The app
| Term | Definition |
|---|---|
| Apogee Shell | The main application interface. Everything you see and interact with — the launcher, spaces, apps, settings — is part of the shell. |
| Oasis | The local AI engine that runs on your device. It powers the chat, personas, and agent capabilities. |
| IDN (Intelligence Delivery Network) | The peer-to-peer network formed by devices running the app. Each device is a node that creates intelligence locally and can share it with others. |
How the app runs
| Term | Definition |
|---|---|
| Tauri | The framework that turns the app into a native desktop application. It provides access to the file system, native windows, camera, and local network discovery. Features labeled “desktop only” rely on Tauri. |
| Vite | The build tool that bundles and serves your code during development and production. If you’re building a custom app, Vite handles hot reloading, asset bundling, and fast builds so you can iterate quickly. |
| React | The UI library the shell is built with. Relevant if you’re building custom apps using JSX or contributing to the codebase. |
AI and models
| Term | Definition |
|---|---|
| WebGPU | A browser API that lets the app run AI models directly on your GPU. This is what makes local inference fast without a separate application or cloud service. |
| MLX | Apple’s machine learning framework. Used as an alternative inference backend on Apple Silicon Macs. |
| llama.cpp | A lightweight C++ inference engine. Acts as a fallback backend when WebGPU or MLX aren’t available. |
| GGUF | A file format for AI models. Models downloaded for local use may be in this format. |
| LoRA (Low-Rank Adaptation) | A technique for customizing AI behavior by loading small adapter files on top of a base model. This is how personas can have specialized knowledge or style without downloading a new model. |
| Persona | A saved configuration for the AI’s behavior — including its system prompt, tool permissions, and any adapters loaded. See Personas. |
| Token | The basic unit of text that an AI model processes. Words are broken into one or more tokens. Token counts and tokens-per-second appear in the UI as performance indicators. |
| Model tier | A grouping of AI models by size and memory requirements. The system recommends a tier based on your device’s capabilities. See On-Device AI. |
Networking and collaboration
| Term | Definition |
|---|---|
| P2P (Peer-to-Peer) | Direct device-to-device connections without a central server. Spaces, chat, file sharing, and voice all work this way. |
| WebRTC | The browser technology that makes peer-to-peer connections possible. It handles real-time data, audio, and video between devices. |
| CRDT (Conflict-Free Replicated Data Type) | The data structure used to sync state between collaborators. It ensures that when two people edit at the same time, their changes merge cleanly without conflicts. |
| Edit lock | A short-lived lock (up to 30 seconds) that prevents two people from editing the same field at the same time. Locks release automatically when the editor moves on. |
| mDNS | A protocol for discovering devices on your local network. This is how the app finds nearby spaces and contacts without an internet connection. |
| E2E Encryption | End-to-end encryption where only the sender and recipient can read the content. Direct messages and private file transfers use this. See About webAI. |
| Relay | A backup node that can take over space hosting if the original host disconnects. Keeps spaces alive as people come and go. |
Identity
| Term | Definition |
|---|---|
| ODID (On-Device ID) | Your unique device identifier. Generated locally, used to identify you to contacts and in spaces. No account or sign-up required. |
| Pairing | The process of adding a contact securely by exchanging cryptographic keys. The 6-digit rotating code method establishes a trusted, verified connection. |
App building
| Term | Definition |
|---|---|
| iframe | The sandbox where custom apps run inside the shell. Your app gets its own isolated environment with access to platform APIs through message passing. |
| Canvas Protocol | The lightweight message protocol custom apps use to sync state with the shell and other collaborators. |
| JSX | A syntax extension that lets you write HTML-like markup in JavaScript (used by React). You can upload .jsx files as custom apps and the platform compiles them automatically. |
| postMessage | The browser API that apps use to communicate with the shell from inside their iframe. This is the bridge between your app and the platform’s collaboration, identity, and storage APIs. |