Table of Contents

User guide

Welcome to the practical side of LocalGPT. The interface is playful; the operational boundaries are not. A tiny bit kawaii, a lot of explicit control. 🌸

Start here

  1. Getting started — application shape, providers, storage, and approval boundaries.
  2. Chat and AI Council — direct conversations, teams, roles, provider-qualified models, benchmarks, and human participation.
  3. Projects and workspaces — durable projects, revisions, tracked files, toolchains, and permission assessment.
  4. Embedded planning and games — firmware plans, logical telemetry, GameDirector, ASCII play surfaces, and Minecraft generation.
  5. Documentation inside LocalGPT — HTML, API reference, PDF, GitHub Pages, search, and theme persistence.

What LocalGPT does not do silently

LocalGPT does not treat an AI answer as authorization. Consequential actions—writes, builds, native commands, serial access, flashing, actuator control, or protected peer operations—must cross an explicit application policy and, where required, a fresh user approval.

The application may prepare plans, drafts, diffs, firmware layouts, Council proposals, and capability reports without pretending those artifacts have already been executed.

Tip

Begin with one provider and one project. Add Council members, runtime classes, and toolchains after the basic route works. Cute complexity is still complexity.