taglo

TwinLab challenges premature diagnoses when a fault disappears before an investigator arrives. A normal reading under low demand may not explain a complaint that occurred under high demand. In a condition-coverage scenario, two simulated explanations remain plausible after an invented low-demand reading; a second high-demand reading changes their compatibility under explicit assumptions. A handoff separates the reported symptom, observations, alternatives, and missing conditions. Separately, a simulated one-zone room lets users run checks, update probabilities across four modeled states, and see which check is recommended by an information-gain calculation. A sensitivity view shows how broader error assumptions can leave multiple explanations compatible. The application uses Python for simulation and inference, and an HTML/JavaScript interface. An optional Meta model integration can select from permitted simulated checks; its output is validated and falls back to the local policy. The scenario, measurements, costs, and tolerances are illustrative, not a validated diagnosis or real equipment control. Public building data shown for context does not validate the simulator.