Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AI Agent for Reverse-Engineering Legacy Finite-Difference Code and Translating to Devito

Yinghan Hou, Zongyou Yang

TL;DR

The paper addresses modernizing large-scale legacy Fortran finite-difference codes by translating them into the Devito framework using an integrated AI agent built on GraphRAG and LangGraph. It introduces a five-layer architecture, a knowledge-base pipeline, and a Neo4j-based Devito knowledge graph, leveraging multi-modal retrieval, Fortran static analysis, and K3-Trans-derived translation mappings to ensure accuracy and reproducibility. A quality-driven iterative optimization loop, reinforced by Pydantic-structured outputs and Ruff formatting, guides conversion with stringent validation through static analysis and G-Eval evaluation, achieving high precision and fast responses. The approach offers a scalable, robust path for automated scientific code modernization, enabling reliable, efficient translation of legacy Fortran to Devito with strong retrieval quality and translation fidelity that support reproducible, domain-specific transformations.

Abstract

To facilitate the transformation of legacy finite difference implementations into the Devito environment, this study develops an integrated AI agent framework. Retrieval-Augmented Generation (RAG) and open-source Large Language Models are combined through multi-stage iterative workflows in the system's hybrid LangGraph architecture. The agent constructs an extensive Devito knowledge graph through document parsing, structure-aware segmentation, extraction of entity relationships, and Leiden-based community detection. GraphRAG optimisation enhances query performance across semantic communities that include seismic wave simulation, computational fluid dynamics, and performance tuning libraries. A reverse engineering component derives three-level query strategies for RAG retrieval through static analysis of Fortran source code. To deliver precise contextual information for language model guidance, the multi-stage retrieval pipeline performs parallel searching, concept expansion, community-scale retrieval, and semantic similarity analysis. Code synthesis is governed by Pydantic-based constraints to guarantee structured outputs and reliability. A comprehensive validation framework integrates conventional static analysis with the G-Eval approach, covering execution correctness, structural soundness, mathematical consistency, and API compliance. The overall agent workflow is implemented on the LangGraph framework and adopts concurrent processing to support quality-based iterative refinement and state-aware dynamic routing. The principal contribution lies in the incorporation of feedback mechanisms motivated by reinforcement learning, enabling a transition from static code translation toward dynamic and adaptive analytical behavior.

AI Agent for Reverse-Engineering Legacy Finite-Difference Code and Translating to Devito

TL;DR

The paper addresses modernizing large-scale legacy Fortran finite-difference codes by translating them into the Devito framework using an integrated AI agent built on GraphRAG and LangGraph. It introduces a five-layer architecture, a knowledge-base pipeline, and a Neo4j-based Devito knowledge graph, leveraging multi-modal retrieval, Fortran static analysis, and K3-Trans-derived translation mappings to ensure accuracy and reproducibility. A quality-driven iterative optimization loop, reinforced by Pydantic-structured outputs and Ruff formatting, guides conversion with stringent validation through static analysis and G-Eval evaluation, achieving high precision and fast responses. The approach offers a scalable, robust path for automated scientific code modernization, enabling reliable, efficient translation of legacy Fortran to Devito with strong retrieval quality and translation fidelity that support reproducible, domain-specific transformations.

Abstract

To facilitate the transformation of legacy finite difference implementations into the Devito environment, this study develops an integrated AI agent framework. Retrieval-Augmented Generation (RAG) and open-source Large Language Models are combined through multi-stage iterative workflows in the system's hybrid LangGraph architecture. The agent constructs an extensive Devito knowledge graph through document parsing, structure-aware segmentation, extraction of entity relationships, and Leiden-based community detection. GraphRAG optimisation enhances query performance across semantic communities that include seismic wave simulation, computational fluid dynamics, and performance tuning libraries. A reverse engineering component derives three-level query strategies for RAG retrieval through static analysis of Fortran source code. To deliver precise contextual information for language model guidance, the multi-stage retrieval pipeline performs parallel searching, concept expansion, community-scale retrieval, and semantic similarity analysis. Code synthesis is governed by Pydantic-based constraints to guarantee structured outputs and reliability. A comprehensive validation framework integrates conventional static analysis with the G-Eval approach, covering execution correctness, structural soundness, mathematical consistency, and API compliance. The overall agent workflow is implemented on the LangGraph framework and adopts concurrent processing to support quality-based iterative refinement and state-aware dynamic routing. The principal contribution lies in the incorporation of feedback mechanisms motivated by reinforcement learning, enabling a transition from static code translation toward dynamic and adaptive analytical behavior.
Paper Structure (44 sections, 2 equations, 8 figures, 3 tables)

This paper contains 44 sections, 2 equations, 8 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background and problem description
  3. GraphRAG
  4. Embeddings
  5. K3-Trans
  6. Methodology and Implementation
  7. System architecture overview
  8. Knowledge base construction pipeline
  9. Devito code repository analysis and preprocessing
  10. Multimodal document processing and MinerU integration
  11. Structure-aware semantic segmentation
  12. Entity relationship extraction and graph construction
  13. Embedding, community detection and graph optimization
  14. Neo4j migration and index building
  15. GraphRAG retrieval system design
  16. ...and 29 more sections

Figures (8)

  • Figure 1: LangChain RAG from scratch pipeline overview.
  • Figure 2: System architecture for Fortran-to-Devito pipeline.
  • Figure 3: Devito knowledge base
  • Figure 4: Community hierarchy by theme categories.
  • Figure 5: Retrieval-Augmented Generation (RAG) pipeline
  • ...and 3 more figures