Jev Images Tutorial

Verification status

This request follows official documentation reviewed on October 20, 2026. Jev AI Hub has not executed it against the hosted API. Images and Box are experimental; formats, limits, and behavior may change.

Send an image

Save a JPEG as parcel.jpg, set TYPESAFE_API_KEY, and run in a shell with POSIX quoting:

curl --fail-with-body https://api.typesafe.ai/v1/systemone \
  -H "Authorization: Bearer $TYPESAFE_API_KEY" \
  -H "TypeSafe-Danger-Zone-Images: true" \
  -F 'request={"model":"jev-1.14.0","state":{"scene":"asset-7f9c-parcel"},"assets":{"asset-7f9c-parcel":{"type":"image","part":"photo"}},"questions":{"parcel_present":{"type":"noul","instructions":"Is a delivery parcel visible in `scene`?"}}};type=application/json' \
  -F 'photo=@parcel.jpg;type=image/jpeg'

The request part comes first. Let cURL set the multipart boundary. Use unique asset references to avoid matching ordinary words accidentally.

Check results

Read answers.parcel_present.noul as a probability. Test a visible parcel, an empty scene, occlusion, and a similar-looking object. Record returned model and usage; choose a review threshold from labeled images rather than one successful photo.

Documents and gateways

The documented image formats are PNG and JPEG. Render PDF pages or extract their text first. For text classification, start with the classification example.

Python 0.7.5 provides Image assets; JavaScript 0.7.0 provides image helpers and requires Node.js 22+. The SDKs handle asset serialization. Verify gateway support for images, Box, multipart upload, and compression independently of text compatibility.

Continue with Box for localization and the migration guide for SDK changes.