Open-core model¶
Repster is open-core: the SDK and CLI are open source (Apache-2.0); the cloud control plane is proprietary.
Open source vs. Cloud¶
Open source (repster) |
Repster Cloud |
|---|---|
@rp.flow() decorator and flow discovery |
Serverless execution (isolated worker per run) |
Local runs and scheduler (rp run, rp dev) |
Managed scheduling (crons fire without your machine) |
| Run history, logs, retry | Encrypted secrets vault |
| Local web UI | Authentication, orgs, roles, API keys |
Deploy client (rp deploy, rp secrets push) |
Deployment registry and hosted run history |
| Metered, transparent billing (upcoming) |
Everything in the left column runs entirely on your machine, forever, for free.
What's open source¶
repster (Apache-2.0): everything you install:
@rp.flow()decorator and flow discovery- Local runner and scheduler
- SQLite run history
- Archive packaging
- CLI (
rp/repster), including all cloud commands (login,deploy,secrets) - HTTP client for the cloud API
You can read, fork, and audit exactly what runs when you execute rp deploy or rp run --cloud.
What's proprietary¶
repster-cloud-service: the hosted control plane:
- Deployment registry and artifact store
- Secrets vault
- Cloud execution and worker infrastructure
- Run state, logs, and scheduling
- The Repster Cloud UI
- Billing
The control plane is never distributed; it runs as a hosted service.
Why this split¶
Repster Cloud is the product. The SDK is open because trust sits at the code boundary: your pipeline code runs through the open-source repster.runner contract, identical in local and cloud execution. Everything the CLI transmits to Repster Cloud is visible in open-source code.
The cloud launcher (the code that gets a run onto a worker) lives in the proprietary control plane because workers run in Repster's infrastructure. There's no code running in customer infrastructure that would need to be inspectable.