LIGHT implements a Bayesian framework for reconstructing the three-dimensional galaxy and underlying dark matter fields informed by large-scale structure. The software uses Hamiltonian Monte Carlo (HMC) sampling via NumPyro (ascl:2505.005) to infer the true galaxy density distribution and its magnitude distribution given incomplete galaxy catalog data, addressing biases in statistical analyses that combine galaxy catalogs with other observations. LIGHT’s models incorporate spatial correlations of galaxies on cosmological scales and enable construction of improved three-dimensional priors in redshift and sky position for host galaxies of gravitational-wave events. LIGHT supports dark siren cosmology, where gravitational-wave distances are combined with reconstructed galaxy fields to improve inference of cosmological parameters, and includes example configurations and scripts.