sam.pikesley.org

Markodon

Force your old tweets through a Markov Chain like coffee through a civet

It's spouting nonsense on Mastodon as you read this

Install this

git clone https://github.com/pikesley/markodon
cd markodon
export MARKODON=$(pwd)
make install

You'll need Python 3 (I built it with 3.10, but I now have it working with 3.8 elsewhere)

Prepare your tweets

Download your Twitter archive

If that's still a thing by the time you're reading this

Prepare the data

Unpack your archive and copy the JSON to here:

cd /tmp/
unzip /path/to/twitter-datestamp-long-alphanumeric-string.zip
cp data/tweets.js ${MARKODON}/
cd ${MARKODON}/

Extract just the tweets from the JSON:

make extract

Sanitise the tweets file:

make sanitise

This does the following to the archive:

  • Removes the enclosing quotes from each tweet
  • Removes embedded newlines
  • Removes embedded quotes and so on (', ", * etc)
  • Removes all RT tweets
  • Removes all wordle tweets
  • Removes all @username words
  • Removes all URLs
  • Removes any truncated words from the end of a tweet
  • Replaces HTML-encoded strings (& etc) with their plain-text equivalents

So now you have a cleaned, sanitised list of your tweets at tweets.txt.

Get set up

You need to create a Mastodon application, with the write scope. Once you've done this, you just need Your acccess token (you can safely ignore all the Oauth stuff)

You need the conf-file:

cp conf-example.yaml conf.yaml

And then fill in:

  • Your server URL (e.g. https://mastodon.me.uk)
  • Your token from your application
  • Optionally, the max-toot-length

Send a toot

And now, to generate and send a toot,

make toot

Hacking on it

If you want to play with the (hastily-written over a weekend) code, there's a Docker image

make build
make run

To run the tests:

make all

These comments are generated from replies to this Mastodon post