3 years! - Rasa Community Update

Today, Rasa is 3 years old! Looking back to 2016, so much has happened, and 2019 was a fantastic year for the company and our community.  A few highlights from the past 12 months:

  • We’ve seen our community growth explode to 6 times as many downloads and 4 times the number of forum members - developers build contextual assistants across all use cases and industries
  • We hosted over 250 developers from all over the world in San Francisco for our Rasa Developer Summit
  • We released Rasa Open Source 1.0 with tons of new features coming from our research team and launched an entirely new product, Rasa X
  • We launched Rasa Masterclass which helps developers get started with building contextual AI assistants
  • Our fast growing number of contributors help educate developers all over the world and make our code base better every day

The Rasa community is growing fast

6 x downloads, 4 x forum members (in 12 months)

We are thrilled that developers across the world want to use our open source framework and be part of the Rasa community, helping each other on their journey to build contextual assistants. In 2019, our downloads exploded to over 2 million (up from 300,000) and a huge number of developers joined our community forum (more than 8,000 now, up from 2,000).

Adobe built a voice-enabled assistant using Rasa

So what is our community building? A super wide range of contextual assistants that spans from fun projects to commercial assistants in customer support, sales & marketing, and internal process automation - across all industries. Here are just a few examples:

  • Hey Sensei: Adobe's voice-enabled assistant allows users to sift through hundreds of millions of Adobe Stock assets using an intuitive conversational interface. Users can start with a broad query and use spoken commands to explore options and refine the request, to find just the right image.
  • Djingo: Orange’s Djingo let’s human agents focus on handling the tough problems instead of performing routine and repetitive troubleshooting tasks. Customers are able to resolve issues 24/7, via a personalized conversational interface.
  • Joke Bot: Greg Stevens built Jokebot, a demo assistant that can respond to users with corny jokes, geek jokes, or funny quotes, all sourced from open APIs. The assistant's source code is available on GitHub to use as a starting point for implementing custom actions.

You can see more assistants in our brand new community showcase, and if you have an assistant you’d like to see featured, send us a note.

Rasa Developer Summit in San Francisco was sold out!

For the first time, we hosted the Rasa Developer Summit in San Francisco - where over 250 developers from all over the world shared their experiences in building contextual assistants - big thanks to our speakers from companies such as Adobe, Lemonade,  and United Healthcare. Missed the summit? Presentation slides and recordings can be found here. Expect to see more Rasa hosted events in 2020!

Helping developers build contextual assistants that go beyond basic FAQs with new research, better tooling and more education

Every developer who has tried to build a chatbot or voice assistant would probably agree that it’s easy to build a demo, but hard to actually ship and improve an assistant in production. In fact, that’s why we founded Rasa. This year, we released Rasa Open Source 1.0 with a ton of research and shipped an entirely new product, Rasa X.

In 2019, we releases Rasa Open Source 1.0 and Rasa X

Earlier in 2019, we released Rasa Open Source 1.0. In this release, we combined the NLU and Core libraries into a single package, together with a slick new CLI (remember the old one?) We’ve now already reached version 1.5, with a whole host of new features like retrieval actions, integrating knowledge graphs, our new TED dialogue model, better default hyperparameters for the supervised embeddings pipeline, better conversational embeddings from ConveRT. We’ve also squashed countless bugs, and sped up performance. And there is so much good stuff coming soon! Tracker sessions, a brand new NLU model, letting external events trigger conversations, and detecting conflicts in training stories.

To build great conversational AI, you need to improve your assistant based on real conversations. You can’t anticipate -- but need to know -- how users will talk to your assistant, and there is no substitute for learning from their conversations as soon as possible. That’s why we released an entirely new product, Rasa X, to give every developer the ability to review the conversations people have with your assistant and to turn those into new training data. You load your Rasa assistant in via Integrated Version Control, review conversations with testers and real users, and then continually make incremental improvements based on what you learn.

But research and great tools are not enough. Understanding how those tools work under the hood is very important as well. That’s why we launched the Rasa Masterclass - a series of videos helping developers to build contextual AI assistants. With each episode, we aim to help developers master everything from the fundamentals of NLP, dialogue management policies, to practical implementation and deployment in production. We have already published 10 episodes and we are very stoked to see that the Masterclass has been a big hit with the community - over 10,000 views on YouTube and counting.

Contributors across the world help educate developers and grow the community

Rasa Meetup in Hyderabad, India - Just one example of many community meetups across the world

We would not have been able to grow as fast without our contributors - so a BIG thanks to them! Over 370 people have contributed so far - from helping other Rasa developers on our community forum, over organizing local meetups, to building complex features. It is super hard to pick examples because we had some many people contributing this year, but we wanted to highlight a few:

  • From Flask to Sanic, Rasa contributor Jan Šafařík helped us to replace the core framework of the Rasa SDK. Jan showcased incredible communication throughout the process which meant we could iterate much faster on this huge task. We appreciate the quality of Jan’s work and the time he set aside to support us.
  • Community grows to the next level in India, thanks to Rasa Hero Yogesh Kothir. Yogesh not only prepared, organized and hosted 2 community workshops on how to get started with Rasa in Hyderabad and a fireside chat in Bangalore, but he also supported others hosting their workshops in Mumbai, Mathura and Delhi. In October, our own Daksh Varshneya took part in a Fireside Chat hosted by Yogesh, and we were fortunate enough to connect in person with this great Rasa evangelist.
  • A jack of all trades, we'd like to highlight Rasa Hero Julian Gerhard, who has been supporting the community on multiple levels this year. From answering questions on the forum, to providing extensive feedback on upcoming features and Rasa X, and submitting code contributions. We are particularly thankful for his efforts alongside Jacob Zelko to get Rasa working on RPi 4! We wonder if there is anything that Julian can’t do, and we appreciate the he decided to do them with and for our community!

We also launched the Rasa Contributor Program, made it much easier to get started with contributions and created a dedicated Slack channel for contributors, heroes and superheroes to collaborate and connect in real-time. Stay tuned for more!

What’s next?

The short answer is: more of all the above! We will continue to invest in community, our open source framework, better products for developers, and more developer education. Conversational AI, especially level 3 and beyond, is not solved at all - it’s still early and it needs a combined effort to build better assistants. We’re excited to have so many brilliant folks join along for the ride - thank you so much for your support. Onwards and upwards!