<p>Understanding soil nutrient dynamic mechanisms and vegetation-soil feedbacks in karst bauxite ecosystems requires comprehensive analysis of restoration chronosequences. This study conducted an integrated assessment of soil-vegetation system evolution in a karst-type bauxite mining area in Zunyi, Guizhou Province, China. Combining field surveys with laboratory analyses, we employed multivariate approaches including random forest modeling, structural equation modeling (SEM), and resilience quantification to elucidate restoration trajectories. Four principal findings emerged: First, open-pit mining induced severe soil degradation, reducing the Integrated Fertility Index (IFI) to low levels (&lt; 0.4). Vermicompost amendment facilitated remarkable soil recovery within five years, achieving high fertility status (0.79). All measured nutrient parameters consistently exceeded Grade II thresholds during this remediation period. Second, total nitrogen (TN) emerged as the predominant influencing factor of soil nutrient dynamics (<i>β</i> = 0.95, <i>p</i> &lt; 0.001), with SEM revealing significant direct effects of restoration duration on pH, and total phosphorus (TP), which subsequently mediated organic matter accumulation (SOM) and nutrient availability (AN). Third, critical phase transitions occurred in system interactions: soil nutrient conservation capacity (SNC) and soil water retention capacity (SWC) shifted from trade-off (Year 1–3) to synergistic relationships (Year 3–5). Fourth, temporal succession of key determinants was observed: TN, soil moisture (SMC), and clay fraction (SC) dominated early restoration (Year 1), transitioning to AN, available potassium (AK), and cation exchange capacity (CEC) at mid-stage (Year 3), with SOM becoming predominant in mature systems (Year 5, <i>p</i> &lt; 0.05). These results establish a chronosequence framework for mining ecosystem restoration, providing mechanistic insights for targeted ecological management in karst regions.</p>