ℹ️ Previously known as "Open UI" — The app was renamed to Open Relay starting with v2.0.0. The Xcode project files still use the original "Open UI" name internally, but the GitHub repo, App Store listing, and all display names are now Open Relay.
A beautiful, native iOS & iPadOS client for Open WebUI.
Chat with any AI model on your self-hosted Open WebUI server — right from your iPhone or iPad. Open Relay is built 100% in SwiftUI and brings a fast, polished, native experience that the PWA can't match.
Open Relay connects to your Open WebUI server and lets you have conversations with any AI model you've configured — it's like having ChatGPT on your phone, but pointed at your server and your models.
🗨️ Streaming Chat with Full Markdown — Real-time word-by-word streaming with complete markdown support — syntax-highlighted code blocks (with language detection and copy button), tables, math equations, block quotes, headings, inline code, links, and more. Everything renders beautifully as it streams in.
🖥️ Terminal Integration — Enable terminal access for AI models directly from the chat input, giving the model the ability to run commands, manage files, and interact with a real Linux environment. Swipe from the right edge to open a slide-over file panel with directory navigation, breadcrumb path bar, file upload, folder creation, file preview/download, and a built-in mini terminal.
@ Model Mentions — Type @ in the chat input to instantly switch which model handles your message. Pick from a fluent popup, and a persistent chip appears in the composer showing the active override. Switch models mid-conversation without changing the chat's default.
$ Skills — Type $ in the chat input to browse and apply your Open WebUI skills directly from the composer.
/ Prompt Library — Type / in the chat input to browse and search your Open WebUI prompt library with slash commands.
📐 Native SVG & Mermaid Rendering — AI-generated SVG code blocks render as crisp, zoomable images with a header bar, Image/Source toggle, copy button, and fullscreen view with pinch-to-zoom. Mermaid diagrams (flowcharts, state, sequence, class, and ER) also render as beautiful inline images.
🌐 Rich HTML Embeds — Tools that return interactive HTML (audio players, video, cards, dashboards, charts, forms, SMS composers, and more) render inline in chat as live, interactive webviews.
📞 Voice Calls with AI — Call your AI like a phone call using Apple's CallKit — it shows up and feels like a real iOS call. An animated orb visualization reacts to your voice and the AI's response in real-time. Calls default to loudspeaker with a speaker toggle to switch to earpiece.
🧠 Reasoning / Thinking Display — When your model uses chain-of-thought reasoning (like DeepSeek, QwQ, etc.), the app shows collapsible "Thought for X seconds" blocks. Expand them to see the full reasoning process.
📚 Knowledge Bases (RAG) — Type # in the chat input for a searchable picker for your knowledge collections, folders, and files. Works exactly like the web UI's # picker.
🛠️ Tools & Workspace Management — All your server-side tools show up in a tools menu. Toggle them on/off per conversation. Tool calls are rendered inline with collapsible argument/result views. Manage your models, knowledge bases, prompts, skills, and tools directly from the app via the Workspace section.
🧠 Memories — View, add, edit, and delete AI memories (Settings → Personalization → Memories) that persist across conversations.
🎙️ On-Device TTS (Marvis Neural Voice) — Built-in on-device text-to-speech powered by MLX. Downloads a ~250MB model once, then runs completely locally — no data leaves your phone. You can also use Apple's system voices or your server's TTS with voice selection.
🎤 On-Device Speech-to-Text — Voice input with Apple's on-device speech recognition, your server's STT endpoint (fully supports live mic and voice calls), or an on-device Qwen3 ASR model for offline transcription.
📎 Rich Attachments — Attach files, photos (library or camera), paste images directly into chat. Share Extension lets you share content from any app into Open Relay. Images are automatically downsampled before upload to stay within API limits.
📁 Folders & Organization — Organize conversations into folders with drag-and-drop. Folders support full project workspace configuration — set a name, system prompt, default model, and attached knowledge bases. Pin chats. Search across everything. Bulk select, delete, and Archive All Chats in one tap.
💬 Channels — Collaborative, topic-based chat rooms where multiple users and AI models interact, grouped by Direct Messages, Groups, and Channels.
🗂️ Archived & Shared Chats — Browse all archived chats and restore them individually or all at once. Manage all shared conversations — copy share links or revoke access — from the chat list menu.
📊 Token Usage — Tap the ⓘ info icon in the assistant action bar to see per-message token stats.
🔗 Multi-Server Support — Save multiple Open WebUI server connections and switch between them instantly from Settings or the server connection screen.
🔐 Full Auth Support — Username/password, LDAP, SSO, and auth proxy support (Authelia, Authentik, Keycloak, oauth2-proxy, etc.). Tokens stored in iOS Keychain. Custom headers on sign-in.
🎨 Deep Theming — Full accent color picker with presets and a custom color wheel. Pure black OLED mode. Tinted surfaces. Live preview as you customize.
♿ Accessibility — Independently adjust message text size, conversation title size, and UI element scaling (buttons, icons, spacing) with live preview and quick presets.
⚡ Quick Action Pills — Configurable quick-toggle pills for web search, image generation, or any server tool. One tap to enable/disable without opening a menu.
🔔 Background Notifications — Get notified when a generation finishes while you're in another app.
📝 Notes — Built-in notes alongside your chats, with audio recording support.
🏠 Home Screen Widgets & Shortcuts — Start a chat directly from a home screen widget or your iPhone's Action Button using Shortcuts integration.
📱 iPad Native Layout — Full persistent sidebar, centered reading width, 4-column prompt grid, and terminal as a persistent panel. Optimized for both iPhone and iPad.
🌍 Landscape Mode — Full landscape support on iPhone.
- Default model picker synced with your server — redesigned as a native bottom sheet with search and filter pills
- Send on Enter toggle (Enter sends vs. newline; hardware keyboard on iPad always uses Enter-to-send)
- Streaming haptics — feel each token as it arrives
- Temporary chats — conversations not saved to the server for privacy
- TTS engine selection with per-engine configuration and voice selection
- STT engine selection with silence duration control and language selection
- Cloudflare protection — automatically handles Cloudflare-protected endpoints
- iOS 18.0 or later
- Xcode 16.0 or later (Swift 6.0+)
- A running Open WebUI server instance accessible from your device
git clone https://github.com/ichigo3766/Open-Relay.git
cd Open-Relayopen "Open UI.xcodeproj"Xcode will automatically fetch all Swift Package dependencies on first open. This may take a minute.
- In Xcode, select the Open UI target in the project navigator
- Go to Signing & Capabilities
- Select your Development Team
- Update the Bundle Identifier if needed (e.g.,
com.yourname.openui)
- Select an iOS 18+ simulator or a connected device
- Press ⌘R (or click the
▶️ Play button) - On first launch, enter your Open WebUI server URL and sign in
- SwiftUI — 100% SwiftUI interface
- Swift 6 with strict concurrency
- MVVM architecture
- SSE (Server-Sent Events) for real-time streaming
- CallKit for native voice call integration
- MLX Swift for on-device ML inference (Marvis TTS + Qwen3 ASR)
- Core Data for local persistence
Special thanks to Conduit by cogwheel — Cross-Platform Open WebUI mobile client and a real inspiration for this project.
This project is licensed under the MIT License. See LICENSE for details.