Local-first repository insights for commits, contributors, churn, language mix, and repo health.
ginsights is a single-binary Go CLI that turns local Git history into a polished Repository Atlas dashboard and JSON snapshot. It is built for quick repo handoffs, local audits, and shareable reports that do not require uploading code or connecting a SaaS account.
- Static HTML and JSON output that work offline.
- A full-width commit activity calendar, latest-change strip, contributor table, code-frequency view, hot files, languages, and repository health in one report.
- A local-first overview that shows net line change, commit volume, authors, touched files, primary language, and health signals without ranking people.
- Metric provenance labels so local Git data stays separate from optional GitHub API data.
- Optional disposable cache for faster repeated runs on larger repositories.
- Explicit GitHub connector only when
--github-api owner/nameis requested.
Homebrew:
brew tap marlonjd/tap
brew install ginsightsIf Homebrew refuses the tap as untrusted, approve it explicitly:
brew trust marlonjd/tap
brew install ginsightsShell installer:
curl -fsSL https://raw.githubusercontent.com/MarlonJD/ginsights/main/scripts/install.sh | bashThe shell installer builds from source and installs to ~/.local/bin/ginsights by default. More options, including --install-dir, --ref, tap maintenance, and source builds, are in docs/INSTALL.md.
ginsights serve . --port 43117
ginsights build . --out report
ginsights json .Useful options:
ginsights serve . --since 2026-07-01
ginsights build . --out report --no-cache
GINSIGHTS_GITHUB_TOKEN=... ginsights build . --out report --github-api owner/name
ginsights cache-clear .GitHub Insights is useful, but it is hosted and mixes local Git facts with GitHub server-side analytics. ginsights makes the local part fast, reproducible, and shareable without uploading code or requiring a token.
It is different from raw git log scripts because it provides a stable dashboard, JSON output, health signals, cache behavior, and metric provenance. It is different from SaaS engineering analytics because it does not rank people or require a hosted service.
The full rationale is in docs/WHY.md.
Core mode is local/offline. GitHub Traffic data such as views, clones, referrers, and popular content cannot be inferred from local Git history. When --github-api owner/name is used, API-sourced metrics are labeled as github_api and connector failures degrade into the report instead of breaking local analysis.
go test ./...
go run ./cmd/ginsights doctor .
go run ./cmd/ginsights build . --out /tmp/ginsights-report
scripts/capture-ui-screenshots.sh --update-readmeWhen using Codex, start with AGENTS.md, then open the active execution plan that matches the task.