Gitpulse now reads a .gitpulse.json file at the repo root, letting maintainers set custom publication titles and subtitles alongside per-project analysis preferences.
GitPulse now deploys to Vercel using a CI-first pattern where GitHub Actions runs the analytics build and Vercel just hosts the output — keeping all secrets in the runner. Site URLs are also now auto-detected from platform environment variables on Vercel, Netlify, and Cloudflare Pages, eliminating manual GITPULSE_SITE_URL configuration.
Gitpulse is now installable as a regular npm package. A new reusable GitHub Actions workflow lets any repo publish its editorial story feed to Pages with a single YAML block.
Clicking any PR or commit in the feed now opens a portal-rendered side drawer instead of navigating away. The feed stays frozen in place; closing the drawer returns you exactly where you were. The panel is deep-linkable and works with the browser back button.
The PR detail page gets a visual overhaul: richer metadata up top, a subtitle with the PR title and author, and story content moves into switchable tabs with a proper diff viewer for code changes.
Maintainers can now publish GitHub Releases from the Actions tab without touching the command line — the workflow validates versions, gates on tests, bumps all workspace package.json files, and creates a release with auto-generated notes.
The homepage now displays release editions above PR features, and a new /releases/ section lists every release with hero cards, compact rows, and full detail pages with changelogs.
The analyzer pipeline now processes GitHub releases end-to-end, matching each release to its constituent PRs and generating AI-written editorial copy. Hash-based caching prevents redundant LLM calls on unchanged data.
Gitpulse now tracks releases as a first-class entity alongside stories. This foundation adds the data shape, write-time validation, and restore-from-site plumbing for Phase B's analyzer to write into.
Story links now include a keyword-rich slug derived from the headline, replacing opaque IDs with readable URLs that search engines can parse.
Gitpulse now ships proper SEO metadata on every page: canonical URLs, OpenGraph images, Twitter cards, JSON-LD structured data, robots.txt, and an XML sitemap. Stories should surface better in search results and share attractively on social platforms.
Author names on story pages and PR feeds now display as plain @-prefixed text, removing the previous conditional link behavior for author profiles.
Commits pushed directly to a branch now display the author's GitHub handle and a working profile link, instead of falling back to a plain display name.
Gitpulse's site now respects system color preferences with a theme toggle, and carries a proper branded logo with a recognizable pulse glyph.
GitPulse can now skip commits it has already analyzed. On each daily run, the action fetches the manifest from the deployed site, restores prior stories locally, and only invokes the LLM for brand-new commits.
Gitpulse now displays stories in a proper editorial layout — grouped by day with a hero feature, two-column fixes section, and collapsible housekeeping drawer. The masthead swaps to a publication name on scroll.
GitPulse can now analyze multiple commits at the same time instead of one-by-one, with a configurable concurrency limit that defaults to 10 parallel workers.
GitPulse now distinguishes between merged PRs and direct pushes, enriches stories with linked issues and author info, and displays a size indicator so reviewers instantly gauge change scope.
Gitpulse just got significantly smarter. A commit to the repository lifts the entire PR-analysis pipeline from sister project gitsky — adding structured category scoring, editorial headlines, technical descriptions, and support for both OpenAI and Anthropic API protocols.
GitPulse can now automatically generate readable editorial stories from raw git commits — feeding commit metadata to an AI model and outputting structured stories with headlines and body text.
Gitpulse stories now surface in a chronological feed with individual detail pages. The site reads from JSON files, with stub fixtures included for design verification.