Inputs and parameters

A pipeline reads two kinds of thing from outside itself. Inputs are data: a dataset another pipeline published, a list of URLs or product identifiers you maintain, a file. Parameters are settings with a type and a default: a region, a date, a maximum count. Both live on the Inputs page of the pipeline (How it works › Inputs), and both can be changed for a single run in the run dialog.

1. Declared defaultset by Jason in the pipeline
2. Your defaultset on the Inputs page
3. Per-run valueset in the run dialog
4. Runuses the most specific

The Inputs page

The page shows one block per declared input or parameter, in the order the pipeline declares them, each with the description from the statement of work. There is no commenting here; to change what an input means, comment on the statement of work.

Dataset inputs

A dataset input is shown as a dataset card: name, row counts and current version. The pipeline reads that dataset’s selected revision on every run.

To read a different revision for one run, or a different worksheet, use the run dialog: see Per-run overrides below. To point the pipeline at a different dataset permanently, ask Jason.

Read the product list from @catalogue_2026 instead of @catalogue_2025 from now on.

Value inputs

A value input holds data you maintain by hand: a list of URLs, a list of store ids, a single threshold. Each block shows the current value: scalars as text, URLs as pills, lists as a grid of items.

Including and excluding items

Every list item has a switch. Turn one off and it stays in the list but is left out of what the run receives (tooltip Included in runs — click to exclude / Excluded from runs — click to include). This is the quickest way to pause one retailer or one URL without editing anything.

Editing values

1. Press Edit on the block. The value becomes editable and the block gains Upload, Cancel and Save.
2. Change the items. Lists of simple values use rows with Move row up, Move row down, Remove row and an add-row control. Scalars use a field typed by the value’s declared type: whole number, decimal, true/false, or text.
3. Save. The toast Input saved confirms it. The new value is used by the next run.

Lists of objects

A list of objects (for example { "sku": "…", "retailer": "…", "url": "…" }) opens in a dedicated editor when you press Edit.

Option	Values	Default	What it does
View	Table · JSON	Table	Table shows one row per object with a column per key. JSON shows the same list as text you can edit directly.
Add row	Button	—	Appends an empty row.
Paste rows	Text box, then Add rows	—	Paste tab- or comma-separated rows copied from a spreadsheet; each becomes an object.
Import file	File picker	—	Loads rows from a CSV or similar file.
Columns	New column name field; Remove column on each header	—	Add or drop keys across every row.
Rows	Move row up · Move row down · Remove row	—	Reorder or drop rows.
Save / Cancel	Buttons	—	Save writes the list back to the input.

Uploading a file into an input

Upload (visible while editing) opens Import “” from file. The file is read and reshaped to fit the input: a column of URLs becomes a list of URLs, a sheet becomes a list of objects. The result reads Imported 42 items from stores.csv, or Import failed with the reason.

Parameters

A parameter is declared by the pipeline with a name, a type, a description and usually a default. Parameters appear on the Inputs page, in the statement of work’s Parameters strip, and in the run dialog.

Parameter defaults

Each parameter block shows its current default and where it comes from:

• Your default · the pipeline declares <value> when you have overridden the default on this page. Reset returns to the declared value; a toast confirms it.
• The declared default applies to every run unless changed in the Run dialog when you have not overridden it.

Press Edit, set a value with the field for its type, and Save (toast Default saved). The override is stored on the pipeline, survives new versions, and applies to every run, scheduled or manual, until you reset it. File parameters cannot be given an operator default; set them per run.

A pipeline declares region = US because that is what it was built against. Your team only sells in the UK. Set Your default to GB once, and every scheduled run uses it without touching the pipeline.

Field types

The same fields appear on the Inputs page and in the run dialog. Required parameters carry a red *. Labels are the parameter’s name in title case with acronyms kept upper-case (Start URL, Product ID).

Type	Field	Notes
string	Text box	Placeholder shows the declared default as ex: <default>.
url	Text box	As string; validated as a URL.
int	Number field, steps by 1	Honours the declared minimum and maximum.
float	Number field	Honours the declared minimum and maximum.
date	Relative or fixed date with a timezone	See Relative dates.
bool	Checkbox
enum	Select	The options the pipeline declares.
list[string], list[int], list[float], list[url]	Multi-line box, One per line	Each line becomes one item of the declared inner type.
file	Upload or pick a file	Opens **Upload file for “"**. The hint shows **Default: (synthetic sample)** when the pipeline shipped with a sample, or **No file selected**. **Remove file** clears it.

A type the app does not recognise renders as Unknown param type; tell Jason and it will correct the declaration.

Relative dates

A date parameter can be a fixed calendar date or a relative one that resolves when the run starts.

Option	Values	Default	What it does
Kind	Relative · Fixed	As declared	Relative dates are today, yesterday or tomorrow. Fixed dates are a calendar date.
Relative value	today · yesterday · tomorrow	As declared	Resolved at the moment the run is requested.
Fixed date	Date picker	As declared	Used as-is on every run.
Calendar timezone	A timezone	The pipeline’s calendar timezone	Defines when “today” starts. Choose the calendar timezone that defines this relative date.
Restore	Link	—	Returns to the declared default.

The field warns you when it cannot resolve a relative date reliably:

• No calendar timezone is configured. Until one is set, relative dates use UTC.
• This pipeline has an invalid calendar timezone. Choose a valid IANA timezone.
• This pipeline has an invalid date default. Choose a relative or fixed date.

Scheduled pipelines almost always want a relative date. since = yesterday with the calendar timezone set to Europe/London means every 06:00 run collects the previous London day, including across daylight-saving changes.

Per-run overrides

Run Pipeline on the pipeline page opens the run dialog, where every parameter and dataset input can be set for that run only. Nothing you set there changes the pipeline or your defaults.

• Parameters. The Run settings panel shows every parameter with its field, prefilled with your default (or the declared default). Change any of them. The run’s Summary afterwards lists each parameter that differed from its default as <label>: <value> (default <x>).
• Dataset inputs. Each dataset input shows a Dataset revision card (Price observations · v12 · Prices · 48,102 rows). Click it to open the dataset browser and choose another dataset, revision or worksheet for this run. Optional inputs can be left unbound with Leave unbound. The dialog remembers the bindings you used last time.
• Value inputs. Shown read-only (14 of 16 items on) with Values are edited on the Inputs page. and an Edit inputs shortcut that closes the dialog and opens the Inputs page.
• Saved revisions. A step that can reuse its own earlier output offers Build new or Use saved. Use saved skips the step and takes the saved rows in its place; the panel says so.

Export in the dialog header downloads the exact parameters, item limit, pipeline version and dataset bindings as a file; Import applies one you exported earlier, so a run can be reproduced precisely. The same Export exists on a run’s Inputs panel. Item limits and the source-cache choice are covered in Running a pipeline.

A required dataset input with no binding stops the run before it starts: the dialog highlights the block that needs a decision (One block needs a decision before the pipeline can run.) instead of submitting.

What’s next