Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLM Agents for Automated Dependency Upgrades

Vali Tawosi, Salwa Alamir, Xiaomo Liu, Manuela Veloso

TL;DR

The paper tackles automatic Java dependency upgrades by framing the task as sequential bug-fixing guided by migration documentation. It introduces a multi-agent LLM framework (Summary Agent, Control Agent, Code Agent) built on Meta-RAG change localization within the ALMAS architecture, enabling scalable code localization and editing. Evaluated on three synthetic, industry-inspired repositories, the approach achieves high precision (notably $71.4\%$ in one upgrade) and substantially fewer tokens and steps than state-of-the-art baselines, demonstrating efficient automated maintenance. The results support the framework’s potential to reduce manual effort in library upgrades and encourage broader adoption, with future work spanning real-world datasets and enhanced hybrid methods for even stronger performance.

Abstract

As a codebase expands over time, its library dependencies can become outdated and require updates to maintain innovation and security. However, updating a library can introduce breaking changes in the code, necessitating significant developer time for maintenance. To address this, we introduce a framework of LLM agents to be used in combination with migration documentation to automatically recommend and apply code updates and ensure compatibility with new versions. Our solution can automatically localize updated library usages in live Java codebases and implement recommended fixes in a user-friendly manner. The system architecture consists of multiple key components: a Summary Agent, Control Agent, and Code Agent. To validate our approach, we apply the framework on an industrial use case by which we create three synthetic code repositories with major Upgrade changes and benchmark our approach against state-of-the-art methods. Results show that our approach not only performs upgrades using fewer tokens across all cases but also achieves a precision of 71.4%, highlighting its efficiency and effectiveness compared to state-of-the-art methods.

LLM Agents for Automated Dependency Upgrades

TL;DR

The paper tackles automatic Java dependency upgrades by framing the task as sequential bug-fixing guided by migration documentation. It introduces a multi-agent LLM framework (Summary Agent, Control Agent, Code Agent) built on Meta-RAG change localization within the ALMAS architecture, enabling scalable code localization and editing. Evaluated on three synthetic, industry-inspired repositories, the approach achieves high precision (notably in one upgrade) and substantially fewer tokens and steps than state-of-the-art baselines, demonstrating efficient automated maintenance. The results support the framework’s potential to reduce manual effort in library upgrades and encourage broader adoption, with future work spanning real-world datasets and enhanced hybrid methods for even stronger performance.

Abstract

As a codebase expands over time, its library dependencies can become outdated and require updates to maintain innovation and security. However, updating a library can introduce breaking changes in the code, necessitating significant developer time for maintenance. To address this, we introduce a framework of LLM agents to be used in combination with migration documentation to automatically recommend and apply code updates and ensure compatibility with new versions. Our solution can automatically localize updated library usages in live Java codebases and implement recommended fixes in a user-friendly manner. The system architecture consists of multiple key components: a Summary Agent, Control Agent, and Code Agent. To validate our approach, we apply the framework on an industrial use case by which we create three synthetic code repositories with major Upgrade changes and benchmark our approach against state-of-the-art methods. Results show that our approach not only performs upgrades using fewer tokens across all cases but also achieves a precision of 71.4%, highlighting its efficiency and effectiveness compared to state-of-the-art methods.

Paper Structure

This paper contains 13 sections, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Preprocessing
  5. Main process
  6. Control Agent
  7. Code Agent
  8. Process Workflow
  9. Upgrades Use Case
  10. Synthetic Dataset
  11. Results
  12. Discussion
  13. Conclusion

Figures (1)

  • Figure 1: Diagram of the multi-agent LLM framework for automated Java dependency upgrades, showing the process of code summarization, change localization, and implementation, with iterative compilation and testing, and handover to a human.