Product Updates → July-October 2019

We have some huge product updates - Release of Slang's WebSDK, Improved NLP Engine, Improved Scalability, Multiple Geolocations and Improved Discovery

It’s been quite a while (sheesh, the last one was in June, we ought to this more frequently!) since we shared the previous product update with you. We have been working hard for the last five months to add some cool new features to Slang. Hope you have more fun using them than we did building them.

Slang’s Web SDK

Now add Voice to your website/PWA using Slang’s Web SDK! ( We have added Slang in our website. GIF shows: User asking: “What is 42”. Slang replies back with an answer and also takes you to the page linked to ‘42’)

Woot Woot! Slang now has Web SDK. A lot of clients we spoke to wanted support for web. So we went ahead and built Slang for Web. Earlier your users could talk to your apps, but now they can talk to your websites as well. We heard PWAs are the new sexy, so we went ahead and made sure Slang works with them. You get all the features and components that make Slang awesome for apps now on the web. We made sure it works on iOS browsers too :-). We are currently accepting beta customers for our WebSDK. Don’t forget to reach out to us if you want to see how Slang will look on your website.

Improved NLP engine

When we started Slang, we originally used a third party NLP engine. While we had an abstraction layer around it to add our own extensions, it allowed us to get going quickly. But soon we ran into limitations (eg composite entities, lists, performance, etc) and realised we could not reply on a generic NLP engine but something that is more suited for the kind of capabilities our customers would need and also for us to provide the best possible VAX platform.

Towards this, we started the journey of building our own NLP engine, based on top of one of the best NLP engine but allowing us full fidelity and control.

It took us almost 6 months but we finally are ready for making it our default NLP engine.

We call it Vyasa internally (any resemblance to the engineer Ved who built it is purely coincidental. Or at least that’s what he tells us)

Improved Scalability

Diagrammatic Representation of our Infrastructure before and after the update

We realised early on that our old infra won’t be able to match the scale of our clients. We took a step back and decided to revamp our infra to make sure we match our customer’s scale and deliver them the best performance possible.

In the past five months, our engineering team was hard at work and revamped our backend to make it more robust, reduce latency and increase the number of requests we could handle by 4x.

Multiple geolocations

We now have our servers at multiple geolocations — USA and India.

Alas, after a lot of discussions and fights…err..debates, we decided to set up a server in the land of the dreams. We now have servers in the USA in addition to India. This addition will provide our customers in the west(that’s right, we will be making those announcements soon, follow us on Twitter to keep an eye out) the lowest latency and the best experience possible when their users try out Slang.

Improved Discovery

New Onboarding Screen Built in the framework

After the release of our first customer in prod ‘Madurai Kavalan’ (Yay!), we looked at the usage data and realised that the discovery of what users can do with voice was still a challenge for them. To solve this issue, we added an onboarding experience for voice to Slang’s framework, which helps the user to get exposed to voice and removes the extra amount of work for the developers.

More updates coming

Stay tuned as we continue working with more developers and also dog fooding Slang internally constantly. If you are interested in seeing some specific features do let us know at or leave a comment here.