Skip to content
dbx tools

v0.9.119

Compared HEAD and the working tree with the latest local tag, v0.9.118. Generated-file churn and test-only changes are excluded.

Breaking Changes

  • Removed DatabricksSandbox and related exports from @dbx-tools/appkit-mastra. Use @dbx-tools/appkit-sandbox and its sshWorkspaceSandbox(runtime) integration instead.
  • Mastra no longer accepts sandbox: "databricks" or the former Databricks options object. Pass a native Mastra sandbox or resolver; sandbox: true now selects Monty.

New Features

  • Added user-scoped Databricks Sandbox management, SSH execution, Mastra tools, HTTP/WebSocket forwarding, and React lifecycle, terminal, service, and health controls. Command execution requires system OpenSSH.
  • Added packaged Sandbox recipes and shared read-only skill mounts. Databricks AI Tools skills now use a daily cache and an in-memory mount instead of writing to the user’s Workspace.
  • Added memory-first caching with asynchronous persistence. AppKit uses Lakebase when available and project-scoped disk otherwise; model discovery isolates cached entries by workspace and credential identity and can reuse retained data after refresh failures.

Fixes

  • Mastra chat reconnects background turns after page reloads, refreshes completed history, and shows active-response progress.
  • Sandbox routes are user-namespaced, enforce ownership, and support nested service paths. Terminal recovery after app restarts and mobile service/stat views are improved.
  • Demo development preserves AppKit configuration while using Bun client HMR without restarting backend services.
  • Fixed root-directory listing in the in-memory filesystem, Python Node-runtime cache compatibility, and false error logging during successful compilation.

Validation

  • The configured release validation phase is already complete, as reported. No tests or other validation commands were rerun; no validation counts are claimed. No files were modified.