Reports
ginkgo report turns a finished run into an HTML report. It comes from the
ginkgo-report plugin. See Plugins
to install it.
Create A Report
Section titled “Create A Report”ginkgo report # the most recent runginkgo report <run_id> --open # a specific run, opened in the browserThe run must be finished, with a status of succeeded, failed or
cancelled. A run that is still running or pending is an error.
By default, the report is a folder, .ginkgo/reports/<run_id>/. It holds
index.html, and an assets/ folder for the stylesheet and fonts. Pass
--single-file to write one HTML file that opens offline instead.
What The Report Contains
Section titled “What The Report Contains”The report includes:
- the run summary and the run parameters
- the task graph
- each task’s status and timing
- failure detail, with the end of each failed task’s log
- the error of each failed attempt that a retry then recovered from
- the files and folders each task produced
- rendered pages, such as notebooks
Summary
Section titled “Summary”The Cache hits card shows how many tasks came from the cache, and what share of the run that is.
A run that did not succeed says so first. The summary opens with the run’s error, such as the failure that left the workflow without a result. The Tasks card counts the tasks that succeeded, failed and were skipped.
Failures
Section titled “Failures”The Failure section lists each skipped task, with why it never ran, such as
blocked by analyze[omega]. If many tasks were skipped, it groups them by
cause and shows a count for each.
Errors and logs show a path inside the project relative to the project root.
The Peak memory column in the Tasks table shows how much memory each
task used, next to the amount it declared. For example, 48.0 KiB / 64Mi. A
task that came from the cache shows the peak from the run that produced its
result, and names that run, as in 48.0 KiB (run 9e1f3e6c). See
Measured Usage for how memory is measured.
The Attempts column shows how many attempts a task used, out of the number
its retries allow. A task with retries=1 that succeeded on the first
attempt shows 1 of 2 allowed.
Rendered Pages
Section titled “Rendered Pages”A task that renders a page for people to read, such as a notebook task’s HTML, gets a card in the Rendered pages section. The card’s title is the task’s label. A page reused from a cache hit says which run rendered it. See Notebook Tasks.
Notebook Pages Offline
Section titled “Notebook Pages Offline”A marimo notebook page loads its display code from the internet, so it is blank without a connection. The report therefore also makes a static copy of each notebook page, which works offline. It keeps the notebook’s text, figures and output, shows tables as their first 200 rows, and leaves out interactive elements such as sliders.
- The folder report links both: open for the interactive marimo page, and static copy for the offline one.
- The single-file report embeds only the static copy. To share the interactive pages, share the folder report.
Options
Section titled “Options”--single-filewrites one HTML file,.ginkgo/reports/<run_id>.html, that opens offline. It embeds notebook pages as static copies, and leaves out logs.--hide-codeleaves each cell’s code out of the static copies of notebook pages. By default, the code is folded under a “code” toggle.--openopens the report in a browser when it is built.--max-log-lines Nlimits the log lines shown for each failed task. The default is 80.
Writing The Report Elsewhere
Section titled “Writing The Report Elsewhere”--out <path> writes the report somewhere other than .ginkgo/reports/. For
a folder report it names a directory. With --single-file, a path ending in
.html is the file itself, as in --out report.html.
Ginkgo writes over an earlier Ginkgo report, but stops if the destination
holds other files. Pass --force to write anyway. It replaces only the files
the report writes, and removes nothing else.
See Also
Section titled “See Also”- Tasks And Flows: notebook tasks render marimo notebooks to HTML, which appear in reports.
- Caching And Provenance: how run outputs are stored and reused.
