TypeSafe AI’s Jev 1.14.0 adds image understanding to its typed decision API. The preview release introduces a box question that locates a described object, accepts PNG and JPEG assets, and can expand context budgets to approximately 262,000 tokens when an opt-in header is supplied. Existing text decision types remain available, so Jev 1.13 users mainly need to upgrade for vision or the larger context window.
What Jev 1.14 changes
Jev is a structured decision model rather than a general prose chatbot. An application sends state and typed questions, then receives typed values, probabilities and confidence. Jev 1.14.0 keeps the existing choice, score and noul question types and adds box.
As an Amazon Associate I earn from qualifying purchases.
The release is available through the moving jev-preview alias. TypeSafe AI presents the vision and long-context functions as preview capabilities; pin jev-1.14.0 while testing so a later preview change does not silently alter behavior.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteThe new Box question
A box question asks Jev to find a described target in an image. The response provides four normalized edges—top, bottom, left and right—as fractions from 0 to 1, plus confidence. A value of 0.25 for left, for example, means the left edge is one quarter of the image width from the left side.
#1 Best Overall
Because a box response is shaped around coordinates, pair it with a presence check when the target may be absent. Your application can then reject a coordinate result when the presence decision says the object is not visible, rather than treating every returned rectangle as a detection.
Conceptual request shape
{
"assets": {
"scene": "<uploaded PNG or JPEG reference>"
},
"state": {
"image": "scene"
},
"questions": [
{
"type": "box",
"name": "target_box",
"question": "Where is the red suitcase?"
}
]
}The exact field definitions should follow the Jev 1.14 API schema. The important change is that an image is uploaded once in the top-level assets map and referenced from state, instructions or criteria.
How to send images
Image requests require the opt-in header TypeSafe-Danger-Zone-Images: true. Jev accepts PNG and JPEG still images. Independent coverage reports a limit of up to 16 MiB per image and a supported side range of 256 to 1,568 pixels, while TypeSafe AI’s multimodal documentation puts the whole request-body cap near 32 MiB. Keep the combined assets and other request fields below that body limit.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minute- Upload or otherwise provide each PNG or JPEG in the request’s top-level
assetsmap. - Reference the asset from the state, instructions or criteria instead of embedding the same image repeatedly.
- Send
TypeSafe-Danger-Zone-Images: truewith the request. - Ask a focused question and validate the returned coordinates and confidence before using them downstream.
TypeSafe AI warns that multimodal behavior can be less reliable than pure-text calls. Simple, unambiguous questions and application-side validation are therefore more important than they are for text-only decisions.
Jev’s 262K-token context mode
Jev 1.14 can raise both per-request and per-question budgets to approximately 262,000 tokens, but only when the opt-in header TypeSafe-Danger-Zone-Long-Context: true is sent. The reported default budgets are 64,000 tokens per request and 32,000 tokens per question. The larger ceiling is not automatic merely because the model is Jev 1.14.
Long context does not remove the separate multimodal body limit. A request can have a very large token budget and still fail if its serialized body approaches the near-32 MiB cap. Measure payload size, leave headroom for metadata, and handle limit errors explicitly.
Rank #4
![MixPad Free Multitrack Recording Studio and Music Mixing Software [Download]](https://m.media-amazon.com/images/I/71ltIxIuz1L._SL160_.jpg)
- Create a mix using audio, music and voice tracks and recordings.
- Customize your tracks with amazing effects and helpful editing tools.
- Use tools like the Beat Maker and Midi Creator.
- Work efficiently by using Bookmarks and tools like Effect Chain, which allow you to apply multiple effects at a time
- Use one of the many other NCH multimedia applications that are integrated with MixPad.
Jev 1.13 versus Jev 1.14
| Capability | Jev 1.13 | Jev 1.14.0 |
|---|---|---|
| Availability | Earlier Jev line | Preview; available as jev-preview; pin jev-1.14.0 for a fixed target |
| Question types | choice, score and noul |
Those types plus box |
| Image handling | Not part of the documented release | PNG/JPEG assets referenced through the top-level assets map |
| Image header | Not applicable | TypeSafe-Danger-Zone-Images: true |
| Context budget | Earlier limits | About 262,000 tokens per request and question with the long-context header; reported defaults are 64,000 and 32,000 |
| Long-context header | Not applicable | TypeSafe-Danger-Zone-Long-Context: true |
| Request size | Not stated for this comparison | Official multimodal documentation reports a near-32 MiB whole-body cap |
| Operational reliability | Text-only behavior | Vision and long context are preview features and require output validation |
Is Jev 1.14 ready for production?
It is best treated as an evaluation and controlled-rollout release, not a drop-in production default. Pin jev-1.14.0, gate image and long-context headers behind configuration, log confidence and validation failures, and retain a text-only path for requests that do not need the new features. Test absent targets, crowded scenes, low-quality images, oversized payloads and questions near the context ceiling before widening deployment.
For applications that only use typed text questions, upgrading is optional: those question types continue to work. The case for Jev 1.14 is strongest when an application needs normalized image regions, image-referenced criteria or substantially larger decision context.
Practical failure checks
- No image result: confirm the image header is exactly
TypeSafe-Danger-Zone-Images: trueand that the asset is referenced by its map key. - Context limit error: confirm
TypeSafe-Danger-Zone-Long-Context: true, then check both token budgets and serialized body size. - Unexpected box: add a presence question, inspect confidence and reject coordinates outside the 0–1 range.
- Unstable behavior after testing: replace
jev-previewwith the pinnedjev-1.14.0identifier.
Frequently Asked Questions
What does Jev’s Box output represent?
It returns normalized top, bottom, left and right image edges as 0–1 fractions, together with confidence.
Can Jev 1.14 analyze any image format?
The documented image support covers PNG and JPEG still images; other formats are not established for this release.
Do existing Jev text questions require migration?
No. Choice, score and noul questions continue to work; migration is primarily for image questions or the opt-in larger context.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
