Sugar Activity on Demand

Sugar Activity on Demand addresses a key barrier in Sugar Labs: the difficulty learners face in creating their own activities due to the need for programming knowledge. This limits access to constructionist learning, where students learn best by building meaningful projects. This project introduces an AI-assisted system that allows learners to describe an activity in natural language and receive a fully functional, installable Sugar activity. The system combines a model-agnostic LLM pipeline, a retrieval-augmented knowledge base built from real Sugar Activity source code, and an AST-based validator to ensure that generated Python code is both correct and safe. Activities are generated with proper GTK3 interfaces, toolbar integration, and Journal support, ensuring they are fully compatible with the Sugar ecosystem. Beyond generation, the project emphasizes learning through a Reflective Studio, where learners explore, understand, and modify the generated code. Using scaffolded challenges, inline annotations, and reflection prompts, the system follows the constructionist Use → Modify → Create progression, enabling learners to gradually take ownership of their creations rather than passively consuming generated output. Deliverables AI-powered activity generation pipeline (prompt → plan → code → installable bundle) RAG-based knowledge system using real Sugar Activity examples AST-based security validation and safe code generation framework Interactive Reflective Studio with guided editing, annotations, and challenges Integration into Sugar’s Home View with a seamless “Create with AI” entry point Support for iterative refinement, version history, and sharing via the Sugar Journal This project aligns with Sugar Labs’ constructionist philosophy by transforming AI from a passive generator into an active learning partner, empowering learners not just to use software, but to understand and create it.