AI Skill Hub 强烈推荐:Localight 苹果端模型演示 是一款优质的AI工具。AI 综合评分 8.2 分,在同类工具中表现稳健。如果你正在寻找可靠的AI工具解决方案,这是一个值得深入了解的选择。
这是一个基于SwiftUI开发的开源演示应用,旨在展示苹果在iOS 26和iOS 27版本中引入的Foundation Models(基础模型)能力。它为开发者提供了集成Apple Intelligence功能的参考实现,适合iOS开发者和AI研究员学习端侧模型调用。
Localight 苹果端模型演示 是一款基于 Swift 开发的开源工具,专注于 Apple Intelligence、iOS开发、端侧AI 等核心功能。作为 GitHub 开源项目,它拥有活跃的社区支持和持续的版本迭代,代码完全透明可审计,支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流,都能提供稳定可靠的解决方案。
这是一个基于SwiftUI开发的开源演示应用,旨在展示苹果在iOS 26和iOS 27版本中引入的Foundation Models(基础模型)能力。它为开发者提供了集成Apple Intelligence功能的参考实现,适合iOS开发者和AI研究员学习端侧模型调用。
Localight 苹果端模型演示 是一款基于 Swift 开发的开源工具,专注于 Apple Intelligence、iOS开发、端侧AI 等核心功能。作为 GitHub 开源项目,它拥有活跃的社区支持和持续的版本迭代,代码完全透明可审计,支持本地部署以保护数据隐私。无论是个人使用还是集成到企业工作流,都能提供稳定可靠的解决方案。
# 克隆仓库 git clone https://github.com/timokoethe/Localight cd Localight # 查看安装说明 cat README.md # 按 README 完成环境依赖安装后即可使用
# 查看帮助 localight --help # 基本运行 localight [options] <input> # 详细使用说明请查阅文档 # https://github.com/timokoethe/Localight
# localight 配置说明 # 查看配置选项 localight --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量(覆盖配置文件) export LOCALIGHT_CONFIG="/path/to/config.yml"
Localight is a native SwiftUI chat app for iOS 26 and iOS 27, powered entirely by Apple’s on-device Foundation Models. Designed as a practical demonstration, Localight provides fast and private AI chat without an app server. Once the operating system has downloaded and prepared the model assets, response generation works offline.
Localight showcases how to integrate Apple’s on-device language model into a native iOS experience using SwiftUI and the Foundation Models framework.
[!WARNING] Localight is a demonstration app and is not production-ready. Model output may be inaccurate, incomplete, or misleading.
Each feature links to its detailed specification in docs/features. The matrix describes features implemented by Localight, not the complete API availability of the Foundation Models framework. For example, some token-counting APIs are available in later iOS 26 point releases, while Localight demonstrates per-message response usage in its iOS 27 implementation.
| Feature | iOS 26 | iOS 27 |
|---|---|---|
| [On-device responses](docs/features/on-device-responses.md) | ✅ | ✅ |
| [Markdown model responses](docs/features/markdown-model-responses.md) | ✅ | ✅ |
| [Response streaming](docs/features/response-streaming.md) | ✅ | ✅ |
| [Session prewarming](docs/features/session-prewarming.md) | ✅ | ✅ |
| [Editable model instructions](docs/features/editable-model-instructions.md) | ✅ | ✅ |
| [Adjustable model temperature](docs/features/adjustable-model-temperature.md) | ✅ | ✅ |
| [Current context usage](docs/features/current-context-usage.md) | ❌ | ✅ |
| [Per-message token usage](docs/features/per-message-token-usage.md) | ❌ | ✅ |
| [Single-image attachments](docs/features/single-image-attachments.md) | ❌ | ✅ |
| [Typed generation error alerts](docs/features/typed-generation-error-alerts.md) | ❌ | ✅ |
| [Model availability fallback](docs/features/model-availability-fallback.md) | ✅ | ✅ |
| [Clear chat session](docs/features/clear-chat-session.md) | ✅ | ✅ |
| [Local-only, non-persistent chat](docs/features/local-only-non-persistent-chat.md) | ✅ | ✅ |
Simulator builds are useful for compilation and basic UI inspection, but the on-device model may report the simulator as ineligible. Test generation behavior on supported hardware.
The deployment target remains iOS 26, so a single app built with the iOS 27 SDK can run on both supported system versions. The project separates SDK availability from runtime availability:
LOCALIGHT_IOS27_SDK and compile both implementations.LocalightApp uses #available(iOS 27.0, *) to select the iOS 27 implementation while retaining the iOS 26 fallback.The SDK-specific compilation conditions currently match iOS 27.x. When adopting a later major SDK, update the conditional SWIFT_ACTIVE_COMPILATION_CONDITIONS entries in the target build settings so the iOS 27 implementation remains enabled.
git clone https://github.com/timokoethe/Localight.git
cd Localight
open Localight.xcodeproj
Alternatively, build the current SDK path from the command line:
xcodebuild \
-project Localight.xcodeproj \
-scheme Localight \
-configuration Debug \
-sdk iphonesimulator \
-destination 'generic/platform=iOS Simulator' \
-derivedDataPath /tmp/localight-derived-data \
CODE_SIGNING_ALLOWED=NO \
build
| iOS 26: Text chat | iOS 26: Session settings |
|---|---|
| <img src="docs/assets/ios-26-text-chat.png" alt="Localight text chat on iOS 26 showing a user prompt and an on-device model response" width="240"> | <img src="docs/assets/ios-26-session-settings.png" alt="Localight session settings on iOS 26 showing response streaming, model instructions, context size, and temperature" width="240"> |
| **iOS 27: Multimodal chat with token usage** | **iOS 27: Session settings and context usage** |
| <img src="docs/assets/ios-27-multimodal-chat.png" alt="Localight chat on iOS 27 showing an image prompt, an on-device model response, and per-message token counts" width="240"> | <img src="docs/assets/ios-27-session-settings.png" alt="Localight session settings on iOS 27 showing response streaming, token usage, model instructions, context usage, and temperature" width="240"> |
aiskill88点评:极具前瞻性的端侧AI实现参考,是掌握Apple Intelligence生态开发的高质量样板工程。
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。
总体来看,Localight 苹果端模型演示 是一款质量优秀的AI工具,在同类工具中具备一定竞争力。AI Skill Hub 将持续追踪其更新动态,建议收藏备用,结合自身场景选择合适时机引入使用。
| 原始名称 | Localight |
| 原始描述 | 开源AI工具:A SwiftUI showcase app for Apple’s Foundation Models on iOS 26 and iOS 27. Disco。⭐10 · Swift |
| Topics | Apple IntelligenceiOS开发端侧AI |
| GitHub | https://github.com/timokoethe/Localight |
| License | MIT |
| 语言 | Swift |
收录时间:2026-07-04 · 更新时间:2026-07-04 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。