Glossary¶

A¶

Attendee A standard participant in a Meet room. Can join, speak, share screen, and send reactions. Cannot manage other participants or start recording without owner approval.

C¶

Celery A distributed task queue used by the Meet backend for async operations (email notifications, recording processing) and by the Summary service for transcription and summarization jobs.

Composite recording A recording mode that captures the entire room view as a single video file — the mixed output of all participants, as seen on screen. Produced by LiveKit Egress.

D¶

DID (Direct Inward Dialing) A phone number assigned to a SIP trunk that allows external callers to dial into a Meet room via PSTN.

DTLS-SRTP Datagram Transport Layer Security — Secure Real-time Transport Protocol. The encryption standard used by WebRTC to protect media streams in transit. All Meet media is encrypted with DTLS-SRTP.

E¶

Egress (LiveKit Egress) A LiveKit component that processes room output: recording to files, streaming to RTMP endpoints, or generating composite views. Meet uses Egress for meeting recording.

J¶

JWT (JSON Web Token) A signed token used to authenticate clients with the LiveKit server. The Meet backend generates JWTs for users, encoding their identity, the room they can join, and their permissions.

L¶

LiveKit The open-source WebRTC infrastructure library and server that powers Meet's real-time media. Developed by LiveKit Inc. under the Apache 2.0 license.

LLM (Large Language Model) An AI model used by the Summary service to generate meeting summaries from transcripts. Configurable — can use OpenAI, a local model, or any OpenAI-compatible API.

Lobby A waiting room feature. When enabled, new participants wait for the host to admit them before entering the meeting.

M¶

MCU (Multipoint Control Unit) A legacy conference server architecture that mixes all streams server-side into one stream per participant. High CPU cost, lower scalability. Meet uses an SFU instead.

Garage An open-source S3-compatible object storage server designed for self-hosted, geo-distributed deployments. Garage is lightweight (runs on 1GB RAM), actively maintained, and recommended for self-hosted Meet instances. For large-scale deployments, MinIO/AIStor or Ceph RGW are alternatives.

Moderator A participant role between Owner and Attendee. Can mute participants, admit from lobby, and start/stop recording.

O¶

OIDC (OpenID Connect) An authentication protocol built on OAuth 2.0. Meet uses OIDC to authenticate users via an external identity provider (Keycloak, Authentik, Google, etc.).

P¶

Participant Any person connected to a Meet room. Has a role (owner, moderator, attendee), media tracks, and metadata.

PSTN (Public Switched Telephone Network) The global telephone system. Meet can connect to PSTN via LiveKit's SIP bridge, allowing phone dial-in and dial-out.

Push-to-talk A mode where holding a key (V) temporarily unmutes your microphone. Useful in formal meetings.

R¶

Room A virtual meeting space in Meet. Has a unique URL, an owner, and configuration settings. Rooms are non-persistent by default — state clears when the last participant leaves.

RTP (Real-time Transport Protocol) The network protocol used to transmit audio and video streams. LiveKit uses RTP over UDP for media delivery.

S¶

SFU (Selective Forwarding Unit) A conference server architecture that receives media streams from senders and selectively forwards them to receivers without decoding or re-encoding. LiveKit is an SFU. More scalable and CPU-efficient than an MCU.

Simulcast A WebRTC technique where a browser encodes and sends the same video at multiple quality levels simultaneously. The SFU (LiveKit) delivers the appropriate quality to each receiver based on their bandwidth.

SIP (Session Initiation Protocol) The signaling protocol used in telephony. LiveKit's SIP bridge translates between SIP and WebRTC, enabling phone participants to join Meet rooms.

STT (Speech-to-Text) Technology that converts spoken audio to text. Used by the Summary service to generate meeting transcripts. Powered by Whisper or compatible engines.

SVC (Scalable Video Coding) An advanced video codec mode (available in VP9 and AV1) where a single stream contains multiple quality layers. More efficient than simulcast for spatial and temporal scalability.

T¶

TURN (Traversal Using Relays around NAT) A relay protocol that allows WebRTC clients behind restrictive NAT or firewalls to connect by routing media through a relay server. LiveKit supports TURN configuration.

V¶

VAD (Voice Activity Detection) Technology that detects when a participant is speaking. Used by Meet's metadata collector to track speaking patterns and assign transcription segments to speakers.

W¶

WebRTC Web Real-Time Communication — a browser standard for peer-to-peer audio, video, and data communication. All browser-based media in Meet uses WebRTC. LiveKit is a WebRTC SFU.

Whisper OpenAI's open-source speech recognition model. Used by the Summary service for meeting transcription. Available in multiple model sizes (tiny, base, small, medium, large).