Core Concepts¶
You can run HiveMind without ever reading this section — the Quick Start gets you talking to a satellite in ten minutes flat. But at some point you'll wonder why a new device can't do anything until you allow it, or how a question finds its way to the right server, or what actually crosses the wire. This is where those answers live: the ideas behind the mesh, not the commands to run it. Skim it once and every other page clicks into place.
New here? Read these two first
Start with Mesh Topology (what the pieces are) and Security & Permissions (how devices trust each other). The rest you can read on demand.
What's in this section¶
| Page | What it answers | For |
|---|---|---|
| Mesh Topology | What are hivemind-core instances, satellites, and relays? How do messages travel up and down the mesh? | Everyone |
| Protocol & Message Types | What does HiveMind actually send over the wire, and in which direction? | Curious users · developers |
| Security & Permissions | How are connections encrypted, and how do you control what each device may do? | Everyone running hivemind-core |
| Database Backends | Where are client credentials stored? SQLite, JSON, or Redis? | Operators |
| Auto Discovery | How do satellites find hivemind-core automatically on a local network? | Everyone |
| Plugin Architecture | The five plugin families that make every layer swappable. | Developers · advanced operators |
The one-paragraph version¶
hivemind-core is the brain: it runs skills, an LLM, or another AI backend. Satellites are the lightweight devices you talk to — a microphone, a browser tab, a terminal. They connect to hivemind-core over an encrypted, permissioned protocol that is agnostic to both the transport (WebSocket, HTTP, MQTT…) and the payload (OVOS messages by default). hivemind-core instances can connect to other instances, forming a mesh: requests one instance can't answer escalate upward, and announcements broadcast downward. Every other page in this manual is a detail of that sentence.
Next: Mesh Topology