Deploy and run workflows

Manage installed Definitions, revisions, and Run inputs.

Deploy a Definition into an Environment to create a Pypline. Start a Run when you want to execute it. These are separate operations: many Runs can use the same deployed revision.

Deploy a Definition

After connecting to a server, use an Environment with a configured registry and compatible runner pool:

Terminal

pyplines validate ./workflow.yaml
pyplines deploy ./workflow.yaml --environment development --as example
pyplines info development/example

--as supplies the installed name without changing the source Definition. If omitted, it defaults to the Definition's name. Names are unique within an Environment.

Deployment resolves every Action and captures its contract and image digest. A registry tag changing later does not silently change an existing revision.

Update and inspect revisions

Deploying to an existing name creates another revision. The CLI reads the current revision before submitting an update. For an explicit concurrency check, pass the revision you expect to replace:

Terminal

pyplines deploy ./workflow.yaml --environment development --as example --revision 1
pyplines history development/example
pyplines info development/example --revision 1
pyplines export development/example --revision 1 --output ./original.yaml

The update fails if another deployment has changed the expected revision. Use --revision 0 when you require a new installation. Export writes the portable Definition, not execution evidence, and refuses to overwrite an existing output file.

Start a Run

Terminal

pyplines run start development/example --inputs ./inputs.json --wait

Omit --inputs if no inputs are required. By default the Run captures the latest revision; --revision 1 explicitly selects an older one. Existing Runs retain their revision when a new version is deployed.

--wait observes execution, including approval waits, with a default timeout of 1,800 seconds. --timeout changes the local wait limit. A timeout stops observation, not execution; do not start another Run merely because the wait ended.

Repeat requests safely

Deployment and Run creation accept --key for an idempotency key. Automation should persist a unique key for one logical request and reuse it only to retry that same request after an uncertain response. Reusing a key with changed content or a different actor is rejected. Creating a new key requests new work.

Promote a procedure

Deploy the same reviewed Definition into another Environment, with that Environment's registry, context, secrets, platform, and policies configured separately. Deployment resolves Actions again for that target. Verify its resulting digests; the same version label in two registries is not proof of identical content.

Use monitoring to follow execution and recovery when a Run cannot proceed.

Build a Definition

Compose existing Actions into a portable operational workflow.

Inputs, context, and secrets

Bind operational parameters and environment-specific values without embedding credentials in YAML.