The classify schema

Before a set can be mapped, each company’s page has to become grounded JSON — what it does, who it’s for, how it’s positioned. That’s classification, and stratless does it for you. You rarely need to call it directly, but understanding it helps you feed analyze well.

Two paths

• Send texts (the simplest path). You pass { name, text } per company and stratless classifies each server-side inside the analyze call. Nothing extra to do.
• Send companies (pre-classified). If you’ve already classified — or want to cache and reuse the classification across runs — pass ClassifiedCompany[] directly. This skips server-side classification and is the cheapest path.

The public schema

The classify artifact is public and free. Fetch the current schema:

curl https://api.stratless.com/schema

It returns the JSON Schema for a ClassifiedCompany — the fields stratless reads from a page (what it does, positioning, audience, page type, and so on). Use it to:

• validate a set you built yourself before sending it to analyze, or
• prompt your own model to classify pages into the exact shape stratless expects.

Grounding matters

Classification is extractive — it reads what the page says, it doesn’t invent. Garbage in, garbage out: feed it a blocked page or a cookie wall and the classification reflects that (and analyze will surface it in dropped with the page_type that explains why). Clean text from a real product page gives the best map. See the recipes for getting clean text from each scraper.