This paper presents a system for automatically configuring the fastest approximate nearest-neighbor algorithm for a given dataset and desired precision. The system evaluates and selects the best algorithm and parameter values through a cross-validation approach, aiming to minimize the predicted search cost. Two primary algorithms are compared: the hierarchical k-means tree and multiple randomized kd-trees. The hierarchical k-means tree is constructed by recursively clustering data points using k-means, while the randomized kd-trees use a priority queue to explore the tree in order of distance to the query point. The paper also introduces an optimized version of the kd-tree algorithm that uses multiple randomized trees. Experiments on various datasets show that these algorithms can achieve significant speedups over linear search, with the best performance depending on the dataset and required precision. The system is designed to be user-friendly, requiring minimal input, and provides a public domain library for further research and comparison.This paper presents a system for automatically configuring the fastest approximate nearest-neighbor algorithm for a given dataset and desired precision. The system evaluates and selects the best algorithm and parameter values through a cross-validation approach, aiming to minimize the predicted search cost. Two primary algorithms are compared: the hierarchical k-means tree and multiple randomized kd-trees. The hierarchical k-means tree is constructed by recursively clustering data points using k-means, while the randomized kd-trees use a priority queue to explore the tree in order of distance to the query point. The paper also introduces an optimized version of the kd-tree algorithm that uses multiple randomized trees. Experiments on various datasets show that these algorithms can achieve significant speedups over linear search, with the best performance depending on the dataset and required precision. The system is designed to be user-friendly, requiring minimal input, and provides a public domain library for further research and comparison.