Example apps for the Livepeer live runner

Example apps that run on the Livepeer live runner — go-livepeer's new way to run any app on the network. Each app is a plain HTTP / WebSocket / video service: an orchestrator running the live runner hosts it, and a client calls it through the orchestrator with the livepeer-gateway SDK.

The point is to swap the compute without changing your app — permissionlessly, no lock-in. Your app stays a plain service with little or no Livepeer-specific code, so you're never tied to us. And the network is permissionless: anyone can run or extend it, no one gatekeeps what you deploy, and no single party can take your app down. Write the app once; move the compute freely.

Note

Live runners ship in mainline go-livepeer since v0.9.0. The Python SDK still comes from the ja/live-runner branch until it is published to PyPI.

How it works

Your app is a plain service that clients reach through the orchestrator — the SDK handles discovery / session / payment, and, on-chain, a remote signer settles it. The client never talks to your app directly.

flowchart LR
  client["Client<br/>(livepeer-gateway SDK)"]
  orch["Orchestrator<br/>proxy · discovery · payment"]
  app["Your app<br/>HTTP / WebSocket / trickle"]
  signer["Remote signer<br/>(on-chain)"]

  client -->|"single-shot: discover → call · persistent: + reserve/release"| orch
  orch -->|"forwards your endpoints, unchanged"| app
  app -.->|"dynamic: register_runner · static: runners.json"| orch
  signer <-.->|"micropayment tickets"| orch

Loading

Communication schemas

The orchestrator is a transparent reverse proxy: every endpoint you expose is passed through to your app unchanged, so you write an ordinary service and it runs on the network as-is. The transports supported today:

• HTTP request/response — the common case. (hello-world, tiles, api-proxy)
• HTTP + SSE — streamed / token responses. (vllm)
• Trickle — continuous realtime video in/out. (echo)
• WebSocket — long-lived bidirectional sessions. (external: scope)

Need a schema that isn't here? Open an issue.

Examples

Example	Goal	Registration	Mode	Transport	Pricing
hello-world	The simplest app: one request, one response	dynamic	single-shot	HTTP (JSON)	fixed
tiles	Capacity fan-out — one call per tile	dynamic	single-shot	HTTP (base64 PNG)	fixed
api-proxy	Pass calls through to a hosted API — the operator holds the key, callers pay per call	static	single-shot	HTTP (JPEG bytes)	fixed
echo	Realtime video, transformed and echoed back	dynamic	persistent	trickle	— (offchain only)
vllm	Drop-in OpenAI API; the client stays unmodified	static	persistent (single-shot by nature)	HTTP + SSE	hour

Start with hello-world (the smallest end-to-end path); the others each layer on one new idea. More will follow, including a full example that exercises every feature. Each is self-contained and runs offchain (free, no wallet); most also run on-chain (paid) — see each README.

This set stays minimal and curated: it covers each value of the axes above (registration, mode, transport, pricing), not one example per app. New examples land here only when they fill a gap in that table — apps built on the runner belong in their own repos, listed under External examples.

Registration

How the app attaches to the orchestrator:

• Dynamic — the app self-registers via the SDK (register_runner) and heartbeats; the orchestrator drops it when heartbeats stop. Best for apps that come and go. (hello-world, echo)
• Static — the orchestrator is configured with the app's URL in a runners.json and health-polls it; the app needs no SDK. Best for fixed, long-running deployments. (vllm, api-proxy)

The arrow flips — dynamic, the app announces itself; static, the orchestrator is told about a passive app:

flowchart LR
  subgraph Dynamic
    direction LR
    a1["App<br/>(embeds SDK)"] -->|"register_runner + heartbeat"| o1["Orchestrator"]
  end
  subgraph Static
    direction LR
    o2["Orchestrator<br/>(reads runners.json)"] -->|"health-poll"| a2["App<br/>(no Livepeer code)"]
  end

Loading

Runner modes

Chosen at registration (above); defaults to persistent — set on both register_runner(...) and in runners.json. The examples set it explicitly.

• Persistent — a held-open session the client reserves and releases, billed per second of wall-clock (or once, with fixed pricing). Best for realtime / streaming. (echo, vllm)
• Single-shot — one request in, one response out; the orchestrator reserves a session per call and releases it when the response returns, so the client manages no session at all. Best for batch / request-response. (hello-world, tiles, api-proxy)

Note

The vllm example is single-shot by nature but stays persistent for now: it meters per second across a reserved session, and true per-token billing is brokerage for the gateway/signer layer.

Calling your app

The client side depends on the runner's mode:

• Single-shot — discover → call: find the app via runner_selector, then one call_runner. The orchestrator reserves a session for the call and releases it when the response returns; on the paid path call_runner answers the 402 payment challenge inline. (hello-world, tiles, api-proxy)
• Persistent — discover → reserve → call → release: reserve a session (reserve_session), call it — call_runner, streamed frames, or a WebSocket, depending on transport — then release it (stop_runner_session), which settles payment on-chain. (echo, vllm)

Each example's client.py shows its exact calls — grep # Livepeer: to find them.

External examples

Apps that integrate the live runner and live in their own repos — production deployments and standalone examples alike. This table is links-only: the code, CI, and support stay with the author.

Project	What it is	Transport
daydreamlive/scope	Real-time AI video with downloadable LoRA models	WebSocket + trickle
livepeer/api-proxy	Attach several API endpoints dynamically — key storage and request stats for operators	HTTP

Built one? Open a PR that adds a row. To make your repo easy to find, follow the community convention:

• Name the repo <app>-livepeer-runner (e.g. comfyui-livepeer-runner).
• Add the livepeer-runner GitHub topic.
• Mention Livepeer in the repo description.

Running the examples

Each example is self-contained and its README has the run commands. Everything below is the shared setup they all build on — the examples spin up a local orchestrator (and, on-chain, a signer) via the compose files here, so you set this up once, not per example.

Prerequisites

• Docker for the end-to-end demos. They use the mainline livepeer/go-livepeer:v0.9.0 release image, so there is nothing to build.

• Python 3.12+ and uv for the client.

• The livepeer-gateway SDK from the ja/live-runner branch (not yet on PyPI):

  pip install "git+https://github.com/livepeer/livepeer-python-gateway@ja/live-runner"

Shared components

The orchestrator and signer services are defined once at the repo root and pulled into each example with Docker Compose extends, so examples don't duplicate them:

• compose.orchestrator.yml — the offchain orchestrator (-useLiveRunners).
• compose.onchain.yml — adds a remote signer and re-points the orchestrator on-chain.

On-chain (paid) setup

On-chain runs add a remote signer that holds the payer wallet and mints probabilistic micropayment tickets; the orchestrator redeems the winning ones. Shared across examples:

• Wallets stay outside the repo — *_KEYSTORE_DIR points at go-livepeer keystores (mounted read-only); only the address + password come from .env.
• .env is per example and gitignored — copy .env.example and fill in RPC, network, keystore paths, accounts, and pricing (it holds the keystore password).
• Runner price is a plain USD amount: the app advertises PRICE (e.g. 0.01). currency and unit default to usd / hour, so only price is required. With hour the orchestrator converts it to wei via the price feed and meters the session per second. Apps with bounded per-call work can register unit="fixed" to bill the price once per session (see hello-world, tiles). The signer caps what it pays at MAX_PRICE_PER_UNIT, compared per billing unit: one second of runtime when metered (0.000111USD is about 0.40 USD/hour), the whole session price when fixed.
• Payments are probabilistic — on a short run you'll rarely see a redemption; that's expected.

Verifying discovery

Before running a client, confirm the orchestrator actually advertises your runner with the expected price by calling /discovery directly:

curl -sk https://localhost:8935/discovery | jq

Each entry lists its runners with an app, version, capacity, and price_info. The orchestrator republishes your USD price converted to wei — price in wei with currency: wei and unit: seconds (720p-pixel-seconds for per-pixel video, fixed for once-per-session pricing). Check that your app appears and the price is non-zero.

Conventions

• Apps bind to 127.0.0.1 by default (safe for local runs). In a container the compose files pass --host=0.0.0.0 so the orchestrator can reach the app.
• Orchestrators serve a self-signed TLS cert; the SDK skips verification.

Contributing

Issues and PRs welcome — see CONTRIBUTING.md.