Conductor
A single tool runs from the Viewer’s Process tab on the file in front of you. Conductor is for the next step: chaining tools, models, notebooks and other Relay apps into a workflow you can run again, putting it on a schedule so it runs without you, and seeing every run in one list. Those are its three sections: Workflows, Schedule and Runs.
Workflows
Section titled “Workflows”A workflow is a chain of steps, with data coming in at one end and a result going out at the other. Build it once, then run it on any file, or on every file that arrives in a folder.
The list shows each workflow’s first few steps and a status: Draft (no steps yet), Manual (no schedule), Active or Paused. Filter it with All workflows, Active and Paused, or search by name. Coming up, on the right, lists the next scheduled runs in your local time.
Starting one
Section titled “Starting one”New workflow creates an empty workflow called Untitled workflow and opens it. Click the name to rename it, and the line under it to describe what the workflow is for.
Building it
Section titled “Building it”Steps are listed top to bottom in the order they run. Add a step opens Choose the next step: search, or narrow the list to Relay apps, Processing or Models. A step that cannot follow the one before it, because it needs a kind of input that step does not produce, is greyed out with the reason.
Click a step to set it up in the panel on the right: what feeds it (Input), its parameters, and what it produces (Output). A source step such as Image Input takes a file: pick one with Choose from Drive, or drag it in from the Data app. Add a step after this inserts the next step below it, and Remove step takes it out.
Along the top: whether everything is saved, Save (⌘S), and Run now, which becomes Cancel run while the workflow runs. Undo and redo sit above the steps. When to run, at the head of the chain, and the Schedule tab both open this workflow’s schedules; Runs opens the run history.
A workflow needs a source and every step connected before it will run, and most need an output step as well. Run now tells you what is missing.
Branches: the Graph view
Section titled “Branches: the Graph view”The step list is one straight chain. For branches, or a step that takes more than one input, switch to Graph: a canvas with a searchable palette on the left, grouped into Source, Processing, Model and Output. Drag a step onto the canvas and connect its output to the next step’s input.
Steps from other Relay apps
Section titled “Steps from other Relay apps”Besides processing tools, models and notebooks, a workflow can hand work to the rest of Relay. These are under Relay apps in the step chooser:
| Step | What it does |
|---|---|
| Get the latest results | Picks the newest file in a Drive folder each time the workflow runs. |
| Find reference data | Searches Commons and saves the matching datasets, with citations and source links, as a JSON file in your Drive. |
| Analyze weekly results | Summarises a CSV or TSV table (row counts, missing values, numeric statistics) and saves the summary as a CSV. |
| Run a notebook | Runs one of your published notebooks on the previous step’s result. |
| Capture a view | Saves a PNG of a Viewer pane. The workspace has to be saved, open and visible when the step runs. |
| Check a condition | Checks a value in a CSV or JSON result, or its row count. If the check fails, the run stops there. |
| Create a presentation | Makes an editable Studio presentation that refers to the result, optionally starting from one of your presentations as a template. |
| Review the presentation | Pauses the run until you approve it in Runs. Only approval lets later steps continue. |
| Send to Slack | Posts a message to a channel in your connected Slack workspace. The channel must be enabled in Relay’s Slack settings. |
Where results go
Section titled “Where results go”Finish the chain with one of two output steps:
- Display Output shows the result in a new Viewer pane.
- Export Output writes each result to a Drive folder you choose
(Folder, default
pipeline-results) as PNG or TIFF. The file is named after the input it came from (scan-07.tifin,scan-07.pngout), and each run writes into its own subfolder, stamped with when it ran. Leave Folder at its default in a Project and results go to the Project’s own folder instead.
Running over many files
Section titled “Running over many files”Three ways, depending on how often it happens:
- Once, now: ask Ray (run the nuclei workflow over every file in plates/2026-09) and approve the run when it asks.
- Whenever a file arrives: a schedule that watches a folder. See Automation.
- From a script or cron job: the command line.
Schedule
Section titled “Schedule”Schedule lists every schedule. Opened from a workflow, it shows just that workflow’s; All schedules shows the rest.
- Agenda lists each schedule with how often it runs, its timezone and when it runs next, and marks one whose last run failed. Edit, pause, resume or delete a schedule from its row.
- Week puts each active schedule’s next run on a weekly calendar. The arrows and This week move between weeks.
New schedule sets a workflow or notebook to run on the clock or whenever a new file appears in a Drive folder, with a plain-language repeat menu and your timezone. The details are on the Automation page.
Every workflow run is listed here, newest first, whether you started it, Ray did, or a schedule did overnight. Model jobs, file conversions and uploads are listed too. Each row shows when the run started, its status (Queued, Running, Needs review, Completed, Failed or Cancelled) and how long it took.
View run opens it in place: each step, how far it got, and why any step failed. Depending on the kind of run, the same panel offers Open in Viewer, Run again after a failure, Cancel while it is going, Details, and Delete run.
A run waiting at a Review the presentation step shows Review instead. Open it to preview the presentation, then choose Approve & continue or Reject & stop.