Overview
A full-stack platform for collecting and managing question-and-answer data behind a sign-in wall. Here I can answer questions in a few formats: 1-5 scale, yes/no, or free text. They are organized by topic, and every answer is encrypted before it's stored. The eventual goal of the collected data is to use it for training machine learning models or to use as LLM context for more educated guesses.
Architecture & Implementation
The interface fetches unanswered questions from the backend and supports three answer types: a 1-5 scale, yes/no, and free text - with controls to submit, skip, or rotate to a different question, and a live feed of the most recent answers that refreshes as soon as one is saved and supports deleting an answer.
Adding a new question lets the author pick a topic and configure its answer type, and a statistics panel surfaces totals - how many questions exist, how many are answered, and the scope of topics covered - alongside the timestamp of the most recent activity.
Every annotation endpoint sits behind a signed-in session checked on the server, and answers are encrypted before being written to storage rather than kept as plain text. A dark mode toggle persists the visitor's preferred theme in localStorage.
contact@ericjansson.com
