Building a VMM from scratch, #0: Why BoxLite is replacing libkrun (blog.boxlite.ai)

BoxLite, a platform designed for running AI agents in isolated micro-virtual machines (micro-VMs), is transitioning from using libkrun—a dynamic library for managing VM processes—to a custom-built Virtual Machine Monitor (VMM). This shift reflects BoxLite's evolving needs, as libkrun's limitations hinder its ability to support advanced features essential for AI applications, such as hot-plug mounts, live resizing, and memory snapshots. By developing its own VMM, BoxLite aims to enhance performance and reliability, aligning more closely with its specific use case of running lightweight, efficient AI workloads. This move is significant for the AI/ML community as it highlights the importance of tailored infrastructure in effectively deploying AI agents. The new VMM will provide greater control over virtual environments, enabling BoxLite to support advanced features that were previously unachievable with libkrun. With phases outlined for development, including initial boot processes and device management, the initiative is positioned as a collaborative, open-source project. Developers and researchers can contribute to the ongoing evolution of BoxLite, paving the way for improved AI agent deployment in various computing environments, from personal laptops to cloud infrastructure.