HestonModel

Heston stochastic volatility as a composable step block (Seam 5):

dS = r S dt + sqrt(v) S dW1
dv = kappa (theta - v) dt + xi sqrt(v) dW2 ,   corr(dW1, dW2) = rho

Full-truncation Euler: the drift and diffusion of v use max(v, 0) while the state may drift slightly negative, which keeps the scheme bias small. Every parameter — including rho — is a differentiable input, so one adjoint sweep returns the price plus its sensitivity to v0, kappa, theta, xi and rho next to the spot / rate Greeks. The max floor sits on a measure-zero set, so the variance-parameter adjoints match a finite bump only approximately (a few percent); the spot / rate / strike / rho adjoints are exact to Monte-Carlo noise.