Low-code platforms use model-driven development to manage code at a high abstraction level, with advanced editing features like visual languages, code templates, and AI assistance. These platforms rely on structured documents represented in dynamic data structures, enabling richer interactions compared to traditional text-based programming. For example, OutSystems uses a desktop IDE to edit applications as well-formed directed graphs, and plans to shift to a web-based, AI-assisted environment offering collaborative editing and offline capabilities. This transition introduces the need for a new design paradigm to manage concurrent and speculative edits.

Our proposed solution is a transactional framework with three types of transactions: write transactions for regular edits, read-only transactions for background tasks, and discardable transactions for speculative edits. We base our method on versioned boxes, as used in software transactional memory, to minimize memory use and efficiently manage model modifications. Our hybrid approach reduces memory footprint by storing modified fields externally, resulting in up to a 30% decrease in memory usage compared to naïve versioned boxes.

We apply this framework to MORPHEUS, a cloud-based IDE, aiming for efficient memory usage to lower operational costs by maximizing the number of models per server instance. This work is the first step towards a low-code model editor with offline capabilities, supporting collaborative editing and Undo Redo functionality.