Pocket Stream Archive

(Your own personal Way-Back Machine)

Save an archived copy of all websites you star using Pocket, indexed in an html file. Powered by the new headless Google Chrome and good 'ol wget.

Quickstart

archive.py is a script that takes a Pocket export, and turns it into a browsable html archive that you can store locally or host online.

Runtime: I've found it takes about an hour to download 1000 articles, and they'll take up roughly 1GB. Those numbers are from running it signle-threaded on my i5 machine with 50mbps down. YMMV.

Dependencies: google-chrome --headless, wget, python3

brew install Caskroom/versions/google-chrome-canary
brew install wget python3
# OR on linux
wget -q -O - https://dl-ssl.google.com/linux/linux_signing_key.pub | sudo apt-key add -
sudo sh -c 'echo "deb [arch=amd64] http://dl.google.com/linux/chrome/deb/ stable main" >> /etc/apt/sources.list.d/google-chrome.list'
apt update; apt install google-chrome-canary python3 wget

Archiving:

1. Download your pocket export file ril_export.html from https://getpocket.com/export
2. Download this repo git clone https://github.com/pirate/pocket-archive-stream
3. cd pocket-archive-stream/
4. ./archive.py ~/Downloads/ril_export.html

It produces a folder pocket/ containing an index.html, and archived copies of all the sites, organized by timestamp. For each sites it saves:

• wget of site, e.g. en.wikipedia.org/wiki/Example.html with .html appended if not present
• sreenshot.png 1440x900 screenshot of site using headless chrome
• output.pdf Printed PDF of site using headless chrome

You can tweak parameters like screenshot size, file paths, timeouts, etc. in archive.py. You can also tweak the outputted html index in index_template.html. It just uses python format strings (not a proper templating engine like jinja2), which is why the CSS is double-bracketed {{...}}.

Live Updating: (coming soon)

It's possible to pull links via the pocket API instead of downloading an html export. Once I write a script to do that, we can stick this in cron and have it auto-update on it's own.

For now you just have to download ril_export.html and run archive.py each time it updates. The script will run fast subsequent times because it only downloads new links that haven't been archived already.

Publishing Your Archive

The pocket archive is suitable for serving on your personal server, you can upload the pocket archive to /var/www/pocket and allow people to access your saved copies of sites.

Just stick this in your nginx config to properly serve the wget-archived sites:

location /pocket/ {
    alias       /var/www/pocket/;
    index       index.html;
    autoindex   on;
    try_files   $uri $uri/ $uri.html =404;
}

Make sure you're not running any content as CGI or PHP, you only want to serve static files!

Urls look like: https://sweeting.me/pocket/archive/1493350273/en.wikipedia.org/wiki/Dining_philosophers_problem

Info

This is basically an open-source version of Pocket Premium (which you should consider paying for!). I got tired of sites I saved going offline or changing their URLS, so I started archiving a copy of them locally now, similar to The Way-Back Machine provided by archive.org.

Now I can rest soundly knowing important articles and resources I like wont dissapear off the internet.

My published archive as an example: sweeting.me/pocket.

Security WARNING

Hosting other people's site content has security implications for your domain, make sure you understand the dangers of hosting other people's CSS & JS files on your domain. It's best to put this on a domain of its own to slightly mitigate CSRF attacks.

It might also be prudent to blacklist your archive in your robots.txt so that search engines dont index the content on your domain.

TODO

• body text extraction using fathom
• auto-tagging based on important extracted words
• audio & video archiving with youtube-dl
• full-text indexing with elasticsearch
• video closed-caption downloading for full-text indexing video content
• automatic text summaries of article with summarization library
• feature image extraction
• http support (from my https-only domain)