ZLens 0.3.0: your AI doesn't know your business. Auto ontology drafts it.

Ask an AI how many food orders you had and it writes SQL in seconds, but it can count food items instead of orders: 20,081 instead of 13,759. It doesn't know your business: which table is real, what one row is, how tables join, and where the traps are. That is your business model, and writing it by hand takes weeks, so most teams never do. ZLens 0.3.0 drafts it for you: auto ontology, from a dbt project or straight from your warehouse with no dbt at all. Click Draft the model and it proposes one table per business thing with what one row is, checks every join on your data, and turns the traps into rules measured on your data (13,759 food orders, with the query that measured it). You review it, and nothing is certified until you do. Our research measured it: on one NBA dbt project, 40 new questions it had never seen, Ask got the right rows 31 times with the drafted model and only 10 times with no model (25 with a model built by hand). And it scales to thousands of models, one area at a time. In this video: 0:00 An AI answer that counts the wrong thing 0:14 What your AI doesn't know · your business model 0:28 Auto ontology: draft, check, measure, decide 0:57 Our research: 31 vs 10 of 40 · big projects · get ZLens Recorded on ZLens 0.3.0 with the Jaffle Shop sample and an existing public dbt model repo (over 4,000 models, on an empty local copy of its tables). The opening SQL is an illustration of a common mistake; both counts (20,081 food items, 13,759 food orders) are real on the Jaffle Shop sample. The research result is from the ZLens 0.3.0 release notes: one NBA dbt project, 40 new questions, measured so far. Start here: npm create zlens@latest, then the Auto ontology playlist (5 videos) and Migrate from Tableau and Power BI (4 videos). ZLens: https://zlens.app Music: HeyGen music library. Voice: AI voice generated with HeyGen. Questions or ideas? https://github.com/duckcode-ai-labs/z...