werkrevisor/docs/prs/2026-10-07-PR#45-recent-reviews-per-pr.md

PR#45: the statistics show the recent reviews per pull request

WARNING: This document describes only the change applied in this PR. It may already be outdated once the next PR is merged. Historic PR documentation is not maintained along with new PRs; treat it as a snapshot.

The Problem

The list "Recent reviews" on the statistics page shows one row per review run. A PR with many review rounds fills the list, and nobody sees what the PR cost in total. E.g. mi/werkrevisor#44 needed several rounds for about 0.16 USD; each row shows only the cost of one round.

The verdicts appear only as color names, e.g. YELLOW. A reader who does not know werkrevisor does not know what a color means.

Non-Goals

• No change of the stored records or of the sums.
• No tooltips in the review comments in Gitea: Markdown has no tooltips.

The Scenarios

Scenario#45.01: a pull request with several review runs has one row

So that the reader sees what a PR cost in total.

• Given mi/werkrevisor#44 had 5 review runs
  • and the runs cost 0.16 USD together
• When a reader opens the statistics
• Then "Recent reviews" shows one row for mi/werkrevisor#44
  • and the row shows the time, the verdict and the model of the newest run
  • and the row shows the cost 0.16 USD, 5 reviews and 0.032 USD per review

Verified by

• UsageApiIntegrationTest: recent reviews have one row per pull request

Scenario#45.02: a verdict explains itself

So that a reader who does not know werkrevisor understands the colors.

• Given the statistics show a verdict, e.g. YELLOW
• When the reader points at the verdict
• Then a tooltip says in one sentence what the verdict means

Verified by

• UsageApiIntegrationTest: each verdict has a tooltip of one sentence

The Solution

• "Recent reviews" has one row per PR, for the 20 PRs with the newest review runs.
• The time, the verdict and the model come from the newest run of the PR.
• The cost is the sum of all runs of the PR; two new columns show the number of reviews and the cost per review.
• A local review of a PR has its own row.
• Each verdict on the statistics page has a tooltip with the meaning from the README.
• werkrevisor 0.21.5.

Prerequisite PRs

• PR#43: werkrevisor 0.21.3, the statistics show an unknown model cost as a lower bound, merged.