Dynamic ChatUIs

Chat Interfaces are now dynamically rendered based on the logic of your pipeline. Use this e.g. to have ChatGPT ask questions back to you before diving into the answer.

Define logic in the backend, don't worry about the UI

When building RAG applications, we realized that quite often you want to have some dynamic processing of the user's question. For instance, you might want to filter for a certain set of data before searching within said database.

The problem though is that these interactive, modular components can be extremely complicated to design, and you want your user to have a smooth conversational experience.

Today, we're excited to introduce dynamic UI elements, which are generated on-the-fly from the backend as needed. For instance, in the following pipeline, you can see that a user question was routed through the "RAG with Selection" swimlane:

The selection widget is configured to help the user select from the insurance they want to analyze:

In this case, this is what the user can see as the selection widget is triggered:

In combination with our state management, the conversation can remember this selection from the user for as long as wanted.

Further improvements

We now also support:

internationalization: pick which language to display in the chat interfaces. As of today, we support English, German, Italian and Dutch. If you're missing a language, please contact us at support@kern.ai
branding: choose from different color sets or choose your company's logo to be displayed in the interface
application switch: quickly jump inbetween different projects, e.g. between a vanilla ChatGPT and your specialized RAG application
user roles: we differentiate now between engineers and users; an engineer can configure the backend pipeline, a user can consume the application. An engineer can still also use the full pipeline via API.

Go to newsletter