Running a pipeline

A run executes one version of a pipeline: it fetches, extracts, processes, checks and publishes. Runs start in four ways: you press Run Pipeline, the schedule fires, an integration receives a webhook, or Jason runs the pipeline as part of its work. This page covers starting runs yourself and reading the list of everything that has run.

1. Run Pipelineor a card's Run
2. Configuredialog: inputs, limit, cache
3. Queuedwaiting for a worker
4. Runningsteps execute
5. Verifiedchecks pass
6. Publisheda new dataset revision

Starting a run

From the pipeline page

Run Pipeline in the hero opens the run dialog, where you can adjust inputs and limits before the run starts. The button is disabled with Pipeline has no active revision until the first version is activated, shows a spinner while a run you started from this page is in flight, and opens Monthly collection cap reached instead of the dialog when the workspace has no rows left.

The dialog runs the version you are currently viewing. If you have selected an older version in the version chip, that version runs.

From a card or the list

The Actions menu on a pipeline card has Run, and the bulk bar on the Pipelines list has Run for every selected pipeline. Both start immediately with the pipeline’s defaults: default item limit, your parameter defaults, the current dataset revisions. See The pipeline page.

From Latest runs

New run on the Latest runs page opens Run pipelines, a checklist of every pipeline in the workspace. Each run uses the pipeline’s default settings. Tick the ones you want (Select all / Clear), and every ticked pipeline starts. Rows read Last run 2 hours ago · 48,102 rows, Never run, or No active revision yet (disabled). The toast confirms Started 3 runs.

The run dialog

The header reads Run pipeline · v2. With anything to configure, Import and Export sit beside it. The body is a split view: the pipeline’s step diagram on the left in configure mode (each block overlaid with its decision, blocks that still need one outlined in red, input edges labelled with row counts), and a panel on the right. Pipelines without a diagram show a flat form instead.

Run settings

The panel opens on Run settings: Settings shared across the pipeline. Pick a block on the canvas or below to decide what it does.

• Parameters. Every parameter with its field, prefilled with your default. Field types are described in Inputs and parameters.
• Blocks with decisions. A list of every input and restartable step with its current state; red text marks one that still needs a choice. Click a row to open that block.
• Attached triggers. When an integration can start this pipeline, a banner names it (Webhook attached · This pipeline can also be started automatically using this integration.) with a link to the integration.

Dataset bindings

Selecting a dataset input on the canvas (or in the list) shows its Dataset revision: a card such as Price observations · Retail / Prices · v12 · Prices · 48,102 rows, or Choose a dataset revision when nothing is bound yet. Click it to open the dataset browser.

Option	Values	Default	What it does
Select a dataset	Workspace datasets grouped by folder, searchable	The dataset the pipeline declares	Which dataset this input reads.
Dataset revision	Every revision of that dataset	The declared dataset’s selected revision, else what you used last time	Which revision the run reads.
Dataset worksheet	The worksheets in that revision	The declared worksheet	Which table, when the dataset has several.
Preview	The first rows	—	A look at what the run will receive.
Leave unbound	Button	—	Optional inputs only: run without this input.

Bindings are remembered: the next run of this pipeline offers the same choice again. Submitting with a required input unbound does not start the run; the dialog highlights the block instead.

Restartable steps

A step that can reuse its own saved output shows a Saved revision choice:

• Build new. Runs this step from scratch. (or Runs this step and saves a new revision.)
• Use saved. Picks a saved revision; This step is skipped and the saved rows take its place. Useful when an expensive early step is fine and only later steps changed.

Value inputs

Value inputs are read-only here: 14 of 16 items on and Values are edited on the Inputs page. Edit inputs closes the dialog and takes you there.

Item limit

The footer shows the run’s scope: Up to 1,000 items per map step or No limit on items per map step. Click it to change it.

Option	Values	Default	What it does
Scope	Limit · No limit	Limit with the pipeline’s default when it has one, else No limit	Whether each map step is capped.
Items	A whole number, at least 1	The pipeline’s default (10 if there is none and you switch to Limit)	Each map step takes at most this many input rows. Rows a step expands from one input are kept whole.

In practice: a map step that visits one page per product takes the first n products. If one product page yields twenty offers, all twenty are kept: the cap counts inputs, not outputs. No limit processes the full input. A run’s Summary records the scope as Full input or Cap 1,000 inputs.

Use a small limit for a trial after a change (“does the new column look right on 25 products?”), then run with No limit or the default. Rows from a capped live run are published like any other, so do not use a cap as a way to preview without publishing; use a Jason test run for that.

Source cache

The second footer item reads Source cache · Use fresh cache or Source cache · Fetch live sources.

Option	Values	Default	What it does
Source cache	Use fresh cache · Fetch live sources	Use fresh cache	Use fresh cache: Reuse saved source responses while they are still fresh. — fresh as defined by the pipeline’s cache setting. Fetch live sources: Check sources again for this run. — every page is fetched again for this run.

Unchanged content can still reuse OCR and extraction. Fetching live does not delete saved responses; it bypasses them for this run only. The pipeline’s freshness setting is documented in Settings reference.

Import and Export

Export run config downloads a file with the exact parameters, output limit, pipeline version, and pinned dataset revisions used by the run. Import run config takes such a file (paste or upload, under 1 MB) and applies it: Nothing changes until you apply it. The dialog rejects a file for another version (This config is for pipeline v1. Open that version before importing it.), unknown parameters or inputs, and bindings to datasets that are no longer available.

Starting it

Run pipeline starts the run. You are taken to the new run’s Steps view, where the diagram animates as work happens. When a run you started completes and the pipeline has exactly one dataset, the page moves on to that dataset.

Run with region = GB, the product list bound to Product catalogue v7, Limit 50 items, and Fetch live sources. The Summary afterwards reads: Live run #8 started by you against v2 (Cap 50 inputs). It completed in 2m 14s, fetched 50 pages and collected 118 rows, published to Live product offers as v9.

Run kinds

Every run has a kind, shown as a pill on its title and in the runs rail.

Pill	Started by	Publishes?
Live run	You, a schedule, a webhook, or Jason running the live pipeline	Yes, when it succeeds. Runs the live pipeline and publishes data when it succeeds.
Resumed run	Resume on a failed or stopped live run	Yes. Reuses completed work from an earlier live run and publishes data when it succeeds.
Test run	Jason, while building, changing or repairing	No. Tests pipeline changes before they go live, without publishing data.
Debug run	Troubleshooting	No. Runs the pipeline for troubleshooting without publishing data.
Newsletter	A newsletter’s schedule	Creates a newsletter issue rather than a dataset revision.

Live and resumed runs are numbered in order (Live run #8), and a number never changes once given. Test and debug runs keep their rows on their own Outputs tab, marked not published, and never touch a dataset.

Runs started by webhook

An integration of type webhook can start a pipeline from outside Jsonify: a request to the integration’s address starts a live run with the pipeline’s defaults. Such runs show webhook as their starter in the rail and on the Latest runs page, and a failure on one is eligible for automatic repair in the same way as a scheduled failure. Setting one up is described in Ingest and serve API.

The Latest runs page

Latest runs in the left navigation lists every run in the workspace across all pipelines.

Option	Values	Default	What it does
Search	Free text, Search runs…	Empty	Matches run titles.
Status	All · Completed · Failed · Cancelled · Active	All	Filters by outcome. Active is queued, running or stopping.
Kind	Live runs (Runs that publish data when they succeed.) · All runs (Live runs, resumed runs, test runs, debug runs, and newsletters.)	Live runs	Whether Jason’s test runs and other non-publishing runs are shown. Kind pills appear on rows only under All runs.
Started from	Anywhere · Manual (Someone pressed Run in the app.) · Schedule (A timer fired on the pipeline’s cadence.) · Webhook (An inbound request to an integration.) · Jason (Jason, in web, Slack, or CLI.)	Anywhere	Filters by what started the run.
Pipeline	All pipelines or one pipeline	All pipelines	Limits the list to one pipeline.
Sort runs	Newest first · Oldest first · Longest duration · Most rows · Most pages · Pipeline name · Status	Newest first	Orders the list.
New run	Button	—	Opens Run pipelines (above).
Pagination	50 per page	50	Pages through the list.

What the page shows

• Active runs. A section headed 3 active runs with a card per run: pipeline, status (Queued / Running / Stopping), live elapsed time and estimate, pages fetched, Stop, and Open run.
• Previous runs. Grouped under Today, Yesterday and Earlier when sorted by time. Each row: pipeline icon, run title, status, rows, pages, duration, an icon for what started it, how long ago, by whom, and an Open run arrow. A run whose pipeline has been deleted reads Pipeline deleted.

Deleting runs

Hover a previous run for its checkbox (Select all / Clear select a page). The toolbar becomes 4 selected with Delete and Clear. Delete asks Delete 4 runs? — This permanently removes the runs and their logs. Datasets they produced are not affected. Active runs cannot be selected; stop them from their cards first.

Empty states: No runs yet · Run a pipeline and it’ll show up here., or with filters No runs match these filters · Try a different status, pipeline, or search.

What’s next