> ## Documentation Index
> Fetch the complete documentation index at: https://docs.omni.co/llms.txt
> Use this file to discover all available pages before exploring further.
# June 5, 2026
> More Vibe Apps, AI Evals, More Dashboard Improvements, Markdown Column Promotion, and Sub-Agents
**Stay in the loop!** Subscribe to the [demo RSS feed](/demos/rss.xml) to be notified when we post new demos.
Demos highlight what we are working on or experimenting with, but are not a guarantee of release. Let us know your thoughts at [support@omni.co](mailto:support@omni.co).
## AI Customer Evals
*`Joel Dodge · AI Modeling`*
GA release of the AI hub eval framework — create prompt sets, run evals against different model branches (shared, fastest, smartest), and compare results side-by-side with an AI judge scoring confidence.
## AI Tracing & Observability
*`Joel Dodge · AI Modeling`*
Internal tool for super/support users to inspect every LLM call Blobby makes - view full system prompts, tool definitions, parallel tool calls, and nested sub-agent spans in a collapsible trace drawer.
## AI Chat UI Cleanup
*`Gustav Staprans · AI`*
Bottom bar declutter (topic chip + subtle model/connection display) and replacement of the split-view sidebar with inline header tabs for chat history, fixing nested-sidebar issues and enabling future left/right panel positioning.
## Dashboard Agent Tab Cleanup
*`Trip Tate · Dashboards AI`*
Blobby can now delete unused workbook tabs on command, with protection against removing tabs that are the source of linked/cross-tab calculations.
## Sub-Agent Orchestration
*`Trip Tate · AI`*
Proof-of-concept showing the main agent invoking a sub-agent via tool call to offload complex work (e.g. dashboard creation), keeping the main context clean and enabling iterative replanning when queries fail mid-flow.
## History Filter by File
*`Conrad Slimmer · Modeling`*
New filter on the version history view to show only commits where a specific file changed — speeds up binary-search debugging when tracking down when a model or view was modified.
## Color Calculation Fix
*`Jared Hardy · Design UX`*
Bug fix for incorrect light/dark contrast calculation on custom chart colors — dynamic color previews now update correctly as users change positive/negative palette settings.
## Dashboard Advanced Layout Alignment Helpers
*`Jared Hardy · Dashboards`*
New alignment guides (red snap lines) for items within the same container layer, including half-size snapping, plus in-progress ruler/column-count indicators to help size and align elements without the properties panel.
## AI Model Suggestions
*`Jade Khiev · AI Modeling`*
AI-generated suggestions for improving topic AI context, now with diff view (not full rewrites), sentiment-aware quality signals, scheduled daily runs at 2 AM, and upcoming ignore/reject feedback loop.
## Apps Data Storage Restructure and Generation Improvement
*`Nathan Agrin · Apps AI`*
Major structural rewrite — apps now stored like dashboards (own data table), query generation delegated to existing query tools, Blobby can self-fix rendering issues from screenshots, improved security envelope, and apps now appear in the content/permission system.
## Markdown Column Promote
*`TJ Banghart · Workbook Modeling`*
New "Promote" button on markdown column vis configs to save the current formatting back to the workbook as the model default, with a "Revert to model definition" option to undo.
## Top N Control
*`TJ Banghart · Modeling Workbook`*
Users can now change the Top N value directly in the viz without it being locked to the modeled default, including optional min/max bounds — works on both workbooks and dashboards.
## Dashboard API V2
*`Luke Bowerman · Dashboards`*
New breaking-change PATCH endpoint for programmatically editing dashboards — supports targeting a specific draft or published doc, partial updates (send only what changed), explicit null to delete elements, and AI-generated history summaries.
## AI Tool Disable Controls
*`James Davidson · AI`*
Per-model setting to selectively disable specific Blobby tools (Viz, user memory, conversation learning) — reduces token cost and enables behavioral governance; integrated with the eval framework.
## AI Skill Permissions
*`James Davidson · AI Administration`*
Skills can now be conditionally shown/hidden based on existing access grants — users without the required grant don't see the skill in the UI, effectively adding content-level permissions to skills.
## Omni AI Native Markdown Context Demo
*`Nathan Agrin · AI Modeling`*
Testing enabling Markdown files directly in the model. This is analogous to a CLAUDE.md file. The goal here is to start testing how we can push Omni more into the AI first world. In a future world you could imagine someone working with an OMNI.md file directly on the filesystem, talking to a headless Omni through Claude.