0.6.0

Bug fixes

• Fix format of Django API view response data #608
• Prevent non-string type values from breaking comparison algorithms #655

Improvements

• Enhancement of the decimal percent of similarity in levenshtein_distance (via @addelll) #618
• Add Django Setting to allow ChatterBot Statement objects to be used instead of Django Statement model (via @web-maker) #653

Breaking changes

• Change logic adapters to just return one value #587
• The confidence value is no longer returned from the generate_response method (it's available on the statement object now) #588
• Remove confidence parameter from output adapter process methods #589
• Require session id for generate_response method #598

Conversations

• Create conversation model #594

Corpus

• Add data to English dialog corpus (via @vkosuri) #516
• Migrated corpus module to a new separate repository (https://github.com/gunthercox/chatterbot-corpus) [#654]

0.5.5

Features

• Allow custom file paths to ChatterBot corpus files [https://github.com//pull/549]
• Support for Turkish math words (via @yusuf-celik) [https://github.com//pull/558]
• Return confidence value on statement object [https://github.com//pull/565]
• Convert clean_whitespace utility into a preprocessor [https://github.com//pull/569]
• Allow seed word to be set for Twitter training adapter [https://github.com//pull/578]
• Fix issue whre ^ ends up being used as bitwise XOR instead of ** for powers (via @Martmists) [https://github.com//pull/582]

Bug fixes

• The nltk.data.find argument needs to be a path, not just a filename (via @Gaboose) [https://github.com//pull/548]
• Add missing serialize method to Response model [https://github.com//pull/557]
• Fix comparison in levenshtein (via @gballardin) [https://github.com//pull/579]
• Fix "empty generator" bug in ChatBot.filters (via @thesprockee) [https://github.com//pull/580]
• Fix display formatting for exponents [https://github.com//pull/584]

Maintenance

• Only control write permission in the ChatBot class [https://github.com//pull/561]
• Disconnect storage from adapter class [https://github.com//pull/552]
• Add tests for examples [https://github.com//pull/586]

Deprecation

• Remove deprecated code [https://github.com//pull/542]

Documentation

• Document training path methods [https://github.com//pull/550]
• Add more examples and expand tutorial in documentation [https://github.com//pull/564]

0.5.4

Features

• Tighter Django integration [https://github.com//pull/471]
• Add save method to statements [https://github.com//pull/515]
• Add created_at field to statements [https://github.com//pull/536]
• Add ordering parameter to filter method [https://github.com//pull/535]

Deprecation

• Deprecate get_default and update_default session methods [https://github.com//pull/541]

Bug fixes

• Remove tested constraint for response objects in filter method [https://github.com//pull/511]
• Correct unicode characters in German corpus [https://github.com//pull/521]

Maintenance

• Move adapter type exception to adapter class [https://github.com//pull/512]
• Rename SessionManager to ConversationSessionManager [https://github.com//pull/519]
• Add warnings about module rename [https://github.com//pull/538]

Documentation

• Add link to ChatterBot live example created by @vkosuri [https://github.com//pull/517]

0.5.3

Bug fixes

• Change default occurrence count to from 0 to 1 in Response model [https://github.com//pull/491]
• wsgi should point example_app (via @vkosuri) [https://github.com//pull/505]
• Convert response unicode into JSON object (via @vkosuri) [https://github.com//pull/498]

Maintenance

• Remove 'has_storage' method [https://github.com//pull/492]
• Deprecate OutputFormatAdapter [https://github.com//pull/500]
• Run json database in-memory when testing [https://github.com//pull/496]
• Do not update extra data from kwargs [https://github.com//pull/499]
• Update migration documentation (via @rajasimon) [https://github.com//pull/501]
• Remove the deprecated approximate sentence match adapter [https://github.com//pull/502]
• Remove the deprecated sentiment adapter class [https://github.com//pull/503]
• Update bootstrap version to 4-alpha in example app [https://github.com//pull/506]

Increasing Integration with Django

Several changes have been made to help make Django a first class citizen in ChatterBot. These changes improve both the performance of ChtterBot when using the popular Python web framework. The changes also help to make it more convenient for developers to use Django and ChatterBot together.

• Give Django Statement & Response models the same attributes as ChatterBot Statements & Responses [https://github.com//pull/488]
• Make sure default Django input and output adapters work with the Statement & Response models [https://github.com//pull/490]

0.5.2

Efficiency

• Cache results from can_process method in math adapter [https://github.com//pull/461]

Bug fixes

• Fix grammar issues in russian conversation corups (via @alxmamaev) [https://github.com//pull/462]
• Handle the case that a chat session does not exist [https://github.com//pull/480]
• Fix Hipchat session management (via @vkosuri) [https://github.com//pull/485]

Testing

• Move Django tests out of the example app [https://github.com//pull/481]

0.5.1

Features

• Additional mathematical operation support thanks to @le-corentin [https://github.com//pull/449]

Bug fixes

• Fixed unicode characters in the spanish corpus (via @luksireiku) [https://github.com//pull/440]

Deprecation

• Deprecate matching logic adapters [https://github.com//pull/456]

0.5.0

ChatterBot 0.5 Release Notes (Major Release)

Backwards incompatible changes

• Split utils folder into a single file [https://github.com//pull/418]
• Rename 'context' to 'chatbot' [https://github.com//pull/419]
• Clean up or remove NLTK wrapper classes [https://github.com//pull/420]
• Change adapter import paths [https://github.com//pull/422]

Bug fixes

• Update scope of external package imports [https://github.com//pull/426]
• Remove check for ajax request in Django view [https://github.com//pull/428]
• Prevent possible edge case when adding inserting logic adapters [https://github.com//pull/434]

Features

• Add session management [https://github.com//pull/424]
• Remove dependency on Textblob [https://github.com//pull/435]

0.4.14

Bug fixes

• Handel the case that a Mongo DB statement does not have an in_response_to attribute [https://github.com//pull/415]

Features

• Add training class for Ubuntu corpus [https://github.com//pull/405]

0.4.13

Features

• Add support for integration with the Microsoft Bot Framework [https://github.com//pull/381]
  • (A huge thanks to @vkosuri for working to add support for integration with Microsoft's Bot Framework).
• Add Django management command to train chat bot [https://github.com//pull/401]
• Add a logic adapter to return a specific response to a specific input. [https://github.com//pull/409]
• Add a logic adapter to return a default response when no good response is known [https://github.com//pull/409]
• Remove dependency on fuzzywuzzy [https://github.com//pull/410]

Bug fixes

• Fix adapter issue with unicode literals [https://github.com//pull/411]

0.4.12

Simplification of matching-type logic adapters

This is a cool internal change that occurred for this release. By pulling several statement comparison methods out into their own module, much of the code for several matching style logic adapters became nearly identical. This made it possible to reduce a large amount of repeated code.

In a future release, all of the matching style logic adapers (ClosestMatch, ClosestMeaning, SentimentAnalysis, ApproximateSentenceMatch) will be replaced by a single MatchingAdapter class that will allow the comparison function to be set as a parameter.

• Modify synset_distance comparison function [https://github.com//pull/385]
• The comparison function now returns a percent so it can be easily compared to other comparison methods.
• This change allowed the codebase for the closest match adapter and the closest meaning adapter to be combined.
• Move sentiment matching to it's own function [https://github.com//pull/386]
• Don't override get method on ApproximateSentenceMatchAdapter [https://github.com//pull/387]
• Add check for if logic adapter string parameters need to be imported [https://github.com//pull/395]

Additional updates

Bug fixes

• Add check to prevent nltk_data download loop (via @vkosuri) [https://github.com//pull/369]
• Correct JSON formatting in corpora [https://github.com//pull/394]
• Datetime parser updates [https://github.com//pull/397]

Features

• Support extra_data passed to django API view [https://github.com//pull/374]
• Split up response generation and learning parts of get_response so that the learning process can be externally controlled if needed [https://github.com//pull/365]
• Add check for logic adapter agreement [https://github.com//pull/375]
• Add Mailgun input adapter [https://github.com//pull/376]
• Remove twitter storage adapter in favor of trainer [https://github.com//pull/378]
• Add ability to set per-adapter parameters [https://github.com//pull/389]
• Add math words to Italian corpus (via @davideboschetto) [https://github.com//pull/393]

Maintenance

• Update logic methods and documentation [https://github.com//pull/364]

Corpus data

• Add German language (via @ThomasKoscheck) [https://github.com//pull/370]