jevstation

JevStation runs **Jev**, a System One model: you ask typed questions about some content and get probabilities back, never prose. **Tools** - `jev_decide` – ask your own Choice / Score / Noul questions about a state - `jev_run_question_set` / `jev_list_question_sets` – reuse question sets saved in the playground - `jev_route_model` – decide whether a prompt needs a small, medium or frontier model - `jev_guard_tool_call` – check an agent tool call (execute / confirm / block) before running it - `