extract pdf tables · column detection by text positions · per-page csv or merged · runs locally
higher tolerance = sloppier rows/cols · raise if your table is being split too aggressively