demand-driven

Kotlin Multiplatform ViewModels and reactive state management

Reference implementation of Application Management Framework for demand-driven orchestration based on Kubernetes and ROS 2

Large Result Offloading for MCP servers. Writes oversized tool responses to JSONL files, returning compact descriptors with extraction recipes.

Large Result Offloading (LRO) -- a token-aware strategy for managing large tool outputs in LLM context windows. Includes empirical evaluation and formal specification.