Dex Semantics - Let your agent build analytics with 25% fewer errors
by•
Dex Semantics adds declarations to your analytical models, which capture row meaning, join nature, how numbers add up, and decisional assumptions.
Thanks to this, Dex scores 80% on ADE-bench with Claude Sonnet 5.5, compared to 73.3% without Dex (25% fewer tasks failed).
Declarations live in your models as code, not a third-party tool. Dex then helps your agent convert declarations into your Semantic Layer of choice.
Replies
What ide do your team use to build this beauty?
keeping how numbers add up right next to the model is the right place for it, a summed ratio looks totaly plausible on a dashboard and nobody notices
when a declaration goes stale after a schema change, does Dex flag it before the agent relies on it??? feels like the obsolete assumption from your quote all over again
@thamibenjelloun 100%, that's what the `dex maintain` commands are all about!
You can run the maintain commands in agentic workflows or in CI, to make sure that a schema change gets flagged. Your agent can then check the declared assumptions vs the schema drift and update the data models (and declarations) accordingly!
Fixing drift before it reaches production is the part I'd care about most, that's usually where analytics quietly breaks. Does Dex flag the change for a human first or fix it on its own? Good luck with the launch!
The declarations-as-code approach is what caught my eye — keeping row meaning, join nature, and how numbers add up inside the model itself rather than in a separate tool feels like the right call for drift. Curious about the remaining 20% on ADE-bench: which kinds of tasks still fail even with Dex, and are they mostly ambiguous business logic, or things the declarations can't capture yet? Also, does the CLI work outside dbt, e.g. with plain SQL models?