In today's service computing environments, user needs and expectations are constantly changing. New services emerge while old ones become obsolete and need to be replaced. In such settings, composite services need to be adaptive to changes in user requirements and the environment. This paper proposes a conceptual framework for modeling compositional adaptation for services founded on a requirements monitoring facility. This facility helps maintain adherence between user requirements changes and the dynamics of service composition structure and quality attributes. Specifically, user requirements are represented as goals and soft goals, service composition structure is represented with a CSP-like grammar, and the adaptation mechanism is based on AI planning. The proposed approach is evaluated in a service simulation environment of real-world supply-chain adaptation scenarios.