period-tracker | web app that tracks period | Dataset library

Running a bot offline without being dependent on any cloud based service is totally feasible. The app would need to establish the following.

For brevity I am using Online Bot to refer to a bot app that is based on an online service. Offline Bot for a bot app that is based on an offline or on-device bot development platform.

• Knowledge-base (containing Intents, Entities and Contexts) has to be saved within the app. Entities contribute (though in kilobytes maybe) to your app size. For example if your bot is to recognize a country name, you would need to store all the country names within the bot's knowledge-base.
• Natural Language domain-specific information like dictionaries, thesaurus and additional word information. These may not be passed on to your deployment/published app after a model is generated.
• Language Model generated by the app would reside within the app package. This is going to be the main role player in your app size.

Low bandwidth data connectivity shouldn't be much of a problem as the overhead of communicating with services like the following is not that high:

• DialogFlow
• LUIS
• WIT.AI
• IBM Watson

The above services enable APIs directly via small JSON data exchanges. You can check how much overhead is too much of an overhead for you by looking at their APIs and returned data.

• Online bot would enjoy remote updates to their knowledge-base
• Your online bot data is not secure as anything your users say/type goes to third-party services for response generation.
• Response speed of your bot depends on the performance of the online service you use. Typically shouldn't be a problem on normal networks.
• Your app is useless when not connected to the internet.

If you are designing a completely offline bot you would need to rely on frameworks or platforms that are designed to fulfill the aforementioned features.

• OSCOVA - On-Device bot development platform that I personally have used to create an Android Bot for general customer queries in Xamarin.
• SIML - An XML based knowledge-base authoring language with Machine Learning features.
• BotSharp - Bot development platform by SciSharp
• Xatkit - Java-based chatbot creation platform.

• Your offline bot app would need a new package update (in App Store or Play Store) whenever you wish to add more to its knowledge-base (KB). Maybe you could send a KB directly to the app when connected online.
• Your offline bot size is going to be a bit larger in comparison to a bot that uses online services.
• Your bot is always-on as it doesn't require internet to perform tasks or respond to its users.
• Conversation with your bot is extremely secure as the data is not shared to third-party service like in the case of online bots.