ReMarkov is a Python library for generating text from existing samples using Markov chains. You can use it to customize all sorts of writing from birthday messages, horoscopes, Wikipedia articles, or the utterances of your game's NPCs. Everything works without an omnipotent "AI" - it is dead-simple code and therefore fast.

Check out the examples and feel free to contribute!

Installation

pip3 install remarkov

Example

Scrape the Wikipedia page for "Computer Programming" and generate a new text from it:

./tools/scrape-wiki.py Computer_programming | remarkov build | remarkov generate

You can also use remarkov programmatically:

from remarkov import create_model

model = create_model()
model.add_text("This is a sample text and this is another.")

print(model.generate().text())
# "This is a sample text and this is a sample text and this is a sample text ..."

Development

Make sure you run pytest as module. This will add the current directory to the import path:

python3 -m pytest

This project uses black for source code formatting:

black .

Generate documentation for the project (this uses the original pdoc at pdoc.dev):

git checkout gh-pages
pdoc -t pdoc/template -o public/docs <path_to_remarkov_module>

Run type checks using mypy:

mypy -p remarkov

Publishing is done like this (don't forget to bump the version in setup.py):

pip3 install twine # optional

git tag -a <version>
git push --tags

python3 setup.py clean --all
python3 setup.py sdist bdist_wheel
twine check "dist/*"
twine upload "dist/*"