Does the slower planner take over steering directly?

It does not fly the drone. When confidence falls below threshold it sends one-use guidance, Jev keeps steering while waiting, and only decisions that actually consume returned guidance are marked differently.

How do I keep experiment spend down on large fleets?

Start with one drone and the planner off, since larger fleets multiply decisions and planner calls cost more than Jev calls. A lower-cost planner model is supported where availability allows.

Can I try the simulation without provider credentials?

Use the local controller toggle to verify the app with rule-based decisions and no credentials. Switch to Live API for Jev plus OpenRouter planning, choosing direct TypeSafe routing when you hold that key.

How do I read the confidence and telemetry charts?

Green marks a standalone Jev decision, purple marks a Jev decision that consumed new guidance, bars mark planner arrival or failure, and exports include seed, probabilities, timing, and outcomes.

How do I run the autonomy lab locally?

It needs Node.js 22.13 or newer. Run npm install then npm run dev and open the printed local URL; the app starts with development mocks, so it works without provider credentials.

What did one large run cost?

In one successful 15-drone run, Jev made 201 decisions across roughly 5.65 million tokens for about $0.24, approximately $0.0012 per decision. Turning off System 2 is the main way to keep experiments cheap.

Is a seeded scenario a deterministic replay?

No. The seed controls the obstacle layout, but live provider latency can still change a trajectory, so a seed is repeatable geometry rather than a deterministic asynchronous replay.

CommunityJev Demos & Experiments

Jev Reflex Autonomy Lab

Browser drone-fleet lab where TypeSafe Jev steers each drone as fast reflexes and an optional planner offers one-use strategy advice when confidence drops, without pausing control.

See highlights

Open github.com

Stars: 23
Language: TypeScript
License: MIT
Published by: Community
Author: khordoo
Last verified: Oct 2026

khordoo/jev-reflex-autonomy-lab

User repository on GitHub

View on GitHub

Real-time drone swarm autonomy simulation using Jev for fast System 1 reflex decisions and collision avoidance, with optional System 2 reasoning for strategic guidance

Forks: 3
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

An interactive multi-drone autonomy simulation powered by TypeSafe Jev, exploring a simple question: what happens when fast, typed System 1 reflexes can ask a slower System 2 reasoning model for advice without giving up control? The fleet size is configurable from one to fifteen drones. System 2 can be turned off entirely, making it easy to compare Jev operating independently against Jev augmented with strategic guidance.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Controls one to fifteen drones independently with per-drone decisions, confidence, latency, and collision outcomes.
Keeps Jev steering during planner latency and treats asteroids, debris, objects, and drones as sensed contacts.
Runs locally with mocks by default and switches to live Jev plus OpenRouter planning when keys are configured.
Exports full mission JSON with observations, probabilities, revisions, failures, and final outcomes for analysis.

Watch out

The flight environment stays simulated, seeds fix geometry rather than async replay, and live success varies with latency and model choices. 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.

More Jev resources like this

laya-server

Self-hosted Docker service and web UI that packages Laya structured-decision inference behind a compatible System One endpoint.