Edge microservice deployment encounters three primary challenges: inter-service dependency constraints that impede optimal resource provisioning, mobility-driven service quality degradation, and geographically heterogeneous load distribution patterns. To address these issues, we propose a GraphSAGE-enhanced reinforcement learning framework for microservice deployment optimization, aimed at reducing deployment response time and balancing edge server loads. The framework leverages GraphSAGE’s inductive feature aggregation capabilities to efficiently extract and encode topological characteristics of microservice dependencies and edge server interconnections. Subsequently, deployment decisions are optimized through a Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm incorporating dual Critic networks to mitigate Q-value overestimation. Experimental results demonstrate that, compared to existing methods such as DQN, SAC, and TRPO, our proposed approach achieves significant improvements in average response time and load balancing, reducing average response time by up to 28.10% and improving load balancing by 68.66%. Notably, the proposed method exhibits excellent performance stability under resource-constrained conditions and large-scale application scenarios.