Open Intelligent UI
Jev chooses text, tables or interactive visuals before a chat model creates the answer.
Added to Jevfast
How it uses Jev
Before the chat model answers a new user turn, middleware sends the latest conversation context to Jev with one Choice question about presentation. Jev selects text, a table, a chart, a diagram, a map, a calculator, or another listed presentation. The router maps text and tables to their designated renderers, maps other choices to Open Generative UI, and restricts the available rendering tools accordingly. The chat model still supplies the answer and UI content.
What Jev decides
Conversation presentation routing (illustrative request)Example answers · not a recorded Jev response · Source ↗
Question 1 · visualization
YOUR APP
INSTRUCTION
Select the best presentation for the latest request, respecting requested formats and choosing text when a visual adds no value or required data is missing.
STATE
The latest four nonempty human and assistant messages, with each message text limited to 6000 characters. Jev is asked to select a presentation for the latest user request in that context.
JEV · CHOICE
- text
- table
- bar_chart
- line_chart
- scatter_plot
- distribution
- part_to_whole
- flowchart
- static_diagram
- interactive_diagram
- animated_route
- map
- calculator
App workflow
- Collect turn context
Find the latest human turn and build a bounded context from the last four nonempty human and assistant messages.
- Choose presentation
Send that context with the single visualization Choice question to Jev.
- Constrain rendering
Map the selected format to a renderer and remove rendering tools that do not match it.
- Generate the answer
Pass the presentation decision and format-specific instructions to the chat model, which supplies the answer and UI content.
Why it is interesting
The project separates presentation selection from answer generation. Jev makes a bounded choice from criteria that describe when each format helps, while middleware constrains the rendering tools and supplies format-specific instructions to the chat model.