A system reconstruction method, system, and medium based on AI

Abstract

This invention relates to an AI-based system reconstruction method, system, and medium. The method includes: extracting metadata from the old system; analyzing and training data query statements to a vector library based on retrieval enhancement architecture; generating a new database as needed; using an Excel template as a basis, organizing the functions of the old system based on a large language model, and supplementing information as needed to generate a business function list; generating a new system interface and editable low-code configuration data based on a visual language model, the function organization results, and the business function list; intelligently supplementing and correcting deviations in the generated new system interface and editable low-code configuration data based on the business rules of the low-code platform, optimizing the development of the system interface and function configuration; synchronizing and migrating data between the old and new systems to complete the system reconstruction. This invention can significantly reduce the cost of reconstructing old systems, shorten the reconstruction cycle, and provide a new paradigm for the current domestic IT innovation transformation and the strategy of eliminating and retaining old systems.

Description

Technical Field

[0001] This invention relates to the field of system reconfiguration technology, and specifically to an AI-based system reconfiguration method, system, and medium. Background Technology

[0002] With the continuous development of information technology, many organizations' existing legacy systems have become obstacles to their digital and intelligent transformation. On the one hand, due to outdated technical architecture and vendor service disruptions, these legacy systems pose serious cybersecurity and data security risks that are difficult to rectify. On the other hand, inadequate source code management and a shortage of experienced programmers specializing in older programming languages ​​make it difficult to implement domestic IT innovation upgrades and unified identity and access control systems. Furthermore, these legacy systems are difficult to integrate with other systems at the data and application levels, resulting in severe application and data silos.

[0003] Current solutions generally involve refactoring the old system through complete system development. However, the cost and time-consuming nature of such redevelopment and modification severely hinder the progress of old system reconstruction. The common drawbacks are as follows: 1. Unclear data structure: Old systems often have missing or incomplete database design documents. New developers need a lot of time to sort out the data structure and clarify the data relationships in order to understand the business logic and realize the data migration of the old system.

[0004] 2. High development costs: Every function of the old system needs to be developed using the latest mainstream programming languages ​​and development architectures. Even with the use of AI programming and other auxiliary tools, the development costs are still high.

[0005] 3. Limited reconstruction costs: For many state-owned enterprises and government departments, a single unit may have built and used hundreds of old systems. Given the tight timeframe and insufficient funding for information technology innovation transformation, it is obviously difficult to complete the reconstruction of old systems with an investment of 30-40% of the old system construction cost, which will make it difficult to advance the project. Summary of the Invention

[0006] This invention provides an AI-based system reconstruction method, system, and medium, aiming to fully utilize the new productivity of AI, such as large language capabilities and AI multimodal models, and combine them with traditional low-code platforms, workflow engines, and other tools. By intelligently sorting out the system's data structure and data relationships, generating a new interface after intelligently analyzing the interface elements of the old system through multimodal models, and using natural language to describe business rules and logic to automatically generate business logic code, the system reconstruction is completed, reducing workload.

[0007] This invention provides an AI-based system reconstruction method, the method comprising the following steps: Step S10: Access the old system based on multi-source heterogeneous data access technology, extract metadata from the old system, analyze and train data query statements to the vector library based on retrieval enhancement generation architecture, and generate a new database as needed with reference to the old system database to provide a data foundation for subsequent system reconstruction. Step S20: Based on the EXCEL template, sort out the functions of the old system based on the large language model, and complete the information as needed to generate a business function list, providing functional constraints and configuration basis for AI reconstruction; Step S30: Based on the visual language model, the functional analysis results, and the business function list, generate a new system interface and editable low-code configuration data. Step S40: Based on the intelligent completion and correction of business rules on the low-code platform, the deviations of the generated new system interface and editable low-code configuration data are optimized to improve the development of system interface and function configuration. Step S50: Synchronize and migrate data between the old and new systems to complete system reconstruction.

[0008] A further technical solution of the present invention is that step S10 includes: Step S101: Read the old system database structure through database connection technology to obtain the old system database metadata, which includes table, field, primary key, foreign key and index information; Step S102: Automatically identify and organize the metadata using a large language model; Step S103: Train the data query statements into the vector library based on the retrieval enhancement architecture for subsequent querying and retrieval based on natural language parsing; Step S104: Export database logs from the old system database, extract the business operation SQL statements from the old system, and use AI combined with intelligently organized data structures to perform reverse training of the business operation SQL statements. Step S105: Vectorize and store the training results of the completed data structure and data corpus; Step S106: Refer to the old system database and generate a new database as needed.

[0009] A further technical solution of the present invention is that step S103 includes: Step S1031: Based on the intelligent data structure analysis, extract the metadata of the data tables and table ER relationships of the business system; Step S1032: For each data table, sample and train the field set data. Using the XXL-JOB scheduled task framework, connect to the data source periodically, sample the enumerated values ​​of the key fields of the specified table, and define the standard expression. Step S1033: Construct an atomic semantic training mechanism based on a single-table scenario. With business combinations as the core, generate a mapping relationship between semantic questions and corresponding execution logic by analyzing high-frequency query patterns. For typical business tables, extract common business combination scenarios of core fields.

[0010] A further technical solution of the present invention is that step S104 includes: Step S1041: Perform complex business logic operations in the old system and enable the database logging function; Step S1042: Export the database logs of complex business logic operations and complete the business scenario description; Step S1043: Submit the business scenario and database log documents to the large model for analysis. The large model analyzes the data operation objects, data operation order and data operation type in the business scenario to form business logic reasoning.

[0011] A further technical solution of the present invention is that step S20 includes: Step S201: Design a business function review template. The business function review template maintains information such as function description, associated business domain, form operation, view operation, form interface screenshot, and view interface screenshot, based on the functional module. Step S202: For the functions in the old system that need to be refactored, complete the information and take screenshots of the pages according to the subsystem, module and function dimensions, and record them in the Excel template; Step S203: Summarize all functions that need to be refactored into the business function list. The function list includes the following elements: first-level menu, second-level menu, view tabs, menu description, view function operation description, view data column operation description, view search, form function operation description, business domain description, view screenshot, and form screenshot.

[0012] A further technical solution of the present invention is that step S30 includes: Step S301: Import the completed business function list, which includes: subsystem name, module name, function name, function description, interface screenshot, and main operations. Step S302: For each function, match the corresponding data table according to the business rules described in natural language; Step S302 specifically includes the following steps: Step S3021: Extract the business rules described in natural language for each function from the "Business Function List"; Step S3022: Based on the business rules of the natural language description of each function, and based on intent recognition, match the data table or multiple tables with the highest confidence in the database DDL already stored in the vector library, thereby matching a data table for each function page.

[0013] Step S303: Input the old system function images captured for each function into the VL model, and let the VL model identify the type, position and relationship of the interface elements; the interface elements include form fields, view components and operation buttons, wherein the form fields include field name, type, default value and validation rules, the view components include table columns, sorting rules, filter conditions, charts and layout information, and the operation buttons include add, edit, delete and query buttons; Step S304: Compare and analyze the VL model with the generated interface element data, and fill in the missing information; Step S305: For each function, call the low-code platform component library, render the visual interface based on the interface element data, and use a microservice architecture to split the interface modules to generate a new system interface. Based on the new system interface, generate editable low-code configuration data. The new system interface includes form service, view service, and component rendering service.

[0014] Following step S305, the following is also included: Step S306: For the data dictionary implicit in the old system, AI is used to automatically detect and generate the data dictionary to provide metadata for the drop-down component of the page; Step S306 specifically includes: Step S3061: Access the corresponding tables and fields in the original database through the data tables and data fields corresponding to the interface elements to detect the old system data structure and sample data; Step S3062 involves using AI to identify the business meaning and value range of fields, and generating a standardized data dictionary. Specifically, this includes performing SQL function analysis by analyzing the values ​​of all data fields to generate a standardized data dictionary.

[0015] A further technical solution of the present invention is that step S40 includes: Step S401, Interface element correction: Adjust the order of form fields, component styles, button positions, and add custom components; Step S402, Business rule configuration and code generation: Improve field validation logic and data association rules. For business operation buttons, support manual description of business rules in natural language through the interface. Combine the old system business SQL statements that have been sorted out with AI code generation tools to generate business logic code. Step S403, Permission Configuration: Configure the functional permissions and data permissions of the business system.

[0016] A further technical solution of the present invention is that, after step S40, it further includes: Step S501: For business systems containing workflows, configure the workflow as needed. The workflow configuration includes: basic process information configuration, task node configuration, process flow direction configuration, and process role configuration.

[0017] To achieve the above objectives, the present invention also proposes an AI-based system reconfiguration system, the system comprising a memory, a processor, and an AI-based system reconfiguration program stored in the memory, wherein the AI-based system reconfiguration program is executed by the processor to perform the steps of the method described above.

[0018] To achieve the above objectives, the present invention also proposes a computer-readable storage medium storing an AI-based system refactoring program, which, when run by a processor, executes the steps of the method described above.

[0019] The beneficial effects of the AI-based system reconstruction method, system, and medium of this invention are: This invention, through the aforementioned technical solution, can intelligently organize data assets and refactor business system code for a large number of existing legacy systems, especially traditional C / S architecture systems (such as systems developed using tools like VC, PB, and DELPHI) and old MVC architecture Java systems, even in situations where source code is lacking or vendor service is interrupted. It can also complete system-wide IT innovation transformation, data security transformation, personal information protection transformation, unified identity and unified permissions transformation, and commercial encryption transformation in one go. This significantly reduces the cost and shortens the transformation cycle of legacy systems, providing a new paradigm for the current IT innovation transformation and legacy system retention strategy. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the first embodiment of the AI-based system reconstruction method of the present invention; Figure 2 This is a detailed flowchart of step S10; Figure 3 This is a detailed flowchart of step S103; Figure 4 This is a detailed flowchart of step S104; Figure 5 This is a detailed flowchart of step S20; Figure 6 This is a detailed flowchart of step S30; Figure 7 This is a detailed flowchart of step S302; Figure 8 This is a detailed flowchart of step S40; Figure 9This is a schematic diagram of the hardware architecture of the AI-based system reconfiguration system of this invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0022] This invention proposes an AI-based system reconstruction method. The technical solution adopted by this AI-based system reconstruction method mainly makes full use of the new productivity of AI, such as large language capabilities and AI multimodal models, combined with traditional low-code platforms, workflow engines and other tools. It intelligently sorts out the system data structure and data relationships, generates a new interface after intelligently analyzing the interface elements of the old system through multimodal models, and uses natural language to describe business rules and logic to realize the automatic generation of business logic code, thereby completing the reconstruction of the old system.

[0023] Specifically, such as Figure 1 As shown, the first embodiment of the AI-based system reconstruction method of the present invention includes the following steps: Step S10, Intelligent sorting and training of old system metadata: Based on multi-source heterogeneous data access technology, access the old system, extract the old system metadata, generate architecture analysis based on retrieval enhancement and train data query statements to the vector library, refer to the old system database, generate a new database as needed, and provide a data foundation for subsequent system reconstruction.

[0024] The core of multi-source heterogeneous data access technology is to achieve unified access to data from different databases (such as MySQL, Oracle, SQL Server, etc.) and file types in legacy systems through the JDBC / ODBC universal database connection protocol and an interface adaptation layer that adapts to various heterogeneous data sources. It supports automatic identification of data source types, establishment of connection pools to optimize connection performance, and efficient reading of metadata such as database table structures, field attributes, and index relationships, providing a stable data input channel for subsequent data analysis and migration.

[0025] The Retrieval Enhancement Generative Architecture (RAG) is used for training and retrieving data query statements from legacy systems. It transforms extracted business query statements into vectors and stores them in a vector library. During subsequent refactoring, matching query logic can be quickly retrieved based on natural language parsing results, providing accurate support for the data query functions of the new system, avoiding redundant query statement development, and improving generation efficiency and business adaptability.

[0026] Step S20: Manual review and standardization of old system business functions: Based on the EXCEL template, the functions of the old system are reviewed based on the large language model, and information is supplemented as needed to generate a list of business functions, providing functional constraints and configuration basis for AI reconstruction.

[0027] Large Language Modeling (LLM) is a core technology throughout the entire refactoring process, primarily undertaking three key roles: First, it automatically organizes the business logic of the legacy system based on metadata and sample data, identifying business relationships between tables and fields; second, it parses the "functional descriptions" in Excel configurations, accurately matching the corresponding business domain data sources and database objects; and third, through natural language understanding capabilities, it assists in generating a data dictionary and verifying the rationality of business rules, reducing the complexity of manual organization. This embodiment, based on natural language processing technology, automatically identifies business entities, entity attributes, and relationships between entities from the legacy system's metadata, business descriptions, and sample data. It provides structured semantic support for data source matching, business rule generation, and data dictionary construction, improving the accuracy of AI-driven refactoring.

[0028] Step S30, Intelligent code generation based on standardized function analysis results: Based on the visual language model, function analysis results and the business function list, a new system interface and editable low-code configuration data are generated.

[0029] The Visual Language Model (VL model) focuses on the accurate identification and supplementation of interface elements. By inputting screenshots of the old system interface, it automatically identifies the type, position, and hierarchy of components such as buttons, input boxes, tables, and dropdown lists. Its core value is to correct the deviations of AI-generated interface elements and supplement information such as custom components and component order not reflected in the Excel configuration, ensuring that the new system interface is consistent with the core interactive experience of the old system.

[0030] Step S40, Intelligent completion and correction of business rules based on low-code platform: Based on the deviation of the new system interface and editable low-code configuration data generated by the intelligent completion and correction of business rules based on low-code platform, optimize the development of system interface and function configuration.

[0031] Low-code development technology is the core supporting technology for refactoring, providing a visual editor and component library (based on React), supporting drag-and-drop adjustment of the interface layout. It automatically renders the visual interface by parsing AI-generated interface element data, and adopts a microservice architecture to decouple the interface from functionality by separating form services, view services, and other modules. It supports real-time preview, configuration data editing and export, significantly shortening the development cycle.

[0032] Step S50: Synchronize and migrate data between the old and new systems, synchronizing the data from the old system to the new database to complete the system reconstruction.

[0033] In this embodiment, the core of the data migration technology is to achieve seamless data migration between the old and new systems. It supports data format conversion and compatibility adaptation between heterogeneous databases (such as MySQL / Oracle / SQL Server). At the migration execution level, a "batch parallel migration + breakpoint resume" mechanism is adopted, combined with database connection pooling and batch write optimization to improve the efficiency of massive data migration. The migration process provides a visual progress panel to display the real-time migration progress of each data table, supports the separate export and secondary migration of abnormal data, and requires no manual intervention in the core processes throughout.

[0034] This embodiment of the AI-based system reconstruction method, through the above-mentioned technical solutions, can intelligently sort out data assets and complete business system code reconstruction for a large number of existing old systems, especially traditional C / S architecture systems (such as systems developed with tools such as VC, PB, and DELPHI) and old MVC architecture Java systems, in situations where source code is lacking or vendor service is interrupted. It can also complete the system's domestic IT innovation transformation, data security transformation, personal information protection transformation, unified identity and unified permission transformation, and commercial encryption transformation in one go. This can greatly reduce the cost of transforming old systems and shorten the transformation cycle, providing a new paradigm for the current strategy of domestic IT innovation transformation and the elimination and retention of old systems.

[0035] Furthermore, based on Figure 1 The first embodiment shown presents a second embodiment of the AI-based system reconstruction method of the present invention, as follows: Figure 2 As shown, this embodiment is similar to Figure 1 The difference between the first embodiment shown and the present embodiment is that, in this embodiment, step S10 specifically includes the following steps: Step S101, Source Data Access: Read the old system database structure through database connection technology to obtain the old system database metadata, which includes table, field, primary key, foreign key and index information.

[0036] This embodiment reads the old system database structure, including information such as tables, fields, primary keys, foreign keys, and indexes, through database connection technology (JDBC / ODBC, etc.).

[0037] Step S102, Intelligent data structure organization: The metadata is automatically identified and organized using a large language model.

[0038] Step S103: Train the data query statements into the vector library based on the retrieval enhancement architecture for subsequent querying and retrieval based on natural language parsing.

[0039] like Figure 3 As shown, step S103 specifically includes the following steps: Step S1031, Extract metadata: Based on the “Intelligent data structure sorting”, extract the metadata of data tables and table ER relationship of the business system.

[0040] Step S1032, Field set data sampling and training: Perform field set data sampling and training for each data table. Using the XXL-JOB scheduled task framework, connect to the data source periodically, sample enumerated values ​​for key fields of the specified table (such as deduplicated school names, class types, etc.), and define standard expressions (such as mapping "Nantou Primary School" to "Nantou Primary School, Nanshan District, Shenzhen").

[0041] The writing format example is as follows: { "table": "table name", "tabdes": "table description", "fldattr": [ { "fldname":"", "flddes": "", "fldvalue": [{ "fldvalue": "Nantou Primary School", "flcvalue": "Shenzhen Nantou Primary School, Shenzhen Nanshan District Nantou Primary School"} ]} ]}.

[0042] Step S1033, Atomic-level and Field Business Combination Training: Construct an atomic semantic training mechanism based on single-table scenarios, with business combination as the core. By analyzing high-frequency query patterns, generate the mapping relationship between semantic questions and corresponding execution logic. For typical business tables, extract common business combination scenarios of core fields.

[0043] To improve the accuracy of natural language to SQL conversion, this embodiment constructs an atomic semantic training mechanism based on a single-table scenario, focusing on learning and modeling the business attributes of fields and their common combination patterns. This process centers on "business combinations," automatically generating a mapping relationship between semantic queries and corresponding execution logic by analyzing high-frequency query patterns. For typical business tables (such as student information tables, grade tables, attendance tables, etc.), common business combination scenarios of their core fields are extracted, for example: (1) Average grades for each grade; (2) Statistics on the number of classes in primary schools in Nanshan District; (3) How many students are there in the primary schools in Nanshan District? ......

[0044] Step S104, Source System Business Logic Review and Training: Export database logs from the old system database, extract the business operation SQL statements from the old system, and use AI combined with intelligently reviewed data structures to perform reverse training of the business operation SQL statements.

[0045] The biggest obstacle to refactoring old systems is that users are unable to clearly describe the business logic of complex business systems for various reasons. If the business logic of a complex business system lacks detailed design documents and historical code is lost, it is extremely difficult to refactor the complex business logic functionality.

[0046] This embodiment exports database logs from the source system database, extracts the business operation SQL statements from the old system, and uses AI combined with intelligently organized data structures to perform reverse training on the business operation SQL statements. For example... Figure 4 As shown, step S104 specifically includes the following steps: Step S1041: Perform complex business logic operations in the old system and enable the database logging function.

[0047] Step S1042: Export the database logs of complex business logic operations and complete the business scenario description.

[0048] Step S1043: Submit the business scenario and database log documents to the large model for analysis. The large model analyzes the data operation objects (such as tables), data operation order, and data operation types (query, modification, deletion, etc.) in the business scenario to form business logic reasoning.

[0049] Step S105, Result Vectorization: Vectorize and store the training results of the completed data structure and data corpus.

[0050] Step S106, Generate a new database: Refer to the old system database and generate a new database as needed.

[0051] Furthermore, based on Figure 1 The first embodiment shown presents a third embodiment of the AI-based system reconstruction method of the present invention, as follows: Figure 5 As shown, this embodiment is similar to Figure 1 The difference between the first embodiment shown and the present embodiment is that, in this embodiment, step S20 includes the following steps: Step S201, Business Function Review Template Design: Design a business function review template. The business function review template maintains information such as function description, associated business domain, form operation, view operation, form interface screenshot, and view interface screenshot, based on the functional module.

[0052] In this embodiment, a standardized configuration template based on Excel is used to define unified configuration content according to functional module dimensions. The template has built-in field format validation rules to ensure that the sorted information is standardized and complete, while providing structured input for AI parsing, thus achieving an efficient connection between "manual sorting" and "intelligent parsing".

[0053] Step S202, Business Function Review: For the functions in the old system that need to be refactored, complete the information and take screenshots of the pages according to the subsystem, module and function dimensions, and record them in the Excel template.

[0054] Step S203, Generate Business Function List (Excel): Summarize all functions that need to be refactored into the business function list. The function list includes the following elements: first-level menu, second-level menu, view tab, menu description, view function operation description, view data column operation description, view search, form function operation description, business domain description, view screenshot, and form screenshot.

[0055] Furthermore, based on Figure 1 The first embodiment shown presents a fourth embodiment of the AI-based system reconstruction method of the present invention, as follows: Figure 6 As shown, this embodiment is similar to Figure 1 The difference between the first embodiment shown and the present embodiment is that, in this embodiment, step S30 includes the following steps: Step S301, Import Business Function List: Import the business function list that has been sorted out. The business function list includes: subsystem name, module name, function name, function description (natural language description), interface screenshot, main operations (such as: add, modify, delete, batch modify, batch delete, next step, etc.).

[0056] Step S302, Natural Language Parsing and Data Source Matching: For each function, match the corresponding data table according to the business rules described in the natural language. For example... Figure 7 As shown, step S302 specifically includes the following steps: Step S3021: Extract the business rules described in natural language for each function from the "Business Function List".

[0057] Step S3022: Based on the business rules of the natural language description of each function, and based on intent recognition, match the data table or multiple tables with the highest confidence in the database DDL already stored in the vector library, thereby matching a data table for each function page.

[0058] Step S303, Identify form / view interface elements based on VL model: Input the old system function images captured for each function into the VL model, and the VL model identifies the type, position, and relationship of the interface elements; the interface elements include form fields, view components, and operation buttons, wherein the form fields include field name, type, default value, and validation rules, the view components include table columns, sorting rules, filter conditions, charts, and layout information, and the operation buttons include add, edit, delete, and query buttons.

[0059] Step S304, VL model interface screenshot recognition and page element completion: Compare and analyze the VL model with the generated interface element data (such as custom buttons not configured in the EXCEL template in the screenshot), and complete the missing information (such as component position information and order not shown in the EXCEL template).

[0060] Step S305, New system interface generation: For each function, call the low-code platform component library (based on React), render the visual interface according to the interface element data, and use a microservice architecture to split the interface modules to generate a new system interface. Based on the new system interface, generate editable low-code configuration data. The new system interface includes form service, view service, and component rendering service.

[0061] This embodiment supports real-time interface preview and generates editable low-code configuration data. It can intelligently match interface elements based on a recommendation algorithm, taking into account functional scenarios, general design specifications, and user configuration habits. Input components with corresponding formats are recommended according to field types (such as date types), reducing manual configuration costs.

[0062] This embodiment is based on a low-code platform's workflow engine, providing a graphical configuration interface that supports the visual construction of business processes (such as approval processes). Users can configure workflow nodes by dragging and dropping, define node roles, and set flow rules (such as conditional branches and rejection logic), enabling the flow and approval of form data without writing code, thus adapting to the workflow requirements of different business scenarios.

[0063] In this embodiment, the method further includes the following after step S305: Step S306: For the data dictionary implicit in the old system, AI is used to automatically detect and generate the data dictionary, providing metadata for the dropdown component of the page.

[0064] Step S306 specifically includes: Step S3061, detect the original system data structure and sample data: access the corresponding tables and fields in the original database through the data tables and data fields corresponding to the interface elements to detect the old system data structure and sample data.

[0065] Step S3062, AI identifies the business meaning and value range of fields, and generates a standardized data dictionary: A standardized data dictionary is generated by identifying the business meaning and value range of fields through AI; specifically, this includes: analyzing the values ​​of all data fields and performing SQL function analysis such as "DISTINCT", "MIN()", and "MAX()" to generate a standardized data dictionary.

[0066] This embodiment supports manual editing and export of the data dictionary.

[0067] Furthermore, based on Figure 1 The first embodiment shown presents a fifth embodiment of the AI-based system reconstruction method of the present invention, as follows: Figure 8 As shown, this embodiment is similar to Figure 1 The difference between the first embodiment shown and the present embodiment is that, in this embodiment, step S40 includes the following steps: Step S401, Interface element correction: Adjust the order of form fields, component styles, button positions, and add custom components; Step S402, Business rule configuration and code generation: Improve field validation logic and data association rules. For business operation buttons, support manual description of business rules in natural language through the interface. Combine the old system business SQL statements that have been sorted out with AI code generation tools to generate business logic code. Step S403, Permission Configuration: Configure the functional permissions and data permissions of the business system.

[0068] This embodiment allows for manual configuration of business system function permissions and data permissions. This includes configuring information such as users, roles, and authorizations. Alternatively, it can achieve permission control through integration with a unified identity and unified permission system.

[0069] In this embodiment, for business systems containing workflows, workflows are configured as needed. Specifically, in this embodiment, after step S40, the following is also included: Step S501: For business systems containing workflows, configure the workflow as needed. The workflow configuration includes: basic process information configuration, task node configuration, process flow direction configuration, and process role configuration.

[0070] Among them, the basic process information configuration is based on the system's built-in workflow engine.

[0071] Task node configuration: Refer to the old system for process node configuration within the workflow.

[0072] Process flow configuration allows you to configure the flow direction as needed within the process.

[0073] Configure process roles as needed within the process.

[0074] The beneficial effects of the AI-based system reconstruction method of this invention are: This invention, through the aforementioned technical solution, can intelligently organize data assets and refactor business system code for a large number of existing legacy systems, especially traditional C / S architecture systems (such as systems developed using tools like VC, PB, and DELPHI) and old MVC architecture Java systems, even in situations where source code is lacking or vendor service is interrupted. It can also complete system-wide IT innovation transformation, data security transformation, personal information protection transformation, unified identity and unified permissions transformation, and commercial encryption transformation in one go. This significantly reduces the cost and shortens the transformation cycle of legacy systems, providing a new paradigm for the current IT innovation transformation and legacy system retention strategy.

[0075] To achieve the above objectives, this invention also proposes an AI-based system reconfiguration system, such as... Figure 8 As shown, the system includes a processor 1001, a CPU, a network interface 1004, a user interface 1003, a memory 1005, a communication bus 1002, and an AI-based system reconfiguration program stored on the processor. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0076] Those skilled in the art will understand that Figure 8 The system structure shown does not constitute a limitation on the system and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0077] like Figure 8 As shown, the memory 1005, which serves as a computer storage medium, may include an operating device, a network communication module, a user interface module, and an AI-based system reconfiguration program.

[0078] exist Figure 8 In the system shown, the network interface 1004 is mainly used to connect to the network server and communicate with the network server; the user interface 1003 is mainly used to interact with the user terminal and receive user input instructions; and the processor 1001 can be used to call the AI-based system reconstruction program stored in the memory 1005.

[0079] To achieve the above objectives, the present invention also proposes a computer-readable storage medium storing an AI-based system refactoring program, which, when run by a processor, executes the steps of the method described above, which will not be repeated here.

[0080] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. An AI-based system reconstruction method, characterized in that, The method includes the following steps: Step S10: Access the old system based on multi-source heterogeneous data access technology, extract metadata from the old system, generate an architecture based on retrieval enhancement and train data query statements to a vector library, and generate a new database as needed, referring to the old system database, to provide a data foundation for subsequent system reconstruction. Step S20: Based on the EXCEL template, sort out the functions of the old system based on the large language model, and complete the information as needed to generate a business function list, providing functional constraints and configuration basis for AI reconstruction; Step S30: Based on the visual language model, the functional analysis results, and the business function list, generate a new system interface and editable low-code configuration data. Step S40: Based on the intelligent completion and correction of business rules on the low-code platform, the deviations of the generated new system interface and editable low-code configuration data are optimized to improve the development of system interface and function configuration. Step S50: Synchronize and migrate data between the old and new systems to complete system reconstruction.

2. The AI-based system reconstruction method according to claim 1, characterized in that, Step S10 includes: Step S101: Read the old system database structure through database connection technology to obtain the old system database metadata, which includes table, field, primary key, foreign key and index information; Step S102: Automatically identify and organize the metadata using a large language model; Step S103: Train the data query statements into the vector library based on the retrieval enhancement architecture for subsequent querying and retrieval based on natural language parsing; Step S104: Export database logs from the old system database, extract the business operation SQL statements from the old system, and use AI combined with intelligently organized data structures to perform reverse training of the business operation SQL statements. Step S105: Vectorize and store the training results of the completed data structure and data corpus; Step S106: Refer to the old system database and generate a new database as needed.

3. The AI-based system reconstruction method according to claim 2, characterized in that, Step S103 includes: Step S1031: Based on the intelligent data structure analysis, extract the metadata of the data tables and table ER relationships of the business system; Step S1032: For each data table, sample and train the field set data. Using the XXL-JOB scheduled task framework, connect to the data source periodically, sample the enumerated values ​​of the key fields of the specified table, and define the standard expression. Step S1033: Construct an atomic semantic training mechanism based on a single-table scenario. With business combinations as the core, generate a mapping relationship between semantic questions and corresponding execution logic by analyzing high-frequency query patterns. For typical business tables, extract common business combination scenarios of core fields.

4. The AI-based system reconstruction method according to claim 3, characterized in that, Step S104 includes: Step S1041: Perform complex business logic operations in the old system and enable the database logging function; Step S1042: Export the database logs of complex business logic operations and complete the business scenario description; Step S1043: Submit the business scenario and database log documents to the large model for analysis. The large model analyzes the data operation objects, data operation order and data operation type in the business scenario to form business logic reasoning.

5. The AI-based system reconstruction method according to claim 4, characterized in that, Step S20 includes: Step S201: Design a business function review template. The business function review template maintains information such as function description, associated business domain, form operation, view operation, form interface screenshot, and view interface screenshot, based on the functional module. Step S202: For the functions in the old system that need to be refactored, complete the information and take screenshots of the pages according to the subsystem, module and function dimensions, and record them in the Excel template; Step S203: Summarize all functions that need to be refactored into the business function list. The function list includes the following elements: first-level menu, second-level menu, view tabs, menu description, view function operation description, view data column operation description, view search, form function operation description, business domain description, view screenshot, and form screenshot.

6. The AI-based system reconstruction method according to claim 5, characterized in that, Step S30 includes: Step S301: Import the completed business function list, which includes: subsystem name, module name, function name, function description, interface screenshot, and main operations. Step S302: For each function, match the corresponding data table according to the business rules described in natural language; Step S302 specifically includes the following steps: Step S3021: Extract the business rules described in natural language for each function from the "Business Function List"; Step S3022: Based on the business rules of the natural language description of each function, and based on intent recognition, match the data table or multiple tables with the highest confidence in the database DDL already stored in the vector library, thereby matching a data table for each function page. Step S303: Input the old system function images captured for each function into the VL model, and let the VL model identify the type, position and relationship of the interface elements; the interface elements include form fields, view components and operation buttons, wherein the form fields include field name, type, default value and validation rules, the view components include table columns, sorting rules, filter conditions, charts and layout information, and the operation buttons include add, edit, delete and query buttons; Step S304: Compare and analyze the VL model with the generated interface element data, and fill in the missing information; Step S305: For each function, call the low-code platform component library, render the visual interface based on the interface element data, and use a microservice architecture to split the interface modules to generate a new system interface. Based on the new system interface, generate editable low-code configuration data. The new system interface includes form service, view service, and component rendering service. Following step S305, the following is also included: Step S306: For the data dictionary implicit in the old system, AI is used to automatically detect and generate the data dictionary to provide metadata for the drop-down component of the page; Step S306 specifically includes: Step S3061: Access the corresponding tables and fields in the original database through the data tables and data fields corresponding to the interface elements to detect the old system data structure and sample data; Step S3062 involves using AI to identify the business meaning and value range of fields, and generating a standardized data dictionary. Specifically, this includes performing SQL function analysis by analyzing the values ​​of all data fields to generate a standardized data dictionary.

7. The AI-based system reconstruction method according to claim 6, characterized in that, Step S40 includes: Step S401, Interface element correction: Adjust the order of form fields, component styles, button positions, and add custom components; Step S402, Business rule configuration and code generation: Improve field validation logic and data association rules. For business operation buttons, support manual description of business rules in natural language through the interface. Combine the old system business SQL statements that have been sorted out with AI code generation tools to generate business logic code. Step S403, Permission Configuration: Configure the functional permissions and data permissions of the business system.

8. The AI-based system reconstruction method according to claim 7, characterized in that, The process following step S40 also includes: Step S501: For business systems containing workflows, configure the workflow as needed. The workflow configuration includes: basic process information configuration, task node configuration, process flow direction configuration, and process role configuration.

9. An AI-based system reconfiguration system, characterized in that, The system includes a memory, a processor, and an AI-based system refactoring program stored in the memory, the AI-based system refactoring program being executed by the processor to perform the steps of the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an AI-based system refactoring program, which, when executed by a processor, performs the steps of the method as described in any one of claims 1 to 8.

Images