Automations: what runs, and how each answer is made
Automations is three panes side by side: your automations on the left, the chosen one's recipe in the middle, and its details and runs on the right. An automation is what the API calls a workflow: the address names it as ?workflow=<slug>, so a link opens the same one.
The list
Automations lists each automation you have saved, one row each: its health as a dot, its name, and its last run in words at the right ("30 min ago", Running, Failed, No run yet), with its version under the name. The title row counts the health of them all, such as "3 healthy · 1 waiting":
| Health | Means |
|---|---|
| Ready | It feeds no bridge or destination: its results are readable over the API |
| Healthy | Every bridge it feeds has read its results and checked in |
| Waiting | A bridge has not read its results or checked in yet |
| Failing | The newest run's deliveries failed |
| Unknown | The bridges' check-ins could not be read |
A value every row shares is said once above the rows: "All deliver to Yamcs." (or "All are read over the API."). Past 20 automations a search box narrows the list by name. + New automation at the foot of the list opens the templates to start from.
The recipe
Choose an automation and its recipe fills the middle pane, read-only, as five cards joined by a line:
| Card | Holds |
|---|---|
| Price | What a run prices: passes or links, over each run's window, or for an audit the last 7 days against your pass log |
| Using | The models each pass is priced with |
| Calculate | The steps that turn what the models say into numbers, in order |
| Decide | What it recommends, and whether each recommendation is sent as advice or held for your approval (on a plan without Act, each is shown as advice); Recommends nothing when it has no such step |
| Deliver to | The API, and each system or destination its results go to |
Edit in the pane's title row opens the builder on it, a page of its own (?view=builder&workflow=<slug>), with Automations as the way back. Share copies the link or the API call behind the view.
Details and runs
The right pane, Details, reads as label and value: Version, Delivers to, Plan it needs (Any plan, or Network for one that sums capacity; on a plan without Act, one that recommends reads "Advice on Explore · approvals need Operator + Act"), Health with its reason, Last run, and Trial when the newest run recorded recommendations without sending them.
Run now, next 48 h runs the published version over the next 48 hours, the window Passes looks ahead over, or an audit over the last 7 days of live telemetry. It delivers, so it asks once first, naming where results go. It is not offered on Explore, which runs no delivering run; a refusal shows under the button with a link to compare plans.
Runs, under the details, lists the automation's newest runs, newest first, each a time and what happened in plain words: "Pass yield v2 priced 42 passes", "held 2 recommendations for you", "failed", "(test run)". With no automation chosen, the right pane is Runs for all of them: the newest run of each, and the runs started here, refreshed while the page is open. Escape or the close button clears the choice.
New automation
+ New automation opens the templates maintained by Constellation, one row each with the question it answers, named as the API names them. The one that fits your fleet and your plan is already chosen: Pass yield, unless you have it, then the next one your plan can save.
- Open in the builder opens the chosen template as an unsaved copy named for it.
- You have one: this makes another marks a template you already have a copy of; opening it makes a second copy.
- A plan name beside a template, such as Network, is the plan it needs when yours is below it.
- A template this deployment cannot run is dimmed and says why in a few words; its tip gives the reason.
In the builder, Test prices the window and delivers nothing. The API tests only a saved version, so while you have unpublished edits the button reads Publish and test: it publishes them as the next version first. The builder guide and the template catalog go further.
In the API
| On screen | Call |
|---|---|
| The list | GET /predictions/workflows |
| An automation | GET /predictions/workflows/{slug} |
| Its newest run, Runs | GET /predictions?workflow=<slug> |
| Run now | POST /predictions with the window; see Long runs |
| The templates | GET /predictions/workflows?templates=true |
| Health | The bridges' last reads, GET /predictions/feed/readers, and the newest run's deliveries |
Next: audit a past week of passes.