Declarative Data Services: Structured Agentic Discovery for Composing Data Systems

Shanshan Ye, Duo Lu

arXiv:2605.20690 · 2026-05-22 공개 · arXiv · PDF

agentic-discovery data-systems llm-search composition-knowledge runtime-attribution skill-patching operator-dag intent-decomposition

Abstract

Agentic discovery has shown that LLM-driven search can find novel algorithms, designs, and code under benchmark conditions. Translating the paradigm to multi-system data backends surfaces a harder problem: the search space is heterogeneous, the verifier is whether a deployed stack actually runs, and composition knowledge is unevenly captured in pretraining. Unbounded agentic discovery, a coding agent iterating on failure-log feedback, fails to converge consistently on a working stack even when iteration and explicit composition knowledge are added. We propose Declarative Data Services (DDS), an architecture for structured agentic discovery of data-system compositions from declarative user intent. The framework owns four typed contracts at successive layers (intent, operator DAG, per-system skills, runtime attribution) that decompose the global search into bounded sub-searches; sub-agents search each typed space, while the framework provides the channels by which knowledge flows forward as inline skill citations and errors route backward as typed signals. As a proof of life on a trading-backend workload, DDS converges where unbounded discovery does not; runtime failures become skill patches that the next deployment cites inline. We position this as an early prototype reporting lessons from real-world data-system composition.

한국어 요약

한 줄 요약

**[Agentic Discovery / 데이터 시스템]** DDS는 declarative intent에서 멀티시스템 데이터 백엔드 조합을 4 typed contract(intent·operator DAG·system skill·runtime attribution) 계층으로 분해, unbounded agentic discovery가 수렴 못 하는 trading 백엔드에서 수렴·실패가 inline skill patch가 됨.

핵심 기여도

핵심 아이디어

멀티시스템 데이터 백엔드의 agentic discovery는 unbounded 코딩 에이전트로 수렴 불가능하며, intent→operator DAG→per-system skill→runtime attribution의 4 typed contract로 탐색을 계층적으로 분해하고 inline skill citation·typed signal로 지식·오류를 routing해야 수렴 가능하다.

기술적 접근법

주요 결과

의의 및 한계

**의의**: Agentic discovery를 벤치마크 너머 real-world 멀티시스템으로 확장, 4-layer typed contract라는 architectural 원리 제시, 실패가 학습으로 전환되는 closed-loop, declarative interface의 사용자 친화성. **한계**: 초기 prototype·trading 백엔드 단일 사례, typed contract 정의 자체의 도메인 의존성, sub-agent 학습·관리 cost.

실용적 활용