Use Cases / image-object-box
Jev Image Object Box (beta)
Image inputs and Box questions are beta and dreamer-ready, not recommended for production. Teams still want a schema-safe way to ask “where is the product in this photo?” without parsing free-form vision prose.
Try a related template free on this host — 5 anonymous real-API runs with no signup — then ship the same schema with a prepaid jv_live_ key. See is Jev free? and playground.
Why Jev
On Jev 1.14 you can attach named image assets and ask a box question. The model returns a single bounding box { top, bottom, left, right } plus confidence — typed output your code can draw or gate on. Treat results as experimental.
How to define the decision
Send assets: { "photo": { "type": "image", "format": "png", "data_base64": "…" } } and state that references `photo`. Question: { "type": "box", "instructions": "Find the primary product in photo" }. See /docs/api/#box.
When to use Jev
- Prototyping vision gates before a production detector
- Learning the assets + box request shape on this host
- Side-by-side evals against a dedicated vision model
When not to use Jev
- Production safety, medical, or access-control pipelines
- Pixel-perfect localization SLAs
- Video or multi-object detection
Failure modes and human fallback
- Upstream or network errors: retry with backoff; do not invent a decision.
- Low confidence: route to a human queue or safer default — never auto-act on money/destructive tools.
- Schema drift: keep a golden fixture set in CI so label changes fail the build before production.
- Official: Image inputs and Box questions are beta and dreamer-ready, not recommended for production.
- Prepaid metering still applies when you call the live API — homepage preview does not.
- Prefer a dedicated vision stack for production localization.
API example
Same prepaid jv_live_ balance as tools. Output tokens are free on this host's metering model.
curl https://jevtypesafe.org/api/v1/decide \
-H "Authorization: Bearer jv_live_YOUR_KEY_HERE" \
-H "Content-Type: application/json" \
-d '{"state":"Customer: I've been waiting five days for my refund.","questions":{"action":{"type":"choice","instructions":"Decide the next step","criteria":{"allow":"safe","review":"unsure","block":"unsafe"}}}}'FAQ
- Is Image / Box production-ready?
- No. Image inputs and Box questions are beta and dreamer-ready, not recommended for production.
- How do I try without spending tokens?
- Open the homepage playground Image · Box (beta) local preview, or call /api/v1/decide with a jv_live_ key when you are ready to meter.
- What does the box response look like?
- answers.<name> = { type: "box", box: { top, bottom, left, right }, confidence }. Coordinates are normalized 0–1.
Related use cases
Ready to build?
Free anonymous tool runs, then prepaid packs and a jv_live_ key — independent host, not affiliated with TypeSafe AI.