# BoostAI

This data is stored at question level, topic level, and session level

a sequenced list of 10 questions with a plain-English justification for each choice, covering difficulty gradient, question type, and topic rationale. The agent must reason about misconception patterns

and must be able to re-route mid-session when a pupil answers unexpectedly well or poorly

attempt records, solve mode (independent / step-by-step / Solve Together / handwritten), time on task, score per question, topic and subtopic tags, class ave. score for that topic

Output must be structured JSON with question ID, topic, difficulty level, and a plain-English justification for every item in the plan