DarkGrid AI

Knowledge Base

The knowledge base is what makes DarkGrid useful offline: instead of relying only on what the model memorized during training, it retrieves real passages from your local sources and answers from them, with citations.

There are two kinds of sources, searched differently and then combined.

1. Offline Wikipedia (ZIM archives)

Kiwix publishes Wikipedia (and many other sites) as single-file ZIM archives that include a built-in full-text search index. That means DarkGrid can search across a huge Wikipedia archive instantly without having to re-index it on your machine.

Downloading an archive

From Settings → Knowledge Base, choose a Wikipedia archive:

ArchiveSizeNotes
Simple English Wikipedia~1 GBRecommended default — small, plain-language
English Wikipedia (medical)~5 GBHealth/first-aid focused
English Wikipedia (full, no images)~50 GBComprehensive; needs lots of disk

DarkGrid never downloads the giant full archive automatically. You opt in, and the app verifies you have the disk space first.

You can also drop a .zim file you obtained elsewhere into the kb/zims/ folder in your data directory.

2. References (your own folders)

Add folders of your own documents — field manuals, first-aid guides, repair manuals, foraging references, personal notes — and DarkGrid indexes them so it can answer from them by meaning. Supported file types within a folder are PDF, plain text (.txt), Markdown (.md), and Word (.docx); other files are skipped.

Adding a reference

Settings → Knowledge Base → Add folder, then choose one or more folders. You can add as many references as you like.

Files are read in place — nothing is copied into the app. Removing a reference drops its indexed text; the files themselves stay where they are.