The phase transition phenomenon in random constraint satisfaction problems (CSPs) is a pivotal topic at the intersection of computational complexity theory and theoretical computer science. We propose an extended random CSP model (RB \(^\textrm{mix}\) model) by introducing a hierarchical hybrid structure combining multi-length local constraints and all-different global constraints, systematically investigating the impact of multiscale heterogeneous constraints on computational hardness phase transitions in NP-complete problems under the theoretical framework of exact phase transitions in RB model. Theoretically, we rigorously prove the existence of a sharp satisfiability phase transition using first- and second-moment methods, deriving analytical expressions for critical threshold parameters. Numerical experiments demonstrate that the hybrid constraint mechanism preserves the sharp phase transition of RB model while significantly enhancing theoretical inclusiveness. Moreover, all-different global constraint integration maintains critical phase transition characteristics. These results establish a universal framework for phase transition analysis under heterogeneous constraints, providing new theoretical tools for understanding computational hardness boundaries of NP-complete problems, optimizing constraint design, and generating algorithm testing benchmarks.