Discover & Connect Marketplace Compute Services

This page explains how to find and broker Orchestrator AI & Video Offerings via your Gateway for Livepeer Application consumption.

Page is under construction.
Feedback Welcome!

This page should be streamlined.

Options are:

• on-chain vs off-chain
• ai vs video vs dual

What it explains:

• Discovery
• Filtering
• Routing

This page is a work in progress. Needs a lot of edits.

Show TODO

TODO: - [ ] Overview, - [ ] Edit (A LOT) - [ ] Streamline, - [ ] Format, - [ ] Style, - [ ] Link to github / otherresources

How to Discover & Connect Marketplace Offerings

This page explains how Gateways can discover and offer Orchestrator Services, available on the Marketplace, to Applications & Builders.

Discovery Process

Gateways discover orchestrators through the orchestrator pool in discovery/discovery.go discovery.go:64-111

1. Query orchestrators: Get OrchestratorInfo from each orchestrator
2. Filter capabilities & price: Match against required capabilities & pricing limits & optionally rank results
3. Expose results: Make matching orchestrator services available through HTTP endpoints
4. Route requests: Forward incoming requests to selected orchestrators

Find All Orchestrators & Offerings

The /getNetworkCapabilities endpoint in server/handlers.go exposes all available orchestrator offerings

func (s *LivepeerServer) getNetworkCapabilitiesHandler() http.Handler {
    return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
        // Returns orchestrators with their capabilities, pricing, and hardware
    })
}

Response Data Structure

The response uses NetworkCapabilities structure from common/types.go types.go:166-176

type NetworkCapabilities struct {
    Orchestrators []*OrchNetworkCapabilities `json:"orchestrators"`
}

type OrchNetworkCapabilities struct {
    Address            string                     `json:"address"`
    LocalAddress       string                     `json:"local_address"`
    OrchURI            string                     `json:"orch_uri"`
    Capabilities       *net.Capabilities          `json:"capabilities"`
    CapabilitiesPrices []*net.PriceInfo           `json:"capabilities_prices"`
    Hardware           []*net.HardwareInformation `json:"hardware"`
}

Gateway

Gateways expose offerings describing:

1. Supported Models

Example:

• stable-diffusion-v1.5
• depth-anything
• controlnet_lineart
• ip_adapter

2. Supported Pipelines

• Daydream-style real-time style transfer
• ComfyStream workflows
• BYOC containers
• Custom inference graphs

3. Pricing

A Gateway may price:

• $0.004 / frame
• $0.06 / second
• $0.0005 / inference token

4. Regions

Example:

• us-east
• eu-central
• ap-southeast

5. Reliability Scores

Gateways may publish:

• Availability %
• Average inference latency
• Failover configuration