Jev AgentA source-verified directory of 487 Jev projects — browser agents, model routers, SDKs and MCP servers — each one showing the exact decision it hands to Jev. Curated by the Awesome Jev community, mirrored here under MIT.
Give Jev the current state and a few options. Get the Agent’s next step.
Core call · setup in the full example
const nextStep = choice("What next?", { login: null, askUser: null, retry: null });
const { answers } = await client.systemOne({ state: "Login form is ready", questions: { nextStep } });
console.log(answers.nextStep.choice);Node.js 20+ · Set TYPESAFE_API_KEY on your server. This page shows code and makes no model requests.
Tiny Jev-like decision head built atop Qwen2.5-0.5B that can be trained and run locally on Apple Silicon MacBooks.
Attaches a parallel decision head to an open 0.5B model to answer discrete questions directly from token activations.
Runs entirely offline on edge hardware with minimal memory footprint, offering an accessible base for custom judgment modeling.
Nano replica of Jev featuring parallel decisions, dynamic candidate sets, and an end-to-end training and evaluation pipeline.
Evaluates multiple questions and dynamic candidate spaces concurrently in a single forward pass, logging navigation choices.
Provides open-source training infrastructure and game environment benchmarks for investigating non-generative discrete decisions.
An experimental NES Mario controller that gives Jev structured emulator RAM and telemetry instead of screenshots.
Reads motion, enemies, terrain and recent controls, then selects a predefined legal action.
Records model state and actions without screenshot input.
A browser driving simulator where Jev chooses among locally generated paths and speeds.
Jev reads road and traffic state and selects motion; geometry, collision prediction and braking remain local.
Makes candidate trajectories and decision probabilities visible during a simulation.
A MuJoCo drone experiment deriving scene features from camera buffers for Jev tactical advice.
Uses range sectors, obstacle heights and target state to choose maneuvers and assess risk or target loss.
Separates tactical judgments from local guidance, reflexes and flight control.
Two LIBERO tasks, one control engine. Each demo loads its own JSON task definition. Videos follow simulation time, with decision and physics-preview waiting omitted.
Jev returns a structured decision for the local program; consult the source for the exact decision policy.
Adds structured choices or scores to the workflow; performance and cost benefits have not been independently verified.
Open reproduction of Jev decision layer on open-weights models, serving the System One schema by inspecting option logits directly.
Implements the /v1/systemone endpoint (Choice, Score, Noul) by calculating candidate token log-probabilities without text generation.
Enables private and local hosting of Jev-compatible decision services on GPUs without relying on proprietary cloud APIs.
RoboJEV is a small, inspectable robotics laboratory. JEV receives **structured simulator state, not images**, selects an immediate intent, then selects X/Y/Z directions and a gripper command. A Cartesian controller executes the action using real MuJoCo contacts. Each task has independent physical success checks; model answers cannot declare success.
Jev returns a structured decision for the local program; consult the source for the exact decision policy.
Adds structured choices or scores to the workflow; performance and cost benefits have not been independently verified.
A Jev harness for StarCraft shareware’s Strongarm mission, pausing the game during state reads and inference.
Selects commands from structured game state and executes them with mouse and keyboard input.
The author provides a victory recording and verification report; this is a bounded mission experiment, not a real-time competitive benchmark.
Snake autoplayer driven by TypeSafe Jev, executing one discrete System One decision per tick with code-enforced legal moves.
Supplies board state, fruit location, and legal direction choices to Jev, selecting the optimal move direction each tick.
Demonstrates fast discrete choice in constrained environments while deterministic logic prevents illegal collisions.
A browser-based 1v1 shooter where Jev reads structured match state and chooses movement, aim, and firing.
Each decision tick asks about movement, view, aiming, firing, and jumping, with a heuristic fallback on API failure.
Makes structured decisions visible in an interactive game loop.
Laya (open source, local) vs TypeSafe Jev (API): two AI models race in Snake and fight in a Mortal-Kombat-style arena. Every move is a real model decision.
Jev returns a structured decision for the local program; consult the source for the exact decision policy.
Adds structured choices or scores to the workflow; performance and cost benefits have not been independently verified.
A browser-based top-down driving simulator using Jev for lane and speed choices, with an optional chat-model comparison.
Jev answers lane, speed, hazard and pedestrian-yield questions; local rules apply the actions.
Runs controllers on the same seeded course, with separate local emergency-braking logic.
Multi-drone autonomy lab demonstrating TypeSafe Jev reflex decisions with optional System 2 strategy guidance.
Jev returns a structured decision for the local program; consult the source for the exact decision policy.
Adds structured choices or scores to the workflow; performance and cost benefits have not been independently verified.
Uses Jev to chain predefined skills for English-language goals in a ManiSkill robot-arm simulation.
Chooses a skill and target from privileged object and gripper state and about thirty primitives.
Separates skill selection from Python motor control.
A RuneBench extension using Jev and a bounded rs-sdk action catalog for RuneScape tasks.
Chooses goal actions, tick interventions and the next polling interval from game state.
Provides action-distribution views, run logs and burst or tick control modes.
Observable browser stealth game where Jev makes typed guard judgments while deterministic code owns the physics world.
Queries Jev during gameplay ticks or events to determine guard alert states, feeding discrete decisions back into the game loop.
Demonstrates using Jev for game AI behavior trees, achieving dynamic situational awareness within deterministic constraints.
A browser Doom experiment comparing Jev-controlled players from the same initial state.
Jev reads structured health, ammunition and targets and selects tactical controls executed by a local controller.
Exposes model decisions alongside explicitly labeled offline and fallback policies.
This project evaluates a fixed model. It records rewards and decisions without training or updating model weights. A seeded random policy provides a local baseline.
Jev returns a structured decision for the local program; consult the source for the exact decision policy.
Adds structured choices or scores to the workflow; performance and cost benefits have not been independently verified.
An island-airport simulator using Jev for routes, yielding, emergency broadcasts and landing order.
Turns aircraft and nearby traffic state into questions whose answers drive simulated actions.
Shows requests, answers and simulation state, with a separate mock path available.
A Python Flappy Bird game where code simulates candidate routes and Jev picks one.
For each new pipe, Jev selects a route and the game executes its scheduled flaps.
Visualizes candidate trajectories, probabilities and the selected path.
TypeSafe's Jev model plays Vampire Survivors on Steam: BepInEx plugin + Python brain + live decision dashboard. Native Linux only.
Jev returns a structured decision for the local program; consult the source for the exact decision policy.
Adds structured choices or scores to the workflow; performance and cost benefits have not been independently verified.
A ViZDoom Agent that uses Jev to choose movement, targets and firing from structured game state.
Jev selects tactical goals and actions while local geometry handles fine aiming.
Separates game ticks from network inference so the engine need not wait on every frame.
A MuJoCo arm demo where local code proposes candidate moves and Jev chooses the target, grasp or release, and completion.
Chooses among candidate targets, grasp or release, and a done Noul.
Candidate motions are generated in a local physics copy; Jev only chooses. Author-reported timings were not retested here.
A community-maintained directory and radar for Jev, highlighting open-source projects with verified code and clear decision architectures.
Jev is TypeSafe's decision model. This directory groups projects by how they use choices, scores and probability judgements, with implementation links to help you assess fit for your own task.
TypeSafe ↗We prioritize open-source projects with clear Jev implementation evidence. Repositories without verifiable source code are excluded from the main directory.
Directory data and design from Awesome Jev by logicrw, used under the MIT License.
