According to CoinMeta, TypeSafe has released version 1.14 of Jev, which adds visual capabilities that can identify people in camera footage or compare whether two photos show the same dog. The new version's box positioning function can locate the position of a specified object within an image and return the coordinates of the rectangular area. CEO Diogo Almeida indicates that the older version of Jev 1.13 already had image recognition capabilities, but they were not made available due to concerns about accuracy. Both versions now support image input, and images can be used as criteria and options for judgment; however, box positioning requires version 1.14. In experimental mode, the maximum context per request has increased from 64,000 to approximately 262,000 token, which is four times the previous amount. Jev 1.14 is already available through `jev-preview` and is expected to become the default stable version on October 21st. TypeSafe reminds users that the visual and long-context functions are still in the experimental stage, which may result in increased latency and reduced accuracy. box positioning will return a coordinate box even if no object is specified, so developers need to determine whether the target exists to avoid incorrect operations.