AI & agents

Inference that never leaves your machine

The most private assistant is one that never phones home. Nolocron can run models on-device, so search and chat can work without any data leaving your machine.

Problem

Why it matters

Cloud AI means your most personal context leaves your control on every request. Custody should extend to inference, not stop at storage.

Capabilities

What Nolocron provides

  • On-device GGUF embeddings via node-llama-cpp, accelerated on Apple Silicon.
  • A local chat path (Gemma) that plugs into the same governed agent runtime.
  • No API keys required — full on-device execution.
  • Model paths and choices are settings you control; changes take effect on the next run.

Workflow

How it works

  1. Turn on local models and point Nolocron at a GGUF file.
  2. Embed your archive on-device for local search.
  3. Chat with a local model through the same governed surface.
  4. Keep everything — data and inference — on your machine.

Evidence

Product proof points

  • No keys, no network round-trip — inference stays local.
  • Designed for the same governance and receipts as cloud models.
  • Opt-in and early — local chat is a compatibility path today.