Community · E2 · artifact verified
Give Jev eyes over cameras and streams
A bridge connecting images, video streams and RGB-D cameras to Jev's judgment engine: identify what matters in a frame, estimate risk, score a situation or judge many visible objects at once - typed answers over the visual world, 46 stars in its first day.
01 · Role in the system
What Jev does here
The session layer turns what a camera sees into Jev questions and turns the typed answers back into scene state: Nouls gate whether something notable is present, Choices caption which object or option applies, and the same loop works on a single image, a live stream or an RGB-D depth camera. Because the judgment engine never generates text, visual decisions arrive as calibrated types the calling code can threshold - spot what matters, estimate risk, score the situation - rather than captions to parse. The README is bilingual (English and Chinese), and the project drew 46 stars on its first day, the fastest adoption of any community case in this directory.
02 · Control boundary
Where Jev sits
Jev as the judgment engine behind a vision bridge: frames become typed questions, Nouls gate detection and Choices caption what applies, and calling code thresholds calibrated answers instead of parsing generated text.
Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.
03 · Known limits
What this evidence does not prove
- One-day-old project; the vision-to-question mapping is early and may change fast.
- Frame understanding depends on the upstream captioner feeding the state; no published accuracy benchmark.
- 46 stars in a day is adoption signal, not an evaluation.
04 · Attribution
Public sources
This is a Community record: the project was published by a third-party community author.
- CharlesFeng0314 ↗Community · github · public · checked 2026-10-01