Run detail
The runs rail, the run head with its status pill and actions, and the Summary, Outputs, Steps, Activity and Logs tabs of a single run.
A run page tells you what one execution of the pipeline did: which version ran, who started it, what it fetched, the rows it produced, where each step got to, and what went wrong if anything did. Runs open inside the pipeline page (Runs tab), so the pipeline’s hero stays at the top and the runs rail sits on the left.
Open a run from the Runs tab (newest first), from a card’s View Runs, from last ran in the hero, or from the Latest runs page.
- Summarywhat happened, in a sentence
- Outputsthe rows, per output
- Stepsthe diagram, live
- Activityevery fetch and delivery
- Logsthe worker's own record
The runs rail
| Element | What it shows |
|---|---|
| Filter runs… | Filters the rail by title or by the parameter values shown inline. No runs match. when nothing does. |
| Groups | Active (queued, running, stopping), then Today, Yesterday, and one group per earlier date. |
| Row | A status icon, the run title (Live run #8 — Sep 15, 06:00), and a meta line: a ticking clock while live; afterwards 48,102 rows · 2m 14s · by timer. by names a person, timer (the schedule) or webhook. |
| Left edge | Coloured by kind: emerald for live, blue for resumed, violet for test, slate for debug, teal for newsletter. |
Use the ↑ and ↓ keys to move to the newer or older run. The section you are reading stays selected where it makes sense (Steps, Activity, Logs); an open output or step is dropped because the next run may not have it. The keys are ignored while you are typing in a field.
The rail collapses and resizes like the other rails; collapsed rows show their title and meta as a tooltip.
The run head
The head carries the run title with a status pill, a detail line under it, and the actions on the right. Its background follows the state: green for completed, red for failed, slate for cancelled, orange for queued, amber for a run that looks stuck, and the brand colour while active.
| Pill | Meaning |
|---|---|
| Queued | Waiting for an available runner. This run will start automatically. |
| Running / Running — extract_offers | Executing; the current step is named when known. |
| Stopping… | Winding down — the run is being stopped and will finish shortly. |
| Looks stuck | An active run with no activity for 30 minutes: No activity for 34 minutes during extract_offers. You can stop the run, or ask Jason to investigate. |
| Completed | Finished and, for live runs, published. |
| Failed | Stopped on an error. The error is at the top of Logs. |
| Cancelled | Stopped before finishing, with the reason on the detail line: Stopped by Maya Okafor, The worker processing this run was lost., Stopped by the stall watchdog., Stopped after hitting its time limit., or Stopped externally. |
Actions
| Action | When | What it does |
|---|---|---|
| Notify bell | While the run is active, in browsers that support push notifications | Notify me when this finishes: sends a browser notification when the run finishes or is stopped. The toast reads I’ll notify you when this finishes — you can close the tab. Click again to stop. If notifications are blocked for the site the bell is crossed out with instructions. |
| Export | When an output with rows is open | Downloads that output as CSV or JSON. |
| Stop | Active runs | Asks the workers to stop. The run passes through Stopping… to Cancelled; the button stays visible but disabled meanwhile. |
| Resume | Failed or cancelled runs that completed at least one stage | Opens Resume run? and starts a new Resumed run from that stage. See Run lifecycle. |
| Fix with Jason | Failed runs | Opens a new chat scoped to this pipeline with the run, its error and its id, asking Jason to read the logs and activity, diagnose the problem, repair the pipeline if needed, and re-run with the failed run’s settings. |
| Ask Jason | Runs that look stuck | The same, for a run with no activity for 30 minutes. |
| Delete run | Finished runs, in the menu beside the last action | Removes the run with its logs, fetches and outputs. Datasets it published are untouched. There is no confirmation step; you land on the neighbouring run. |
Summary
The Summary opens with a sentence:
Live run #8 started by timer on Sep 15, 2026 against v2 (Full input). It completed in 2m 14s, fetched 312 pages and collected 48,102 rows, and published Live product offers as v9.
A failed run’s sentence ends It failed after 41s in extract_offers, having fetched 12 pages and collected 0 rows; 2 of 6 steps completed. The error is at the top of the logs. with an Open logs link. A cancelled run says when and why it was stopped; a queued one that it is waiting; a running one where it is and how long is left.
Configuration
| Fact | What it shows |
|---|---|
| Kind | The run kind pill. |
| Status | The status pill. |
| Version | v2, linking to that version’s statement of work. |
| Started by | A person, timer or webhook. |
| Started, Finished, Duration | Timestamps; Elapsed and Time left while running. |
| Rows collected, Pages fetched | Totals for the run. |
| Steps | 6 of 6 completed, or 2 of 6 completed · failed in extract_offers, linking to Steps. |
| Scope | Full input or Cap 1,000 inputs; through <step> when the run stopped at a chosen step; cache-only mode or warm-cache mode when set. |
| Publication | Private · does not publish on test and debug runs. |
| Recovery | On resumed runs: Reused stage extract_offers from run #7. |
Inputs and Outputs
Inputs lists each dataset input as Product catalogue v7 · 1,240 rows (a link) and each parameter whose value differed from its default as Region: GB (default US). Outputs lists Datasets: Live product offers (v9), deliverable cards, and one card per output: a dataset card when published, or offers · 118 rows · not published for a test, debug or stopped run, which opens the run’s own copy of the rows.
Outputs
Outputs shows the rows the run produced, one worksheet tab per output when there are several (Deliverables 2, offers 48,102, matches waiting with a spinner for one still being written). A single output shows the banner Added to dataset Live product offers as v9 · 48,102 rows with View dataset.
The table is the run’s rows. Click a row to open Row details:
Previous row (←), Next row (→), Open in full page and Close (Esc) sit in the header. While a run is active the tab reads The results will be available once the run has completed.; a run that produced nothing reads No data extracted in this run. The Deliverables tab lists Files this pipeline built on this run. with preview and delete.
Steps
Steps is the pipeline diagram with this run’s facts on it: row counts on the edges, a pulse on the step that is running, and a camera that follows the active step while the run is live. Selecting a block opens it in the panel on the right; Escape or × returns to follow mode.
| Panel tab | What it shows |
|---|---|
| Details | The same reading panel as the pipeline’s Steps page (what the step reads, saves and which rules apply), plus this run’s facts for it and Open output. |
| Data / Data (1,240) / Input Data | The rows the step produced: Partial preview while it is running, Intermediate output afterwards, Reused exact dataset revision for a reused stage. Messages: This stage is waiting to start., No rows produced yet., This stage produced no rows. |
| Pages (12) | The pages the step fetched, when it fetched any. |
| Log | The run’s log lines filtered to this step. |
Selecting an input block opens Inputs: Input values captured at run start., with Parameters (2) and Pipeline inputs (1), each with its type and value, and an Export button that downloads the exact run configuration so the run can be repeated. Selecting an output pins it and shows its rows. Before the diagram exists the tab reads Steps appear once the pipeline compiles.
The stage labels under each step summarise its state: Waiting to start, 340 of 1,240 processed, Failed after 12 rows, 1,240 rows, 1,240 rows · exact revision.
Activity
Activity is the run as a feed of events, newest at the bottom, each stamped with the time since the run started (0:01:42).
| Row type | What it shows |
|---|---|
| Page fetch | Host and path, method, size, duration, and flags such as Reused from cache, Served this page, No rows returned, blocked or failed. Expand it for Result 24 rows, Requests, Downloaded, Duration, and a link to the fetched content. |
| Extraction | Rows extracted from a page; No rows extracted is flagged as a warning. |
| Browser session | A recorded browser session you can open where one was kept; otherwise Browser session unavailable. |
| Delivery | Delivered to Live product offers, or Failed. |
| Lifecycle | Dimmed rows for dispatch, start, each step’s completion, stopping, resuming, finishing and failing, plus the pipeline’s own log lines when they are worth surfacing. |
An empty feed reads No activity recorded for this run.
Logs
Logs is the workers’ own record: a dark panel of timestamped lines ([ 12.3s] extract_offers fetched 24 offers from argos.co.uk). Lines from helper agents are indented with ↳. Load previous logs pages back through older lines; while the run is live the panel waits with waiting for next log….
On a failed run the error sits in a red banner at the top of the panel. That banner is what Fix with Jason sends, and it is the first thing to read.
The favicon pulses while a run you have open is active, and the page keeps itself current over a live connection with polling as a fallback. Leave the tab open in the background and glance at it.