05 orchestrUP

Deutsch · Français

orchestrUP

Jobs across several machines, repeatably.

orchestrUP hands work from one master to agents on your own machines — with dependencies, conditions and a history you can read afterwards.

One master, many agents

The master holds the jobs, schedules them and hands them out. The agents run on your machines, fetch work, carry it out and report back — on Linux, on Windows and on the Raspberry Pi.

Diagram: one master hands jobs to three agents, which report back. Master Agent Agent Agent Linux Windows Raspberry Pi
The master does not call out — the agents fetch their work. That is why no agent needs a reachable address.

Building and testing across machines

One ensemble builds on Windows, tests on the Pi and publishes the artefacts — in one place instead of in scattered logs.

Repeatable instead of started by hand

Schedules, conditions, resuming at the failure, pausing single jobs. What is defined correctly once runs the same way again.

Runs where it should

No service in somebody else's hands: one program, one database, your machines. The licence check works without internet, too.

Four kinds of job

What a job does is decided by its tool. The agent decides which of them it permits.

process
A program with arguments — the everyday case.
plugin
A library the agent loads. For work that needs no process of its own.
wasm
A WebAssembly module, walled off from the machine it runs on.
ai
A model is asked — a bought service or one on your own network. Its answer is used further; on request the agent carries out what it suggests.

More than cron on many machines

Jobs can be joined into an ensemble: dependencies, branches by condition, cleanup steps, values passed from job to job. The difference shows when a step fails — the run can be resumed there instead of only started over.