covonaut v1.1.4 enhances AI agent framework with parallel processing, multi-modal support, and file recovery

covonaut, a Go-based agent framework for production use, has released version 1.1.4, introducing a rolling execution pool for parallel processing, multi-modal content handling, and file recovery features. The update improves scheduling by allowing calls to start in order and fill available slots immediately, reducing delays from slow tools. It also supports serial execution for non-parallelizable tasks, ensuring model return order matches execution sequence. The framework now processes files of various types—text, images, PDFs, Jupyter Notebooks, and videos—automatically routing them through appropriate channels. Deleted files are moved to a system-specific recycle bin, preventing permanent loss. Isolation features have been strengthened, separating shell processes and chat interfaces. The update also includes a persistent terminal tool family, retaining session data across calls. Upgrades require only a module version change, with no breaking interface changes, though users must adjust scheduling modes if they prefer conservative execution policies.