Glossary¶

Plain-language definitions of every term used throughout the HiveMind documentation. The user-facing guides favour hivemind-core and satellite; the protocol and mesh references also use the lower-level vocabulary (node, Mind, master/slave) — both describe the same network.

In a nutshell

Terms are grouped into roles, network structure, protocol, credentials & access control, discovery, and OVOS voice vocabulary.
hivemind-core and satellite are the everyday names; Mind, master, and slave are the equivalent protocol-level terms.
Keep this page open in a second tab while reading the rest of the manual.

Roles¶

Term	Definition
Node	Any device or process that participates in a HiveMind network.
hivemind-core	A node that accepts connections, authenticates clients, isolates sessions, and authorizes individual messages. Runs the AI workload (skills, LLM, or audio processing) on behalf of its satellites. See OVOS Skills backend and Persona backend.
Satellite	A user-facing node that connects to hivemind-core but does not accept connections of its own. Ranges from text-only (CLI) to a full local voice stack (voice satellite). See Choosing a Satellite.
Mind	Protocol term for a node that listens for connections and understands natural-language commands — i.e. hivemind-core. A Mind authenticates other nodes, isolates their connections, and authorizes each message.
Master	The node at the top of a hive that receives connections but does not itself connect upward as a client.
Slave	A hivemind-core instance that connects to another hivemind-core instance as a client and accepts bus messages from it; this is how nested hives are formed. A satellite differs from a slave in that it is not itself a hivemind-core instance.
Bridge	A node that links an external service (Matrix, Mattermost, DeltaChat, Twitch, …) to hivemind-core, relaying messages in both directions. See Integrations.
Relay	A satellite that streams audio to hivemind-core for STT/TTS while keeping wakeword detection local. See Voice Relay.

Network structure¶

Term	Definition
Hive	A collection of interconnected nodes — one hivemind-core instance and the satellites (and nested instances) attached to it.
Nested hive	A hivemind-core instance that is also a client of a parent instance, so hives stack into a hierarchy. `ESCALATE` travels up the chain, `BROADCAST` travels down. See Mesh Topology.
`site_id`	A label identifying the physical location or grouping of a node (e.g. a room or building), used for routing and addressing within a hive.
Session	The per-client context a hivemind-core instance maintains for each connected satellite — independent permissions, identity, and conversational state.

Protocol¶

Term	Definition
HiveMind protocol	The encrypted message protocol satellites and hivemind-core instances speak. Transport-agnostic; WebSocket is the default. See Protocol & Message Types and the Protocol Specification.
Handshake	The connection setup that derives a shared session key (PBKDF2-HMAC-SHA256, 100 000 iterations) and establishes an encrypted channel using ChaCha20-Poly1305 or AES-GCM, backed by per-node RSA identities. See Security & Permissions.
BUS message	An OVOS `messagebus` message carried over the hive, the primary payload type for utterances and responses.
Bus message types	The transport verbs that route messages through the mesh: `ESCALATE`, `BROADCAST`, `PROPAGATE`, `QUERY`, `CASCADE`, `INTERCOM` (defined individually below). See Message Types.
`ESCALATE`	A message sent upward to a parent hivemind-core instance when a node cannot handle a request locally.
`BROADCAST`	A message sent downward from hivemind-core to all of its satellites at once.
`PROPAGATE`	A message flooded in all directions across the hive (used for announcements and topology discovery).
`QUERY`	A routed request that expects exactly one answer, which travels back to whoever asked.
`CASCADE`	A scatter/gather request: every reachable node may answer, and a select-callback picks the best reply.
`INTERCOM`	An end-to-end encrypted message addressed to one specific node; relays along the way cannot read it.
Agent protocol	The pluggable AI back-end hivemind-core uses to actually answer natural-language queries — OVOS skills, an LLM persona, or a Google A2A agent. See Plugin Architecture.
A2A (Agent-to-Agent)	Google's open Agent-to-Agent protocol; hivemind-core can use an A2A agent as its back-end via the `hivemind-a2a-agent-plugin`.
Network protocol / transport	The pluggable carrier that moves HiveMind messages between nodes — WebSocket (default), HTTP polling, or MQTT. Transport-agnostic: the same protocol runs over any of them.
Binary protocol	The framing used to carry raw audio (and other binary payloads) over the hive for server-side STT/TTS. See Audio Binary Protocol.
Permission / policy	The deny-by-default rules that decide which message types and skills a given client may use, enforced as a policy chain. A new client is fail-closed: it can send nothing until a type is explicitly allowed. See Security & Permissions.
PBKDF2	The password-stretching function (PBKDF2-HMAC-SHA256, 100 000 iterations) that turns a client's password into the shared session encryption key during the handshake.
AES-GCM	A symmetric authenticated-encryption cipher used (alongside ChaCha20-Poly1305) to encrypt every message on the wire after the handshake.

Credentials & access control¶

Term	Definition
Access key	The per-client identifier a satellite presents to hivemind-core when connecting — the "username" half of its credentials. Issued by `hivemind-core add-client`.
Password	The per-client secret used during the handshake to derive the encryption key (via PBKDF2). The "password" half of a client's credentials.
Identity file	`~/.config/hivemind/_identity.json` on a satellite — stores its access key, password, hivemind-core address, `site_id`, and RSA keys so it can reconnect without re-entering credentials. Written by `hivemind-client set-identity`.
ACL / allowlist	The list of OVOS message types a given client is permitted to send (`allowed_types`). Deny-by-default: anything not on the list is rejected.
`allow-msg`	The `hivemind-core` command that adds a message type to a client's allowlist (its counterpart, `blacklist-msg`, removes one). A brand-new client must be granted at least one type before it can do anything.
`blacklist-skill`	The `hivemind-core` command that forbids a specific OVOS skill for a client; enforced by the OVOS agent policy. (`allow-skill` reverses it.)

Discovery¶

Term	Definition
HiveBeacon	A planned zero-dependency UDP-broadcast auto-discovery transport. The default LAN discovery transport is mDNS/Zeroconf (with UPnP/SSDP as an off-by-default legacy option). See Auto Discovery.
GGWave pairing	An audio-based pairing mechanism that exchanges connection details over sound. Proof-of-concept. See Auto Discovery.

OVOS & voice vocabulary¶

Terms a HiveMind user runs into that come from the OVOS voice stack hivemind-core runs.

Term	Definition
OVOS (OpenVoiceOS)	The open-source voice assistant platform hivemind-core runs skills on. See openvoiceos.org.
ovos-core	The skills/intent engine — the default HiveMind agent.
messagebus / ovos-messagebus	OVOS's internal websocket event bus that skills and services communicate over locally. It is distinct from the HiveMind protocol, which is what hivemind-core and its satellites speak; a BUS message carries an OVOS `messagebus` message across the hive.
Skill	A plugin that handles a category of intents (weather, timers, etc.).
Intent	The parsed meaning of an utterance, used to route it to a skill.
Utterance	A transcribed line of user speech (or typed text).
Wakeword	The trigger phrase (e.g. "hey mycroft") that starts listening.
VAD (Voice Activity Detection)	Detects when speech starts and stops in the microphone stream.
STT (Speech-to-Text)	Transcribes audio into text.
TTS (Text-to-Speech)	Synthesizes spoken audio from text.
Solver	An `ovos-persona` plugin that answers a query (e.g. an LLM backend); personas chain solvers together.
Persona	An `ovos-persona` config — a named chain of solvers — that a Persona backend serves instead of full skills.
OCP (OVOS Common Play)	The OVOS media/audio playback framework.
PHAL	OVOS's Platform Hardware Abstraction Layer — plugins for volume, system control, and similar; relevant to the Home Assistant integration's permission list.