The rapid adoption of blockchain technology, exemplified by Bitcoin and Ethereum, has intensified the need for high-performance systems. Existing optimization approaches primarily focus on two directions: simplifying consensus protocols and parallelizing consensus processes. Simplified protocols reduce consensus overhead but remain constrained by single-leader inefficiencies, particularly in large-scale networks where throughput degrades rapidly. Parallelization enhances throughput but frequently introduces complexity and tightly coupled architectures that lack dynamic adaptability, limiting adaptability. To address these limitations, we propose Streamline, a functionally decoupled multi-leader consensus protocol for permissioned blockchains that decouples transaction distribution from consensus processes. By integrating horizontal parallelism for concurrent consensus instances and vertical decoupling of data distribution, Streamline eliminates single-leader bottlenecks and supports dynamic modularity. Its layered design enables flexible workload adjustments, such as real-time load balancing under varying node resources, ensuring robust adaptability without compromising security. Compared to the latest consensus protocols, our approach demonstrates stable scalability and adaptability, incurring lower performance degradation as the network scales while maintaining more stable throughput and latency. Experimental results show that Streamline’s performance is comparable to state-of-the-art parallel consensus protocols and effectively delays network saturation; under skewed load, its throughput degrades by only 10%.