Analytics for Github Pages

[Artucuno]

Hey! I'm not sure if its already a thing but Github Pages should have a graph in the traffic tab of your Github Repo with the visitors count, country of origin, etc.

Thanks.

(13/09/2024) EDIT:
As of now, there is still no way to view Github pages traffic inside of your Github repo. If you are looking for a service with built-in analytics, I recommend using Cloudflare pages.

Cloudflare Pages also supports Jekyll and has automated builds from Git, so migration from Github Pages is super simple. [Comparison]

[melvincwng]

@Artucuno Hey there, I believe that feature (or at least most of it) is available in GitHub already.

You can follow the steps below:

1. Go to your code repository
2. Look for the 'Insights' tab, click on it
3. Then click on 'Traffic' tab
4. Once there, you will see a variety of metrics like visitors, how many people cloned your repo etc

Hope this answers your question 😁

[port19x]

This does not track website traffic, but repo traffic!

[sourabhshete99]

does it count our own view to our repo. I see on my repo there'e 1 unique visitor, does it mean that 1 me or somebody else ?

[NadimGhaznavi]

Nice!!!! I just ripped out my <script> reference to Google analytics which I couldn't get any useful information from. Frankly, I'm very, very, very happy to not give more of my data to Google :P

[jfullstackdev]

that sounds exactly like Google Analytics,

well you can use Google Analytics for GitHub Pages if you are really into it, I tried it before

[guest271314]

Is there a way to write to a file in the repository with GitHub Actions?

[port19x]

Is there a way to write to a file in the repository with GitHub Actions?

Yes. https://port19.xyz/tech/github-actions-monitoring/

[guest271314]

Then we can write the requested URLs to a file in the repository.

[port19x]

And how would that be triggered?
There is no actions trigger that you can hook into your webiste via a tracking pixel or sth similar.
Even if there was, it would be just as good/bad as google analytics and not server-side.

[guest271314]

Not sure yet. That is one part of what is required.

I've thought of using the GitHub API to write a comment in a Discussion or public Gist section to facilitate writing the request URL.

[rterakedis]

Hello,

@melvincwng - the Traffic stats at the location you specified appear to be directly related to repository traffic (https://docs.github.com/en/rest/metrics/traffic). In other words, when folks view the repository via github.com, stats are tracked.

What I would like to see (and what I think @Artucuno is asking), is a set of similar traffic stats as to what GitHub Pages (project.github.io) traffic is happening:

• Popular Content - what github.io pages are being request)
• Referring Sites - where did traffic to the GitHub Page (*.github.io/pagename.html)

My primary reason for asking this is that a tool like Google Analytics is over-engineered for these basic needs. Also, it's far less privacy-focused. I'm simply looking for equivalent-type of stats for GitHub Pages as what is provided for Github Repos.

UPDATE: I did just manage to figure out that you can get some of this type of reporting if you sign up for a free Cloudflare account and enable web analytics on cloudflare. You can read about it on the Cloudflare Blog

[acalatrava]

I agree with this proposal and I think it should be kind of mandatory. Most of us are privacy focused folks and I’d love to have, not just analytics, but direct access to logs of every request to GitHub pages, I mean server logs. We could then use any software or tool to generate analytics from them.

[joadoumie]

Just wanted to echo the sentiment in this thread that having some built-in mechanism for:

1. Server Logs
2. Web Traffic analytics (basic number of requests etc.)

Would be super useful!

[svnv-svsv-jm]

Oh yes please!

[ghost]

can I haz log?

[Artucuno]

Real

[emmadesilva]

Would be nice to have something super simple, like the visitor counts for different pages. Similar to Popular content on the Traffic data, but for pages hosted on Pages.

[kirkone]

This would be very useful.
The integration in the Traffic page should be good for most usecases.

[hossainalmahdi]

Google analytics can be used instead

[emmadesilva]

Three problems with that:

1. It's absolutely overkill for most small sites
2. It will slow down page loads
3. It will track users

Sure, if you need to know every little detail, go for GA, but if you like me just want some stats over which pages are most popular, and the visitor count, something small on the GitHub Pages server end would be ideal.

[sanxofon]

I'm sure MS already does it, so share it

[unrealsolver]

Google analytics could be opted out. Static server file access stat - can not.

[brandonbrinkley]

Having the logs would be super helpful, even if user info has to be scrubbed. I'd be very interested in knowing what URLs are referring my pages.

[davidmegginson]

Bump! Like a lot of posters, I don't use Google Analytics for ethical reasons (third-party tracking scripts are nasty), but I'd love to have basic server-side traffic info — even just pageviews/day — to know how many people are visiting.

[Skipants]

@davidmegginson There are more ethical analytics tools you can use. Check out https://dataethics.eu/tools/ under "Alternative Webstatistic Tools"

[reukiodo]

Aren't those all still client-side implementations? Server stats are basically already collected, just want exposure to view them.

[emmadesilva]

Aren't those all still client-side implementations? Server stats are basically already collected, just want exposure to view them.

Exactly. GitHub (presumably) already collects all the information we are interested in, we just want to see it.

[dwarfered]

Agreed, though probably no financial incentive for GitHub to add such a feature. Unless they tie it to GitHub Pro or something

[emmadesilva]

Agreed, though probably no financial incentive for GitHub to add such a feature. Unless they tie it to GitHub Pro or something

On the other hand, it should not cost that much to implement the feature, and they shouldn't lose any money by offering it. Considering that GitHub Pages is free I don't think this should be a deal breaker. I'm assuming GitHub wants people to use Pages, so it should be in their interest to have as many good features as possible.

[mvaldesdeleon]

In case any product people read this: The number 1 reason that would make me churn right now is need to have insight into my site's traffic. I've deliberately chosen to not include any cookies nor 3rd party content in my site, and would have to rely solely on server logs to have this kind of insight. I can do without it for now, but eventually that'll be my reason to host somewhere else.

[allan-stewart]

Hear, hear!

[guest271314]

Seems like we can do this using a ServiceWorker, Deno Deploy or Google App Engine if Microsoft doesn't want to do it.

[davidrsch]

Hello, would you mind linking to a tutorial or demo gh-page with this solution implemented?

[guest271314]

Something like

index.html

<script type="module">
  await navigator.serviceWorker.register("./sw.js", {
    scope: "./"
  });
  navigator.serviceWorker.addEventListener("controllerchange", async (e) => {
    // Do stuff when the ServiceWorker has control of the WindowClient
    fetch(location.href, {
      method: "head"
    });
  });
</script>

sw.js

async function log(url) {
  return fetch("https://your/deno/deploy/instance", {
    method: "query",
    body: url
  }).catch((e) => {
    // Handle error or exception
  });
}

self.addEventListener("install", (event) => {
  console.log(event.type);
  event.waitUntil(self.skipWaiting());
});

self.addEventListener("activate", async (event) => {
  console.log(event.type);
  event.waitUntil(self.clients.claim());
});

self.addEventListener("fetch", async (event) => {
  event.respondWith(fetch(event.request));
  // We're not holding up content for analytics
  log(event.request.url);
});

https://docs.deno.com/deploy/kv/manual/

Opening a database

In your Deno program, you can get a reference to a KV database using Deno.openKv(). You may pass in an optional file system path to where you'd like to store your database, otherwise one will be created for you based on the current working directory of your script.

Watching for updates in Deno KV

You can also listen for updates from Deno KV with kv.watch(), which will emit a new value or values of the key or keys you provide.

...

Production usage

Deno KV is available for use in live applications on Deno Deploy. In production, Deno KV is backed by FoundationDB, the open source key-value store created by Apple.

[guest271314]

Or. log *everything*, persistently https://github.com/guest271314/persistent-serviceworker.

[alexvgjm]

No progress on this topic yet?

All the people I know who migrates from Pages is precisely because of the lack of this feature.

Nobody talks about exposing IP addresses or anything similar, but rather something as simple as counting daily visits. I found several threads about the same. What is so problematic, dangerous or complicated about this that these requests have been ignored for more than 7 years?

I guess Microsoft wants to somehow get rid of this service. If not, I don't understand why they don't provide something so seemingly basic.

[Tedpac]

Exactly, I regretted not using Cloudflare Pages from the beginning. Maybe at some point I will migrate there.

[guest271314]

Using Deno Deploy, where it's possible to sign in with GitHub account, and export the code to GitHub.

Server

Deno.serve(() => new Response("", {
  cache: "no-store",
  headers: {
    "access-control-allow-origin": "https://github.com",
    "vary": "origin",
    "pragma": "no-cache",
    "cache-control": "no-cache",
  }
}));

Request from GitHub origin

fetch("https://xxx-zzz-123.deno.dev", {cache: "no-cache"})
  .then((response) => {
    console.log(response.ok);
  }).catch(console.warn);

Analytics on https://dash.deno.com/projects/xxx-zzz-123/analytics/24h

[Taormina]

Github, Microsoft, this is still basic functionality that everyone wants, and you have to already be tracking on your end. Can we get this?

[guest271314]

I think Microsoft hears, but ain't listening. Deno Deploy and Vercel both have "free" tiers, and both provide analytics. In GH Pages you can make a request to either to get statistics for Gh Pages requests.

[collapsinghierarchy]

Would an e2ee solution work for you people? Currently trying to build one and wanted to get a heads-up if it might serve exactly this purpose.

You create a keypair in the browser, add the analytics script (in the same way as it happens with GA) to your static webpage. If now a visitor is opening your page the same stats are gathered, but they are encrypted in-Browser before being streamed to some remote back-end. You can then open the dashboard, pull the encrypted analytics and decrypt them on your local machine and the JS shows the analysis results. It's not ready yet, but your answers might provide valueable feedback!

EDIT: Now i saw that most of you want only to see the logs. That's fine. Can't solve this problem. But an e2ee would at least keep the analytics part private to you. And you don't have to track anyone. Simple metrics should already suffice like the unique visitors / acitvity duration et al.. There is no tracking required for these stats.

[guest271314]

There's options. I would just use a free tier service like Deno Deploy or Vercel to get how many times the script loads.

I think folks are kinda hopin' Microsoft will chime in and say something like "We hear ya", and just slap on one of their existing analytics products to GitHub Pages.

That has not happened.

[Taormina]

There are a million analytics solutions, it's just wild that Microsoft and GitHub have several in-house solutions to this, AND that they have to be doing this tracking internally anyway, but it's not exposed anywhere. I can grab GA, I just don't really want to have to for what should've existed for literal years at this point.

[guest271314]

GitHub is Microsoft. Microsoft owns GitHub last time I checked. So, they know and have been silent on the feature request.

[Taormina]

I mean, yes, Microsoft bought GitHub, but massive corporations don't instantly merge and seamlessly operate in lock step. The whole point of posting here is to create volume and noise and engagement for whatever product manager is deciding what they should work on next.

[guest271314]

Microsoft bought GitHub. That's not a merger. That's an acquisition.

There's plenty of volume here. GitHub management is clearly avoiding and ignoring this feature request. Not for lack of technical capability.

[GReturn]

Bump. I would love this being implemented. It's such a basic thing to have.

[NadimGhaznavi]

Why not some Javascript? You don't have to be "sneaky". You can still respect privacy while collecting the data. You could save to CSV and render using ApexCharts and do it all within your repo.

[NadimGhaznavi]

So the flow is:

Visitor loads page
↓
JS sends repository_dispatch to GitHub API (no auth needed for public repos)
↓
GitHub Action triggers
↓
Action appends entry to metrics.csv
↓
Commit & push

All authentication is handled inside GitHub by the Actions runner using the automatic ${{ secrets.GITHUB_TOKEN }}.
Your visitors never see or handle any tokens.

[NadimGhaznavi]

Sure, it would be great if Microsoft engineered a solution, but I wouldn't hold my breath. The fact that GitHub pages is free is already pretty amazing. Also, you're only collecting page counts. That's the rub, you don't get "by country" metrics. You can get this info:

Data	            Example	          How to Get It
Local timezone	    America/New_York  Intl.DateTimeFormat().resolvedOptions().timeZone
Browser language	"en-US"	          navigator.language

So you can build pseudo location info based on timezone and the language setting the user is using....

[NadimGhaznavi]

FYI: Toronto, Canada is in the same timezone as Quito, Ecuador!

[guest271314]

I'm not a GitHub Actions expert. I'm also not interested in specifics about visit metrics, just count, if that. Can you share the complete code with instructions here?

[NadimGhaznavi]

It's just an idea, I dont' have any code.

[DeCEll-1]

Could this be implemented already?

[ross-spencer]

+1

I also think the OP's original suggestion of a graph in the repository traffic section would be good enough, and clearly only needs to be available to repos making use of GH pages.

[BfdCampos]

Mandatory monthly bump to this thread. Anything yet?

[nurzhansabitov]

Up

[deanalvero]

How is this not implemented yet? It is almost 2026!

[NadimGhaznavi]

LOL... Okay, if nobody picks it up by 2026, I'll spend some time in January doing what I can i.e. basic page count + timezone + browser language metrics.

[guest271314]

Who pays for GitHub Pages?

[guest271314]

So they aren't giving anything away for free, we're questioning why what we've paid isn't enough.

I guess that's one way of looking at it.

Another way to look at it is Microsoft could be paying developers to have their gear here, instead of, for example GitLab.

Is your strategy to keep asking until somebody in Microsoft world says, alright, for 99.99 a month, you can get IP addresses of folks who visit your GH Pages, all sorted and analyzed by Copilot, with recommendations about how to improve traffic?

[Taormina]

It's the only feature I pay for. Everything else I need is Free Tier.

For $100/month, folks won't probably bother using it. Especially since Github currently charges by the seat. We're not even saying we want IP address, and the sort of folks who avoid GA aren't the people who want Copilot recommendations. We're saying we want the absolute basics. How many times did the server serve my page? Basic server-side analytics would be plenty.

[guest271314]

Yeah, I ain'y givin' GitHub (Microsoft) a dime for analytics I can make happen myself for free.

[Taormina]

It's 2026 now!

I initially thought that this thread was about providing a solution within the bounds of what is currently available on GitHub.

This is all we ever wanted.

[jdmarshall]

@NadimGhaznavi What were you proposing to do about this issue?

[NadimGhaznavi]

I initially thought that this thread was about providing a solution within the bounds of what is currently available on GitHub. However, it seems this thread is focused on pressuring Microsoft to make the (scrubbed) server logs available. Given that focus, I have nothing to offer: I can't apply any pressure on The Borg :P

[jdmarshall]

I don’t think that’s consensus. I’d like to have unique visitors and page counts. The logs thing smells more like bargaining to me.

[guest271314]

I don't think there's a difference. There's no such thing as "scrubbed" server logs. If the feds ask Microsoft to turn over data, they'll turn it over. Users ask for the same data Microsoft remains silent.

[CodyCBakerPhD]

I would love this feature :)

Please GitHub, make it so!

[techtanic]

This is really a basic feature that just needs some API and UI changes to display the data that is almost surely already collected internally, we don't even need complex stats just the basic page metrics like; visitor count, that way we developers can know what's popular and focus on improving it.

Also, it helps a motivator seeing the popularity of a site.

This is really causing me to consider moving to Cloudflare Pages

[AqRec]

lmao it's 2026 now GitHub why not just ask your Copilot to achieve this feature?

[guest271314]

It's not a lack of know-how or technology...