In this paper we present ColibriCook: a CBR system for ontology-based cooking recipe retrieval and adaptation. The system's purpose is to participate in the 1st Computer Cooking Contest, organized by the European Conference on Case-Based Reasoning (ECCBR'08), at the University of Trier, Germany. CBR is based on a best-adaptation likeness paradigm between ingredient sets, with a domain ontology providing one-on-one fuzzy ingredient similarity. A number of other machine learning techniques are used to calculate, propagate, compare and adapt other recipe properties.