Chatbot Development

Building Bots with Gupshup

Using Gupshup bot-building tools, IDE Bot Builder and Flow Bot Builder, developers and non-developers can build, test, and deploy sophisticated chatbots.

An assortment of bot-building tools, a full-fledged chatbot development framework, a collection of readymade chatbot templates, and a pre-trained natural language processing (NLP) engine—just what is Gupshup, exactly?

The answer is that it’s all of those things. Although Gupshup is more than a framework, its suite of tools does include a bot-building environment with all the trimmings. However, that aspect of Gupshup is just for developers. Other Gupshup tools enable non-developers to prototype bots and to supply them with basic functionality—even before any actual coding takes place.

All of this makes Gupshup pretty unique and, by all outward appearances, a flexible option for building a chatbot. To get the full story, let’s take a closer look.

What is Gupshup?

Gupshup does a lot, but from a chatbot development standpoint, it consists of two primary components:

• A true bot-building framework, known as IDE Bot Builder (or just Bot Builder)
• Flow Bot Builder, a GUI-based bot-building application geared toward non-developers

A third component is a template-based chatbot-building tool, but that’s aimed at small business owners who are doing everything themselves.

About IDE Bot Builder

The Integrated Developer Environment (IDE) Bot Builder is the flagship Gupshup chatbot product. It consists of a code editor, a publishing tool, and diagnostic features, and you can read all about them in the IDE Bot Builder documentation.

To remove a lot of the hassle of bot development, the framework comes with a JavaScript async library and a terminal for using git commands or installing npm packages. Since Bot Builder runs on AWS Lambda, it has built-in hosting.

About Flow Bot Builder

On the other hand, Flow Bot Builder is for people who aren’t developers but still want to build a bot. The target user might be a marketing professional who has a keen understanding of the business case behind a potential chatbot—a better understanding, perhaps, than the developers who actually know how to build it. This person can use Flow Bot Builder, which relies on a straightforward graphical user interface (GUI), to prototype a chatbot.

Developers can then take the Flow-built bot and build out JavaScript modules and functions. They can essentially start with what a non-coder gives them, access the underlying code, and beef up the bot. In this way, Gupshup tools transcend business units and skill sets.

About Template Bot Builder

To help small businesses take advantage of automated communication, Gupshup offers a Template Bot Builder that lets non-developers with modest needs easily create a chatbot. You choose a template that aligns with your business type and then fill in your information. There are also a few customization options, all of which are available in a simple interface that most people can use.

So how many templates are there? At the time of this writing, Gupshup only has templates for restaurants and spa/massage-type businesses. Functionality is limited to appointment-booking, order-taking, and other straightforward tasks.

Gupshup chatbot examples

Although you can use Gupshup to build almost any kind of bot, the following examples of bots for business offer a pretty good idea of what’s possible via the framework.

Financial services

At Indian stock brokerage Motilal Oswal, countless requests and transactions were burdening internal operations—and the brokerage’s customers. To expedite the process of viewing and interacting with data, Motilal Oswal used Gupshup to build multiple chatbots serving sub-brokers, branch managers, and dealers.

More specifically, the company’s bots help users check ledgers and statements, reducing the amount of time it takes to navigate its website and open new accounts and, therefore, cutting down onboarding time. Gupshup was flexible enough to connect with multiple APIs to help different types of customers.

Customer service

Satellite TV provider Tata Sky also used Gupshup to build a bot that helps customers. Instead of requiring that customers call in or email requests (and burdening internal staff with the need to respond), Tata Sky created a bot that helps customers perform common actions, such as checking account balances, pulling an online statement, and recharging their accounts.

Consumer engagement

When Zee TV, a paid television channel in India, wanted to increase audience engagement for its viewers’ choice event, developers used Gupshup to build a bot that communicated via SMS. The bot greeted users with an SMS message displaying categories and award nominees. From there, it was easy for users to submit their choices.

Health and fitness

Another company that built a consumer-oriented bot with Gupshup is FitCircle, which offers a mobile app for tracking fitness. FitCircle app users interact with the built-in chatbot to define fitness goals and to provide personal health-related information.

Why use Gupshup? The pros and cons

As already noted, the potential for cross-team collaboration is one of the unique benefits of Gupshup. Virtually anybody can build a bot with Flow Bot Builder and send it to developers for further refinement via the IDE Bot Builder.

Another benefit is full integration with the larger Gupshup ecosystem of tools. Chatbots, after all, are just one focus of Gupshup. The company also offers a Gupshup CRM platform for keeping track of customer conversations and the Single API that allows you to connect with over 30 different conversation channels.

With all of these tools working together, you can build your bot (IDE Bot Builder), connect it to the channels of your choice (Messaging API), and then track conversations and manage leads (CRM). Gupshup handles everything. Your bot is also hosted on AWS Lambda, so you always benefit from the highest quality, scalable architecture.

As for cons, the “full” experience of the Gupshup ecosystem might not be full enough for your needs. A framework like Amazon Lex, for example, integrates with a larger array of cloud computing services from Amazon Web Services (AWS). The same goes for bots built with the Microsoft Bot Framework, which gives you total integration with Microsoft Azure Cognitive Services, among other options.

Another potential issue is natural language processing. Gupshup offers what it calls “NLP on the fly,” an NLP engine that you don’t have to train. It’s an API that seeks out the closest match between a user’s input and up to 200 variations that you specify. Some observers have found “NLP on the fly” lacking, noting that it isn’t built upon an actual model and doesn’t scale particularly well.

If that’s a problem for you, Gupshup bot-building tools integrate with Wit.ai, which is the Facebook chatbot framework. Wit.ai is known for having very capable NLP, so you might do well to use it instead of the Gupshup offering.

Key features and integrations

In addition to Flow-to-IDE Bot Builder iteration and the ability to utilize others’ NLP tools, Gupshup offers something pretty unique—WhatsApp Business integration. WhatsApp, of course, is famously cagey about chatbots and only allows organizations to deploy them in a limited fashion. That, and you’ve got to get approved by WhatsApp before you can onboard your bot.

Gupshup might be able to accelerate all of that for you. For starters, the Gupshup Messaging API plays nicely with WhatsApp. That’s a rarity. What’s more, you can actually apply for access to WhatsApp Business via Gupshup. Ostensibly, this tells WhatsApp that Gupshup is your solution provider and helps you in the application process.

Assuming all goes well, you can use Gupshup tools to build your bot, test it, and deploy it to WhatsApp fairly quickly. For anyone who wants to get their bot on the world’s most widely used messaging application, that’s a pretty big deal. Additional integrations include:

• Facebook Messenger
• Slack
• Google Assistant
• Google Hangouts
• Kik
• Telegram
• Cisco Webex Teams
• Twitter
• Viber

Just remember that you should always deploy your bot to a dedicated domain in addition to messaging channels. That helps users find your bot even when they’re not logged in to Slack, Viber, or wherever they might normally find it. A .BOT domain name is a good choice because it’s simple to remember and easily connects users to your bot through the domain. And when you build a bot with Gupshup, you are able to obtain a .BOT domain name without going through a separate validation process. Gupshup has a direct link with .BOT to validate its customers and quickly get them through the domain name registration process.

Gupshup pricing model

Like most chatbot frameworks, Gupshup charges according to how much your bots are used. At the time of this writing, there are three pricing tiers:

• Free. This tier covers up to 100,000 API calls per month.
• Paid. After your first 100,000 API calls, you pay $1 for each additional 1,000 API calls.
• Custom. Organizations with large volumes of API calls are asked to get in touch with Gupshup to discuss a custom price.

In summary, you can start using Gupshup without paying anything. When your bot starts getting significant and consistent usage, the paid tier kicks in.

Supported programming languages

Gupshup offers bot SDKs for Java and Node.js. If you want to create a bot in a different language, head over to Gupshup and associate a callback to your bot code on whichever server you’re hosting the code. So, when a user starts communicating with your bot, Gupshup sends the info to your bot via the callback URL.

It’s a bit of a workaround, but it is a way to use Gupshup and code your bot as you see fit.

How machine learning works in Gupshup

As noted earlier, Gupshup “NLP on the fly” might not compare well to the more sophisticated models offered by Dialogflow, Amazon Lex, Wit.ai, and others. Nevertheless, it is an artificial intelligence engine that’s likely to be sufficient for certain kinds of bots. Here’s how it works:

1. You create a JSON with different variations of user intents and the adjacent entities.
2. The API identifies the closest match between user input and the intent variations you’ve defined.
3. The API then matches that input/intent to the entities you defined in the JSON.

With this NLP, you’re not really training a model. You’re just putting in some variations with the assumption that your user will input something that matches what you’ve provided. Currently, Gupshup only supports 200 variations. If you need more—and many chatbots do—consider taking advantage of Gupshup Wit.ai integration.

How to get started with Gupshup

All you have to do is sign up and start building! Since Gupshup is free for your first 100,000 API calls, you can start exploring the IDE Bot Builder with virtually zero risk.

Be sure to review the Gupshup documentation before you begin so that you know what’s possible—and what isn’t—with the IDE Bot Builder. And consider exploring other Gupshup tools, like the CRM and Messaging API, before you get started. They could help you unlock new opportunities for your chatbot, and that’s something you’ll want to know about sooner rather than later.