openjev

Calibrated 151M Non-Autoregressive Decision Engine beating TypeSafe Jev & Laya on LocalLLaMA/typed-decisions (77.10% acc, 0.0636 Brier, 0.0144 ECE)

🚀🚀 A 0.8B JEV-like multimodal model playing GUI games directly from raw pixels.

Super fast Jev like server, model agnostic, working with any OpenAI compatible endpoint

OpenJev: an independent Jev-inspired System One decision API based on TypeSafe.ai concepts. Choice, score and noul primitives, local mock server, Python and TypeScript SDKs. Real inference planned; not affiliated with TypeSafe AI.

Local multimodal decisions on your Mac. Jev-compatible typed probabilities with Qwen, llama.cpp and Metal.

Local multimodal decisions on your Mac. Jev-compatible typed probabilities with Qwen, llama.cpp and Metal.

Openjev experiments

MCP server exposing TypeSafe's Jev through the public OpenJEV API. Send one shared context and several independent questions; get typed judgments and probabilities back — choice, score, or yes/no — not prose. Built for DeepSeek Harness; runs unchanged in any MCP client.

Sub-10ms calibrated semantic decision engine & runtime for edge, embodied control, agent simulation, and typed dynamic workflows.

A .NET client library for the [TypeSafe AI](https://typesafe.ai) System One API (`https://api.typesafe.ai`).

Add as much tabular data as you want to Jev models using adaptive ensembles that learn to query only the rows and columns needed.

Random numbers subject to Jev approval. Python CLI and library.

Jev (TypeSafe System One) operating a web page through its DOM — no WebMCP required

Typed decisions for AI agents as an MCP tool: yes/no probability (noul), choice, and score in one fast request. Backed by the TypeSafe System One model.

Rank Pi Agent Skills for the current task with TypeSafe Jev

High-throughput synthetic and pretraining dataset sifter for TypeSafe Jev. Rust streaming core, Parquet and JSONL I/O, typed Choice/Score/Noul judgments, speculative fan-out, 24.0 rows/sec measured.

Jev-powered context compaction plugin for opencode: prune stale tool calls and results before every request, keep the rest verbatim.

IntentFlags — semantic feature flags for websites, powered by Jev. Infers what a visitor is trying to do and hands it to your React code as a flag.

Semantic "Did you mean?" for any CLI — wraps commands and uses TypeSafe's Jev to match subcommand typos by intent, not edit distance.