JEVia.

JEVia is a local agent experiment. Choose a workspace, configure your model key, and give it a task. Jev helps make decisions before and during execution: whether to use the network, whether there is enough context, and whether a file change should proceed.

The model returns structured scores; the program applies thresholds. Language models write the content, while tool access and path checks remain under program control. The decision panel shows the question, answer, and threshold for each step.