CADENZA in Action: Breaking the Monolith with Intent-Dependent Plan Spaces for Semantic Queries
For database systems processing semantic queries, CADENZA provides a novel optimization approach that breaks the monolithic model paradigm, enabling flexible trade-offs between accuracy and efficiency.
CADENZA optimizes semantic queries by decomposing intents into steps and selecting physical implementations under user preferences, achieving better quality-latency-cost trade-offs than monolithic model calls.
Semantic query processing engines execute semantic operators, whose behavior is specified by natural-language intents, via model inference over multimodal data. Most existing optimizers optimize the operators at the granularity of monolithic implementations -- such as LLMs and embedding models -- forcing a trade-off between expensive model calls and cheaper alternatives that fail to capture intent-dependent semantics. We present CADENZA, a semantic operator optimizer that compiles an intent into decomposed steps, selects concrete physical implementations for each step, and tunes their parameters under user-specified quality-latency-cost preferences. In this demonstration, users interact with CADENZA through a web interface over multimodal databases, exploring how an intent is decomposed into alternative plans, how each plan is optimized, and how different preferences yield different winning plans.