How to detect a real-time signal's local maxima and minima?

Most toolkits require all data before detection of maxima and minima happens. What if I have a real-time signal "feed" and I need a detector that measures whether the gradient slightly behind the data point at current time? So this is a gradient checking function to determine if the next point belongs to a local maxima or minima region.

Does anyone know a function or toolkit that is capable of doing this?

[Adam Danz]

Detecting an instantaneous maxima (or minima) only depends on whether the most recent sample is higher or lower than the other samples within a lagged time window. For example, if the data were to monotonically increase, every sample would be a local maxima. Maybe a visual inspection of the data would be helpful to suggest ideas.