Week of July 11–17, 2026
Scenario Import 🔥
You can now import scenario data into a Context directly from a JSON or CSV file. If your test data already lives in a spreadsheet, a database export, or another system, bring it in directly — no manual re-entry.
The import maps incoming fields to your existing Context structure. Extra fields are ignored; missing fields stay blank. If you have a lot of scenarios to set up, this is the fast path.
GPT-4o as the default evaluator
Upgraded the default evaluation model to GPT-4o. The qualitative judgment is better — especially on nuanced tone, accuracy, and helpfulness checks where the previous model was sometimes too lenient or too rigid. If you've had evaluations that felt slightly off, this should improve them.
In-app help
A help panel is now accessible from any page in the app. It surfaces the documentation in context — when you're setting up Context Mapping, the Context Mapping guide is one click away. When you're writing a qualitative check, the Assertions docs are right there. You don't have to leave what you're doing to look something up.
Scenario sync
Scenarios now stay in sync automatically when you change the structure of a Context. Add a field, and your Scenarios get a blank slot for it. Remove a field, and the corresponding data is cleaned up. The Context definition and the Scenario data now track each other.