Analyze the download statistics of any public GitHub repository's releases — and track how they grow over time, with zero backend, no sign-in, and nothing to host.
Everything runs client-side in a single index.html. Your data never leaves your browser.
GitHub's API only exposes a cumulative download count — a single number per asset, with no history. Every other static release-stats tool is therefore stuck showing a one-time snapshot, and the tools that do show history require a backend, a login, and only start tracking after you sign up.
Release Radar solves the history problem without a backend:
Every time you analyze a repo, a timestamped snapshot is saved locally in your browser. Revisit later and Release Radar computes real measured velocity (not estimated from release dates) and plots true downloads-over-time. Export/import the history as JSON to back it up, sync across machines, or share it.
Snapshots store per-asset counts, so history is re-read under the current classification rules rather than frozen at write time: installer-only history and update-channel history reach back to your very first snapshot, and correcting a platform later fixes the whole series at once.
An app that ships auto-updates publishes latest.yml and .blockmap files next to its installers, and every running client re-downloads them — on many repos they are 80% of the "total downloads" number. Release Radar counts installers by default and splits the rest into what it actually is:
- Update checks —
latest*.yml, fetched every time an installed app polls the update feed. - In-app updates —
.blockmap/.zsyncdifferential payloads, fetched only when a client actually installs an update. - Verification — checksums and signatures.
Blockmaps are paired to their installer by filename, so every release also gets a direct-download vs. in-app-update split, per platform: 15 downloads of App.exe against 10 of App.exe.blockmap means 10 people updated through the app and 5 came from the link. Flip the Installers only / All files toggle to switch every number on the page between the two definitions.
All of it is auto-detected — repos without an update feed never see any of this, and their numbers are identical in both modes.
Versions that gained downloads since your last snapshot are flagged with a green ▲ and the exact delta — so you can see at a glance which releases are still being pulled. Expand a release and the same tracking goes one level deeper: track download growth for every release asset over time, and see exactly how many new downloads each binary gained since your last visit. Assets that have gone quiet get a discreet "last seen" hint instead, showing the last window in which growth was detected.
See a repo's star growth as a cumulative chart, plus stats like average stars/day and the day it gained the most. Requires a token with access to the repo (GitHub restricts this data to admins/collaborators as of June 2026), so it works out of the box for your own repos.
Most-downloaded release, dominant platform, download momentum (recent vs. older releases), and latest-version adoption — generated automatically, no configuration.
Benchmark total downloads across several repositories side by side.
Per-version, per-asset breakdown with expanded platform detection (Windows, macOS, Linux, Android/APK, snap, flatpak, msix, and more), grouped so the installers sit above the update traffic. Detection matches whole filename tokens, not substrings — Ollama-darwin.zip is macOS, not Windows — and reports architecture where the name carries it.
An archive that names no platform at all (App-1.2.3-arm64.zip could be anything) is labelled Unknown rather than guessed. Click the badge to assign one: the choice is stored as a filename pattern, so it applies to that asset across every release, past and future, including your snapshot history.
Export to CSV, JSON, or PDF.
Narrow the releases table and the download trend chart down to a search term or a date range — useful for zooming into a specific version or release window on repos with a long history.
Pick any two snapshots you've taken — down to the exact time of day — and see the total downloads gained between them, the average per day, and a per-version breakdown sorted by growth. Works for any window: 7am vs. 1pm the same day, or three weeks ago vs. today.
Dark / light mode, KPI cards, version adoption chart, platform split donut, recent-repo shortcuts, and ?repo=owner/repo deep-links that auto-load.
- Open the live app.
- Type a repository as
owner/repo(e.g.ollama/ollama) and hit Analyze. - (Optional) Add a GitHub Personal Access Token to raise the rate limit from 60 to 5,000 requests/hour. By default it stays in memory for the session only; check "Remember on this device" to save it in
localStorageso you don't have to re-enter it. Never sent anywhere except the GitHub API. - Revisit periodically — the more snapshots you accumulate, the more accurate the measured velocity.
You can also deep-link straight to a repo:
https://luqp2.github.io/github-release-analytics/?repo=ollama/ollama
- 100% client-side app. No backend, no sign-in — your GitHub token and snapshots never leave your browser.
- Anonymous page-visit analytics via GoatCounter (goatcounter.com), a privacy-friendly, cookieless tool. It only counts visits — no personal data and no cross-site tracking.
- Snapshots are stored in your browser's
localStorageand never leave your machine unless you explicitly export them. - Tokens are kept in memory only by default, and sent exclusively to
api.github.com. If you opt in to "Remember on this device", the token is saved inlocalStorage— only enable this on a device you trust.
Single static index.html — Chart.js for charts and html2pdf.js for PDF export, both loaded from a CDN. No build step, deployable on any static host (GitHub Pages).
There is no test runner, so the asset classifier ships with its own corpus of real release filenames: open the app at #selftest and read the console.
MIT