- Remove support for Python 2
- Allow for 280-character tweets
- Raise a RunTimeError if model fails to compose anything.
- refactor TwitterMarkov.compose to use make_short_sentence.
- Add max-len and state-size options to cli.
- change default state size to 2
- ignore tweets with blacklisted words when replying
- refactor checking module
- expand docs
- bump required twitter_bot_utils
- Remove one per line option
- Expand docs
- Add option to create a corpus from a file with one text per line
- Fix a double-logging bug
- Small update for changed twitter bot utils API
- Replace cobe with markovify. This simplifies learning, since it's just adding to a text file corpus.
- Also changes Twitter_markov API, removing catalyst argument and replacing 'brain' keyword arguments with 'corpus' or 'model'.
- Replace twittermarkov learn with twittermarkov corpus
- Add checking.generator function
- Restore similarity/blacklist checker for generated text
- rename class from Twitter_markov to TwitterMarkov
- Rework cli tools, renaming twittermarkov learn -> twittermarkov corpus
- Combine two command line tools into one command with subcommands.
- Handle errors in learning more cleanly.
- Twitter_markov class no longer extends Tweepy.API, so an existing API object can be passed in
- Cleaned up code around brain naming.
- Expanded readme.