RepoTransAgent: Multi-Agent LLM Framework for Repository-Aware Code Translation

2025-08-25

AI Summary

Existing code translation methods suffer from three key limitations in real-world codebases: weak contextual understanding, inflexible prompt design, and insufficient error correction capability. To address these challenges, this paper proposes a repository-level multi-agent code translation framework that decomposes the task into three synergistic submodules: context-aware retrieval, dynamic prompt construction, and iterative refinement. We introduce a novel reflection-based error correction mechanism that integrates retrieval-augmented generation (RAG), adaptive prompt engineering, and multi-agent collaboration to enable semantic-aware end-to-end translation. Empirical evaluation across six open-source projects for Java-to-C# translation demonstrates that our approach achieves a compilation success rate of 55.34% and a test pass rate of 45.84%, significantly outperforming state-of-the-art methods.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Multiagent Systems: Agent Communication
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Repository-aware code translation is critical for modernizing legacy systems, enhancing maintainability, and enabling interoperability across diverse programming languages. While recent advances in large language models (LLMs) have improved code translation quality, existing approaches face significant challenges in practical scenarios: insufficient contextual understanding, inflexible prompt designs, and inadequate error correction mechanisms. These limitations severely hinder accurate and efficient translation of complex, real-world code repositories. To address these challenges, we propose RepoTransAgent, a novel multi-agent LLM framework for repository-aware code translation. RepoTransAgent systematically decomposes the translation process into specialized subtasks-context retrieval, dynamic prompt construction, and iterative code refinement-each handled by dedicated agents. Our approach leverages retrieval-augmented generation (RAG) for contextual information gathering, employs adaptive prompts tailored to varying repository scenarios, and introduces a reflection-based mechanism for systematic error correction. We evaluate RepoTransAgent on hundreds of Java-C# translation pairs from six popular open-source projects. Experimental results demonstrate that RepoTransAgent significantly outperforms state-of-the-art baselines in both compile and pass rates. Specifically, RepoTransAgent achieves up to 55.34% compile rate and 45.84% pass rate. Comprehensive analysis confirms the robustness and generalizability of RepoTransAgent across different LLMs, establishing its effectiveness for real-world repository-aware code translation.

Problem

• Addressing insufficient contextual understanding in code translation
• Overcoming inflexible prompt designs for repository scenarios
• Improving error correction mechanisms for accurate translation

Innovation

• Multi-agent LLM framework for repository-aware translation
• RAG for contextual information gathering
• Reflection-based mechanism for error correction