CrustAI — Your Private, Local AI Assistant

🦀 v0.1.0 — Open Source & MIT License

██████╗██████╗ ██╗ ██╗███████╗████████╗ █████╗ ██╗<br>██╔════╝██╔══██╗██║ ██║██╔════╝╚══██╔══╝██╔══██╗██║<br>██║ ██████╔╝██║ ██║███████╗ ██║ ███████║██║<br>██║ ██╔══██╗██║ ██║╚════██║ ██║ ██╔══██║██║<br>╚██████╗██║ ██║╚██████╔╝███████║ ██║ ██║ ██║██║<br>╚═════╝╚═╝ ╚═╝ ╚═════╝ ╚══════╝ ╚═╝ ╚═╝ ╚═╝╚═╝<br>Your Private, Local<br>AI Assistant

Always on. Always yours. Never in the cloud.<br>Connect your favorite messaging apps to a local AI — zero subscriptions, zero telemetry.

🚀 Get Started<br>⭐ Star on GitHub

Node.js ≥ 20<br>Ollama<br>Telegram<br>WhatsApp<br>Discord<br>Slack<br>MIT License

// Why CrustAI

Everything you need.<br>Nothing you don't.

Built for developers and privacy enthusiasts who want a real AI assistant without giving up control of their data.

🔒

100% Private

All conversations stay on your machine. No data sent to external AI services. No telemetry. No tracking. Ever.

🧠

Local LLM Power

Powered by Ollama — supports llama3.2, tinyllama, phi3, mistral and any other compatible model.

📱

Multi-Platform

One assistant, four platforms. Telegram, WhatsApp, Discord and Slack — all managed from a single config file.

🧬

Long-term Memory

CrustAI remembers facts about you across conversations using the /remember command and local SQLite storage.

🗣️

Offline Voice

Speech-to-text and text-to-speech support in Portuguese (pt-BR) — works completely offline after setup.

REST API

Built-in Fastify REST API for custom integrations. Connect any tool or service to your local AI assistant.

// Installation

Up and running<br>in 5 steps.

No Docker required. No cloud accounts. Just Node.js and Ollama on your own machine.

Clone the repository

Download CrustAI to your machine using Git.

terminal

git clone https://github.com/DaveSimoes/CrustAI.git

cd CrustAI

Install dependencies

Install all required Node.js packages.

terminal

npm install

Start Ollama and pull a model

Start the Ollama server and download your preferred AI model. tinyllama works great on modest hardware.

terminal

# Start Ollama server (keep this terminal open)

ollama serve

# In a new terminal — pull a model

ollama pull tinyllama # lightweight (600MB)

# or

ollama pull llama3.2 # powerful (2GB, needs 8GB RAM)

Configure the project

Copy the example config and add your Telegram bot token from @BotFather.

config/config.yml

model: tinyllama

ollama_url: http://localhost:11434

language: pt-BR

telegram:

enabled: true

token: YOUR_BOT_TOKEN_HERE

allowed_user_ids: []

Launch CrustAI!

Start the assistant and open your Telegram bot to begin chatting.

terminal

npm start

# Expected output:

✓ Ollama connected (tinyllama)

✓ Memory store ready (./data/memory.db)

✓ Telegram ready

✓ REST API ready (http://localhost:3000)

🦀 CrustAI is ready. Your shell awaits.

// Integrations

One assistant.<br>Every platform.

Enable only the platforms you use. Each adapter is independently configurable in config.yml.

✈️

Telegram

Full bot integration via @BotFather token

💬

WhatsApp

Connect via QR code scan — no business account needed

🎮

Discord

Full Discord bot with server and channel support

💼

Slack

Slack app integration for teams and workspaces

🌐

REST API

HTTP API always running on localhost:3000

🎙️

Voice

WebSocket voice server with offline STT/TTS

// Usage

Built-in commands.

Type these commands directly in any connected messaging platform.

/ping<br>Check if the bot is alive and responding

/help<br>Show all available commands

/model<br>Show which AI model is currently running

/remember<br>Store a fact in long-term memory

/forget<br>Erase all stored facts about you

/clear<br>Clear the current conversation history

// Privacy First

Your data.<br>Your rules.

CrustAI was built with privacy as its core principle — not an afterthought.

Stays on your machine<br>All conversations processed locally. Nothing leaves your hardware.

No external AI APIs<br>Zero calls to OpenAI, Anthropic, Google or any cloud AI service.

No telemetry<br>No usage tracking, no analytics, no crash reports sent anywhere.

Open source<br>Every line of code is auditable. MIT license — use it however you want.