TableVault: Managing Dynamic Data Collections for LLM-Augmented Workflows

2025-06-22

AI Summary

Managing LLM-augmented data workflows introduces critical challenges—including concurrency conflicts, irreproducibility, version chaos, and fragmentation across heterogeneous data sources. To address these, this paper proposes the first dynamic data management platform specifically designed for LLM-enhanced workflows. The platform unifies database-style versioning, fine-grained concurrent transaction control, and declarative workflow orchestration to jointly manage structured data and LLM-generated artifacts—enabling consistent storage, provenance tracking, versioning, and composable reuse. Its key innovation is a transparent, dynamic coordination layer that preserves ACID guarantees while accommodating the inherent non-determinism of LLM outputs. Experimental evaluation demonstrates significant improvements in traceability, reproducibility, and execution reliability under multi-user workloads. The platform establishes a foundational infrastructure for AI-native data governance, bridging the gap between traditional data management principles and the operational realities of generative AI systems.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Natural Language Processing: (Large) Language Models

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Large Language Models (LLMs) have emerged as powerful tools for automating and executing complex data tasks. However, their integration into more complex data workflows introduces significant management challenges. In response, we present TableVault - a data management system designed to handle dynamic data collections in LLM-augmented environments. TableVault meets the demands of these workflows by supporting concurrent execution, ensuring reproducibility, maintaining robust data versioning, and enabling composable workflow design. By merging established database methodologies with emerging LLM-driven requirements, TableVault offers a transparent platform that efficiently manages both structured data and associated data artifacts.

Problem

• Managing dynamic data collections for LLM workflows
• Ensuring reproducibility and robust data versioning
• Integrating database methods with LLM-driven requirements

Innovation

• Supports concurrent execution for workflows
• Ensures reproducibility and data versioning
• Merges database methods with LLM needs