Foundation Studio is a native macOS chat app built with SwiftUI, SwiftData, and Apple's FoundationModels framework.
It is designed as a private, on-device AI workspace inspired by the layout and UX of Google AI Studio, while staying fully native to macOS.
- 100% on-device LLM workflow with Apple Intelligence
- No API keys, no server backend, no network dependency for model inference
- Fast SwiftUI desktop UX with persistent chat history
- Practical reference app for FoundationModels + SwiftData integration
- NavigationSplitView app shell with persistent sidebar chat list
- New chat creation and chat deletion
- Chat persistence using SwiftData models (
ChatThread,Message,PromptRecord) - FoundationModels integration via
SystemLanguageModel.default - Model availability awareness and status badge
- Streaming assistant responses in the chat surface
- Per-thread system instructions and temperature control
- Auto-generated conversation title from first exchange
- Markdown-style assistant rendering (headings, lists, code blocks, block quotes, inline formatting)
- Message utilities: copy, share, markdown/plain toggle, generation metrics (TPS, token count, duration)
- SwiftUI (macOS app UI)
- SwiftData (
@Model,@Query, relationship persistence) - FoundationModels (
SystemLanguageModel,LanguageModelSession) - Observation (
@Observable) for model service state
- macOS 26+ (Apple Silicon)
- Xcode with FoundationModels support (WWDC 2025+ SDK line)
- Apple Intelligence enabled on device
FoundationStudio/
FoundationStudio/
FoundationStudioApp.swift
ContentView.swift
Models/
ChatThread.swift
Message.swift
PromptRecord.swift
Services/
FoundationModelService.swift
Views/
SidebarView.swift
ChatView.swift
MessageBubbleView.swift
MessageInputView.swift
MarkdownContentView.swift
ModelStatusBadge.swift
- Open
FoundationStudio/FoundationStudio.xcodeprojin Xcode. - Select the
FoundationStudioscheme. - Run on an Apple Silicon Mac with Apple Intelligence available.
- Create a new chat and send your first prompt.
The app reads SystemLanguageModel.default.availability and surfaces status in the UI.
Handled states include:
- Available
- Device not eligible
- Apple Intelligence not enabled
- Model not ready
- Other unavailable states
When unavailable, user-facing errors are shown in-thread as system messages.
ChatThread: conversation metadata, system instructions, temperature, relationshipsMessage: role-based chat messages with optional generation metricsPromptRecord: prompt/response logs and generation metadata for review/reuse
This repository includes the original seed prompt used to bootstrap the app design and implementation flow:
- See
prompt.md
That prompt captures the intended UX direction, feature order, and architectural constraints for iterative agentic development.
Foundation Studio is intended to run inference on-device using Apple Intelligence Foundation Models.
- No API keys required
- No cloud relay required for core chat inference
- Local SwiftData persistence
Always validate your own privacy and compliance requirements before production use.
- Settings surface expansion (top-p, max tokens where exposed)
- Structured output flows with
@Generable - File/image attachment workflows
- Prompt library and reusable templates
- Export/import for threads and prompt records
- Command palette and menu extras
- This project is an independent implementation inspired by Google AI Studio UX patterns.
- Google AI Studio and related marks are trademarks of their respective owners.
Issues and pull requests are welcome. For larger changes, open an issue first to discuss scope and approach.
Licensed under the MIT license.