Skyportal SRE – an open-source AI infrastructure engineer (github.com)

SkyPortal has introduced an open-source AI infrastructure engineer SDK, named Skyportal SRE, currently in its alpha stage. This SDK provides a Python interface for interacting with the SkyPortal API, allowing users to manage AI-related tasks such as monitoring servers and managing chat interactions with a server-side agent. Installation is straightforward using either pip or poetry, and it requires Python 3.11 or higher. The SDK features a persistent terminal client, enabling interactive operations like creating and managing agent chats, polling for command approvals, and obtaining system status updates. The significance of Skyportal SRE lies in its ability to streamline AI infrastructure management, particularly for machine learning workflows. By offering an intuitive command center and CLI, it enhances observability and control over experiment metadata management, greatly benefiting data scientists and AI engineers. Key technical features include handling API authentication, agent command execution, and support for self-hosted environments over HTTPS, ensuring secure connections. As the project evolves, ongoing documentation and community engagement through GitHub releases highlight the commitment to iterating this valuable tool for the AI/ML community.