With the advent of the era of LLMs, the demand for exploring multimodal data continues to grow. Queries over such data often involve hybrid workloads that combine structured attribute filtering with high-dimensional semantic matching. Efficient index advising for such hybrid query workloads is crucial for improving query performance. However, existing index advisors typically optimize individual types of queries in isolation and lack the capability to recommend scalar and vector indexes simultaneously for hybrid workloads. To address this limitation, this paper proposes HQIA, a hybrid query index advisor based on RIBE. Within this framework, we design a feature representation method tailored for hybrid queries, which integrates query cost and attribute distribution information. By incorporating physical query plans with histogram features, the method gains the ability to represent the predicate coverage of queries, thereby improving the distinguishability among different queries and ultimately enhancing workload compression and clustering. Building upon this foundation, we extend RIBE’s index advising mechanism to support hybrid queries. Experimental results show that HQIA achieves superior tuning effectiveness and significant performance improvements over existing methods in hybrid query scenarios.