RoboJev
Turn natural-language standing orders into bounded robot-arm decisions.
What Jev decides
Which target, motion, speed, avoidance, and next primitive fit the scene.
RoboJev uses Jev to choose actions for a WidowX arm and a MuJoCo simulation from compact scene reports and allowed primitives.
- Jev role
- Choose robot-arm primitives and target behavior from a bounded action set.
- Input
- A compact scene report, available primitives, task text, and standing orders.
- Output
- Typed selections composed into limited arm setpoints.
- Decision frequency
- On meaningful scene changes or a silence timeout during control.
- Fallback
- Code checks preconditions and limits; documented physical-arm coverage is narrow.
How it works
Overview
The controller converts scene state and standing orders into small typed questions, then composes Jev answers into arm commands under code-enforced limits. The README reports repeated paper-cup pick-and-place cycles on real hardware and broader behavior in simulation. It explicitly says the real-arm result is fitted to one cup and table, and several safety and perception scenarios were not tested on the physical arm.
- Language
- Python
- Author
- alee792
Sources
Related projects
Jev Autopilot
A Jev autopilot flies a simulated drone across a generated Three.js city and lands on a target pad.
What Jev decides
Throttle, yaw, pitch, roll, whether to commit to landing, and whether to cut the motors.
Mobile Jev
Choose bounded actions on a real Android device, execute through Mobilerun, and verify fresh device state.
What Jev decides
Whether to open an app, tap, type an exact supplied span, scroll, navigate, wait, finish, or report that the task is blocked.
typesafe-computer-use
A native macOS computer-use loop built from local OCR, accessibility data, and bounded Jev actions.
What Jev decides
Which computer action, visible item, site, or off-screen accessibility control to use next.