The pipeline page

A pipeline is the repeatable definition of a dataset. Jason builds it, workers run it, and the pipeline page is where you read what it does, run it, and ask for changes. This page covers the list of pipelines, the cards that summarise each one, and the pipeline page from the hero down to its five sections.

1. Pipelines listevery pipeline and dataset
2. Pipeline pagehero, sections, rail
3. RunRun Pipeline or a schedule
4. Outputsdatasets, files, integrations

The Pipelines list

Open Pipelines in the left navigation. The list holds everything in the workspace that produces or holds data: every pipeline, the datasets pipelines produce, and standalone datasets that nothing produces (uploads, imports, samples).

Toolbar

Option	Values	Default	What it does
Search	Free text, placeholder Search pipelines and datasets…	Empty	Matches a pipeline’s name, description or slug, or the name of any dataset it publishes. Standalone datasets match by name. While you search, folders are flattened so matches from every folder appear.
Filters	A button with a sliders icon. It opens a panel with two lists, Show and Pipeline status, and shows a count when either is set.	Both on their first option	See the two rows below.
Show	Everything · Pipelines · Standalone datasets · Connected datasets	Everything	Which kinds of item are listed. Each option shows its count. Standalone datasets are uploads, imports and samples that no pipeline produces; Connected datasets are datasets a pipeline produces.
Pipeline status	Any · Active Revision · Active runs · No Active Revision · Failed	Any	Filters pipelines by state, with a count on each option. Active Revision means a reviewed version is live; No Active Revision means the build has not finished; Active runs means a run is queued, running or stopping; Failed means the last run failed. Choosing anything but Any hides datasets.
Sort	Last Updated · Date Created · Name	Last Updated	Orders the cards.
Clear filters	Button	Hidden	Appears when a search or filter is active and resets all of them.
View options	Large grid · Small grid · List	Large grid	Changes the layout. Your choice is remembered per page in this browser.
New Pipeline	Button with a menu	—	Opens the new-pipeline composer. The menu also offers Upload data, Open playground and New folder. The Add pipeline tile at the end of the Pipelines section does the same as the button.
Pagination	25 · 50 · 100 per page	50	Shown only when there are more items than the smallest page size.

Sections and folders

Items sit in three collapsible sections, each with a count and a one-line description:

Section	Description in the list	Default
Pipelines	Collect and refresh data on a schedule or on demand	Expanded
Standalone datasets	Data you uploaded, imported, or ingested; no pipeline produces these	Expanded
Connected datasets	Output a pipeline produces; each one updates when its pipeline runs	Collapsed, unless Show is set to Connected datasets

When the current folder has subfolders or a parent, a Folders section appears above them.

• Folder tiles show the folder name, <n> subfolders · <n> items, and a few item icons. The tile menu offers Rename folder and Delete folder. Inside a folder, a Back to tile returns to the parent.
• Drag to move. Drag one or more selected cards onto a folder tile to move them. Or select cards and use Move to folder in the bulk action bar, which opens a folder picker with an inline New folder field. A connected dataset stays in its pipeline’s folder; move the pipeline instead.
• Lasso select. Drag on empty panel space to draw a selection rectangle around several cards.

The New Pipeline composer

New Pipeline opens a dialog titled New pipeline where you describe what to collect. Everything you type becomes the opening message of a new chat with Jason.

Option	Values	Default	What it does
Brief	Multi-line text, placeholder Describe what you want to collect. Type @ to reference a dataset, or attach files.	Empty	What you want. Drop files onto the box to attach them.
@ mentions	Type @ and a word	—	Opens a Datasets picker of ready datasets. Arrow keys move, Enter or Tab inserts the reference, Escape closes.
Attach	File picker (CSV, PDF, image and more)	None	Attached files show as removable chips. Jason reads them as part of the brief.
Link dataset	Checklist of ready datasets in this workspace	None	Links datasets without mentioning them in the text. Linked datasets show as chips you can unlink.
Suggestions	Pills	—	Ideas based on your workspace. Click one to fill the brief.
Start building	Button, also ⌘↵ / Ctrl+Enter	Disabled until the brief has text	Starts a new chat with Jason and opens it.

I want to build a pipeline: track the price and promotion of every product in @product_catalogue across Amazon, Argos and Currys. One row per product per retailer per day. Refresh every morning.

The build itself happens in the chat. See Chatting with Jason and Your first pipeline.

Pipeline cards

Each pipeline is a card. The card shows the pipeline icon, a clock and the cadence when it is scheduled, the name, its output counts (for example 1,274,880 rows), a status line, and a timestamp: a live clock while the last run is running, otherwise how long ago it ran. Next run appears under the status when a schedule is set.

The status line and the card colour follow the pipeline’s health:

Card colour	Status text	When
White	Not run yet	The pipeline has never run.
White	Building / Blocked	The first version is still being built, or the build is waiting on something.
Blue	Queued / Running / Stopping	A run is in flight. Updating shows while Jason is changing the pipeline.
Emerald	Finished successfully	The last run completed and published.
Brand purple	Repaired	The last publish came from a version that an automatic repair produced after the previous failure.
Amber	Awaiting publication, No recent successful run, Last run cancelled, Needs attention	Something to look at: the run has finished but has not published yet, a scheduled pipeline has missed two slots, or the last run was stopped.
Red	Needs attention	The last run failed.
Amber, blue or red, with an alert icon	This pipeline is degraded and needs healing to resume. and the heal status	The latest live run failed. This replaces the status line and the next run. See Degraded pipelines.

The same cards appear on the Home page in the Pipelines section (up to six, running pipelines first, then Show N more pipelines), and pipelines that failed or need attention also appear in Home’s Needs attention list with the failure time and reason.

Hover a card for its checkbox and its Actions menu:

• Run. Starts a run with the pipeline’s default settings and opens the run page. Shown only when the pipeline has an active version and is not degraded.
• View Runs. Opens the pipeline’s newest run.
• Pin to sidebar. Adds the pipeline to the pinned items in the left sidebar. On a pinned pipeline it reads Unpin from sidebar. The toast confirms Pinned to sidebar or Unpinned from sidebar. See The sidebar.
• Delete. Asks Delete "<name>"?, then deletes the pipeline. Its datasets stay.

Dataset cards in the other two sections have Pin to sidebar in their menu as well.

In List view the same information is laid out in columns: icon, name, counts, cadence or On demand with the next run, then status and time.

Bulk actions

Select cards with their checkboxes, the lasso, or (select all) in the subtitle. The bulk bar shows <n> selected with:

• Run. Runs every selected pipeline that has an active version, each with its defaults. Datasets and unbuilt pipelines are skipped. Degraded pipelines are refused, and the toast counts the ones that did not start.
• Copy. Opens Migrate pipelines for the selected pipelines (below).
• Delete. Confirms Delete <n> item(s)? This cannot be undone., then deletes the selected pipelines and standalone datasets.
• Move to folder. Moves the selection into a folder.

Runs started from a card or the bulk bar use the pipeline’s default settings. To pick a dataset revision, change a parameter or set an item limit, open the pipeline and use Run Pipeline. See Running a pipeline.

Migrate pipelines

Migrate pipelines copies or moves pipelines to another workspace you can access. Open it for a selection with Copy in the bulk bar, or for every pipeline in the workspace from Settings › General › Danger Zone › Migrate <n> pipelines (see Workspace settings). Cards no longer have their own Change workspace entry.

Option	Values	Default	What it does
Change method	Copy (keep the original) · Move (remove the original)	Copy	Whether the source workspace keeps the pipeline.
Destination workspace	Workspaces grouped by organisation	None	Where the pipelines go. Members of that organisation will be able to see them.
Dataset history	Structure (no rows) · Current (active rows) · All versions (full history)	Current	How much of each linked dataset travels with the pipeline.
Pipeline version history	Checkbox	Off	Include earlier versions and their changelog entries.
Completed run history	Checkbox	Off	Include past runs with their outputs, fetch evidence and provenance, as read-only history. Requires All versions and turns on version history.

A manifest on the right previews what will be transferred. Schedules, credentials, integrations, billing and active work stay in the source workspace; transferred pipelines arrive with their schedules off until you enable them. After the transfer the dialog reads, for example, 2 pipelines copied; choose Stay here or Open destination.

Empty states

• No pipelines yet. Create your first data extraction pipeline, or upload a file to start from your own data. with a New Pipeline button.
• No matching items. Filters or a search excluded everything; Clear filters resets them.

The pipeline page

Open any card. The pipeline page has three layers: the hero, up to five section tabs, and the content of the selected section. Runs open inside the same page, so the hero and tabs stay put while you read a run.

Hero

• Breadcrumbs. Workspace › Pipelines › the pipeline. The pipeline crumb is a dropdown listing every pipeline in the workspace, so you can switch without going back to the list. On a run, a trailing crumb names the run.
• Icon. Click it to pick another icon. The colour is fixed per pipeline.
• Title and description. Click to edit; changes save when you leave the field. Placeholders are Untitled Pipeline and Add a pipeline description.
• Pills. Running while a run you started from this page is executing; Scheduled (a clock) when a schedule is set.
• Stats line. last ran 3 hours ago (red when the last run failed; links to that run), next run in 6h 12m or next run due on scheduled pipelines, and takes 14min to run (the average of the last five completed runs). The next run is not shown while the pipeline is degraded, because scheduled slots are skipped until it is healed.
• Edit pipeline. Opens a composer addressed to Jason, with suggestion pills drawn from the pipeline’s specification, steps, parameters and last run. Placeholder What should change? Jason picks it up straight away.; send with Send to Jason or ⌘↵. A new chat scoped to this pipeline opens in the Jason panel with your request already sent.
• Run Pipeline. Opens the run dialog. Disabled with the tooltip Pipeline has no active revision until the first version is activated, and with Needs healing. Open Changelog to autoheal now. while the pipeline is degraded. See Running a pipeline.

Add the retailer’s stock status next to the price, and flag rows where the promotion ended in the last 24 hours.

Each active run of the pipeline gets its own banner above the hero: Run 8f3a2c is active · v2 · 4m 10s · 312 pages · about 9 minutes left, with View run, which opens that run’s Steps. A pipeline can have up to four runs queued or running at once, so you may see up to four banners. The run you are reading has no banner of its own.

When the workspace has no rows left this month, Run Pipeline opens Monthly collection cap reached instead of the run dialog. Adjust monthly cap takes you to usage settings. See Usage.

Before the first version

A pipeline whose build has not finished shows a holding page instead of the sections:

You see	Meaning	Button
Jason is still building this pipeline	The build chat is active. The first version appears after Jason finishes testing it and the independent review accepts it.	Open build conversation
This pipeline isn’t ready yet — the build stopped	The chat stopped before a version was activated.	Open build conversation
This pipeline isn’t ready yet — no reviewed version	There is no build chat to return to.	Open Jason

Degraded pipelines

A pipeline is degraded when its latest live run failed, whoever started that run. A degraded pipeline does not run again until it is healed:

• Run Pipeline is disabled with Needs healing. Open Changelog to autoheal now., and Run disappears from its card menu. Starting a run any other way is refused with Pipeline needs healing. Use Autoheal now in Changelog.
• Scheduled slots are skipped. Each skipped slot asks Jason to heal the pipeline instead, when automatic healing is on.
• A banner under the hero reads This pipeline is degraded and needs healing to resume. followed by the heal status, with Learn more, which opens the Changelog. The Changelog entry in the rail carries a ! badge.

Heal status in the banner	Card colour	Meaning
Automatic healing will retry in 3h 20min. (or …will retry shortly.)	Amber	Jason tried and the pipeline is waiting out the 24-hour cooldown, or is waiting for its first attempt. The retry happens at the next scheduled slot if that comes sooner. Autoheal now in the Changelog skips the wait.
Automatic healing is in progress.	Blue	Jason is working on a repair.
Automatic healing is turned off.	Red	Enable Auto Heal is off in Settings › Advanced. Nothing heals the pipeline until you turn it back on or ask Jason with Fix with Jason.
Automatic healing is paused.	Red	The Jsonify team has marked the pipeline Permanently broken, usually because the source is gone. The Changelog records the decision.

On the Pipelines list and on Home, a degraded card shows an amber alert icon in its corner, the same sentence in place of its status line, and Learn more. Clicking it opens the pipeline’s Changelog. In List view the cadence column reads Needs healing. How healing works, and what Autoheal now does, is in Verification and repair.

Section tabs

Under the hero, up to five tabs switch the body of the page. Each tab keeps its own place in the URL, so links land on the right section.

Tab	Shows	Link
How it works	The specification rail and its pages: statement of work, pipeline, inputs, integrations, changelog, settings.	#pipeline
Latest outputs <n>	The files and datasets this pipeline publishes, and the integrations that receive them. The count is files plus datasets. Hidden until the pipeline has an output.	#outputs
Analytics <n>	The dashboards that read this pipeline’s datasets, embedded in the page. Always shown; the count is the number of dashboards.	#analytics
Sources	The sites read to build the latest datasets. Shown once the pipeline has a dataset.	#sources
Runs <n>	The newest run, with the runs rail down the side.	#results

The How it works rail

The left rail lists the pages of the specification under the heading Specification. It collapses with Collapse how it works, resizes by dragging its edge, and becomes a select on small screens.

Entry	What it is	Page
Statement of Work	The plain-language contract: objective, scope, parameters, what the pipeline produces and the rules it follows. Select text to comment and send changes to Jason.	Statement of work and steps
Pipeline	The step diagram of the version you are viewing, with a reading panel for each block. Shown once the version has a diagram.	Statement of work and steps
Inputs	Dataset inputs, editable value lists and parameter defaults.	Inputs and parameters
Integrations	What starts or feeds the pipeline, where its rows go, and the alert rules that watch it.	Pipeline integrations
Changelog	Versions, automatic repair attempts and human decisions, with the heal status of the pipeline. Shows ! while the pipeline is degraded.	Versions and changelog
Settings	Schedule, Cache and Advanced (automatic repair).	Settings reference

If a link asks for a page the pipeline does not have (for example Pipeline before a diagram exists) you land on the Statement of Work. A link with nothing specific lands on Latest outputs when the pipeline has published anything, otherwise on the Statement of Work.

Latest outputs

Latest outputs shows what the pipeline publishes, always at each dataset’s active revision. When it has more than one output, a rail on the left lists them as cards in three groups:

• Files. One card per file the pipeline builds on each production run, with its type badge, name and size.
• Datasets. One card per output dataset, with its row count and Version <n>, Building dataset… or No data yet.
• Integrations. One card per integration wired to those datasets, with the service, whether it feeds or receives the data, and Live when it is live. A long list ends with Showing 50 integrations · View all.

A pipeline with a single output shows it without the rail. Dashboards are not listed here; they have their own Analytics tab.

Dataset view. The table of the dataset’s active revision, with the same browser as the dataset page: worksheets, sort, search and paging. Two buttons sit at the right of the section tab bar: an Export menu with Excel workbook (every worksheet) and CSV of this table (the table you are looking at), and Open dataset, which opens the dataset page on the same revision, worksheet, page, search and sort. There is no revision picker here; to read an earlier revision, use Open dataset and the dataset’s own version list. See Dataset pages and Dataset versions.

File view. Select a file card to preview the file in place, with Download on the tab bar.

Deliverables view. While the rail is still loading, a single Deliverables entry lists every file as a card you can preview or delete. Delete asks Delete every version of "<name>"?. Until the first run: Files appear here after the next production run.

Integration view. An integration card shows its type, what it does, its status, which dataset it Feeds or Receives, when it was Updated, and Open integration.

A pipeline with an active version but nothing published shows Nothing published yet — run the pipeline to create its first output. Before the build finishes it reads Outputs appear here once the build finishes and the first run completes.

Analytics

Analytics embeds the dashboards that read this pipeline’s datasets, so the charts sit next to the data they come from. One dashboard fills the section; several get a rail on the left with one entry per dashboard (Showing 50 dashboards · View all when there are more). Filters and links inside the dashboard work as they do on the dashboard page; a link to another dashboard opens that dashboard’s own page. See Dashboards.

A pipeline with no dashboard shows No dashboard yet and Add a dashboard to chart this pipeline’s data. It updates with every run., with Add dashboard. There is no form: Add dashboard opens Jason in a pop-out with a new chat that asks for a dashboard or set of linked dashboards to explore this data, starting with the pipeline’s first dataset. When Jason finishes, the dashboard appears in the tab and the Jason panel docks to the side.

Sources

Sources lists every site the pipeline read to build its latest datasets, one card per domain, with the rows and URLs that came from it.

Option	Values	Default	What it does
Search	Free text, Search sources…	Empty	Filters the cards by domain.
Sort	Sort: Rows · Sort: URLs · Sort: Domain	Sort: Rows	Orders the cards.

Each card’s footer names the datasets that site feeds; clicking one opens the dataset filtered to rows from that source. The tab appears once the pipeline has a dataset. Until a run publishes a dataset with per-row source attribution, the section reads Sources appear here once a run publishes a dataset with per-row source attribution.

Runs

Runs opens the newest run inside the pipeline page, with a rail of every run on the left. Filter the rail, use the ↑ and ↓ keys to move between runs, and read each run’s Summary, Outputs, Steps, Activity and Integrations. A pipeline that has never run shows No runs yet · Run the pipeline to publish its datasets. The run page is documented in Run detail.

The pipeline page refreshes itself: every 15 seconds while the tab is visible, every 3 seconds while Jason is working on the pipeline, and every 4 seconds for the runs list while a run is in flight. You do not need to reload.

What’s next