Back to projects

Eris

2025
iOS Design & Development
View projectView code
Eris

About the project

Eris is an AI chat application that runs entirely on your device using Apple's MLX framework. Named after the dwarf planet that challenged our understanding of the solar system, Eris challenges the notion that AI must live in the cloud. All conversations remain private on your device, with no telemetry or analytics.

Main features

  • 100% Private - All conversations on your device
  • Works without internet connection
  • Support for multiple models: Llama, Qwen, DeepSeek, Gemma, Phi
  • Apple Shortcuts integration for automation
  • Syntax highlighting for 100+ programming languages
  • Full Markdown support
  • Native Dark Mode and Light Mode
  • Native design with SwiftUI

Technical challenges

  • Optimize AI models to run on mobile devices
  • Efficient memory management with 2-8GB models
  • Implement real-time inference with Metal and MLX
  • Create a smooth experience during model loading

Tech stack

SwiftSwiftUIMLXSwiftData

Color Palette

Background

#0A0A0A

Main background

Surface

#111111

Cards, elevated elements

Border

#262626

Borders, separators

Foreground

#FFFFFF

Primary text, accent

Muted

#A3A3A3

Secondary text

Typography & Iconography

Geist

Sans-serif (Vercel)

Main typography for UI, headings and body text on landing

RegularMediumSemiboldBold

Private AI Chat

72px • 700

Your conversations. Your device.

20px • 400

Geist Mono

Monospace (Vercel)

Code blocks, technical data and syntax highlighting

RegularMedium

await model.generate()

16px • 400

SF Pro (iOS)

System Font

Native typography in iOS app for maximum readability

RegularMediumSemiboldBold

Eris.

34px • 700

Start a conversation

17px • 400

SF Symbols

SF Symbols

Regular / Hierarchical • Apple's native iconography for iOS consistency

https://developer.apple.com/sf-symbols/
  • Over 5,000+ symbols available
  • Native support for Dynamic Type
  • Weights that scale with text
  • Multicolor and hierarchical modes

Supported AI Models

Llama 3.2

General Purpose

Meta

1B3B

Qwen 2.5

Multilingual

Alibaba

0.5B1.5B3B

DeepSeek-R1

Reasoning

DeepSeek

1.5B 4bit1.5B 8bit

Gemma 2

Lightweight

Google

2B

Phi 3.5 Mini

Compact

Microsoft

3.8B

Technical Details

Apple's MLX Framework

MLX is an array framework for machine learning on Apple Silicon, designed to be efficient and easy to use.

  • Runs models directly on Apple Silicon
  • Optimized for Neural Engine and GPU
  • No server dependencies
  • Support for quantized models

Total Privacy

No telemetry, no analytics, no tracking. Your conversations never leave your device.

  • Zero data sent to servers
  • Local storage with SwiftData
  • No account required
  • Works completely offline

Shortcuts Integration

Ask Eris

Send a question to Eris and get an instant response

New Conversation

Start a new conversation with the selected model

System Requirements

iPhone

iPhone 15 Pro or later

iPhone 15 Pro, iPhone 15 Pro Max, iPhone 16 series

iPad

iPad with M1 chip or later

iPad Pro M1/M2/M4, iPad Air M1/M2

System

iOS 18.0 / iPadOS 18.0 or later

Storage

2-8 GB per model