Offline and Air-Gapped AI Context

Vibgrate AI Context is designed to run locally and offline. After a single one-time download, it needs no network at all — which makes it a strong fit for air-gapped, regulated, or disconnected environments. This guide explains how to prepare and operate it without ongoing connectivity.

It is for developers and security teams who cannot or will not let their tooling reach the internet during normal work.

Overview

The local context server serves your code map, drift, and version-correct library docs from local artifacts. The only network step is a one-time download of a small local model used for semantic search. Once that is cached, the server runs fully offline.

Step 1 — Warm the local model once

Before going offline, precompute the semantic index so the model is downloaded into the shared cache:

vg embed

vg embed downloads the local ONNX model once into a shared cache and precomputes the semantic index — fully offline after that first run. This is also what makes the next vg ask instant.

Step 2 — Build the code map locally

vg build

The code map is built from your source on your machine; no network is required.

Step 3 — Serve offline

vg serve

With the model cached and the map built, the server runs without any network access. Your assistant reads grounded context entirely from local data.

Version-correct docs offline

The Free Dev Docs Library draws from a hosted catalog and from local ingestion. Catalog fetches need connectivity, but libraries ingested locally from your installed copies are served offline. In an air-gapped setting, rely on local ingestion of what is already installed.

What never leaves the machine

Vibgrate AI Context uploads nothing and needs no account. Your code, code map, and drift stay local. This is the same privacy posture whether you are online or air-gapped.

Troubleshooting

• First vg serve tries to reach the network. Run vg embed while connected so the model is cached before you disconnect.
• A library has no docs offline. Ensure it is installed so Vibgrate can ingest it locally rather than depending on the hosted catalog.

Related

• Starting the local context server with vg serve
• Security and privacy of local AI context
• Keeping the library current