Changelog

1.1.8

Multi-Datasource Support, White Labeling, and Chat Experience Enhancements

Multiple Datasource Instances Per Project

You can now connect multiple instances of the same datasource type or combine different data sources within a single project.

Configure separate production and development Snowflake environments, multiple Looker instances, or mix BigQuery with Snowflake to analyze data across your entire stack. Each instance gets a human-readable alias (like "Production Warehouse" or "Marketing Analytics") that appears in chat, making it clear which datasource you're working with at any time.

Orion will query and analyze across enabled data sources and surface any discrepancies or insights across your data infrastructure.

Slide Generation (Beta)

Generate presentation-ready slide decks directly from your insights. Simply ask Orion to create slides from an analysis, and it will produce an HTML slide deck with navigation controls and professional formatting. You can preview slides in-browser with fullscreen mode and export to PowerPoint format when ready to share.

White-Label Branding

Enterprise customers can now customize Orion with their company branding. Upload custom logos for light and dark modes, configure brand colors, and set company icons that appear throughout the interface and in PDF exports. Contact our team to enable white-label features for your organization.

Improvements

  • Tools now group automatically when multiple tools run in sequence, reducing visual clutter in chat
  • User profile avatars can now be uploaded and display throughout the interface with automatic fallback to initials
  • BigQuery configuration now supports location specification for datasets outside the US region
  • Insight retrieval tool displays metadata including title, creation date, creator, and direct links
  • Sidebar maintains alphabetical project order with auto-scroll to active project instead of reordering
  • Tool components maintain their expanded state during streaming responses
  • Fixed scheduling reliability issues that caused errors when rapidly updating recurring analyses