FSM and Conversation handling
The library also supports the FSM mechanism, which is a mechanism for progressive processing of user input with incorrect input handling.
Let's imagine a situation where you need to collect a user survey, you can ask for all the data of a person at one step, but with incorrect input of one of the parameters, it will be difficult both for the user and for us, and each step may have a difference depending on certain input data.
Now let's imagine step-by-step input of data, where the bot enters dialogue mode with the user.
Green arrows indicate the process of transitioning through steps without errors, blue arrows mean saving the current state and waiting for re-input (for example, if the user indicated that he is -100 years old, it should ask for age again), and red ones show exit from the entire process due to any command or any other meaning cancellation.
Such a mechanism can be implemented in the library through a simple class that implements a certain interface and marked with a specific annotation @InputChain.
@InputChain
object ConversationChain {
object Name : BaseStatefulLink() {
override val breakCondition = BreakCondition { _, update, _ -> update.text.isBlank() }
override suspend fun breakAction(user: User, update: ProcessedUpdate, bot: TelegramBot) {
message {
"Please say your name, because that's what well-mannered people do :)"
}.send(user, bot)
}
override suspend fun action(user: User, update: ProcessedUpdate, bot: TelegramBot): String {
message { "Oh, ${update.text}, hey there" }.send(user, bot)
message { "How old are you?" }.send(user, bot)
return update.text
}
}
object Age : BaseStatefulLink() {
override val breakCondition = BreakCondition { _, update, _ -> update.text.toIntOrNull() == null }
override suspend fun breakAction(user: User, update: ProcessedUpdate, bot: TelegramBot) {
message {
"Perhaps it's not nice to ask your age, but maybe you can tell me anyway."
}.send(user, bot)
}
override suspend fun action(user: User, update: ProcessedUpdate, bot: TelegramBot): String {
message { "Pleased to meet you!" }.send(user, bot)
return update.text
}
}
object Final : ChainLink() {
override suspend fun action(user: User, update: ProcessedUpdate, bot: TelegramBot) {
val state = user.getAllState(ConversationChain)
message {
"I'm not good at remembering, but I remembered you! " +
"You're ${state.Name} and you're ${state.Age} years old."
}.send(user, bot)
}
}
}And after we described the mechanism to start the processing, we just need to call the method and specify the initial step, then the library itself will follow the sequence.
bot.inputListener.setChain(user, Conversation.Name)All links have the same foundation and implementing Link<T> interface, which have such properties:
Key Properties:
- afterAction: An optional action to execute after the main action.
- beforeAction: An optional action to execute before the main action.
- breakCondition: A condition that, if met, will trigger a break in the chain.
- chainingStrategy: Defines how the next link is determined.
- retryAfterBreak: Indicates whether to retry the action after a break condition is met.
Key Functions:
- action: This is an abstract function that must be implemented to define the primary behavior of the link.
- breakAction: An optional function that can be overridden to define behavior when a break condition is met.
There are two types of links they are differentiated by state, stateless and stateful:
Stateless links are represented by the abstract class ChainLink. This class serves as the foundation for creating links that do not maintain any state information between user interactions.
The InputChain mechanism, particularly when employing StatefulLink and its various implementations, offers a sophisticated approach to managing conversational states within applications, such as chatbots. This system automatically stores the result of the action function associated with each state, linking it directly to the user involved in the interaction (or other selected key).
- By default, the outcome of the
actionfunction executed within aStatefulLinkis automatically stored. This storage is linked to the user, ensuring personalized interactions based on past exchanges.
- Developers have the flexibility to override the base implementation to specify custom keys for state association. This could range from identifiers unique to a chat session to any other relevant attribute that suits the application's requirements.
- The foundational implementation,
BaseStatefulLink, categorizes data asStringtypes, utilizingUser.idas the primary key for state association. This approach streamlines data management and retrieval processes.
- Should all
Linkobjects within the InputChain utilize identical keys, the system generates functions that facilitate unified access to all states. This enhancement significantly simplifies the process of retrieving and managing state information across different parts of the application.
To access all states associated with a particular chain for a given user, the following syntax can be employed:
user.getAllState(MyChain).LinkNameThis command retrieves the data linked to the specified LinkName within MyChain, providing a comprehensive overview of the user's interaction history.
Alternatively, for more granular control, states can be accessed directly through the links themselves:
Chain.LinkName.state.get(key)Or, if querying the state within the current chain, Chain.LinkName may be omitted, simplifying the call to:
state.get(key)This methodology enables stricter data management protocols, offering rapid and convenient access to stored states. It enhances the efficiency of state retrieval and manipulation, contributing to a more seamless user experience.
Default BaseStatefulLink implementation uses ConcurrentHashMap, but for serious projects it is recommended to use other solutions :)
It is possible to use the proposed tools with different variations to create quite flexible interaction, if you have any questions contact us in chat, we will be glad to help :)