Avoid relying on LLM for markup in the content suggestions modal

Created on 24 June 2025, 9 days ago

Problem/Motivation

The Content Suggestions submodule currently instructs the LLM to return not just the suggested content (e.g., title), but also the full HTML markup for the modal buttons where you can select the suggestion.

This approach is sometimes not as efficient as it could be, as it relies on the model to successfully understand and deliver the correct markup so that the appropriate event listeners and styling can be added, which is not always the case - specially on "dumber" models.

Proposed resolution

Update the LLM prompt to return a simpler response instead of full HTML. Then the actual markup can be generated based on that response.

User interface changes

No changes for the user

📌 Task
Status

Active

Version

2.0

Component

AI Content Suggestions

Created by

🇵🇹Portugal bbruno Poland

Live updates comments and jobs are added and updated live.
Sign in to follow issues

Merge Requests

Comments & Activities

Production build 0.71.5 2024