A control plane where CI processes and infrastructure resources share one model.
Write a pipeline as a typed program, create and use infrastructure through resource libraries, and keep every resource attached to an owner, a lifetime, and a durable execution history.
Early development — the current implementation is usable for development and evaluation.
Traditional CI knows processes but not the resources they create. Traditional IaC knows desired resources but not the recoverable process that needs them. Graphene joins the two:
- a pipeline is a typed Go program and a durable Temporal workflow;
- machines, agents, containers, Kubernetes objects, artifacts, and other resources are records with state, ownership, and lifecycle;
- an outward-connected agent runs work on a machine without opening an inbound control port;
- ownership gives temporary infrastructure deterministic teardown and lets a resource be transferred to a long-lived stand when it must survive a run;
- state, events, logs, metrics, and traces are available through one management surface.
See the documentation for the model, working examples, CLI and SDK guides, and current limitations.
| Repository | Responsibility |
|---|---|
graphene |
Control-plane server, Management API, and graphenectl |
pipeline |
Go SDK and the self-managing pipeline binary runtime |
agent |
Outbound machine agent and its container host |
library |
Resource libraries for Docker, Kubernetes, Git, and files |
studio |
Standalone operator web and desktop application |
docs |
Product documentation |
examples |
Compiling minimal and full pipelines |
temporal-entity |
Durable entity-workflow chassis used by resource kinds |
