Medication advice is often represented digitally as an isolated recommendation, alert, or relation between a medicine and an action. Such representations may omit why the advice exists, when it applies, which patient or setting qualifies it, and which source or evidence supports it. This article proposes an assertion-centred pharmacy knowledge-graph architecture for representing medication advice as a context-bearing, temporally bounded, provenance-linked knowledge object. The central construct, the Medication Advice Assertion, connects medication identity, action, rationale, intended clinical goal, applicability conditions, temporal scope, source version, evidence item, evidence-strength assessment, contradiction state, responsible agent, and intended-use boundary. The architecture separates extracted associations from normative advice, evidence strength from model confidence, provenance from truth, and technical performance from clinical usefulness. It also proposes qualified relations among assertions, including support, refinement, exception, contradiction, and supersession, so that disagreement is preserved rather than silently overwritten. Evaluation is organized as a sequence of semantic-conformance, identity-resolution, provenance, temporal, applicability, contradiction, technical, workflow, safety, equity, and governance checks. Clinical use would additionally require pharmacist and stakeholder assessment, external validation, human-factors testing, monitoring, and controlled maintenance. The proposed architecture does not establish the correctness of any medication recommendation, clinical benefit, regulatory acceptance, or deployment readiness. Its original contribution is a representational and governance framework for making the why, when, and source of medication advice independently inspectable while preserving professional authority and uncertainty.