Comprehensive management process design system and method based on AI code interpretation

Abstract

The invention relates to a comprehensive management process design system and method based on AI code interpretation, and belongs to the technical field of computer networks. The system comprises a process design console, an AI code interpreter and a business logic execution framework, the flow design console is used for providing a Web interface, so that a user can design and manage a flow in an interactive form, including interactive natural language programming, real-time feedback and preview, flow deployment and management, and historical record and version control; the AI code interpreter is used for converting business requirements and processes expressed by a user in a natural language form into executable codes, and comprises a natural language understanding sub-module, a code generation sub-module and a security verification sub-module; the business logic execution framework comprises a code execution sub-module, a data access sub-module and an exception handling sub-module. According to the method, the technical threshold of process design is lowered, personnel can easily master, and organic combination of intelligence and humanization is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of computer network technology, and specifically relates to a comprehensive management process design system and method based on AI code interpretation. Background Art

[0002] There are already some mature office process engine products in the industry. In addition to basic process management functions, some of them also integrate a series of simplified operation functions such as process customization and visual development. However, even so, these products still have some shortcomings in terms of user experience and intelligence in response to various process requirements of internal comprehensive management:

[0003] 1. Most of the personnel in each unit are non-computer professionals. Although simple processes can be built through visual development, for complex process design and development (such as comprehensive evaluation rules for personnel promotion), they can only provide relevant requirement documents. Without the cooperation of professional technicians, it is difficult to implement the process design in the information system.

[0004] 2. In terms of usage, most current process engines use code files (such as JSON, XML files, etc.) or flowcharts to describe specific processes, and lack interactive intelligent services such as natural language understanding, which limits the user's interaction efficiency and experience.

[0005] 3. The current process engine achieves a certain degree of automation through background configuration, such as file circulation and approval. However, for processes that require data to assist in making decisions, simple configuration is no longer sufficient to meet the requirements, and more intelligent means are needed to improve the quality and efficiency of the process.

[0006] Therefore, how to improve the intelligence level and user experience of the comprehensive management process has become an urgent problem that needs to be solved. Summary of the invention

[0007] 1. Technical issues to be resolved

[0008] The technical problem to be solved by the present invention is how to provide a comprehensive management process design system and method based on AI code interpretation to solve the problem of improving the intelligence level and user experience of the comprehensive management process.

[0009] (II) Technical solution

[0010] In order to solve the above technical problems, the present invention proposes a comprehensive management process design system based on AI code interpretation, which includes: a process design console, an AI code interpreter and a business logic execution framework;

[0011] The process design console is used to provide a simple, friendly and intuitive web interface, allowing users to design and manage processes in an interactive form, including: interactive natural language programming, real-time feedback and preview, process deployment and management, as well as history and version control;

[0012] The AI ​​code interpreter is used to convert the business requirements and processes expressed by users in natural language into executable code, including: natural language understanding submodule, code generation submodule and security verification submodule;

[0013] Natural language understanding submodule: Receives natural language input from users and understands their intentions and instructions, including the specific rules, steps, tasks, scenarios, and conditional judgments that users want to execute;

[0014] Code generation submodule: automatically generates corresponding codes or scripts based on the understood user intent and business logic. These codes can be correctly interpreted and executed by the process engine to complete the business process expected by the user.

[0015] Security verification submodule: ensures that the generated code undergoes strict security verification before execution to prevent the injection of malicious code or other security risks;

[0016] The business logic execution framework includes a code execution submodule, a data access submodule, and an exception handling submodule;

[0017] Code execution submodule: The process code generated by the AI ​​code interpreter is input into the actual process engine for code execution. The executed code corresponds to the business logic, including data processing, calculation, approval, decision-making, and task scheduling;

[0018] Data access submodule: used to make data requests to the databases and API interfaces in the integrated management systems to ensure the correct implementation of business processes;

[0019] Exception handling submodule: During the execution of the process code, the public interface of the process engine is called to view the process running status and results, monitor and handle errors and exceptions, and feed back this information to the process design console module.

[0020] A comprehensive management process design method based on AI code interpretation, the method comprising the following steps:

[0021] Step S1: Data preparation

[0022] S11. Preparation of training data for generative AI models. First, it is necessary to investigate the operating status and relevant historical data of the deployed process engine for actual work scenarios, and collect a large amount of natural language descriptions and corresponding process codes, including: business introduction, process description documents, process engine development specifications and sample codes. Then, data cleaning and preprocessing are performed to remove invalid, erroneous or inconsistent data, and to perform text segmentation, standardization, and entity recognition preprocessing operations to ensure data quality. After that, data annotation is performed, including intent classification, entity annotation, and relationship annotation. The annotation work requires the participation of business personnel to ensure the accuracy and consistency of the annotation. Data enhancement technology is used to generate more training samples, which is achieved through text rewriting, synonym replacement, and sentence structure adjustment methods. Finally, the prepared training data is fed to the AI ​​model for training and fine-tuning.

[0023] S12. Preparation of comprehensive management database: First, conduct a comprehensive inventory of existing specific data resources, including structured data, unstructured data and semi-structured data; then, classify and sort the data according to the nature and purpose of the data, establish standard databases, subject databases, special databases, etc., and form a scientific and standardized data organization structure; then, clean the data, remove errors and redundant information, standardize the data, and ensure the consistency and availability of the data; finally, set data access permissions to ensure the security of sensitive data;

[0024] Step S2: Natural language description of process requirements

[0025] Users enter natural language descriptions through the process design console to express their business process requirements. This step is the basis of the entire process design and requires users to provide detailed information about the process goals, steps, conditions, and required data.

[0026] The process design console provides interactive prompts, suggestions, and reference examples to help users improve process descriptions. At the same time, the system can also provide personalized suggestions and optimizations based on the user's historical operations and common process patterns to further improve design efficiency and quality. In addition, considering the complexity and professionalism of the business, this step also supports the introduction of professional terms and rules to ensure that the process design can accurately reflect the business logic and requirements.

[0027] Step S3: AI code interpreter processing

[0028] After the user completes the natural language description of the process requirements, the AI ​​code interpreter takes over the processing; first, the natural language understanding submodule conducts an in-depth analysis of the user's description, including syntax parsing, keyword extraction, and intent recognition, to ensure that the user's intent and process requirements are fully understood; then, the code generation submodule uses rule templates or generative models to convert the results of the understanding into executable process code, and then generates the corresponding code; the security verification submodule will conduct a comprehensive inspection of the generated code, including static code analysis, dynamic code monitoring, and permission control, to promptly discover potential problems;

[0029] Step S4: Process code execution

[0030] After the AI ​​code interpreter successfully generates the process code, the business logic execution framework is responsible for executing the codes and completing the business process. First, the code execution submodule inputs the generated code into the actual process engine to start the process execution. During the execution process, the submodule is responsible for monitoring the execution status of the code, recording the execution log, and ensuring that the process proceeds smoothly as designed. At the same time, the data access submodule will interact with the database and API interface in the integrated management system according to the needs of the process code to perform data query, update, and transmission operations. During the process execution, if an error or exception is encountered, the exception handling submodule will intervene in time, call the public interface of the process engine for status monitoring and fault diagnosis, and take corresponding handling measures. At the same time, the submodule will also feedback the exception information and processing results to the process design console to help users understand and manage the process status in a timely manner.

[0031] (III) Beneficial effects

[0032] The present invention proposes a comprehensive management process design system and method based on AI code interpretation. The innovative points of the present invention are as follows:

[0033] 1. Flexible process design based on natural language: Generative AI technology allows users to describe business processes in natural language and generate accurate, reliable and executable code language, so that non-computer professionals can also independently design, modify and execute complex processes. At the same time, adaptive code generation solutions are provided for simple and complex process requirements. This "on-demand" strategy can significantly reduce server computing costs.

[0034] 2. Unified use of the integrated management database: Through deep integration with the established integrated management database, the traditional way of accessing data resources through various process engines is moved forward and handled uniformly by the process design end. This shields the tedious work of connecting specific data to different process engines, reducing complexity and security risks.

[0035] 3. Simple and efficient user experience: Provide a clear process design interface, lower the technical threshold of process design through visual process operation and management, timely information feedback and preview, etc., so that personnel can easily get started and realize the organic combination of intelligence and humanity. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A system architecture diagram of the system involved in the present invention;

[0037] Figure 2 Design the technical architecture diagram of the console for the process;

[0038] Figure 3 This is the technical architecture diagram of the AI ​​code interpreter;

[0039] Figure 4 Execute the framework technology architecture for business logic;

[0040] Figure 5 This is a typical usage process of the method involved in the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, content and advantages of the present invention more clear, the specific implementation methods of the present invention are further described in detail below in conjunction with the drawings and examples.

[0042] To solve the above problems, the present invention provides a comprehensive management process design system and method based on AI code interpretation, which uses artificial intelligence technology to allow users to describe business requirements and processes in natural language, and then convert them into executable process codes through the AI ​​code interpreter to realize the automatic processing of business processes. The present invention aims to reduce the process development cost of the comprehensive management system, improve its intelligence level, and enhance work efficiency.

[0043] The system architecture of the present invention is as follows: Figure 1 shown.

[0044] The system involved in the present invention includes three modules: process design console, AI code interpreter, and business logic execution framework. In addition, we assume that the data inventory and sorting work of various systems for comprehensive management has been completed, and the structured, unstructured, and semi-structured data have been classified and managed in a detailed manner, forming a relatively complete, accurate, scientific and practical comprehensive management database such as standard library, theme library, and special library.

[0045] Each module is introduced below.

[0046] 1. Process design console

[0047] The process design console provides a simple, friendly, and intuitive web interface that allows users to design and manage processes in an interactive manner, including:

[0048] (1) Interactive natural language programming. It allows users to create or modify processes through natural language descriptions. It provides an interface similar to a conversational chat that displays the user's natural language input and the code converted by the AI ​​code interpreter. It also allows users to edit and adjust directly on the console.

[0049] (2) Real-time feedback and preview. Provide real-time feedback and preview functions when designing processes, so that users can immediately see the effects of natural language changes. It also includes verification of process logic, performance prediction, and possible errors or warning prompts.

[0050] (3) Process deployment and management. Once the process design is completed, users can directly deploy the process to the business logic execution framework through the console and verify its operation in the test environment. In addition, the console also provides management and monitoring tools that allow users to track the execution of the process, perform troubleshooting and performance optimization.

[0051] (4) History and version control. Save the history and versions of all process designs, and support operations such as version comparison, rollback, and replication of a specific process. This helps team collaboration and continuous process improvement.

[0052] In terms of technical implementation, the technical architecture of the process design console is as follows: Figure 2 As shown in the figure, the B / S (Browser / Server) architecture is used for general Web development: for the browser side, many Web front-end development frameworks can be used for construction, such as Vue, React, etc.; for the server side, microservices are used to integrate the design of various interface calls, monitoring, operation and maintenance functions. Typical back-end frameworks can use Java technology stack's SpringBoot, SpringCloud, etc.

[0053] For specific functional points, you can choose the more mature front-end or back-end components in the industry, for example: for the "interactive natural language programming" function, you can use the open source Dify.ai tool; for the "real-time feedback and preview" function, you can use the open source Monaco editor; for the "process deployment and management" function, you need to use the deployed process engine for association; for the "history and version control" function, you can connect to the Git version control system; for the "interface call" function, you can use the OpenAPI specification to develop various RESTFul interfaces; for the "monitoring" function, you can use lightweight log-based management and analysis methods; for the automated "operation and maintenance" function, you can use the Jenkins tool for automatic code building and continuous integration.

[0054] The process design console not only provides powerful automation capabilities, but also greatly improves the user's design experience and operational efficiency, making process creation and management more efficient and flexible.

[0055] 2. AI code interpreter

[0056] The AI ​​code interpreter is the core module of the system involved in the present invention, which is used to convert the business requirements and processes expressed by users in natural language into executable codes. It includes the following functions:

[0057] (1) Natural language understanding submodule: Receives natural language input from users and understands their intentions and instructions, including the specific rules, steps, tasks, scenarios, and conditional judgments that the user wants to execute.

[0058] (2) Code generation submodule: Automatically generates corresponding codes or scripts based on the understood user intent and business logic. These codes can be correctly interpreted and executed by the process engine to complete the business process expected by the user.

[0059] (3) Security verification submodule: ensures that the generated code undergoes strict security verification before execution to prevent the injection of malicious code or other security risks.

[0060] In terms of technical implementation, the technical architecture of the AI ​​code interpreter is as follows: Figure 3 As shown, it includes three sub-modules: natural language understanding, code generation, and security verification.

[0061] (1) Natural Language Understanding Submodule

[0062] In the system involved in the present invention, the natural language understanding submodule mainly relies on deep learning models, especially generative AI technology based on the Transformer architecture. The Transformer architecture can capture the long-distance dependencies of the input text through the self-attention mechanism. When understanding user input, it not only considers single words or sentences, but also considers the context of the entire conversation or document to provide a deeper level of contextual semantic understanding and higher accuracy. At the same time, this semantic recognition can extract code-related information (slots) from the user's natural language instructions. This information may include operation type, object, time, conditions, etc., to provide necessary parameters for subsequent code generation. In addition, if the user's input needs further confirmation or refinement, the dialogue flow can be driven by the dialogue management component to request more information or options. The dialogue management component can maintain a dialogue state and record the dialogue history and context.

[0063] Since there are some relatively mature open source generative AI models in the industry, in the design and development of comprehensive management processes, some high-quality small and medium-sized models can be selected as basic models on Github (an online code hosting platform) or Hugging Face (an AI community and platform focusing on natural language processing), and then trained and fine-tuned with internal data (including but not limited to: business introduction, process description documents, process engine development specifications, data resource catalog lists, etc.) so that the model can better understand the specific business context and process. After the AI ​​model is trained, it can be assembled and integrated with the server-side components of the "Process Design Console", that is, it can be packaged for engineering and productization. At the same time, this model is deployed on an internal GPU server, which has good protection in network security and data privacy.

[0064] (2) Code Generation Submodule

[0065] The core of the code generation submodule also utilizes generative AI technology. Its code generation model is a later component of the AI ​​model in the natural language understanding submodule. During the model training process, it will be trained on a large number of code libraries related to the process engine, and fine-tune the model for a specific process framework and process code language. It can generate structured code with high accuracy and strong adaptability based on the instructions generated after natural language understanding and the code parameters filled in the slots.

[0066] In order to reduce the computing pressure of the code generation model on the GPU server, two scenarios are considered for code generation: 1) For some simple, common or standardized processes, the code generation model is not used, but predefined templates and rules are used for rapid generation. These templates and rules can be adjusted and optimized according to actual usage. 2) For more complex and customized requirements, the code generation model is used to dynamically generate code.

[0067] After the preliminary code is generated, it needs to be further processed by the synthesizer and optimizer. The synthesizer is responsible for combining the generated code fragments into a complete program, and the optimizer mainly optimizes the generated code for possible deficiencies in performance, readability or maintainability, including code simplification, performance improvement, style unification, etc.

[0068] In the design and development of integrated management processes, in order to collect training data for a large number of code bases, on the one hand, existing codes of existing process engines can be expanded, and on the other hand, mature and common process engine specifications in the industry can be selected for production. For example, process codes based on the BPMN standard (a widely adopted standard for graphically representing business processes, which provides a set of standardized symbols to represent different process steps) are widely available in the open source community. For code synthesis and optimization, mature open source tool chains in the industry can be selected for processing, such as: JSONLint (a JSON code validator that helps optimize JSON data structures), XSLT (can be used for XML conversion, and can also be used to synthesize and optimize XML code), etc.

[0069] (3) Security check submodule

[0070] The security verification submodule adopts the following measures to ensure the security of the final code.

[0071] 1) Static code analysis: Static analysis is the first step to detect code security vulnerabilities and defects. In the present invention, professional static code analysis tools (such as SonarQube) can be introduced to perform pattern matching and process analysis based on the syntax and structure rules of the process code, so as to detect potential security issues or logical errors without running the code.

[0072] 2) Dynamic code execution monitoring: Dynamic monitoring is a security check performed when the code is running. In the present invention, a professional code execution inspection tool (such as Wireshark) can be introduced to execute the generated code in a test environment and monitor its behavior, including access to external resources, network communications, file operations, etc., to ensure that no security threats to the system are caused during the execution process.

[0073] 3) Code execution permission control: Ensure that the code is executed under limited permissions, especially when the code needs to run in a server or cloud environment. It can achieve fine-grained access control and limit the code's access to sensitive resources. In the present invention, the generated code can be executed in a Docker container to create a sandbox environment to isolate the application and limit the application's network access and file system access.

[0074] Through these technologies, the AI ​​code interpreter module will become a powerful, efficient, and interactive system component that can understand complex natural language descriptions, quickly and accurately generate executable code, and provide a good user experience. At the same time, the use of generative AI technology will make the understanding and generation process more accurate and flexible, and better adapt to changing business needs.

[0075] 3. Business logic execution framework

[0076] The business logic execution framework is the part of the system involved in the present invention that is responsible for the operation and implementation of the process code, and includes the following functions.

[0077] (1) Code execution submodule: The process code generated by the AI ​​code interpreter is input into the actual process engine for code execution. The executed code corresponds to the business logic, including data processing, calculation, approval, decision-making, task scheduling, etc.

[0078] (2) Data access submodule: The business logic execution framework has the authority to access specific data across systems. Therefore, if there is a need to obtain data in the process code, it can make data requests to the databases, API interfaces, etc. in the comprehensive management systems to ensure the correct implementation of the business process.

[0079] (3) Exception handling submodule: During the execution of the process code, the public interface of the process engine can be called to view the process running status and running results, monitor and handle errors and exceptions, and feed this information back to the process design console module.

[0080] In terms of technical implementation, the technical architecture of the business logic execution framework is as follows: Figure 4 As shown, it includes three sub-modules: code execution, data access, and exception handling.

[0081] (1) Code execution submodule

[0082] The code execution submodule itself does not contain a process engine, but provides a common interface layer for interacting with various deployed process engines. This interface layer can accurately pass the code generated by the AI ​​code interpreter to the process engine and trigger execution, while receiving execution status and result feedback from the process engine.

[0083] In addition, since different process engines may require codes in different formats or standards, the code execution submodule adopts the adapter pattern in the design pattern to provide a development adaptation conversion layer to convert the generated code into a format acceptable to a specific process engine.

[0084] (2) Data access submodule

[0085] If there is a need to obtain data in the process code (such as calling personnel information, schedule information, etc.), the data access submodule will be responsible for accessing the data. There is a premise here, that is, the data access submodule in the present invention has the authority to access specific data across systems, including interacting with multiple databases and API interfaces, and performing operations such as querying and updating specific data, regardless of whether the deployed process engine has the same authority. In development and implementation, database interface tools (such as JDBC, Hibernate, MyBatis, etc.) and API client libraries (such as Apache HttpClient) can be used to achieve this goal.

[0086] In addition, the data access submodule also has the ability to cache data. By using data caching tools such as Redis and Memcached, it can improve data access efficiency and reduce latency and resource consumption.

[0087] (3) Exception handling submodule

[0088] The exception handling submodule can capture and classify errors obtained in the code execution submodule, including system errors, business logic errors, insufficient resource errors, etc. that occur during the execution of the process engine. In terms of development and implementation, some mature framework libraries (such as Log4j and Logstash) can be used to record and manage error information.

[0089] At the same time, for different types of errors and exceptions, this submodule has corresponding processing strategies, including retry mechanism, rollback operation, resource release, etc., and feeds back error information and processing results to the process design console. In terms of development implementation, you can use mature framework libraries (such as Apache Commons Lang's ExceptionUtils) to call the action trigger interface of the process engine for exception handling.

[0090] Through these technical implementations, the business logic execution framework can provide a stable, efficient, and scalable environment to execute the process code generated by the AI ​​code interpreter to achieve complex integrated management business processes.

[0091] Embodiment 1:

[0092] The present invention is a comprehensive management process design system and method based on AI code interpretation. For the methods involved, the typical usage process is as follows: Figure 5 shown.

[0093] Step S1: Data preparation

[0094] In any AI-driven system or solution, data is a crucial component. In the present invention, data preparation needs to be done in advance, which is mainly divided into two aspects: preparation of generative AI model training data and preparation of comprehensive management database.

[0095] S11. Preparation of training data for generative AI models. This type of data is mainly used to support model training and fine-tuning of AI code interpreters to provide high-quality and high-accuracy natural language understanding capabilities. First, it is necessary to investigate the operating status and relevant historical data of the deployed process engine for actual work scenarios, and collect a large amount of natural language descriptions and corresponding process codes, including but not limited to business introductions, process description documents, process engine development specifications and sample codes, etc.; then, data cleaning and preprocessing are performed to remove invalid, erroneous or inconsistent data, and preprocessing operations such as text segmentation, standardization, and entity recognition are performed to ensure data quality; after that, data annotation is performed, including intent classification, entity annotation, relationship annotation, etc. The annotation work requires the participation of business personnel to ensure the accuracy and consistency of the annotation; in order to improve the generalization ability of the model, data enhancement technology can be used to generate more training samples. This can be achieved through methods such as text rewriting, synonym replacement, and sentence structure adjustment; finally, the prepared training data is fed to the AI ​​model for training and fine-tuning. For example, the domestic high-quality open source model GLM-4-9B can be selected.

[0096] S12, preparation of the integrated management database. This type of data is mainly to enable the designed process code to access the relevant business data. Traditional integrated management data is scattered in different business application systems, and has relatively strict authority control on its visitors. There is a premise in the present invention, that is, to have the authority to access specific data in various databases. As for how to define the scope of "specific data", it is divided by the data set of the business process involved in the integrated management system. The preparation process of the integrated management database is: first, a comprehensive inventory of existing specific data resources is carried out, including structured data (such as administrative records, personnel information, financial and budget information, item information, etc.), unstructured data (such as policy documents, audio and video materials, communication records, etc.) and semi-structured data (such as spreadsheets, e-mails, log files, etc.); then, according to the nature and purpose of the data, the data is classified and sorted, and a standard library, theme library, special library, etc. are established to form a scientific and standardized data organization structure; then, the data is cleaned, errors and redundant information are removed, and the data is standardized to ensure the consistency and availability of the data; finally, data access rights are set to ensure the security of sensitive data.

[0097] Step S2: Natural language description of process requirements

[0098] Users (staff or managers) enter natural language descriptions through the process design console to express their business process requirements. This step is the basis of the entire process design and requires users to provide detailed information about the process goals, steps, conditions, required data, etc. For example, for the "staff leave" process, users can enter: "Develop a business process for staff leave. The applicant first selects the leave type, then the time range and the reason for the leave. Then, the supervisor will approve it, and the approval result will be synchronized to the applicant."

[0099] In order to lower the technical threshold and improve the user experience, the process design console provides interactive prompts, suggestions and reference examples to help users improve the process description. For example, when a user enters a process step, the system can provide relevant operation options, best practice prompts or error corrections in real time to ensure the accuracy and feasibility of the process description. At the same time, the system can also provide personalized suggestions and optimizations based on the user's historical operations and common process patterns to further improve design efficiency and quality. In addition, considering the complexity and professionalism of the business, this step also supports the introduction of professional terms and rules to ensure that the process design can accurately reflect the business logic and requirements.

[0100] Step S3: AI code interpreter processing

[0101] After the user completes the natural language description of the process requirements, the AI ​​code interpreter takes over for processing. First, the natural language understanding submodule conducts an in-depth analysis of the user's description, including syntax parsing, keyword extraction, intent recognition, etc., to ensure that the user's intent and process requirements are fully understood. Then, the code generation submodule uses rule templates or generative models to convert the results of the understanding into executable process code. For example, for the above-mentioned "staff leave" process, the AI ​​code interpreter will extract entities such as "applicant", "approver", "leave time", "leave reason", etc., and will understand the action of "leave", intelligently add a series of rule constraints and conditional judgments, such as "whether the applicant is eligible for leave", "whether the applicant has remaining leave", "whether the leave time range is correct and compliant", "whether the leave reason matches the leave type", etc., and then generate the corresponding code.

[0102] During the code conversion process, the system will consider different process engines and execution environments to ensure that the generated code not only conforms to the business logic, but can also be executed correctly. In order to ensure the security and quality of the code, the security verification submodule will conduct a comprehensive check on the generated code, including static code analysis, dynamic code monitoring, permission control, etc., to promptly identify potential problems. In addition, considering the diversity and variability of the business, this step also supports rapid iteration and modification. Users can adjust the process description or choose different conversion strategies based on real-time feedback to achieve dynamic optimization and personalized customization of process design.

[0103] Step S4: Process code execution

[0104] After the AI ​​code interpreter successfully generates the process code, the business logic execution framework is responsible for executing these codes and completing the business process. First, the code execution submodule inputs the generated code into the actual process engine to start the process execution. During the execution process, this submodule is responsible for monitoring the execution status of the code, recording the execution log, and ensuring that the process proceeds smoothly as designed. At the same time, the data access submodule will interact with the database, API interface, etc. in the integrated management system according to the needs of the process code to perform data query, update, transmission and other operations. During this process, the system will strictly control data access rights and security to ensure data accuracy and confidentiality. During the process execution, if an error or exception is encountered, the exception handling submodule will intervene in time, call the public interface of the process engine for status monitoring and fault diagnosis, and take corresponding processing measures, such as error logging, user notification, retry or rollback, to ensure the stable operation of the process. At the same time, this submodule will also feedback the exception information and processing results to the process design console to help users understand and manage the process status in a timely manner.

[0105] For example, for the above-mentioned "staff leave" process, the generated code will be passed to the actual process engine (for example, in the form of XML code to the Activiti process engine). During the execution process, the leave type data (including adjustment leave, annual leave, sick leave, personal leave, etc.) and the applicant's personnel data (including position rank, leave balance, superior-subordinate organizational relationship, etc.) are needed. At this time, the data access submodule will go to the corresponding database to access this data and feedback to the process engine through the code execution submodule, and the process engine will make conditional rule judgments based on this. Afterwards, in the actual business process application scenario, when the applicant fills out the staff leave form, the process engine behind the service can automatically make the correct process constraints based on the applicant's personnel information.

[0106] The innovative points of the present invention are as follows.

[0107] 1. Flexible process design based on natural language: Generative AI technology allows users to describe business processes in natural language and generate accurate, reliable and executable code language, so that non-computer professionals can also independently design, modify and execute complex processes. At the same time, adaptive code generation solutions are provided for simple and complex process requirements. This "on-demand" strategy can significantly reduce server computing costs.

[0108] 2. Unified use of the integrated management database: Through deep integration with the established integrated management database, the traditional way of accessing data resources through various process engines is moved forward and handled uniformly by the process design end. This shields the tedious work of connecting specific data to different process engines, reducing complexity and security risks.

[0109] 3. Simple and efficient user experience: Provide a clear process design interface, lower the technical threshold of process design through visual process operation and management, timely information feedback and preview, etc., so that personnel can easily get started and realize the organic combination of intelligence and humanity.

[0110] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A comprehensive management process design system based on AI code interpretation, characterized in that: The system includes: process design console, AI code interpreter and business logic execution framework; The process design console provides a web interface that allows users to design and manage processes in an interactive manner, including: interactive natural language programming, real-time feedback and preview, process deployment and management, and history and version control; The AI ​​code interpreter is used to convert the business requirements and processes expressed by users in natural language into executable code, including: natural language understanding submodule, code generation submodule and security verification submodule; Natural language understanding submodule: Receives natural language input from users and understands their intentions and instructions, including the specific rules, steps, tasks, scenarios, and conditional judgments that users want to execute; Code generation submodule: automatically generates corresponding codes or scripts based on the understood user intent and business logic. These codes can be correctly interpreted and executed by the process engine to complete the business process expected by the user. Security verification submodule: ensures that the generated code undergoes strict security verification before execution to prevent the injection of malicious code or other security risks; The business logic execution framework includes a code execution submodule, a data access submodule, and an exception handling submodule; Code execution submodule: The process code generated by the AI ​​code interpreter is input into the actual process engine for code execution. The executed code corresponds to the business logic, including data processing, calculation, approval, decision-making, and task scheduling; Data access submodule: used to make data requests to the databases and API interfaces in the integrated management systems to ensure the correct implementation of business processes; Exception handling submodule: During the execution of the process code, the public interface of the process engine is called to view the process running status and results, monitor and handle errors and exceptions, and feed back this information to the process design console module.

2. The comprehensive management process design system based on AI code interpretation as claimed in claim 1, characterized in that: In the process design console, Interactive natural language programming: allows users to create or modify processes through natural language descriptions, providing a conversational chat interface that displays the user's natural language input and the corresponding code converted by the AI ​​code interpreter, allowing users to edit and adjust directly on the console; Real-time feedback and preview: Provide real-time feedback and preview when designing a process, allowing users to immediately see the effects of natural language changes. It also includes verification of process logic, performance prediction, and possible errors or warnings. Process deployment and management: Once the process design is completed, users can directly deploy the process to the business logic execution framework through the console, and can run and verify it in the test environment. In addition, the console also provides management and monitoring tools to allow users to track the execution of the process, troubleshoot and optimize performance; History and version control: Save the history and versions of all process designs, and support version comparison, rollback, and copy operations for a specific process.

3. The comprehensive management process design system based on AI code interpretation as claimed in claim 2, characterized in that: The process design console adopts B / S architecture for general Web development. For the browser side, it is built using the Web front-end development framework. For the server side, it adopts the microservice approach to integrate the design of various interface calls, monitoring, and operation and maintenance functions.

4. The comprehensive management process design system based on AI code interpretation as claimed in claim 1, characterized in that: The natural language understanding submodule specifically includes: The natural language understanding submodule relies on deep learning models, especially generative AI technology based on the Transformer architecture. In addition, if the user's input requires further confirmation or refinement, the dialogue flow is driven by the dialogue management component to request more information or options. The dialogue management component can maintain a dialogue state and record the dialogue history and context. In the design and development of comprehensive management processes, we select some high-quality small and medium-sized models on Github or HuggingFace as basic models, and then use internal data to train and fine-tune them so that the models can better understand specific business contexts and processes. After the AI ​​model is trained, it is assembled and integrated with the server-side components of the "Process Design Console", that is, it is packaged for engineering and productization. At the same time, this model is deployed on the internal GPU server, which has good protection in terms of network security and data privacy.

5. The comprehensive management process design system based on AI code interpretation as claimed in claim 1, characterized in that: The code generation submodule includes: For some simple, common or standardized processes, the code generation model is not used, but predefined templates and rules are used for rapid generation. For more complex and customized requirements, the code generation model is used to dynamically generate code. After the preliminary code is generated, it is further processed by a synthesizer and an optimizer; the synthesizer is responsible for combining the generated code fragments into a complete program, and the optimizer optimizes the generated code for possible deficiencies in performance, readability, or maintainability; In the design and development of integrated management processes, for the training data of a large number of code bases, on the one hand, the existing codes of the existing process engines are expanded, and on the other hand, the process engine specifications are selected for production; For code synthesis and optimization, select JSONLint and XSLT for processing.

6. The comprehensive management process design system based on AI code interpretation as claimed in claim 1, characterized in that: The safety check submodule comprises: Static code analysis: Static analysis is the first step to detect code security vulnerabilities and defects. Professional static code analysis tools are introduced to perform pattern matching and process analysis based on the syntax and structure rules of process codes, so as to detect potential security issues or logical errors without running the code. Dynamic code execution monitoring: Dynamic monitoring is a security check performed when the code is running. Professional code execution checking tools are introduced to execute the generated code in the test environment and monitor its behavior, including access to external resources, network communications, and file operations, to ensure that there is no security threat to the system during execution. Code execution permission control: By executing the generated code in a Docker container, a sandbox environment is created to isolate applications and limit their network access and file system access.

7. The comprehensive management process design system based on AI code interpretation as claimed in claim 1, characterized in that: The code execution submodule includes: the code execution submodule itself does not contain a process engine, but provides a common interface layer for interacting with various deployed process engines. This interface layer accurately passes the code generated by the AI ​​code interpreter to the process engine and triggers execution, and can receive execution status and result feedback from the process engine; the code execution submodule provides a development adaptation conversion layer by adopting the adapter pattern in the design pattern to convert the generated code into a format acceptable to a specific process engine.

8. The comprehensive management process design system based on AI code interpretation as claimed in claim 1, characterized in that: The data access submodule includes: if there is a need to obtain data in the process code, the data access submodule will be responsible for accessing the data; the data access submodule has the authority to access specific data across systems, including interacting with multiple databases and API interfaces, and performing query and update operations on specific data; in development and implementation, database interface tools and API client libraries are used; in addition, the data access submodule also has the ability to cache data, and improves data access efficiency by using Redis and Memcached data caching tools.

9. The comprehensive management process design system based on AI code interpretation as claimed in claim 1, characterized in that: The exception handling submodule includes: the exception handling submodule can capture and classify errors obtained in the code execution submodule, including system errors, business logic errors, and insufficient resource errors that occur during the execution of the process engine; in development implementation, Log4j and Logstash are used to record and manage error information; at the same time, for different types of errors and exceptions, the submodule has corresponding processing strategies, including retry mechanism, rollback operation, resource release, and feedback of error information and processing results to the process design console. In development implementation, based on Apache Commons Lang's ExceptionUtils, the action trigger interface of the process engine is called for exception handling.

10. A comprehensive management process design method based on AI code interpretation based on the system according to any one of claims 1 to 9, characterized in that: The method comprises the following steps: Step S1: Data preparation S11. Preparation of training data for generative AI models. First, it is necessary to investigate the operating status and relevant historical data of the deployed process engine for actual work scenarios, and collect a large amount of natural language descriptions and corresponding process codes, including: business introduction, process description documents, process engine development specifications and sample codes. Then, data cleaning and preprocessing are performed to remove invalid, erroneous or inconsistent data, and to perform text segmentation, standardization, and entity recognition preprocessing operations to ensure data quality. After that, data annotation is performed, including intent classification, entity annotation, and relationship annotation. The annotation work requires the participation of business personnel to ensure the accuracy and consistency of the annotation. Data enhancement technology is used to generate more training samples, which is achieved through text rewriting, synonym replacement, and sentence structure adjustment methods. Finally, the prepared training data is fed to the AI ​​model for training and fine-tuning. S12. Preparation of comprehensive management database: First, conduct a comprehensive inventory of existing specific data resources, including structured data, unstructured data and semi-structured data; then, classify and sort the data according to the nature and purpose of the data, establish a standard library, theme library and special library, and form a scientific and standardized data organization structure; then, clean the data, remove errors and redundant information, standardize the data, and ensure the consistency and availability of the data; finally, set data access rights to ensure the security of sensitive data; Step S2: Natural language description of process requirements Users enter natural language descriptions through the process design console to express their business process requirements. This step is the basis of the entire process design and requires users to provide detailed information about the process goals, steps, conditions, and required data. The process design console provides interactive prompts, suggestions, and reference examples to help users improve process descriptions. At the same time, the system can also provide personalized suggestions and optimizations based on the user's historical operations and common process patterns to further improve design efficiency and quality. In addition, considering the complexity and professionalism of the business, this step also supports the introduction of professional terms and rules to ensure that the process design can accurately reflect the business logic and requirements. Step S3: AI code interpreter processing After the user completes the natural language description of the process requirements, the AI ​​code interpreter takes over the processing; first, the natural language understanding submodule conducts an in-depth analysis of the user's description, including syntax parsing, keyword extraction, and intent recognition, to ensure that the user's intent and process requirements are fully understood; then, the code generation submodule uses rule templates or generative models to convert the results of the understanding into executable process code, and then generates the corresponding code; the security verification submodule will conduct a comprehensive inspection of the generated code, including static code analysis, dynamic code monitoring, and permission control, to promptly discover potential problems; Step S4: Process code execution After the AI ​​code interpreter successfully generates the process code, the business logic execution framework is responsible for executing the codes and completing the business process. First, the code execution submodule inputs the generated code into the actual process engine to start the process execution. During the execution process, the submodule is responsible for monitoring the execution status of the code, recording the execution log, and ensuring that the process proceeds smoothly as designed. At the same time, the data access submodule will interact with the database and API interface in the integrated management system according to the needs of the process code to perform data query, update, and transmission operations. During the process execution, if an error or exception is encountered, the exception handling submodule will intervene in time, call the public interface of the process engine for status monitoring and fault diagnosis, and take corresponding handling measures. At the same time, the submodule will also feedback the exception information and processing results to the process design console to help users understand and manage the process status in a timely manner.

Images