AI-assisted software development system that generates cross-platform code using intermediary specifications

Abstract

This invention provides an AI-assisted software development system for generating cross-platform code using intermediate specifications. It utilizes AI technology to rapidly generate target code for a given object in a required programming language. The system includes a specification building module, a data processing module, and a generation module, with the generation module electrically connected to the specification building module and the data processing module. The specification building module uses natural language generation technology to build program logic and business logic, forming an intermediate specification template. The data processing module receives and extracts the target object's code, performs content cleaning, semantic parsing, and data extraction, and then generates an intermediate specification document using the logical structure of the intermediate specification template. This document describes the target object's program logic and business logic using natural language. The generation module receives the intermediate specification document and fills in the corresponding code template to generate the target code.

Description

Technical Field

[0001] This invention relates to software development systems, and more particularly to an AI-assisted software development system that utilizes intermediate specifications to generate cross-platform code. Prior Technology

[0002] With the rapid evolution of information technology, information systems built by traditional industries or organizations in their early years, such as the AS / 400 core system used in the traditional financial industry, are generally facing problems such as outdated technology, high maintenance costs, and a shortage of professional maintenance personnel. Because these old information systems are often built using programming languages ​​or frameworks that are gradually being phased out, such as VB, PowerBuilder, Delphi, older versions of .NET, or older Java EE, their architecture and maintenance models are difficult to adapt to the engineering standards required by modern software, such as front-end and back-end separation, cloud deployment, automated testing, and continuous integration (CI / CD). As a result, these old systems are inadequate to meet the needs of modern business operations and cannot meet service requirements.

[0003] In existing technologies, upgrading or migrating an old system to a new platform, such as Node.js, Java Spring Boot, .NET Core, Angular, or Vue.js, often requires software development teams to manually understand the old system's business logic, data flow, UI processes, and data access mechanisms using source code and limited documentation before they can redesign and write the corresponding new system. This process is not only time-consuming but also heavily reliant on senior engineers familiar with the old platform's languages ​​and structures, which exacerbates the current talent shortage. To address this issue, syntax-level code conversion tools have emerged, such as VB to C# transpilers or SQL parsers. However, these tools are mostly limited to statement-level syntax conversion and struggle to handle high-level abstract language translations involving cross-module logic, flow control, and user interface interaction logic. Furthermore, they cannot generate APIs (Application Programming Interfaces), modular architectures, and test cases that conform to modern development frameworks. When faced with a large system and inconsistent quality of original code and documentation, engineers often struggle to grasp the full picture of the system, leading to frequent issues such as missing information, data errors, or even the need to redevelop the architecture during the conversion process.

[0004] Furthermore, when changes or upgrades are needed after the system is built, the current practice involves manually adjusting multiple modules, increasing maintenance costs and error risks. In light of this, the present invention aims to explore how to incorporate the semantic understanding and structural analysis capabilities of existing AI technology to initially extract high-level program logic from the existing system, generate intermediate specifications, and then quickly generate corresponding translated code and test cases using these intermediate specifications. This would address the shortcomings of the aforementioned conventional techniques and allow subsequent system maintenance to be performed through modifications and updates to these intermediate specifications, thereby reducing the overall system development and maintenance costs. Summary of the Invention

[0005] In view of the above problems, the purpose of this invention is to provide a software program translation and maintenance system based on AI technology. Through the automatic translation of AI, the traditional program logic (i.e., code) is presented in an intermediate specification (Spec), and then cross-language program code is generated based on this specification, so as to improve the overall system performance and update speed.

[0006] To achieve the above objectives, this invention discloses an AI-assisted software development system for generating cross-platform code using an intermediate specification. This system uses AI (Artificial Intelligence) technology to rapidly generate target code for a given object, supporting a required programming language. The object is a source system program, and the target code includes at least one of front-end code and back-end code. It comprises: a specification building module for using natural language generation technology to build program logic and business logic to form an intermediate specification template. The program logic defines the corresponding definitions of multiple fields in the program code, and the business logic describes the content requirements and functional requirements of the program code. The intermediate specification template includes an interface design specification to define the page layout, component styles, interactive flow, and API (Application Programming Interface) presented by the front-end code. The application interface (API) is connected to the system to restore the original system program to the page after translation; a data processing module receives the input target object, extracts the code of the target object, performs content cleaning, semantic parsing and data extraction, and generates an intermediate specification document using the logical structure of the intermediate specification template, thereby describing the program logic and business logic of the target object in natural language; and a generation module is electrically connected to the specification building module and the data processing module and has multiple code templates corresponding to different languages. When the generation module receives the intermediate specification document, it fills the corresponding code template according to the intermediate specification document to generate the target code; wherein the target code supports at least the required programming languages ​​Vue, Angular, Java, .NET and Python.

[0007] Furthermore, the AI-assisted software development system includes an automatic testing module electrically connected to the data processing module and the generation module. When the automatic testing module receives the intermediate specification document, it uses natural language processing technology to identify the program logic and business logic within the document and automatically generates at least one test script for automatically testing the target code. After automatically testing the target code, the automatic testing module automatically generates a test report to provide information on test results, test coverage, and code defect analysis. The AI-assisted software development system also includes an intelligent maintenance module electrically connected to the specification building module, the data processing module, and the generation module. The intelligent maintenance module receives the intermediate specification document and, upon receiving a requirement change instruction, automatically updates the intermediate specification document according to the instruction, thereby driving the generation module and the automatic testing module to generate updated target code and test scripts.

[0008] The data processing module utilizes intelligent RAG (Retrieval-Augmented Generation) technology to analyze the target object based on the experience of domain experts, in order to perform content cleaning, semantic parsing, and data extraction.

[0009] In summary, this invention utilizes the generation of an intermediate specification document to translate a target object, such as the original system program or software requirements specification of an old system, into a human-computer-friendly intermediate specification. Then, through a generation module, it translates and generates target code that is cross-language and compatible with both front-end and back-end systems. This achieves an innovative and groundbreaking Code to Spec to Code translation technology architecture, significantly improving the overall system transformation efficiency. Furthermore, the automatic testing module's operation settings ensure that appropriate test cases are provided simultaneously with the target code translation and generation, allowing for rapid testing and readjustment, thus shortening development time and reducing the manpower and time costs of system development. The intelligent maintenance module's operation settings allow the target code to be directly updated through the re-editing or adjustment of the intermediate specification document when modifications or updates are needed, reducing reliance on maintenance manpower, thereby mitigating the risks associated with manpower shortages in the industry and effectively reducing post-maintenance costs. Simple Explanation of the Diagram

[0010] Figure 1 is a structural diagram of a preferred embodiment of the present invention. Figure 2 is a structural diagram of two preferred embodiments of the present invention. Figure 3 is a flowchart of two preferred embodiments of the present invention. Figure 4 is a schematic diagram of the overall system architecture of a preferred embodiment of the present invention. Implementation

[0011] To enable those skilled in the art to clearly understand the contents of this invention, please refer to the following description and accompanying drawings.

[0012] Please refer to Figure 1, which is an architecture diagram of a preferred embodiment of the present invention. As shown in the figure, the AI-assisted software development system 1 for generating cross-platform code using intermediate specifications is designed to quickly generate target code 3 of a target object 2 that supports a required programming language through AI (Artificial Intelligence) technology. It includes a specification building module 10, a data processing module 11, and a generation module 12. The generation module 12 is electrically connected to the specification building module 10 and the data processing module 11 and is provided with a plurality of code templates 120 corresponding to different languages.

[0013] The specification building module 10 is used to build a program logic 1000 and a business logic 1001 using natural language generation technology to form an intermediate specification template 100. The program logic 1000 specifies the corresponding definitions of multiple fields in the program code, such as functions, modules, and data fields. The business logic 1001 describes the content requirements and functional requirements of the program code. The data processing module 11 receives the input target object 2. After extracting the program code of the target object 2, the data processing module 11 performs content cleaning, semantic parsing, and data extraction. Then, it uses the logical structure of the intermediate specification template 100 to generate an intermediate specification file 110, which describes the program logic 1000 and the business logic 1001 of the target object in natural language. When the generation module 12 receives the intermediate specification file 110, it fills the corresponding program code template 120 according to the intermediate specification file 110 to generate the target program code 3.

[0014] Please refer to Figures 2-4, which are respectively architectural diagrams, flowcharts, and overall system architecture flowcharts of two preferred embodiments of the present invention. As shown in the figures, the AI-assisted software development system 1, which generates cross-platform code using intermediate specifications, includes a specification building module 10, a data processing module 11, a generation module 12, an automatic testing module 13, and an intelligent maintenance module 14. The generation module 12 contains a plurality of code templates 120 corresponding to different languages. The generation module 12 is electrically connected to the specification building module 10 and the data processing module 11, the automatic testing module 13 is electrically connected to the data processing module 11 and the generation module 12, and the intelligent maintenance module 14 is electrically connected to the specification building module 10, the data processing module 11, and the generation module 12.

[0015] The AI-assisted software development system 1 uses AI technology to assist a user in quickly generating target code 3 supporting one of the required programming languages ​​from a target object 2, such as a source system program or a software requirements specification. Taking the target object 2 as the source system program, such as a Legacy System, the target code 3 can include at least one of a front-end program and a back-end program, and supports at least various required programming languages ​​such as Vue, Angular, Java, .NET, and Python; in other words, the target code 3 is cross-platform, cross-front-end, and cross-language code, thereby effectively improving the overall performance and update speed of the invention. Furthermore, the operation flow of the AI-assisted software development system 1 can include at least the following steps.

[0016] The user can first build an intermediate specification template 100 through the specification building module 10. In step S1, the intermediate specification template 100 is formed by building a program logic 1000, a business logic 1001 and an interface design specification 1002 through the specification building module 10 using natural language generation technology. The program logic 1000 specifies the corresponding definitions of multiple fields in the program code, such as function, module and data field. The business logic 1001 describes the content requirements and functional requirements of the program code. The interface design specification 1002 defines the page layout, component style, interactive flow and API (Application Programming Interface) connection settings presented by the front-end program code, so that the original system program can be restored to present the page after translation, thereby maximizing the restoration of the UI (User Interface) or UX (User Experience) of the original system program. Accordingly, the intermediary specification template 100 converts data or code into easily understandable specification documents using natural language. This allows it to be processed by computers and easily understood by humans, providing a complete description of the system's logic and design, independent of any specific programming language. Incidentally, while the intermediary specification template 100 can be a basic, general template, the user can also create a customized or proprietary version of the intermediary specification template 100 based on the content of the target object 2; there are no limitations on this.

[0017] Step S2: When the data processing module 11 receives the input target object 2, the data processing module 11 extracts the code of the target object 2 from various system programs such as Informix or AS / 400, and even from information such as subroutines, database table structures, parameter settings and sample data in the system program. Then, it uses an intelligent RAG (Retrieval-Augmented Generation) technology to analyze the target object 2 based on the experience of domain experts. After that, it performs content cleaning operations such as annotation removal, blanking and formatting code through existing formatting tools. Then, it uses a large language model (LLM) to perform semantic parsing and data extraction, and generates an intermediary specification document 110 based on the logical structure of the intermediary specification template 100. Thus, the data processing module 11 achieves the purpose of describing the program logic 1000, the business logic 1001, and the interface design specification 1002 of the target object 2 through natural language. In other words, the present invention uses AI technology to convert the target object 2 from Code to Spec.

[0018] In step S3, when the generation module 12 receives the intermediate specification file 110, it fills in the corresponding code template 120 according to the intermediate specification file 110 to generate the cross-platform target code 3. Incidentally, when the target code 3 is to be generated as the front-end code, the generation module 12 can automatically generate the component, page routing, state management, and API connection code of frameworks such as Vue and Angular according to the intermediate specification file 110 and UI / UX design specifications; when the target code 3 is to be generated as the back-end code, the generation module 12 can automatically generate the controller, service layer, data access layer, and database model of frameworks such as Java or .NET according to the structured JSON, YAML, or Markdown format in the intermediate specification file 110. Thus, through the generation module 12, the present invention regenerates the previously translated Spec into Code that meets the user's needs. It can be seen that the present invention uses AI technology to translate the system program into an intermediate specification, and then automatically generates code across multiple programming languages ​​on the front end and back end based on the intermediate specification file 110, which is the Code-Spec-Code technical architecture. This achieves the benefits of code translation applicable to old program systems, as well as rapid development and continuous maintenance of new systems.

[0019] Simultaneously, when the automatic testing module 13 receives the intermediate specification document 110, in step S4, it uses natural language processing technology to identify the program logic 1000, the business logic 1001, and the interface design specification 1002 in the intermediate specification document 110 and automatically generates at least one test script 130. The test script 130 may include black-box testing, data-driven testing, and boundary testing scripts. This script may be initially generated by the automatic testing module 13 and then manually reinforced with complex boundaries and exception processes. Accordingly, in step S40, when the generation module 12 generates the target code 3, the automatic testing module 13 immediately and automatically tests whether the target code 3 meets a result requirement value. If so, in step S41, a test report is automatically generated to provide real-time feedback on test results, test coverage, and code defect analysis, to ensure the accuracy of the target code 3 output by the generation module 12. Conversely, if the automatic testing module 13 detects that the test result of the target code 3 is lower than the required result value, it will send an error message to the generation module 12, causing the generation module 12 to regenerate the target code 3.

[0020] In step S5, when the intelligent maintenance module 14 receives the intermediate specification file 110, it parses and temporarily stores it. In step S50, when the intelligent maintenance module 14 receives a requirement change instruction 140, it compares and automatically updates the intermediate specification file 110 according to the requirement change instruction 140, and further drives the generation module 12 and the automatic testing module 13 to regenerate and update the target program code 3 and the test script 130, so that the final output target program code 3 is completely modified and updated according to the requirement change instruction 140, so that the user can easily maintain the software and achieve the effect of simplifying maintenance operations.

[0021] The modules described in this invention are implemented through hardware or software supplemented by hardware. For example, the definitions of the specification building module 10, the data processing module 11, the generation module 12, the automatic testing module 13, and the intelligent maintenance module 14 essentially refer to the integration of various hardware devices such as CPUs, microprocessors, memory, or signal transmitters, and the technical features implemented by software programs. Unless otherwise explicitly specified and limited, the terms "setting" and "connection" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections, or even wired electrical connections or wireless signal electrical connections; they can refer to direct connections or indirect connections through an intermediate medium, or connections within two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0022] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of the present invention. Therefore, any equivalent changes and modifications made without departing from the spirit and scope of the present invention should be covered within the patent scope of the present invention.

[0023] 1: AI-assisted software development system 10: Specification Building Module 100: Intermediary Specification Template 1000: Program Logic 1001: Business Logic 1002: Interface Design Specification 11: Data Processing Module 110: Intermediary Specification Documents 12: Generate Module 120: Code Template 13: Automated Testing Module 130: Test Script 14: Intelligent Maintenance Module 140: Requirement Change Order 2: Subject Matter 3: Target code S1~S50: Steps

Claims

1. An AI-assisted software development system for generating cross-platform code using an intermediate specification, which uses AI (Artificial Intelligence) technology to quickly generate target code in a required programming language from a target object, wherein the target object is an original system program, and the target code includes at least one of front-end code and back-end code, comprising: a specification building module for building a program logic and a business logic using natural language generation technology to form an intermediate specification template, wherein the program logic specifies the corresponding definitions of a plurality of fields in the program code, and the business logic describes the content requirements and functional requirements of the program code, wherein... The intermediate specification template includes an interface design specification to define the page layout, component styles, interactive flow, and API (Application Programming Interface) connection settings presented by the front-end code, so that the original system program can restore the page presentation after translation; a data processing module receives the input target object, extracts the code of the target object, performs content cleaning, semantic parsing, and data extraction, and generates an intermediate specification document using the logical structure of the intermediate specification template, thereby describing the program logic and business logic of the target object in natural language; and a generation module is electrically connected to the specification building module and the data processing module and has multiple code templates corresponding to different languages. When the generation module receives the intermediate specification document, it fills in the corresponding code template according to the intermediate specification document to generate the target code; wherein, the target code supports at least the required programming languages ​​Vue, Angular, Java, .NET, and Python.

2. The AI-assisted software development system as described in claim 1 further includes an automatic testing module electrically connected to the data processing module and the generation module. When the automatic testing module receives the intermediate specification document, it identifies the program logic and business logic in the intermediate specification document through natural language processing technology and automatically generates at least one test script for automatically testing the target code.

3. The AI-assisted software development system as described in claim 2, wherein, After automatically testing the target code, the automated testing module automatically generates a test report to provide information on the test results, test coverage, and code defect analysis.

4. The AI-assisted software development system as described in claim 2 further includes an intelligent maintenance module, which is electrically connected to the specification building module, the data processing module, and the generation module. The intelligent maintenance module receives the intermediate specification document and, upon receiving an input requirement change instruction, automatically updates the intermediate specification document according to the requirement change instruction, thereby driving the generation module and the automatic testing module to generate updated target code and test scripts accordingly.

5. The AI-assisted software development system as described in claim 1, wherein, The data processing module utilizes intelligent RAG (Retrieval-Augmented Generation) technology to analyze the target object based on domain expert experience, in order to perform content cleaning, semantic parsing, and data extraction.

Images