wiyco/Review Insights

Analyze PR review workload, merge outcomes, reviewer-author imbalance, and AI involvement with visual reports

Trust Signals

name	description	required	default
github-token	GitHub token with repo read access. Defaults to the workflow token. Required permissions: contents:read and pull-requests:read. For output-mode "comment", also pull-requests:write.	no	${{ github.token }}
repository	Repository to analyze (owner/repo). Defaults to current repository.	no	${{ github.repository }}
since	Start date for the PR creation window (ISO 8601, e.g. 2025-01-01). Defaults to 90 days ago. NOTE: For large repositories with high PR volume, use a shorter range (e.g., 30 days) to avoid hitting GitHub API rate limits.	no	—
until	End date for the historical snapshot (ISO 8601). Defaults to now. Reviews and merge/close state after this timestamp are ignored.	no	—
output-mode	Where to publish results: "summary", "comment", "artifact", or comma-separated combination.	no	summary,artifact
bias-threshold	Finite Pearson residual threshold for activity-adjusted reviewer-author imbalance, clamped to 0.5-10.0 (default: 2.0).	no	2.0
include-bots	Whether to include bot accounts in statistics. When false, bot-authored PRs (and all reviews on them) and bot reviews are excluded. See docs/filtering.md (default: false).	no	false
max-prs	Maximum number of PRs to analyze (default: 500). If additional PRs remain within the requested date range after this cap is reached, outputs are marked as capped rather than complete. IMPORTANT: Values above 1000 on very active repositories may consume significant API quota and increase execution time.	no	500

name	description
report-path	Path to the generated HTML report file
total-prs-analyzed	Number of PRs analyzed
top-reviewers	JSON array of logins tied for the maximum reviewsGiven among users with reviewsGiven > 0, or [] if no active reviewers exist
max-reviews-given	JSON number for the maximum reviewsGiven among users with reviewsGiven > 0, or null if no active reviewers exist
bias-detected	Whether at least one reviewer-author pair was flagged by bias detection (true/false). If the bias model is unavailable, that warning is surfaced in the reports and logs.
partial-data	Whether analysis used a capped or partial PR dataset because pagination hit max-prs or the fixed 10-minute collection budget (true/false)