application - ~190 lines
├── dist/               # Compiled output (generated by `npm run build`)
├── .env                # Your secrets (never commit this)
├── package.json
└── tsconfig.json

---

License

ISC