We study graphical house allocation, where agents are connected by a friendship graph and envy can arise only between neighboring agents. Each agent approves a subset of houses and envies a friend if she receives no approved house while the friend is allocated one she approves. We consider two problems: minimizing the number of envious agents, and, among such allocations, maximizing the number of agents receiving an approved house. We provide a detailed complexity analysis, showing that both problems are solvable in polynomial time when each agent approves at most one house, but become \NPH when agents may approve two houses, establishing a tight tractability boundary. We further present exact algorithms under structural restrictions on the agent graph, including sparsity, small balanced separators, and bounded vertex cover.