Rete

What is Rete?

Rete is a local AI platform for Apple devices that lets you run AI directly on your Mac, iPhone, iPad, and Apple TV—without using a cloud service. After installation, your devices can connect to form a compute mesh and split work to speed up and improve responses.

The core idea is “pay once, own forever,” with no subscriptions and no cloud requirement. Rete provides a chat experience plus supporting tools such as document question answering (RAG), code assistance, and searchable conversation history, all running on your devices.

Key Features

• Local-first AI on your Apple devices (macOS, iOS, Apple TV): conversations and model execution run on-device rather than relying on cloud processing.
• Multi-device “mesh computing” over Wi‑Fi: devices discover each other automatically and can contribute GPU power via Metal.
• Automatic load balancing across devices: Rete routes tasks to the fastest device in the mesh and exposes performance metrics.
• Model support and flexibility: download or import models, including Llama, Mistral, Phi, Gemma, and Qwen.
• Built-in “fine-tuned experts”: ships with 5 fine-tuned expert modes you can switch between.
• Document RAG and search over your files: supports PDFs, code, CSV, and text so questions can be grounded in provided documents using embedding-based context retrieval.
• Markdown rendering with syntax highlighting: improves readability for writing and code-related outputs.
• Memory and personalization across conversations: remembers key facts and learns your preferences and context over time.
• Conversation organization and search: folders plus full-history search across messages.
• Remote mesh via invite code: connect devices across different networks using an invite code so additional devices can join your mesh.

How to Use Rete

1. Install Rete on your Mac, then open the app. Rete scans your hardware and recommends an AI model suited to your machine.
2. Install Rete on your iPhone, iPad, and/or Apple TV. When these devices are on the same Wi‑Fi network, they connect automatically to form a mesh.
3. Start a chat in Rete. Your devices split the work, and you can use features like document RAG, folders, and searchable history as you interact with the AI.

Use Cases

• Personal assistant and research-style Q&A: ask questions in chat and keep a searchable history across conversations while staying local to your devices.
• Document-grounded questions (RAG): upload or reference PDFs, CSVs, or text/code files and ask questions that are grounded in the content using embedding-based retrieval.
• Code help and review workflows: use built-in expert modes and syntax-highlighted outputs to assist with writing and reviewing code.
• Writing support with formatting: generate and refine text while viewing responses with Markdown rendering and highlighted code blocks.
• Faster local inference by scaling devices: for longer or heavier prompts, add an iPad, iPhone, or Apple TV to expand the compute mesh and route work across the available hardware.

FAQ

Do I need a cloud account or subscription to use Rete?
No. The site states Rete runs on your devices, does not use cloud, does not require accounts, and has no subscriptions.

Which devices are supported?
Rete supports macOS on Mac and iOS on iPhone and iPad, plus an Apple TV app.

How do I connect multiple devices together?
Devices connect over Wi‑Fi automatically to form a mesh. For connecting across different networks, you can share an invite code.

What models can Rete run?
Rete can download models from the app or import your own. The site lists Llama, Mistral, Phi, Gemma, and Qwen as supported model families.

What is “Document RAG” in Rete?
It refers to adding documents (such as PDFs, code, CSV, or text) so you can ask questions grounded in your data, using embedding-based search to retrieve relevant context.

Alternatives

• Self-hosted local LLM/chat apps: Instead of a device mesh on Apple hardware, these run models locally on a single machine (or via local servers you manage), typically with a different setup workflow.
• Cloud-based chatbots: These handle inference in the cloud and usually provide easier access across devices, but they differ from Rete’s local-first approach (no cloud, no accounts stated on the site).
• Local RAG toolkits and embeddings pipelines: If your main goal is document-grounded Q&A, alternatives may focus on retrieval, indexing, and embeddings while leaving chat UX and device orchestration to you.
• Developer-focused model runners: Tools that prioritize model management and inference on your hardware can be an alternative if you want more control over model selection and runtime, though they may not provide the same Apple device mesh workflow.