Causa

CAUSA is an evidence-first diagnostic agent for physical systems. It investigates no-flow faults in a tabletop water rig (reservoir → pump → main/bypass branches → outlet) by guiding structured experiments where the operator must commit a prediction before observing each test result. The Problem: When physical equipment fails, technicians guess and swap parts. There is no enforced discipline to predict before testing, no audit trail of reasoning, and no measured verification that a repair actually worked. What CAUSA Does: 1. Selects the most informative diagnostic test using entropy-based reasoning over remaining hypotheses. 2. Forces the operator to commit a written prediction before any observation is revealed — preventing hindsight bias and building a transparent evidence trail. 3. Uses Meta Muse Spark 1.3 to propose the next best physical test based on reported evidence, operator notes, and available procedures. The operator accepts or declines the proposal. 4. Uses Meta Muse Spark 1.3 and SAM 3.1 to identify and segment physical components (valves, pump, tubing) directly from camera frames or uploaded photographs. 5. Requires a functional verification check — measured flow output over a minimum interval — before closing any case. Every prediction, observation, model proposal, perception result, and operator review is recorded in a structured, timestamped, exportable evidence trail. The model proposes; the operator decides. Diagnosis alone does not close a case — measured restored operation does. Current scope: four single-fault causes in a reviewed tabletop water topology. This is not a general-purpose machine diagnostic tool.