We rebuilt the internals of the four web-content endpoints under /global. Routes, authentication and rate limits are unchanged; four details of the contract did change and are worth a look if you already call them.
Research in a few seconds#
POST /global/research/v1 now answers in a few seconds rather than the 10 to 30 we announced at launch. The response shape is the same: a markdown report with numbered citations and sources listing the pages behind it. pages_analyzed now counts the sources the report drew on.
curl https://api.croma.run/global/research/v1 \
-H "Authorization: Bearer $CROMA_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "query": "¿Qué cambios trajo la Ley 2195 de 2022 en materia de responsabilidad de personas jurídicas en Colombia?" }'country is gone#
The page endpoints no longer accept country. If you still send it, it is ignored; there is nothing to fix. Reading a page does not depend on the country it is read from.
effort says what it does#
| Value | What you get |
|---|---|
min | The fastest answer; it may come from a recent copy of the page. |
standard | The default balance. |
max | Always reads the page live, including pages that render themselves in the browser. |
Honest metadata#
With include_metadata: true, metadata carries title, author, published_at and image. A field the page does not declare comes back as null. The fields we could almost never fill (description, site name, keywords, page count) leave the contract.
404 when there is nothing to read#
POST /global/extract/json/v1 and POST /global/generate/json/v1 now return 404 when a page has no readable content, just like extract/markdown, instead of an empty object. The message says which option to change.