Low-code process generation method based on AI large model

Abstract

The invention relates to the technical field of low-code platforms, and provides a low-code process generation method based on an AI large model, and the method comprises the steps: S1, obtaining a business demand description text through a natural language interaction interface; s2, inputting the business demand description text into the trained AI large model, training the AI large model based on the security industry historical process data, and outputting a Markdown format intermediate file comprising a form definition module and a process definition module; s3, analyzing the Markdown intermediate file through a format conversion program, and generating a process configuration file conforming to a target low-code platform specification; and S4, deploying the process configuration file to a self-developed low-code platform, and generating an executable business process application. According to the method, prototype design of a form and a process is converted into a format which can be identified by a low-code platform and is imported, and finally a formal annual authorization entrusting process is generated, so that digital and automatic management of the business process is realized.

Description

Technical Field

[0001] This application relates to the field of low-code platform technology, and specifically to a low-code process generation method based on an AI big model. Background Art

[0002] Currently, the use of low-code systems in the securities industry is quite popular. Low-code platforms have simple and user-friendly interfaces and can be customized according to specific business needs, which improves business response speed. However, due to the complexity of the business and the compliance requirements of the securities industry, there are many form elements to fill in, and the approval process is also complicated. Therefore, there is still a certain degree of complexity and professionalism when defining and configuring low-code processes. In the initial stage of creation, it is difficult to define business requirements completely, intuitively, and accurately. Business personnel cannot generate forms and flowcharts through natural language expressions, and the final results will deviate from the actual business. Summary of the Invention

[0003] To help solve the above technical problems, this application provides a low-code process generation method based on an AI large model, which adopts the following technical solutions: A low-code process generation method based on an AI large model, wherein the method comprises: S1. obtaining a business requirement description text through a natural language interactive interface; S2. Input the business requirement description text into a trained AI model, which is trained based on historical process data from the securities industry and outputs a Markdown-formatted intermediate file containing a form definition module and a process definition module; S3. Parsing the Markdown intermediate file using a format conversion program to generate a process configuration file that complies with the specifications of the target low-code platform; S4. Deploy the process configuration file to the self-developed low-code platform to generate an executable business process application.

[0004] Preferably, the form definition module includes a Markdown structured description of the form field name, data type, validation rules and layout structure.

[0005] Preferably, the process definition module uses a business process diagram described by Mermaid syntax, including approval nodes, conditional branches and data flow relationships.

[0006] To summarize, the method of this application starts from the prototype design of forms and processes, to the conversion into a format recognizable by the low-code platform and import, and finally generates a formal annual authorization and delegation process, thus realizing the digitalization and automated management of business processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1A schematic diagram of an embodiment of a form prototype for describing the business requirements of this application; Figure 2 A schematic diagram of an embodiment of a form definition module for a Markdown format intermediate file of the present application; Figure 3 This is a schematic diagram of an embodiment of a process definition module for a Markdown format intermediate file of the present application. DETAILED DESCRIPTION

[0008] The present invention will be further described below with reference to the accompanying drawings. The structure and principle of the present invention will be very clear to those skilled in the art. It should be understood that the specific embodiments described herein are only intended to explain the present invention and are not intended to limit the present invention.

[0009] Figure 1 This is a schematic diagram of an embodiment of a form prototype for describing the business requirements of this application. Figure 2 This is a schematic diagram of an embodiment of a form definition module for a Markdown format intermediate file of this application. Figure 3 This is a schematic diagram of an embodiment of a process definition module for a Markdown format intermediate file of the present application.

[0010] Combine Figures 1 to 3 It is understood that the method of implementing a low-code process based on AI enhancement in this application may include: S1. Obtain business requirement description text through the natural language interactive interface; S2. Input the business requirement description text into the trained AI big model, which is trained based on the historical process data of the securities industry. First, a large number of user natural language and form Markdown matching samples are collected. The main data comes from the design and description of all processes in the current low-code, as well as manually produced diversified samples. Align the same form structure using different natural language expressions. Detailed annotations are made to the elements described in the natural language through field names, type descriptions, etc., to establish a unified Markdown output standard. Based on the privately deployed Deepseek 33B or 67B model, the model performance is continuously optimized through initial fine-tuning, reinforcement learning, and iterative optimization. The model performance is achieved by enhancing domain keywords and fine-tuning parameters. Finally, a Markdown format intermediate file containing the form definition module and the process definition module is output; S3. Parsing the Markdown intermediate file using a format conversion program to generate a process configuration file that complies with the specifications of the target low-code platform; S4. Deploy the process configuration file to the self-developed low-code platform to generate an executable business process application.

[0011] The form definition module includes a Markdown structured description of the form field name, data type, validation rules, and layout structure. The process definition module uses a business process diagram described in Mermaid syntax, including approval nodes, conditional branches, and data flow relationships.

[0012] In step S3, the format conversion program converts the Markdown form and process prototypes into a format that the low-code platform can recognize and process. Because the low-code platform has its own specific data structure and logic specifications, the conversion program can automatically parse the Markdown field definitions, process logic, and other information, and perform the corresponding format conversion and code generation.

[0013] The design ideas of the format conversion program are as follows: 1) Design a Markdown syntax interpreter to convert Markdown input into an AST syntax tree through an abstract syntax tree (AST); 2) Convert the AST syntax tree into a JSONSchema that can be recognized by the internal low-code platform, and convert basic types such as Input, List, Checkbox, Radio, Textarea, etc. into the corresponding form definitions of the low-code. In this way, the form field names, types, prompt information, required conditions, etc. in Markdown can be mapped to the form elements of the low-code platform to ensure that each field can be correctly created and configured in the low-code platform. Similarly, convert the process nodes, flow conditions, participants and other information in Markdown into the process definition language or visual process design elements of the low-code platform, so that the process can be accurately reproduced on the low-code platform.

[0014] The converted data and process definitions are imported into the low-code platform, where formal forms and processes are generated. The low-code platform provides a visual interface, allowing developers or business personnel to further adjust and optimize the generated forms and processes, such as setting field validation rules and approval permissions for process nodes.

[0015] After adjustments and optimizations on the low-code platform, a formal annual authorization and delegation process was ultimately generated. This process features a complete form interface that accurately collects user input. Furthermore, the process logic strictly adheres to design requirements, ensuring authorization and delegation flow and are reviewed at every stage according to established conditions and rules.

[0016] By prototyping and importing into a low-code platform, we significantly shortened the process development cycle and improved work efficiency. Standardized forms and process design reduce the potential for human error and ensure the accuracy and compliance of the authorization and delegation process. Formal processes run on a low-code platform, facilitating subsequent maintenance and optimization based on business changes, such as adding new fields and adjusting process nodes.

[0017] The following content is the form definition module of the Markdown format intermediate file output by the AI ​​model, which is the same as Figure 2 Corresponding.

[0018] # Annual Authorization Form ## Basic Information |Field name|Field type|Prompt information|Required conditions| | ---- | ---- | ---- | ---- | |Process name|Text input box|Please enter|No| |Applicant|Text input box|Please enter|No| |Situation description|Text input box|Please enter|No| |Related processes|Drop-down selection box|Please select|No| |Authorization attribution|Drop-down selection box|Please select|Yes (marked with *)| |Authorization information|Text input box|Please enter|No| ## Format Contract Subtable |Serial Number|Format Contract|Business|Final Draft of Format Contract|Authorizer|Authorized Party| | ---- | ---- | ---- | ---- | ---- | ---- | |1|Drop-down selection box, prompt "Please select"|Drop-down selection box, prompt "Please select"|File upload box, prompt "Click to upload"|Drop-down selection box, prompt "Please select"|Drop-down selection box, prompt "Please select"| ## Time and number information |Field name|Field type|Prompt information|Required conditions| | ---- | ---- | ---- | ---- | |Application authorization start date|Date selection box|Please select|No| |Application authorization end date|Date selection box|Please select|No| |Authorization Date|Date Selection Box|Please Select|No| |Authorization number|Text input box|Please enter|No| |Official authorization start time|Date selection box|Please select|No| |Official authorization end time|Date selection box|Please select|No| The following content is the process definition module of the Markdown format intermediate file output by the AI ​​model. Figure 3 Corresponding.

[0019] A[Process Initiation] -->|Everyone| B{Add Condition} B -->|Condition 1| C[Department Leader Review] B -->|Condition 2| N[Financial Review] C -->|Designee A| D [Deputy Chief Auditor] D -->|Designee B| E[Company Leadership Review] N -->|Designee A| D [Deputy Chief Auditor] E -->|General Manager| F[Notify the organizer for review].

Claims

1. A low-code process generation method based on AI big model, characterized by: The method comprises: S1. Obtain business requirement description text through the natural language interactive interface; S2. Input the business requirement description text into a trained AI model, which is trained based on historical process data from the securities industry and outputs a Markdown-formatted intermediate file containing a form definition module and a process definition module; S3. Parsing the Markdown intermediate file using a format conversion program to generate a process configuration file that complies with the specifications of the target low-code platform; S4. Deploy the process configuration file to the self-developed low-code platform to generate an executable business process application.

2. The low-code process generation method based on AI big model according to claim 1 is characterized in that The form definition module includes a Markdown structured description of the form field name, data type, validation rules, and layout structure.

3. The low-code process generation method based on AI big model according to claim 1 is characterized in that The process definition module uses a business process diagram described in Mermaid syntax, which includes approval nodes, conditional branches and data flow relationships.

Images