Jev DOOM Agent
A live Jev controller chooses movement, view, fire, and interaction controls for DOOM.
A live Jev controller chooses movement, view, fire, and interaction controls for DOOM.
The live controller sends structured Doom game state to Jev and requests four simultaneous Choice decisions: movement, view, trigger, and interaction. It validates each returned choice and maps the control frame to the game's action; incomplete or invalid answers are rejected.
Choose navigation for this tick.
Structured tactical state for the current Doom tick, including player condition, combat sensor, exploration, visible enemies, entities, linedefs, controls, and policy.
Choose where to look.
Structured tactical state for the current Doom tick, including player condition, combat sensor, exploration, visible enemies, entities, linedefs, controls, and policy.
Choose whether to fire.
Structured tactical state for the current Doom tick, including player condition, combat sensor, exploration, visible enemies, entities, linedefs, controls, and policy.
Choose whether to use a nearby line.
Structured tactical state for the current Doom tick, including player condition, combat sensor, exploration, visible enemies, entities, linedefs, controls, and policy.
The live decision path receives structured game state and an optional system policy.
Send Jev a Choice question for movement, view, trigger, and interaction in one request.
Require all four choices with confidence and probabilities, then translate the frame into the game's action.
It decomposes real-time game control into independent typed axes, then translates the validated frame into the game's action vocabulary in code.