This paper presents an attempt to capture individual and social characteristics of poly-drug use in a generic agent-based model called SimUse. We consider poly-drug use and its social context as a complex adaptive system. We use a generative framework to create an iterative dialogue between qualitative interviews, theoretical constructs and computer simulations. This iterative process allows us, through abductive reasoning, to progressively build an ontological representation of the different factors influencing poly-drug use patterns. The ontological structure includes three levels of influence: neurological effects of drugs, physiological and social responses of users and peer influences in a given societal context. First we introduce our generative framework; then, we describe the overall structure of the model before detailing important aspects of SimUse via UML diagrams: (1) the neurological engine and its physiological and behavioural consequences, (2) evolutions of social representations and instrumental uses associated to different substances by poly-drug users, and (3) social influences among peer networks. Finally, we conclude with an example of how SimUse reacts to external shocks (like drug depletion or disruption of the drug dealing market) and their consequences for public policies decision-makers.