Skip to content

Mimicking Apple Intelligence, this project enables on-device Retrieval-Augmented Generation across all your Apple devices with zero cloud dependence.

License

Notifications You must be signed in to change notification settings

bismansahni/AppleRAGMesh

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AppleRAGMesh

AppleRAGMesh is an on-device Retrieval-Augmented Generation (RAG) system designed for the Apple ecosystem.
It creates a local mesh network between your Apple devices (Mac, iPhone, iPad), allowing one device to query documents stored on any other nearby device — all without internet access.


💡 What It Does

  • Builds a mesh using MultipeerConnectivity (Bluetooth / peer-to-peer Wi-Fi).
  • Enables one Apple device (e.g. Mac) to query files stored on other devices (e.g. iPad, iPhone).
  • Automatically handles document chunking, local embedding, and retrieval.
  • All processing happens on-device — no cloud, no server, no internet.

🎥 Demo (macOS)

Currently demonstrated on MacBook, showcasing:

  • Adding PDF/TXT files.
  • Generating embeddings locally with MiniLM.
  • Querying across the mesh (e.g. asking about a file that's only on another device).
  • LLM answering achieved via CoreML optimized version of Mistral7B

📽️ Watch Demo:
Click here for the Demo Video


📱 Extending to iPhone & iPad

  • The codebase is ready to support iOS and iPadOS.
  • Just add a new target for iPhone or iPad in Xcode.
  • Once connected, you can ask your Mac to query files stored only on your iPad/iPhone — no need to touch the other device.

⚡ Why It’s Useful

Perfect for:

  • When you’ve forgotten which device a file is on.
  • Quickly referencing a document on another device without opening or browsing.
  • Local, fast, secure retrieval in Airplane mode or without Wi-Fi.

No internet required.
No third-party servers.
Just Apple devices talking directly.


🛠 The most stable build ison the stable-version branch.

About

Mimicking Apple Intelligence, this project enables on-device Retrieval-Augmented Generation across all your Apple devices with zero cloud dependence.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages