Databricks CLI v1.17.0 introduced state version 3 alongside bundle and credential enhancements.

Databricks CLI v1.17.0 arrived with direct deployment state version 3, introducing a breaking change that required at least CLI v1.8.0 for compatibility [4]. The release expanded Databricks Asset Bundles by adding AI Gateway service support, file-change triggers for job runs, and revised resource reference handling [4]. It also delivered Docker credential helpers for Databricks Artifact Registry along with improved bundle deployment reporting [4].

Tooling integrations rapidly incorporated these updates throughout September, moving from bundled CLI v1.16.0 to v1.17.0 across sequential releases [2, 5, 6]. These updates fixed an SSH transfer regression affecting files larger than 1 MiB [5] and allowed active SSH sessions to survive transient network disconnects [2]. This improved stability for developers using CLI-driven SSH tunnels to link local IDEs and AI coding agents with remote Databricks compute [9]. By late September, configuration expanded further to permit custom CLI binary paths, corporate CA certificates, and HTTP proxies [1].

In the community, developers explored deployment pipelines and authentication boundaries. Practitioners called for a formal validation step inside Databricks CI processes to catch errors before deployment [3]. Meanwhile, others sought approaches to automate file downloads to local environments in locked-down environments that restrict both personal access tokens and the CLI [7].