We introduce compositional symbolic execution à la carte, a modular approach for building compositional symbolic execution (CSE) tools, powered by combinator-based design of key components, including memory models, management of user-defined predicates, and bi-abductive analysis. Our approach is grounded on meta-theory mechanised in the Rocq proof assistant and allows tool developers to easily build correct-by-construction CSE tools for symbolic analysis of their programming languages of choice, with minimal implementation and proof effort. We show how our approach scales to real-world languages by implementing it in Gillian, a state-of-the-art CSE tool, and using it to capture the intricacies of the memory models of C, JavaScript, and Rust. We evaluate our approach by running all of the relevant previous case studies of Gillian, showing no unexpected behaviours and comparable performance.