openjev

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

OpenJev: open-source, local alternative to Jev. System-One decision models: typed questions in, calibrated probabilities out; general decisions and browser-agent steps.

🚀🚀 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.

Native Swift implementation of OpenJev: Jev-compatible typed decisions (yes/no, choice, score) read from model probabilities, on Apple silicon Macs and in iPhone apps. Apache-2.0.

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.

Core ML model packages for typed AI decisions with OpenJevSwift on iOS and macOS (iPhone and Mac).

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.

Claude Code plugin: OpenJev (Codiv) makes the small typed decisions: rule-enforcing edit hook, skill picker, file discovery, browser navigation, review pre-check. Shadow mode by default.

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

An open-source, model-agnostic decision router for Claude Code and Codex.

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

Free English RAG benchmark for TypeSafe Jev 1.13 — frozen candidate pools, calibration, paired bootstrap CIs, $0 runs. Compared with OpenJev and NVIDIA cross-encoders.

Openjev experiments

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

Selectia - A family of System One-style models fine-tuned from LFM 2.5, designed for one-pass typed decisions with calibrated probabilities. Decision API (Choice, Noul, Score) powered by open LLMs.

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.