Output formats

--format chooses how run, runs view, and runs resume print results on stdout. The summary still goes to stderr.

--format	Prints
text (the default)	answers for people to read, as in Reading the output
json	one JSON line per item, with every probability; --json is short for this
csv	one row per item, for a spreadsheet
md	a Markdown report, for a pull request comment
github	an annotation for each flagged or matched item, for GitHub Actions

decide runs view --format csv > results.csv
git diff main | decide run code-risk --each function --format md > report.md

A CSV row has the item’s source and status, whether it was flagged or matched, a column for each question, and any error. A yes-or-no question’s column holds the probability of yes, a score question’s the score, and a choice question’s the choice, followed by its probability in a column such as queue_probability. A cell that starts with =, +, -, or @ starts with ', so a spreadsheet does not read it as a formula.

The Markdown report lists flagged, matched, and failed items first, and collapses the rest. Each table shows up to 100 items. When a diff has nothing to judge, the report says so, so a pull request comment made from it does not go stale.

With --format github, each flagged item becomes a warning on its file and line, which GitHub shows on the pull request’s changes, and each matched item becomes a notice. The items that --fail-on names are errors instead, since they fail the job. Run decide from the top of the repository so the files’ paths match. In a GitHub Actions job, decide also adds the Markdown report to the job’s summary page. GitHub shows at most 10 annotations of each kind for a step, so the summary is the place to see more. Review pull requests has a workflow to copy.

runs resume prints the items answered before the run stopped as well as the new ones, so its output covers the whole run in every format.