The above diagram illustrates the general concepts of how a Rocket.Chat server communicates with a chatbot framework or platform. These concepts are applicable to external bots.
Bots architecture consists of two main parts: Rocket.Chat host and bot host. These two parts are connected with each other via Rocket Chat SDK methods. Bot host runs externally to the Rocket.Chat host and consists of the bot infrastructure and, typically, external services. The bot infrastructure, in turn, consists of SDK, adapter, and framework.
A bot, being subscribed to room’s messages (all or specific ones) handles these messages and responds back to the user in an interactive manner. The conversational model can be designed and managed by tooling and frameworks on the bot host.
Today, many bots are created using bot frameworks or platforms based on popular programming languages, such as NodeJS and Python.
How Bots Architecture works
Bot creators design the bot’s logic with the help of their favorite framework. This logic defines behavioral interaction with users, giving the bot context-awareness. Context-awareness helps the bot to identify user’s intents, understand what the user is looking for so that they can simulate more personalized human-like conversation.
With the ability to listen to and respond to messages, the bot can be integrated into the Rocket.Chat host, so that it can engage with users in a conversational manner. To perform this connection, bot creators use adapters that can translate the logic defined via the framework’s powers to the format that the Rocket.Chat host understands. Adapters use Rocket.Chat SDK methods to communicate with the Rocket.Chat host.
Users and Bots
Bots in Rocket.Chat require user accounts with the
bot role. Other than having
a specific set of permissions, bot accounts are treated like regular user accounts
within the Rocket.Chat host. However, to prevent possible user confusion, messages
from online bots are marked with a default “Bot” label.
Admins (and only admins) can create the bot account and configure role permissions. This includes setting the credentials a bot uses to connect as well as which user fields the bots have access to for privacy and security. Rocket.Chat team works on improving management views for bots and enhancing the future potential of user-activated bots.
A bot host is a machine where bots run. It can be a bare-metal server, a network cluster, a virtualized environment or cloud-based.
A bot host can manage and run many different bots as well as many instances of any bot.
The bot host typically manages connections to external services as well.
Bot platforms and frameworks
The hosting for bots depends on the platform. For example, Hubot can be launched on any Node.js environment, but you have to set it up yourself (using Heroku or Glitch services). Other bot platforms like Botkit provide their own services for hosting and provisioning bots. They just need a configuration to connect to your Rocket.Chat instance.
Every user in Rocket.Chat has a collection for their messages that can be emitted by the server’s streamer. Bots subscribe to the collection of messages for their own user account, creating a stream that is updated every time a message is sent either directly to them or any room they are joined in.
The SDK is a low-level software module that offers an interface for external consumers to subscribe to message streams, send messages, and call methods on the Rocket.Chat server via WebSockets / DDP.
Bot framework adapter are written using the SDK.
An adapter (also referred as connector or middleware) is a software component that is used to link the framework with the Rocket.Chat host.
An adapter allows bot creators to design and build bot interactions according to the framework’s rules and practices without dealing with the low-level programming details and methods of the communication with the Rocket.Chat host.
Adapters use Rocket.Chat SDK to communicate with the Rocket.Chat host.
Bot frameworks (such as Hubot, Botkit, Rasa or Botpress) provide unique approaches to create conversational interfaces and allow developers to focus on the design and management of complex bot logic.
Most of framworks can be run on your own bot host. There are also frameworks that operate as SaaS (Software as a Service), so they work in the cloud.
Most of the popular frameworks have adapters created by the Rocket.Chat’s community.
In production, hundreds or thousands of users can be carrying out conversations with a bot at the same time. The ability to scale a bot from one conversation to many concurrent ones is essential.
Bot frameworks/platforms support varying degrees of scalability. To ensure that growing number of users will not cause architectural and performance problems, you should choose your bot’s framework wisely. For example, considerng potential scalability issues even before you start implementing your bot’s logic
The value of most production bots relies on their ability to access and manipulate external data and systems (services). External services, like databases, scheduling systems, CRM, NLP can be used to provide data or conversational processing.
The framework simplifies connection of a bot with these services and ensures ease of access to them through the bot host.