The ButlerUseful technology. Clearly explained.

Home / Useful resources

The Butler practical library

Your first local AI setup

A short route from hardware questions to a response you can check.

Reviewed November 14, 2026. Official-documentation-based guidance; not a claim that every installation or hardware combination was hands-on tested.

Your finish line

A local model answers a harmless prompt on your own computer, with a clear idea of what stays local and what could leave it.

  1. Check your hardware. Start with a smaller quantized model and a short context rather than filling every last gigabyte.
  2. Choose LM Studio for a desktop interface, or Ollama for a command-line/API foundation. You do not need both.
  3. Send a test question and inspect memory use. If loading fails, use the memory checklist.
  4. With Ollama working, try the local chat-interface project. If documents are your goal, use the document assistant project instead.

Know what local means

A local interface can still call a cloud model, remote embeddings or web-search tool. Check each provider and connector. Downloads and updates normally need a connection, even when inference can later run offline.

Read for perspective

Open versus closed models: choosing for a workflow is archived April 2026 analysis. Use it for the tradeoffs, not as a current price or model list.