Create a Schema
A schema defines the target structure your clean data should conform to — its fields, types, and definitions. Every dataset you transform is mapped into a schema, so this is where you decide what “clean” looks like.
Open the schema dialog
Open a folder
From the Folders page, open the folder you created.
Start a new schema
Click New Schema. The Add New Schema dialog opens with four ways to get started.
The four ways to create a schema
Pick whichever matches the source you have. Three of the four (everything except presets) use AI to infer fields for you.
1. Upload from existing file
Parse headers from Excel, CSV, PDF, images, Word, or PowerPoint files into a schema you can configure. Transformer reads the column headers (and a small sample of rows) and proposes fields and types that you can then refine.
Choose this when you already have a representative file and want the structure inferred from it.
2. Connect to Data Source
Import a schema directly from a connected database table. Transformer reads the table’s columns and maps their database types into schema field types.
Choose this when your target structure already lives in a database and you want it to stay in sync.
3. Describe to AI
Paste any unstructured data or a plain-language description and let AI determine the fields for you. Transformer extracts a candidate set of fields and types from whatever you provide.
Choose this when you don’t have a clean file yet but can describe — or paste an example of — the data you expect.
4. Use Preset
Start from a predefined schema template for common data structures. Presets give you a ready-made set of fields you can adopt and adjust.
Choose this when your data matches a standard shape and you want a fast start.
Whichever option you pick, you land in a schema editor where you can rename fields, change types, mark required fields, and add definitions before saving with Create Schema.
Next step
Once your schema is saved, you can run data through it: Create a Dataset →