This page is a work in progress.
TODO: Overview, Edit, Streamline, Format, Style, Link to github / other resources
TODO: Overview, Edit, Streamline, Format, Style, Link to github / other resources
How to Publish Offerings to the Marketplace
The Livepeer Marketplace allows Gateways and Orchestrators to advertise compute capabilities, pricing, and performance characteristics. This page explains how operators publish offerings so applications can discover and select them programmatically.What Is an Offering?
An offering is a structured declaration of:- Supported models (e.g., SDXL, ControlNet, depth models)
- Supported pipelines (ComfyStream, Daydream, BYOC)
- Pricing per frame / per second / per request
- GPU tiers and performance metrics
- Regional availability
- SLAs and expected latency
Orchestrators publish compute-level offerings.
Publishing as a Gateway
Gateways expose offerings describing:1. Supported Models
Example:stable-diffusion-v1.5depth-anythingcontrolnet_lineartip_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-easteu-centralap-southeast
5. Reliability Scores
Gateways may publish:- Availability %
- Average inference latency
- Failover configuration
Publishing as an Orchestrator
Orchestrators publish compute offerings showing:GPU Inventory
- 4090
- A40
- L40S
- V100
Model Compatibility
- TensorRT support
- Torch Compile accelerations
Benchmarks
- Frames per second
- Tokens per second
- Latency per model
Marketplace Submission Workflow
Updating Offerings
Operators can dynamically update:- Pricing
- Model support
- Pipeline versions
- GPU availability Performance metrics
Summary
Publishing offerings to the Marketplace enables:- Discoverability
- Fair competition
- Informed selection
- Differentiation and specialization