Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

typesafe-jev-drone-demo

Three.js drone simulator where Jev chooses each of nine candidate movements and estimates whether the route is blocked.

Appprototype
View on GitHub ↗

Overview

Jev Flight Lab is a live drone simulation with a Three.js frontend and a typed Python/FastAPI backend. Python owns the world geometry, position, velocity and checkpoints. For each step it predicts nine short candidate movements (advance, left or right detours, climb, descend, sideways moves and hover), each with swept collision clearance, predicted endpoint and progress toward the checkpoint.

That structured state goes to System One with a Choice question selecting the next action and an independent Noul question estimating whether the direct route is blocked. Python validates the response before moving. The UI shows the live Choice distribution, the obstruction estimate, latency and an Inspect dialog with exactly what Jev received and returned. You can add a barrier to force a fresh decision and switch cameras. The key stays server-side.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich of nine candidate movements should the drone take next?advance · left detour · right detour · climb · descend · sideways left · sideways right · hover · ...
Yes / noIs the direct route blocked?yes · no

How it works

  1. Python computes nine candidate moves with clearance and progress.
  2. Jev chooses a move and estimates obstruction.
  3. Python validates the response and applies the move.
  4. The UI shows probabilities, latency and raw input/output.

Strengths

  • Physics and safety stay in code; Jev only picks among precomputed options.
  • Full transparency via the Inspect dialog.
  • API key never reaches the browser.

Caveats

  • Simulation only.
  • Requires a TypeSafe key.

Getting started

uv sync && npm install && npm run build, set TYPESAFE_API_KEY, then uv run uvicorn backend.app:app --port 8000

At a glance

Type
App
Status
prototype
Language
Python, TypeScript
Runtime
FastAPI backend + Three.js web app
Has code
yes

Use cases and integrations

How curated lists describe it

  • Simulation: Three.js drone simulator with a Python backend where Jev drives the navigation decisions.

    yibie/awesome-jev

Frequently asked questions

What does Jev decide?

Which of nine precomputed movements to take, and whether the direct route is blocked.

Who handles collisions?

Python computes clearance for each candidate before Jev chooses.

Can I see what Jev saw?

Yes, the Inspect dialog shows the exact request and response.

Related projects

  • jev-askable-arm

    Simulated Franka arm that follows plain-English goals by having Jev chain about 30 hardcoded motion primitives, one per step.

    Agent skillprototypePython
  • jev-drone

    Camera-only MuJoCo quadrotor where Jev makes 2.5 Hz tactical judgments while code keeps flight control and a safety veto.

    ProjectprototypePython
  • jev-reflex-autonomy-lab

    Multi-drone simulation where Jev makes fast reflex decisions per drone and an optional slower LLM may advise but never takes control.

    AppprototypeTypeScript
  • DriveJev

    Open 4B System I driving model that scores behaviours like cruise, stop, yield and brake from three camera frames each decision.

    ProjectresearchPython
  • JEV Control Your Roboarm

    Experimental robot loop where Jev chooses the next verified action from perception-generated options for a language goal.

    ProjectresearchPython
  • Jev Robot

    MuJoCo PiPER arm controlled by a local decider-2b model choosing among ten skills via a Jev-style typed-choice interface while targets move.

    Agent skillresearchPython