openroboto-ai/jev-robot-control
Organization repository on GitHub
- Forks
- 3
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Demos & Experiments
Jev Robot Control
Simulation study that drives one xArm7 apple-to-plate task with Jev, a large reasoning model, and a small model through identical incremental controls.
- Stars
- 62
- Language
- Python
- Published by
- Community
- Author
- openroboto-ai
- Last verified
- Oct 2026
About this project
Jev vs GPT-6 Astra vs GPT-4.1 mini: direct Cartesian control of an xArm7 in MuJoCo. One apple. One plate. Each controller chooses an intent, then X/Y/Z movement directions and a gripper command. The shared executor applies one small motion increment and returns physical feedback. This repository includes the code, original recorded responses and trajectories, an offline verifier, and a synchronized three-column replay.
Quoted from the project's README. Read the full README.
Used inJev for Software & DevTools
- Languages
- Python
Highlights
- Issues intent then XYZ direction plus gripper action each cycle, scaled to small motion increments with fixed-orientation kinematics and physics feedback.
- Publishes code, recorded responses, trajectories, an offline verifier with tight state tolerance, and a synchronized three-column replay.
- Replays the published run with only Python and the repo, while new trials need provider keys and write new dated run folders.
- Reports a single seed trial per controller rather than success rates, with costs, cycles, and wall times that future runs can change.
Watch out
One seed trial per controller, not success-rate estimates; model selects direction while the shared executor scales steps and physics. Observation is simulator geometry, not camera perception. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
No. Each controller has a single recorded seed trial with the same scene, physics, and budget, so read costs, cycles, and wall times as one datapoint and rerun seeds before drawing rankings.
Two calls per cycle pick an intent then XYZ direction plus gripper action; the executor scales step size near targets, applies fixed-orientation kinematics, bounds, and physics increments.
Start the read-only replay server and open the local page with no key or GPU; run the offline verifier to re-execute recorded commands and compare joint states within tight tolerance.
Fresh trials need provider keys and write new dated run folders, and the repo holds simulation adapters only, with no model weights or real-robot deployment code included.
Only Python 3.12 and the repository files are needed: clone it, run python incremental_triple_app.py --port 8773, and open http://127.0.0.1:8773/. The server binds to loopback, is read-only, and makes no model calls.
It checks archived source hashes, every recorded question and observation, command validation, costs and counts, and re-executes each recorded motor command in MuJoCo, comparing every saved joint and fruit state (qpos) with a 1e-7 tolerance. No API calls, renderer or learned weights are needed.
Two sequential model calls: first an intent (approach, grasp, lift, carry, lower, release, withdraw or finish), then motor channels where X, Y and Z each select negative, hold or positive and the gripper selects open, hold or close. The model chooses direction, not absolute coordinates or joint torques.
More Jev resources like this
10GitHub stars
Agent bridge that connects any CDP-speaking browser to Jev by reading the DOM instead of the rendered layout, so headless or non-drawing browsers can still drive CLICK, TYPE_TEXT, SELECT, WAIT, DONE, and BLOCKED steps.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
lexmount/jev-browser-bridgePython
lexmount/jev-browser-bridgePython114GitHub stars
Browser agent where Jev picks each action from indexed page candidates while an LLM plans and reads, returning cited answers or schema-checked data with credential controls.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
agent-labs-dev/fastbrowsePython
agent-labs-dev/fastbrowsePython4GitHub stars
Browser-use layer for CLI coding agents: Jev picks the operation and target element in one round trip per step, exposed over MCP with a CLI, a Python API and an HTTP server.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
brnyxx/jev-raPython
brnyxx/jev-raPython2GitHub stars
A Python decision layer for browser and desktop automation: your code observes the surface, builds the candidate list and verifies the result, and the model only picks one id per step.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
aidil2105/jev-browser-pilotPython
aidil2105/jev-browser-pilotPython0GitHub stars
A local Python browser automation UI that turns a URL and a plain-language goal into bounded actions, with Jev picking the next operation from observed candidates and a separate text model used only to fill fields.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
KesavanKing/jev-browserPython
KesavanKing/jev-browserPython1GitHub stars
A Streamlit Tetris where the engine enumerates every legal placement, Jev picks one typed placement ID, and Python verifies rotation, column and exact occupied cells before the piece is allowed to lock.
Jev Demos & Experiments#community#jevbest#catalog-import
From the Jev guides
Original write-ups on what this entry is for.
A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.
Jev Demos & Experiments#gaming#agents#control
Read article
From the Jev community
Posts from builders shipping with Jev right now.