Have you ever had a moment where instead of deleting your test model in Wobby, you delete the one you spent months polishing, and the only thing you can think is "oh no!" (aka the "onosecond")? Well we've got great news for you!

New features

• Versioning
  Every time you (or Steward) make an edit to your semantic layer or agent configuration, Wobby takes a snapshot. You can easily view the list of snapshots, and easily roll back changes by reverting to an earlier snapshot. Pretty handy!

• Recently Deleted
  So, deleted no longer means deleted. When you (or, again, Steward) delete something, it gets put into the "Recently Deleted" bin to give you a chance to undo that.

• Backtrack Steward changes
  If it so happens that Steward misunderstood what you meant, and wreaked havoc on your semantic layer, you can easily revert all changes that Steward has done in one single click.

Improvements

• MS SQL certificate support
  Connect to MS SQL databases that require SSL/TLS certificate verification.

• Nested aggregation queries
  Complex calculations like activation rates and pivot tables with nested aggregations now work correctly.

• Faster AI analyst responses
  Messages show instant loading feedback instead of multi-second delays.

• Clearer data connection errors
  Helpful error messages when data warehouse connection fails, replacing confusing empty states.

• Azure Synapse multi-schema support
  Select from multiple schemas when connecting to Azure Synapse databases.

• Consistent number formatting
  Values in tables and summaries now display consistently throughout the app.

Bug Fixes

• AI analyst conversation stability
  Fixed a crash that could interrupt AI analyst conversations mid-flow.

• Duplicate message prevention
  Fixed an issue where pressing Enter could send the same message twice.

• AI analyst creation fixed
  Resolved an issue that prevented creating new AI analysts.

• Slack integration stability
  Fixed unexpected disconnections for Slack-connected AI analysts.

• Data source deletion cleanup
  Deleting a data source now properly removes all associated models.

• Decimal precision on measures
  Measures configured with decimal places now display correctly.

That's about it! Have a great week, and if you have any questions, hit us up on Slack or the support chat.

We've all had that weird table in a data warehouse that grew with the company, and is an architectural mess. Every time you query it, you need to join tables five levels deep, and then filter using ILIKE to filter out legacy rows from some pre-migration era, which weren't cleaned up because the budget wasn't approved and the person resposible retired.

Happens! (If it didn't happen to you yet, don't worry, eventually it will)

Now, you can save all of this logic in your Wobby Semantic Layer and create models based on a SQL query, not just a physical table, so that your AI analysts can navigate clean data and not get confused. If you're familiar with relational databases, imagine it like a SQL view.

Learn more in our docs →.

Happy New Year! Hope you had a great break - we definitely ate too many cookies. But we're back, caffeinated, and ready to ship.

Here's what's new: audit logs, semantic layer exports, and a fresh coat of paint on settings.

New Features

Audit Logs
Full visibility into who did what and when across your workspace. Perfect for compliance, debugging, or just satisfying your curiosity.

Export Semantic Layer & Agents
You can now export your entire semantic layer and agents in multiple formats. Back it up, share it, do whatever you need.

Agent Deploy Controls
Agents are now private by default. Only Studio users can see them until you explicitly deploy to Explorer. This gives you proper staging before going live.

Improvements

Settings Revamp
We completely rebuilt the settings screens. They're cleaner, faster, and actually make sense now.

Steward AI Agent
Various fixes and improvements to make Steward more reliable when helping you build your semantic layer.

Deprecations

Legacy Data Sources & Semantic Layer Pages
These are now retired. If you were still using them, it's time to switch to the new setup.

SQL-based Agents
Also retired. The old ways are no longer available - but trust us, the new agents are way better.

Hey everyone! It's been a busy week for us, we are joining HCLSoftware. In short, nothing changes for you, we are not going anywhere. Your data, your workflows, and your experience with our platform will continue uninterrupted.

You can read the full press release here.

Major changes under the hood

• Querying syntax refactor
  We have rebuilt the SemQL engine under the hood to increase agent accuracy. Now the agent will make less mistakes when writing SemQL.
  
  There is one main user-facing change to keep in mind — SemQL no longer uses to refer to Semantic Layer entites.
  
  Previously: SELECT from tasks;
  Now: SELECT tasks.task_count from tasks;

• Teams integration stability
  When working with large Teams instances, Wobby would sometimes fail to fetch channel lists. Now it works properly

UI Improvements

• Same chat input for Steward and Analysts
  We have unified the chat input for both Steward and Analysts, which really helps with muscle memory :)

• More space for testing agents
  We hid the name and description editor in the toolbar, so now you have more vertical space to work on agent settings

  

• Better tool use UI for Steward
  It's now easier to see what tools Steward used, what parameters the tool was called with, and how long it took.

  

Bug Fixes

• Fixed Steward hanging.
  Now it works

• Fixed agent interrupts.
  Now agent interrupts when you need it to

• Fixed Steward not showing an input when asking for input.
  Now you can answer Steward's questions

• Metrics now show "error" instead of "null" if something failed.
  Makes more sense that way

• Scroll down button no longer gets in the way.
  That was annoying

So, imagine you're in a crunch, and need to make a quarterly report for your board. Due to nebulous reasons, the only person who is available to do this is an intern with a PhD in statistics, but who started at your company two days ago and doesn't even know how the your office coffee machine turns on.

Due to other nebulous reasons, you have a choice between giving that intern:

• A 500-page manual describing your company's datasets and a computer that only runs calc.exe and a PostgreSQL terminal, or...
• A 20-page intro to your company's PowerBI setup, and a PowerBI account

Realistically, you will give the intern the PowerBI account. Sure, it would still be nice if they had your company's tribal knowledge, but they will more likely succeed with the task and make less errors if they drag and drop measures in PowerBI rather than blindly writing raw SQL.

The problem is, most Text-to-SQL solutions are the 500-page manual case — they give the an extremely smart but nil-experience agent a ton of documentation of dubious quality, tell it to do its best, and then whether it does a good job is anyone's guess.

We wanted to change that, so we'd like to introduce the Wobby Semantic Layer, the operating system for agentic BI.

• Semantic Layer
  A translation layer that transforms your raw data from technical abstractions — tables, columns, joins — into business entities that actually mean something: customers, revenue, conversion rates. Instead of agents guessing what tbl_ord_v2.amt_eur_net means, they work with "net revenue in euros."

• SemQL
  A semantic query language that lets agents query business entities deterministically. No more hallucinated joins. No more invented column names. When an agent asks for "monthly revenue by region," SemQL ensures it gets exactly that — with the exact business logic your team has agreed upon, every single time.

• Steward
  The agent that builds your semantic layer for you. Traditional semantic layers are 6-month PoC projects that become outdated documentation the moment they're finished. Steward monitors your team's conversations, detects knowledge gaps, and proposes updates — turning semantic layer maintenance from a burden into an incremental workflow with immediate results.

If you'd like to try it out, you can check out more information in the docs, or reach out to us, and we will help you get set up. Semantic Layer is currently in beta, and we are still ironing over some rough edges, so we'd love to hear what you think about it.

Thanks for reading, and have a great week!