Deepy is our basic open-source Multiskill AI Assistant. It is a foundation for our eventual open-sourcing of DeepPavlov Dream. With it, you can build simple and sophisticated AI Assistants today. It brings several goal-oriented and chit-chat skills, response selector, a number of annotators, all packed into several distributions that represent simple end-to-end AI assistants. It is our second step after making DeepPavlov Dream AI Assistant available to the public, and we encourage you to talk with it and share your feedback with us.
At the beginning of September 2020, we have launched a public demo of our AI Assistant called Dream, built on top of our original Alexa Prize 2019 DREAM Socialbot. In the beginning of December 2020, we had to disable this public demo because we are participating in Alexa Prize 2020 challenge and challenge rules prohibit running both publicly.
We replaced our original Dream AI Assistant Demo with Deepy 3000, and we will continue open-sourcing parts of Dream as additions to Deepy 3000.
DEEPY IN PICTURES
AIML-based Chit-Chat Skill
INTRODUCING DEEPY DISTRIBUTIONS
Distribution is a set of config files (Docker YAML files and pipeline_conf.json used by DeepPavlov Agent) and accompanying components and services that together form a multiskill AI Assistant built using the DeepPavlov Conversational AI Technology Stack.
Discover different distributions to grow your understanding of how Multiskill AI Assistants are made:
To build your own distribution, you can add custom components to the "thirdparty" directory in the given folder (/annotators for annotators, /skills for skills/, etc.), add them to the docker-compose.yml and agent/pipeline_conf.json, build it with docker-compose up --build.
It's easy, and you can easily build your own Multiskill AI Assistant using our Deepy.
As promised, we will continue releasing more and more sophisticated versions of the Dream AI Assistant Demo within Deepy. There are many great things ahead, including but not limited to deeper integration with Intent Catcher, improvements to Go-Bot-based goal-oriented skills, KBQA-based skills, and more.