OpenJev: An open-source, Jev-compatible System One decision engine (github.com)

OpenJev has launched as an open-source, Jev-compatible decision engine featuring a Claude Code-style REPL (Read-Eval-Print Loop). Designed for fast, typed decision-making using open models, OpenJev eliminates the need for JSON parsing and generates results with confidence probabilities in a single forward pass. It reproduces Jev's interface pattern while being distinct from Jev’s proprietary model. This initiative is significant for the AI/ML community as it democratizes access to decision-making frameworks that operate quickly and efficiently, thus enhancing the development of intelligent agents. Technically, OpenJev introduces three core decision primitives: Choice, Score, and Noul. Each decision can be made using real-time state inputs, asking up to 255 options, and delivering probabilities and confidence in the responses. The system supports various backends, including Hugging Face Transformers, and emphasizes performance benchmarks, latency, and confidence metrics through a streamlined interface. By enabling developers to visualize and experiment with decision processes similarly to how Clawd-Code structures Python interactions, OpenJev aims to foster innovation in AI decision systems and simplifies integrating AI capabilities into applications. The ongoing development phases promise enhancements and further capabilities, ensuring that OpenJev remains adaptable for future advancements.