Developers introduced the OpenJev tool for local work with models

Developers have released OpenJev — a tool for local execution of the Jev interface pattern from TypeSafe. The solution allows using compact language models, such as Qwen3.5-4B, to process structured data without accessing cloud APIs. The main task of the project is to implement typed scoring of options. According to the authors, the OpenJev architecture provides work 5 times faster compared to generative baseline solutions using standard JSON output. The tool is optimized for local execution, which reduces delays in processing requests. The functionality of OpenJev includes support for parallelism with a shared state (shared-state parallelism). This allows for efficient distribution of computational load when working with multiple data streams. The project's source code is published on GitHub and is available for integration into local data processing pipelines. Using local models with typed interfaces allows developers to create predictable automation systems. Unlike standard LLM agents focused on free text generation, OpenJev focuses on strict adherence to the structure of responses, which is critical for integrating AI into software interfaces and application business logic.