Once you have logged into your Slack workspace and installed KNO using this link, you can converse with KNO either one-on-one or in a channel by tagging ‘@kno’. A more detailed description of the steps can be found here.
The KNO engine is based on new and cutting edge technologies, of which the two most important ones are Dialogflow and OpenAI.
KNO uses Dialogflow to understand the intent behind the question. Subsequently, it uses the knowledge graph in the internal database to understand the model and information required to answer the query. After picking the required model, it queries OpenAI with the required parameters of the request and gets the answer.
When we started building out KNO, we wanted to limit writing code for as many platform elements as possible and focus development resources on the conversational engine. The logical choice was to run KNO on a platform - Slack in this case. KNO uses Slack’s native authentication (Oauth2) and authorization (Slack provided keys). It is hosted as a flask app using Gunicorn deployment.
We have two upcoming priorities
In the coming versions, KNO will evolve to understand the ‘context’ of the user and the tone of the query content. It will graduate to include images and videos in not only the questions but also the answers.
KNO can be repurposed as a help widget, an in-app concierge or an enterprise search engine. There is a flexibility in what we train KNO on. We can exclude or include sources and fine tune relative weightage.