ARTICLE AD BOX
We currently have a Node.js application that delegates specific tasks to a Java application (via a .jar file). We have the full Java source code for this .jar.
We are currently transitioning our system to expose these functionalities as "skills" for an AI Agent within an agentic workflow. To maintain a uniform execution environment and simplify the agent's context, we need to convert this entire Java execution path into native TypeScript.
The Problem The Java codebase is quite large and relies heavily on an intricate tree of proprietary internal dependencies. Because these aren't public libraries, we cannot simply swap them out for equivalent npm packages; the dependencies themselves must also be translated from Java to TypeScript.
What we want to avoid We want to avoid a purely manual, file-by-file LLM translation (e.g., copying and pasting individual classes into an LLM), as it is highly error-prone, loses cross-file context, and does not scale for deep dependency trees.
Questions Given the scale of the project and the need to retain business logic across proprietary dependencies:
AST & Transpilation: Are there established Abstract Syntax Tree (AST) approaches or transpilation pipelines (like JSweet, but for modern Node/TS environments) that are highly effective for bulk Java-to-TS conversion when dealing with internal dependencies?
Automated Migration Patterns: What is the recommended architectural pattern or workflow for orchestrating an automated, programmatic migration of a Java codebase to TypeScript while preserving the existing project structure and references?
Agentic Workflows: If using an automated LLM-assisted script (e.g., using an agentic framework designed for refactoring), how can we best supply the complex Java dependency graph to the parser to ensure the resulting TypeScript types and imports align correctly?
Any insights into systematic, scalable approaches for this type of cross-language migration would be highly appreciated.
