Skip to main content
the full stack of tools, infrastructure, and components that power the Livepeer Network at the node, Gateway, and client level. Livepeer’s architecture is modular and developer-facing: you can run an Orchestrator, build a custom AI Gateway, or use APIs to build media apps on decentralised compute.

Architecture layers

The network sits above the protocol: Gateways and Orchestrators handle off-chain compute and routing; the protocol (Arbitrum) handles staking, tickets, and rewards.

Orchestrator node

The Orchestrator node runs go-livepeer (the livepeer binary), available at: https://github.com/livepeer/go-livepeer

Key components

  • Transcoder selection - Internal or external workers; configured via orchSecret and orchAddr for remote transcoders
  • Ticket validation - L2 TicketBroker on Arbitrum for ETH payment redemption
  • Reward claim - Merkle submission to BondingManager each round
  • LPT staking - BondingManager for self-bond and delegator stake
  • Region advertisement - For Gateway routing (latency, capacity)

Deployment modes

  • Bare metal with GPU
  • Containerized
  • Cloud auto-scaling

Tools

  • livepeer_cli - Stake, set fee/reward cut, monitor sessions
  • livepeer_exporter - Prometheus metrics exporter for observability

Worker layer

Workers can be local or remote services attached to an Orchestrator: Configured via Orchestrator config.json or environment variables.

Gateway infrastructure

Gateways manage:
  • Session auth (API key, ETH deposit, or credit check)
  • Job routing to Orchestrators
  • Session logging and retries
Codebases: Features: Rate limiting, region scoring, health probes, fallback Orchestrators, credit tracking (e.g. Postgres/Redis).

APIs

Endpoints:
  • https://livepeer.studio/api - Studio REST
  • https://explorer.livepeer.org/graphql - Explorer GraphQL

CLI and SDKs

CLI: livepeer_cli (shipped with go-livepeer)
  • Stake LPT, bond/unbond
  • Set Orchestrator fees and reward cut
  • Watch active sessions, query protocol state
SDKs:
  • Livepeer JS SDK - Playback, ingest, AI session tools; works in Node.js and browser
  • Python AI pipelines - Used in internal and community projects (e.g. DotSimulate, MetaDJ)

Monitoring and observability

Node software exposes explicit metrics (e.g. Segment success rate, ticket value sent/redeemed, orchestrator swaps); see Job lifecycle for event/transition details.

Deployment examples

See also

References

Last modified on April 8, 2026