ReteAi

Rete turns the computers, phones, and servers you already own into a shared, locally‑run AI platform. After installing the desktop app on macOS, Windows, or Linux, the software scans the hardware and suggests an appropriate model, then lets you add additional devices—including iPhone, iPad, or other machines—via a simple invite code. The devices automatically form a mesh that distributes inference work, so adding more hardware expands the model size and speeds up responses without any cloud services.

The interface provides a full‑featured chat experience with conversation history, markdown rendering, and syntax‑highlighted code snippets. Users can load any compatible model such as Llama, Mistral, Phi, Gemma, or Qwen, or import their own, and switch between built‑in expert personas. Document retrieval‑augmented generation lets the AI answer questions grounded in PDFs, CSVs, or other files through embedding‑based search.

All processing stays on the user’s devices; no accounts, telemetry, or subscriptions are required. The mesh balances load across GPUs using Metal, CUDA, or Vulkan, and a dashboard shows real‑time performance metrics. Memory features retain key facts across chats, while folders and full‑text search help organize and locate past interactions.