Personalize sentences to be recommended based on user's input prompt.

[santanavagner]

Description (Actual Behavior)

The recommendations provided by our /recommend endpoint come from a dataset of human-curated sentences. These human-curated sentences were, by design, phrased in a more abstract way so they could be applied to multiple prompts. However, based on outcomes from a recent user study published at CHI 2025, participants expected the recommendations to be more personalized.

Expected Behavior

Envisioning this personalization of sentences form our JSON sentences dataset, we could use the input prompt provided by the user, request the recommendations from the JSON sentences, and then personalize them considering users' input prompt.
The rephrasing could involved adding domain knowledge terms used by the users, context information, etc.

Possible Fix

An idea for enriching the /recommend endpoint is to add a /personalize endpoint that would receive as input user's prompt, recommendations provided, and retrieve the personalization to the context provided in the users' input.

Steps to Reproduce

• Access the demo.html
• Paste the following prompt: Act as a professional designer with 20 years of experience creating and testing UX interfaces and landing sites for a variety of IT applications. We are in need of more people and an increased budget to be able to keep up with clients' needs. What kind of evidence should I gather to support my demands to gain more resources?
• Hover over the vale integrity

• The recommendation presented is: "Be aware of certain additional rights and resources."
• Considering the context of an IT company and need for resources, the personalized recommendation could be: "Be aware of additional rights and resources for IT companies."

Context

Recommendation.