Locating a signature in a scanned form with Jev 1.14 (TypeSafe) and Microsoft Agent Framework

Описание и применение

Автоматическое определение наличия подписи и её расположения в отсканированных формах с помощью Jev 1.14 и Microsoft Agent Framework, включая экспериментальные функции обработки изображений и возвращения ограничивающих прямоугольников.

Задача

Determining whether a scanned form is signed and locating the signature is difficult with document agents, leading to failures on signed forms, stamps, and handwritten fields.

Результат

The approach enables automated detection of signatures in scanned forms, providing both a probability of presence and precise bounding box coordinates, but remains experimental and should be used cautiously.

Рабочий процесс

• 1. Prepare an image of the scanned form.
• 2. Create a TypeSafe client and send a system_one request with a Noul question for "Is the form signed?" and a Box question for "the signature on the form".
• 3. Include the model "jev-1.14.0" and the header "TypeSafe-Danger-Zone-Images: true".
• 4. Receive the response containing a probability for the Noul answer and a box with top, bottom, left, right fractions and confidence.
• 5. If using the Microsoft Agent Framework, embed the image in a Message and call get_response with the same questions and options.
• 6. Handle pitfalls: check the Noul presence answer before trusting the Box; use separate questions for multiple signatures; crop large images if the target is small.

Доказательства

Jev 1.14 (TypeSafe AI, October 14, 2026) accepts images and a new Box question type that returns the bounding box of one target in an image.

Источник

TypeSafe flags both features as experimental ("Danger Zone"): prototypes only.

Источник

No "not found" answer. Box always returns a box, even when the target is absent.

Источник