A suite of tools for mirroring and hoarding web pages you visit for later offline viewing. I.e. your own personal Wayback Machine that can also archive HTTP POST requests and responses, as well as most other HTTP-level data, which also follows “archive everything now, figure out what to do with it later” philosophy
pwebarc
?Personal Private Passive Web Archive (pwebarc
) is a suite of tools to capture, collect, save, mirror, manage archives of (i.e. hoard), and view web pages and whole websites offline.
In short, pwebarc
’s main workflow is this: you install an extension/add-on into the browser of your choice (both Firefox- and Chromium-based browsers are supported) and just browse the web while it captures and archives everything your browser fetches from the network to your local file system in a way that can be used to reconstruct and replay your browsing session later (by default, the extension has lots of options controlling what data from which tabs should and should not be captured).
pwebarc
does and does not do.pwebarc
.See there for more screenshots.
pwebarc
exists?So, you wake up remembering something interesting, you try to look it up on Google, you fail, eventually you remember the website you seen it at (or a tool like recoll or Promnesia helps you), you go there to look it up… and discover it offline/gone/a parked domain. Not a problem! Have no fear! You go to Wayback Machine and look it up there… and discover they only archived an ancient version of it and the thing you wanted is missing there.
Or, say, you read a cool fanfiction on AO3 years ago, you even wrote down the URL, you go back to it wanting to experience it again… and discover the author made it private… and Wayback Machine saved only the very first chapter.
“If it is on the Internet, it is on Internet forever!” they said. They lied!
Things vanish from the Internet all the time, Wayback Machine is awesome, but
Meanwhile, pwebarc
solves all of the above out-of-the-box (though, the full-text search is currently being done by other tools running on top of it).
Say, there is a web page that can not be easily reached via curl
/wget
(because it is behind a paywall or complex authentication method that is hard to reproduce outside of a browser) but for accessibility (or just simple reading comfort) reasons each time you visit that page you want to automatically feed its source to a third-party app that strips and/or modifies HTML markup in a website-specific way and feeds it into a TTS engine, a Braille display, or a book reader app.
With most modern web browsers you can do TTS either out-of-the-box or by installing an add-on (though, be aware of privacy issues when using most of these), but tools that can do website-specific accessibility without also being a website-specific UI app are very few.
Meanwhile, pwebarc
with some scripts can do it.
Say, there’s a web page/app you use (like a banking app), but it lacks some features you want, and in your browser’s Network Monitor you can see it uses JSON RPC or some such to fetch its data, and you want those JSONs for yourself (e.g., to compute statistics and supplement the app output with them), but the app in question has no public API and scraping it with a script is non-trivial (e.g., they do complicated JavaScript+multifactor-based auth, try to detect you are actually using a browser, and they ban you immediately if not).
Or, maybe, you want to parse those behind-auth pages with a script, save the results to a database, and then do interesting things with it (e.g., track price changes, manually classify, annotate, and merge pages representing the same product by different sellers, do complex queries, like sorting by price/unit, limit results by geographical locations extracted from text labels, etc).
Or, say, you want to fetch a bunch of pages belonging to two recommendation lists on AO3 or GoodReads, get all outgoing links for each fetched page, union sets for the pages belonging to the same recommendation list, and then intersect the results of the two lists to get a shorter list of things you might want to read with higher probability.
Or, more generally, say, you want to tag web pages referenced from a certain set of other web pages with some tag in your indexing software, and update it automatically each time you visit any of the source pages.
Or, say, you want to combine a full-text indexing engine, your browsing and derived web link graph data, your states/ratings/notes from org-mode, messages from your friends, and other archives, so that you could do arbitrarily complex queries over it all, like “show me all GoodReads pages for all books not marked as DONE
or CANCELLED
in my org-mode
files, ever mentioned by any of my friends, ordered by undirected-graph Pagerank algorithm biased with my own book ratings (so that books sharing GoodReads lists with the books I finished and liked will get higher scores)”. So, basically, you want a private personalized Bayesian recommendation system.
“Everything will have a RESTful API!” they said. They lied! A lot of useful stuff never got RESTful APIs, those RESTful APIs that exists are frequently buggy, you’ll probably have to scrape data from HTMLs anyway.
“Semantic Web will allow arbitrarily complex queries spanning multiple data sources!” they said. Well, 25 years later (“RDF Model and Syntax Specification” was published in 1999), almost no progress there, the most commonly used subset of RDF does what indexing systems in 1970s did, but less efficiently and with a worse UI.
Meanwhile, pwebarc
provides some of the tools to help you build your own little local data paradise.
Unlike most of its alternatives, pwebarc
’s main workflow is to passively collect and archive HTTP requests and responses directly from your browser as you browse the web instead of making you to ask some tool or web app to snapshot it for you or forcing you to explicitly snapshot/record separate browsing sessions/tabs, thus
allowing you to archive any HTTP data, not just the pages available via HTTP GET
requests (e.g., it can archive answer pages of web search engines fetched via HTTP POST
, JSON RPC/AJAX data, etc);
downloading everything only once, not once with your browser and then the second time with a separate tool like ArchiveBox;
allowing you to trivially archive web pages hidden behind CAPTCHAs, requiring special cookies, multi-factor logins, paywalls, anti-scraping/curl
/wget
measures, and etc; after all, the website in question only interacts with your normal web browser, not with a custom web crawler;
freeing you from worries of forgetting to archive something because you forgot to press a button somewhere.
In other words, pwebarc
is your own personal Wayback Machine which passively archives everything you see and, unlike the original Wayback Machine, also archives HTTP POST
requests and responses, and most other HTTP-level data.
Technically, pwebarc
is most similar to
Or, to summarize it another way, you can view pwebarc
as an alternative for mitmproxy which leaves SSL/TLS layer alone and hooks into target application’s runtime instead.
In fact, an unpublished and now irrelevant ancestor project of pwebarc
was a tool to generate website mirrors from mitmproxy
stream captures. (By the way, if you want that, pwebarc
’s wrrarms
tool can do that for you. It can take mitmproxy
dumps as inputs.) But then I got annoyed by all the sites that don’t work under mitmproxy
, did some research into the alternatives, decided there were none I wanted to use, and so I made my own.
To highlight the main differences to its alternatives, pwebarc
DOES NOT:
require you to capture and/or collect and/or archive and/or export recorded data explicitly one page/browsing session at a time (the default behaviour is to archive everything completely automatically);
(though, the extension implements optional “limbo mode” which delays archival of collected data and provides optional manual/semi-automatic control if you want it;)
force you to use Chromium (and, yes, this point deserves repeating because you can use pwebarc
with Firefox, Tor Browser, LibreWolf, etc, which is not a small thing, since if you tried using any of the alternatives running under Chromium, you might have noticed that the experience there is pretty awful: it becomes even slower than usual, large files don’t get captured, random stuff fails to be captured at random times because Chromium randomly detaches its debugger from its tabs… none of these problems exist on Firefox-based browsers);
require you to download the data you want to archive twice or more (you’d be surprised how commonly other tools will either ask you to do that explicitly, or just do that silently when you ask them to save something);
require you to store all the things in browser’s local storage where they can vanish at any moment;
require you to run a database server;
share your archived data with anyone by default;
require you to run a web browser to view the data you already archived (in fact, wrrarms
comes with a bunch of scripts which allow you to use other tools for that; e.g., a script to view HTML documents via pandoc
piped into less
in your favorite tty emulator).
The pWebArc
browser extension that captures all HTTP requests and responses your browser fetches, dumps them into WRR
format, and pushes those dumps to the specified archiving server.
The extension is to be considered stable for both Firefox- and Chromium-based browsers (though, it is tested much less on Chromium than on Firefox).
The dumb archiving server that simply dumps everything it gets to disk one file per HTTP request+response.
The dumb archiving server is to be considered stable (it’s so simple there hardly could be any bugs there).
The wrrarms
tool that allows you to display, search, organize, and manipulate archive files.
wrrarms
tool is beta software, it does about 70% of the stuff I want it to do ATM. See the TODO list for more info.
A patch for Firefox to allow the above extension to collect request POST data as-is. This is not required and even without that patch pWebArc
will collect everything in most cases, but it could be useful if you want to correctly capture POST requests that upload files.
See “Quirks and Bugs” section of extension’s “Help” page for more info.
Firstly, pwebarc
is designed to be simple (as in adhering to the Keep It Stupid Simple principle) and efficient (as in running well on ancient hardware):
pWebArc
webextension add-on does almost no actual work, simply generating HTTP request+response dumps, pushing them to the archiving server, and freeing the memory as soon as possible (unless you enable “limbo mode”, but then you asked for it), thus keeping your browsing experience snappy even on ancient hardware (if you ignore the code needed to support Chromium, the UI, and the workarounds for all the bugs in various browsers, the leftovers are also pretty small and simple source code-wise);
dumb archiving server simply compresses those dumps and saves them to disk as-is (it is tiny, taking less than 200 lines of code);
anything that is remotely computationally expensive is delegated to the wrrarms
CLI tool;
meanwhile, on disk, your dumps get stored using the simplest, trivially parsable with many third-party libraries, yet most space-efficient on-disk file format representing separate HTTP requests+responses there currently is (aka Web Request+Response
, WRR
), which is a file format that is both more general and more simple than WARC, much simpler than that mitmproxy
uses, and much more efficient than HAR.
Secondly, pwebarc
is built to follow “capture and archive all the things as they are now, as raw as possible, modify those archives never, convert to other formats and extract values on-demand” philosophy.
Meaning,
pWebArc
webextension add-on collects data as browser gives it, without any data normalization and conversion (when possible),
dumb archiving server simply compresses the dumps the add-on pushes at it and saves them all to disk as-is,
meanwhile, all data normalization, massaging, post-processing, and extraction of useful values is delegated to the wrrarms
, which also does not overwrite any WRR files, ever.
pwebarc
expects you to treat your older pre-pwebarc
archives you want to convert to WRR similarly:
wrrarms import
them into a separate directory, butmitmproxy
(or whatever) dumps alone (on an external backup drive, if you lack disk space on your machine).This way, if wrrarms
has some unexpected bug, or wrrarms import
adds some new feature, you could always re-import them later without losing anything.
Currently, pwebarc
has two main use cases for regular users, in both of which you first capture some data using the add-on and then you either
feed wrrarms
a subset of your archives to generate a static offline website mirror a-la wget -mpk
, which you can then view with your favorite web-browser as normal; except, unlike with wget
you can discover you dislike the result, change some options, and re-generate the mirror without re-downloading anything;
you use wrrarms
to simply maintain a tree of symlinks pointing to latest WRR file for each URL and then read them — by using w3m
, pandoc
, any other HTML reader you want, or feed them to TTS engine, or a Braille display — via some scripts; personally, I prefer this one, because I hate web browsers and prefer to read most things from a TTY;
(TODO: eventually, when that gets implemented, a Wayback Machine-like Web UI for replay).
Alternatively, you can programmatically access that data by asking wrrarms
to dump WRR files into JSONs or verbose CBORs for you, or you can just parse WRR files yourself with readily-available libraries.
Since the whole of pwebarc
adheres to the philosophy described above, the simultaneous use of pWebArc
and wrrarms
helps immensely when developing scrapers for uncooperative websites: you just visit them via your web browser as normal, then, possibly years later, use wrrarms
to organize your archives and conveniently programmatically feed the archived data into your scraper without the need to re-fetch anything.
Given how simple the WRR file format is, in principle, you can modify any HTTP library to generate WRR files, thus allowing you to use wrrarms
with data captured by other software.
Which is why, personally, I patch some of the commonly available FLOSS website scrapers to dump the data they fetch as WRR files so that in the future I could write my own better scrapers and indexers and test them on a huge collected database of already collected inputs immediately.
Also, as far as I’m aware, wrrarms
is a tool that can do more useful stuff to your WRR archives than any other tool can do to any other file format for HTTP dumps with the sole exception of WARC.
See extension’s “Help” page (or the “Help” button in the extension’s UI, which will make it interactive) for a long detailed description of what the extension does step-by-step.
It is a must-read, though reading it from the extension’s UI is highly recommended.
See the “Frequently Asked Questions” section of extension’s “Help” page for the answers to the frequently asked questions, including those about common quirks you can encounter while using it.
See the “Quirks and Bugs” section of extension’s “Help” page for more info on quirks and limitations of pWebArc
when used on different browsers.
See below for a long list of comparisons to its alternatives.
Then notice that pwebarc
is the best among them, and go follow “Quickstart” section for setup instructions. ( •̀ ω •́ )✧
To follow the development:
See the TODO list below for the list of things that are not implemented/ready yet.
See CHANGELOG for the progress log and human-readable description of recent changes (which is much shorter and more comprehensible than the commit log).
If you want to learn to use wrrarms
tool, see its README, which has a bunch of extended and explained usage examples.
Also, a lot of info on that page can be seen by running wrrarms --help
.
See example scripts to learn how to do various interesting things with your archived data.
In the unlikely case you have problems with the dumb archiving server, see its README. Or you can read pwebarc-dumb-dump-server --help
instead.
If you want to build the extension from source, see its README.
If you are a developer, see all the wrrarms
-related links above, and also see the description of the on-disk file format used by all these tools.
If your questions are not unanswered by these, then open an issue on GitHub or get in touch otherwise.
Yes, as of June 2024, I archive all of my web traffic using pwebarc
, without any interruptions, since October 2023. Before that my preferred tool was mitmproxy.
After adding each new feature to wrrarms
CLI tool, as a rule, I feed at least the last 5 years of my web browsing into it (at the moment, most of it converted from other formats to .wrr
, obviously) to see if everything works as expected.
pWebArc
extensionImproved UI:
Roll/unroll popup UI in steps, a-la uBlock Origin. The number of settings pWebArc now has is kind of ridiculous (and I still want more), I find it hard to find stuff in there myself now, so. Alternatively, make a separate “Settings” page, but I kind of dislike that idea, I like having everything on a single page which can be Control+F
ed.
Track navigations and allow to use them as boundaries between batches of reqres saved in limbo mode.
Reorganize tracking- and problematic-related options into config profiles (~25% done), allow them to override each over, and then implement per-host profiles.
Automatic capture of DOM snapshots when a page changes.
Serverless archival into browser’s local storage: for easier bootstrap and to allow using Tor Browser securely with its default config.
Maybe: Dumping straight into WARC, so that third-party tools (i.e. not just wrrarms
) could be used for everything except capture.
wrrarms
toolexport
ing, so that, e.g. you could organize
or export
a reqres containing https://web.archive.org/web/<something>/<URL>
as if it was just a <URL>
.scrub
bing of CSS files.Download the dumb archiving server pwebarc_dumb_dump_server.py
script and run it, it has no dependencies except Python itself, and it’s source code is less than 200 lines of pure Python. It will start saving data into pwebarc-dump
directory wherever you run it from.
Alternatively, install via
and run as
See there for more info.
(This step will eventually become optional, but not yet.)
Install the browser extension/add-on:
On Firefox, Tor Browser, LibreWolf, etc: Install the extension from addons.mozilla.org or see Build from source for Firefox/Tor Browser/LibreWolf/etc.
On Chromium/Chrome/etc: See Installing on Chromium/Chrome or Build from source for Chromium/Chrome/etc.
Unfortunately, this requires a bit more work than clicking “Install” on Chrome Web Store, yes. “Why isn’t pWebArc
on Chrome Web Store?” I’m glad you asked! I’m not a lawyer, but to me it looks like pWebArc
violates Chrome Web Store’s “Terms of Use”. Specifically, the “enables the unauthorized download of streaming content or media” clause. In my personal opinion, any content you web browser downloads while you are browsing the web normally you are “authorized” to download, but given the context of that clause in that document, I feel like Google would disagree.
Meanwhile, pWebArc
tries its best to collect all web traffic you browser generates, which, obviously, includes streaming content.
On Chromium (but not on Firefox), technically, at the moment, pWebArc
does fail to collect streaming media properly because Chromium has a bug that prevents collection of the first few KiB of all audio and video files, and its API design prevents collection of large files in general, but if we are talking about YouTube, then most of the data of those streaming media files Chromium will fetch while you watch a video there will, in fact, get collected even on Chromium.
So, in principle, to conform to those terms, pWebArc
would need to add artificial limitations, which I’m unwilling to do.
(Meanwhile, on Firefox, pWebArc
will just silently collect everything you browser fetches. And addons.mozilla.org’s policies do not restrict this.)
(Also, Chrome Web Store actually requires developers to pay Google to host their add-ons there while Mozilla’s service is free. Meaning, you should go and donate to any free add-ons that do not violate your privacy and sell your data you installed from there. Their authors paid Google so that you could conveniently install their add-ons with a single click. Who else will pay the authors?)
Now load any web page in your browser, the extension will report if everything works okay, or tell you where the problem is if something is broken.
Assuming the extension reported success: Congratulations! You are now collecting and archiving all your web browsing traffic originating from that browser. Repeat extension installation for all browsers/browser profiles as needed.
If you just want to collect everything and don’t have time to figure out how to use the rest of this suite of tools right this moment, you can stop here and figure out how to use the rest of this suite later.
It took me about 6 months before I had to refer back to previously archived data for the first time when I started using mitmproxy to sporadically collect my HTTP traffic in 2017. So, I recommend you start collecting immediately and be lazy about the rest. Also, I learned a lot about nefarious things some of the websites I visit do in the background while doing that, now you are going to learn the same.
Next, you should read extension’s “Help” page. It has lots of useful details about how it works and quirks of different browsers. If you open it by clicking the “Help” button in the extension’s UI, then hovering over or clicking on links in there will highlight relevant settings.
See “Setup recommendations” section for best practices for configuring your system and browsers to be used with pwebarc
.
See the docs of the wrrarms
tool.
Install by running
Start the dumb archiving server by running
Install the add-on and etc as above.
Alternatively, built XPI and Chromium ZIPs can be taken from
see the extension’s README for more info on how to install them manually.
You can add pwebarc_dumb_dump_server.py
to Autorun or start it from your ~/.xsession
, systemd --user
, etc.
You can also make a new browser profile specifically for archived browsing, run Firefox as firefox -ProfileManager
to get to the appropriate UI. On Windows you can just edit your desktop or toolbar shortcut to target
"C:\Program Files\Mozilla Firefox\firefox.exe" -ProfileManager
or similar by default to switch between profiles on browser startup.
It is highly recommended you make separate browser profiles for anonymous and logged-in browsing with separate extension instances pointing to separate archiving server instances dumping data to different directories on disk.
Set the “anonymous” browser profile to always run in “Private Browsing” mode to prevent login persistence there. If you do accidentally login in “anonymous” profile, move those dumps out of the “anonymous” directory immediately.
This way you can easily share dumps from the “anonymous” instance without worrying about leaking your private data or login credentials.
./pwebarc_dumb_dump_server.py --host 127.0.99.1
or similar.about:config
and add 127.0.99.1
to network.proxy.no_proxies_on
.http://127.0.99.1:3210/pwebarc/dump
.You probably don’t want to use 127.0.0.1
and 127.0.1.1
with Tor Browser as those are normal loopback addresses and you probably don’t want to allow stuff from under Tor to access your everyday stuff.
Or, you could run both the Tor Browser, and ./pwebarc_dumb_dump_server.py
in a container/VM and use the default 127.0.0.1
address.
“Cons” and “Pros” are in comparison to the main workflow of pwebarc
. Most similar and easier to use projects first, harder to use and less similar projects later.
Tools most similar to pwebarc
in their implementation, though not in their philosophy and intended use.
Cons:
archiveweb.page
for each browser tab; and thenPros:
pwebarc
currently has.Differences in design:
pWebArc
);pWebArc
will never get (if you want that, pWebArc
expects you to use UserScripts instead).Same issues:
When pWebArc
extension is run under Chromium, a bunch of Chromium’s bugs make many things pretty annoying.
Both pWebArc
and archiveweb.page
suffer from exactly the same issues, which — if you know what to look for — you can notice straight in the advertisement animation on their “Usage” page.
Those issues have no workarounds known to me. But because they exists, I made pWebArc
instead of trying to port archiveweb.page
to Firefox, forking it, and making the fork follow my preferred workflow.
A self-hosted web app and web crawler written in Node.js
most similar to pwebarc
in its intended use.
DiskerNet
does its web crawling by spawning a Chromium browser instance and attaching to it via its debug protocol, which is a bit weird, but it does work, and with exception of pwebarc
it is the only other tool I know of that can archive everything passively as you browse, since you can just browse in that debugged Chromium window and it will archive the data it fetches.
Cons:
Pros:
pwebarc
aims to do on the basic level.Same issues:
Cons:
And then you still need something like this suite to look into the generated archives.
Cons:
mitmproxy
dump files are flat streams of HTTP requests and responses that use custom frequently changing between versions data format, so you’ll have to re-parse them repeatedly using mitmproxy
’s own parsers to get to the requests you want;Pros:
pWebArc
add-on currently does not.Cons:
Pros:
pWebArc
add-on currently does not.And then you still need something like this suite to look into the generated archives.
A web crawler and self-hosted web app into which you can feed the URLs for them to be archived.
So to make it archive all of your web browsing like pwebarc
does:
Cons:
mitmproxy
with archivebox-proxy plugin; or,
POST
requests with it.Pros:
git
repos, etc;Still, probably the best of the self-hosted web-app-server kind of tools for this ATM.
Browser add-ons to capture whole web pages.
Cons:
Pros:
A system similar to ArchiveBox
, but has a bulit-in tagging system and archives pages as raw HTML + whole-page PNG rendering/screenshot — which is a bit weird, but it has the advantage of not needing any replay machinery at all for re-viewing simple web pages, you only need a plain simple image viewer.
Pros and Cons are almost identical to those of ArchiveBox
above.
wget -mpk
and curl
Cons:
wget
and making wget
play pretend at being a normal web browser is basically impossible;curl
, curl
also doesn’t have the equivalent to wget
’s -mpk
options;Pros:
wget -mpk
done right.
Cons:
POST
requests with it;Pros:
A simple web built on top of wpull
, presented to you by the ArchiveTeam, a group associated with the Internet Archive which appears to be the source of archives for the most of the interesting pages I find there.
Cons:
POST
requests with it;Pros:
Stand-alone tools doing the same thing SingleFile add-on does: generate single-file HTMLs with bundled resources viewable directly in the browser.
Cons:
POST
requests with it;Pros:
The crawler behind the Internet Archive.
It’s a self-hosted web app into which you can feed the URLs for them to be archived, so to make it archive all of your web browsing:
Cons:
/save/
REST API URLs (which is not hard, but I’m unaware if any such add-on exists);POST
requests with it.Pros:
A self-hosted wiki that archives pages you link to in background.
ArchiveBox wiki has a long list or related things.
It’s an awesome personal private archival system adhering to the same philosophy as pwebarc
, but it’s basically an abstraction replacing your file system with a content-addressed store that can be rendered into different “views”, including a POSIXy file system.
It can do very little in helping you actually archive a web page, but you can start dumping new pwebarc
.wrr
files with compression disabled, decompress you existing .wrr
files, and then feed them all into Perkeep to be stored and automatically replicated to your backup copies forever. (Perkeep already has a better compression than what pwebarc
currently does and provides a FUSE FS interface for transparent operation, so compressing things twice would be rather counterproductive.)
GPLv3+, some small library parts are MIT.