Skip to content
dbx tools

v0.9.64

Release Notes — Since v0.9.61 Includes HEAD and working-tree changes. Generated churn and test-only changes excluded.

New Behavior

  • Genie and model gateway accept mutually exclusive --model / --model-class selectors. Gateway requests may omit model or use x-dbx-tools-model-class.
  • Mastra uses on-demand skill search and user-scoped workspace caching with a five-minute default TTL. Set workspaceSkillSearch: false to restore eager loading.
  • Release automation accepts --release-notes-instructions, runs autofix before preparing release commits when validation is enabled, and publishes retained release notes in documentation.

Breaking Changes

  • Graphiti removes --model, --embedder-model, --embedder-dimensions, and their environment settings. Use --model-class / MODEL_CLASS (default: chat-fast); embedding models and dimensions come from discovery.
  • Genie homes become profile-scoped, sharing state across models. Previous profile-model homes are not automatically reused.
  • Default model selection no longer favors GPT. Search words such as “summarize” and “embedding” no longer imply capability classes; use explicit classes or model names.

Fixes

  • Default ranking prefers newer models within the leading family; chat endpoint searches no longer impose a balanced-class ceiling.
  • Discovery probes missing embedding dimensions; Graphiti reports an explicit error if dimensions remain unavailable.
  • Workspace skill caches persist across requests and invalidate after mutations. Chat Thoughts omit Markdown tables and collapse excess whitespace.
  • Python binding watchers track reachable source dependencies, avoiding regeneration for unrelated edits.

Validation

  • 98 focused tests passed across nine files; zero failures.
  • bun run version:check and git diff v0.9.61 --check passed.
  • Full builds, Python runtime tests, and live deployment checks were not run. No files were modified.