Compare the full difference
- Implements new feature:
MarkovTextModel.add_new_corpus_data_to_model and its async counterpart. This allows you to add new corpus entries to an existing model without having to regenerate over the whole corpus. This will not work if the stored model is in a compiled state.
Compare the full difference
- Adds a new optional setting
MARKOV_STATE_SIZE to override the default value used by markovify. Note that state sizes need to be consistent, so changing this setting means you should regenerate your text models with the new size.
Compare the full difference
- Adds a cached property
is_compiled_model to MarkovTextModel. This allows a user to check in advance if models are compiled without having to know the internals of markovify.
Compare the full difference
- Bug fix: gracefully handle when
MARKOV_CORPUS_MAX_CHAR_LIMIT is not set.
Compare the full difference
- Adds a migration for data field in MarkovTextModel.
Compare the full difference
- Improves sentence generation performance by loading the text model into a cached property.
- Cached properties are cleared when calling
refresh_from_db.
- Enables storing non-compiled models to the database for use in combining multiple models into a single chain.
- Adds boolean settings parameter
MARKOV_STORE_COMPILED_MODELS to set default behavior.
- This can be overridden when calling
update_model_from_corpus using the store_compiled kwarg.
- Adds ability to auto-detect if a loaded model is compiled or not, and will automatically compile for sentence generation if needed.
- Adds support for the
tries directive when calling generate_sentence.
- Adds support for combining multiple models into one via
MarkovTextModel.combine_models
Compare the full difference
- Remove extraneous dependencies that should only appear in dev-dependencies.