An application development platform and method integrating AI generation and visual orchestration

Abstract

This application provides an application development platform and method that integrates AI generation and visual orchestration, belonging to the field of computer software development. The platform includes: an AI engine for receiving natural language requirements and generating an atomic application skeleton containing data models, application interfaces, business processes, and rules; and an interface engine for providing a visual canvas to present the skeleton and allowing users to edit it through visual operations. The method includes intent parsing and generation steps and visual editing steps. Visual editing includes modifying at least one of the application skeleton's business processes, business rules, and application interfaces. This application solves the problems of high barriers to AI application development and difficulty in deeply customizing AI-generated content. Through a closed loop of AI generation and deep visual editing, it achieves visual modification of the application's deep logic, lowering the development threshold and improving development efficiency and the personalization of the application.

Description

Technical Field

[0001] This application relates to the field of computer software development, and in particular to an application development platform and method that integrates AI generation and visual arrangement. Background Technology

[0002] With the popularization of artificial intelligence technology, the demand for intelligent applications is growing in all walks of life. However, traditional AI application development models usually require developers to have in-depth programming knowledge, a solid foundation in machine learning theory, and experience in deploying complex models, which sets up extremely high technical barriers for non-professionals to participate in AI application innovation.

[0003] To lower the development barrier, some low-code or no-code development platforms have emerged on the market. However, most of these platforms focus on building traditional business applications and lack sufficient support for integrating and developing complex AI functions. When it is necessary to call AI models and process intelligent decision-making logic, these platforms often require users to write additional code or perform complex configurations, which greatly diminishes their "low-code" advantages.

[0004] On the other hand, generative AI technologies, exemplified by large-scale language models, can generate preliminary application code or interfaces based on natural language descriptions, bringing new possibilities to application development. Some existing methods have already achieved the generation of web forms from natural language and can fine-tune the generated results using low-code drag-and-drop editors. However, in these methods, the AI-generated content is often general and preliminary, and the visual editing capabilities are limited to adjusting simple elements such as interface layout. For the complex business processes, business rules, data interface integrations, and even the parameter configuration of the AI ​​model itself behind the application, users still lack effective and unified visual editing tools. This lack of editing capabilities for the "last mile" of deep application logic makes it difficult for AI-generated content to be truly implemented in complex real-world applications, failing to fundamentally solve the problems of high barriers to AI application development and limited personalization. Summary of the Invention

[0005] The purpose of this application is to provide an application development platform and method that integrates AI generation and visual orchestration, in order to solve the technical problems in the existing technology, such as the high threshold for AI application development and the difficulty in effectively and deeply personalizing AI-generated content to meet complex business needs, resulting in low development efficiency and difficulty in application implementation.

[0006] To achieve the above objectives, this application provides an application development platform that integrates AI generation and visual orchestration, comprising: an AI engine for receiving natural language application development requirements from users and generating an atomic application skeleton containing a data model, application interfaces, and basic business logic consisting of business processes and business rules; and an interface engine for providing a visual canvas to present the atomic application skeleton and allowing users to edit the application skeleton through visual operations; wherein, the visual editing includes modifying at least one of the business processes, business rules, and application interfaces of the application skeleton.

[0007] Optionally, the interface engine includes a visual process engine component, and the modification of the business process is accomplished by operating the process engine component on the visual canvas.

[0008] Optionally, the interface engine includes a visual rules engine component, and the modification of the business rules is accomplished by operating the rules engine component on the visual canvas.

[0009] Optionally, the interface engine includes a visual interface aggregation engine component, and the modification of the application interface is accomplished by operating the interface aggregation engine component on the visual canvas.

[0010] Optionally, the visual editing also includes configuring parameters of the large language model used by the AI ​​engine.

[0011] This application also provides an application development method that integrates AI generation and visual orchestration, including: an intent parsing and generation step, which receives the user's natural language application development requirements through an AI engine and generates an atomic application skeleton containing a data model, application interface, and basic business logic composed of business processes and business rules; and a visual editing step, which provides a visual canvas, presents the atomic application skeleton on the visual canvas, and responds to the user's visual operation to edit the application skeleton; wherein, the visual editing step includes modifying at least one of the business processes, business rules, and application interfaces of the application skeleton.

[0012] Optionally, the modification of the business process is accomplished by responding to the user's visual process editing operation on the visual canvas.

[0013] Optionally, the modification of the business rules is accomplished by responding to the user's visual rule editing operation on the visual canvas.

[0014] Optionally, the modification of the application interface is accomplished by responding to the visualization interface aggregation operation performed by the user on the visualization canvas.

[0015] Optionally, the visualization editing step also includes configuring parameters of the large language model used by the AI ​​engine for performing the intent parsing and generation steps.

[0016] Compared with existing technologies, the technical solution provided in this application has the following beneficial effects: 1. Significantly lowers the technical threshold. Through the "natural language generation + visual editing" model, non-programmer users can also lead the development process of AI applications without writing complex code, greatly simplifying the development process. 2. Enables high-quality personalized development. This application breaks through the limitation of existing technologies that can only edit the user interface. Through the ability to deeply edit business processes, rules, interfaces, and even AI model parameters in a visual manner, it ensures that the general prototype generated by AI can be finely customized to meet complex personalized business needs, effectively bridging the "last mile" of AI application implementation. 3. Significantly improves development efficiency. The automatic generation capability of the AI ​​engine significantly reduces the initial development workload, while the unified visual editing environment accelerates the modification and iteration process of applications, realizing a rapid closed loop from requirements to deployment. 4. Enhances the openness and flexibility of the platform. By atomicating software capabilities, supporting access to local large models, and providing users with an extensible component library, the platform can flexibly adapt to and support the development of intelligent agents or applications in all business scenarios, exhibiting extremely high scalability. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This application provides an overall architecture diagram of an application development platform that integrates AI generation and visual orchestration.

[0019] Figure 2 A flowchart illustrating an application development method that integrates AI generation and visual orchestration, provided as an embodiment of this application;

[0020] Figure 3 A signaling interaction timing diagram for an application development method provided in an embodiment of this application;

[0021] Figure 4 This is a schematic diagram of the expense reimbursement process in an embodiment of this application.

[0022] The main reference numerals in the attached figures are explained as follows: 10 - User; 20 - AI Engine; 30 - Atomization Module for Software Capabilities; 31 - Process Engine Component; 32 - Rule Engine Component; 33 - Interface Aggregation Engine Component; 40 - Interface Engine; 41 - Visualization Canvas; 42 - Atomic Component Library; 50 - Deployed Application; 60 - External System API; S101 - Receiving Natural Language Request Step; S102 - Intent Parsing and Generating Atomized Application Skeleton Step; S103 - Mapping Components and Presenting Them on the Visualization Canvas Step; S104 - User Visual Re-editing Step; S105 - Real-time Preview Step; S106 - One-Click Deployment Step; S201 - Employee Submitting Expense Report Step; D201 - Decision on Whether the Amount is Greater Than 5000 Yuan; S202 - Manager Approval Step; S203 - Director Approval Step; S204 - Financial Payment Step. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0024] Example 1

[0025] This embodiment takes the development of a customer relationship management system as an example, and provides a specific implementation of an application development platform and method that integrates AI generation and visual orchestration. It elaborates on the complete process from receiving user natural language requirements to finally deploying the application.

[0026] Figure 1 This is a schematic diagram of the overall architecture of an application development platform that integrates AI generation and visual orchestration, provided as an embodiment of this application. Figure 1 As shown, the platform mainly includes an AI engine 20, a software capability atomization module 30, and a UI engine 40. User 10 can quickly build and deploy pre-deployed applications 50 that meet their specific needs by interacting with the platform.

[0027] Specifically, the AI ​​engine 20, serving as the platform's intelligent entry point, has the core function of receiving and deeply understanding the application development requirements submitted by user 10 in natural language. This AI engine 20 integrates an advanced natural language processing module, capable of performing syntactic analysis, entity recognition, and intent classification on the input text, thereby accurately capturing the user's core needs.

[0028] The software capability atomization module 30 forms the functional foundation of the platform. Its function is to pre-encapsulate complex software functions such as business process management, business rule judgment, and external interface calls into a series of standardized, reusable atomic-level components. In one embodiment of this application, these components include, but are not limited to, a visualization process engine component 31, a rule engine component 32, and an interface aggregation engine component 33. These atomic functional components are included in the atomic-level component library 42 of the interface engine 40, for use by the AI ​​engine 20 during the generation phase or for selection by the user 10 during the editing phase.

[0029] The interface engine 40 provides the main interface for user 10 to interact with the platform, including a WYSIWYG visual canvas 41. This visual canvas 41 not only presents the initial application skeleton generated by the AI ​​engine 20, but more importantly, it provides a complete set of visual editing tools, allowing user 10 to perform comprehensive and in-depth customization of the application through intuitive operations such as dragging, clicking, and configuration. The atomic-level component library 42, as part of the interface engine 40, provides a rich set of pre-built components (e.g., over 60 common components covering forms, tables, charts, layouts, etc.) as well as a user-expandable custom component library.

[0030] The following will combine Figure 2 The process of an application development method that integrates AI generation and visual arrangement, as provided in the embodiments of this application, will be described. Figure 2 This is a flowchart illustrating the method. The entire development process forms a closed loop from requirements to application, mainly including the following steps:

[0031] In step S101, a natural language request is received. User 10, for example, a sales manager without a programming background, inputs a natural language text describing their request through the platform's input interface. Specifically, in this embodiment, user 10 may input: "I need a CRM system to manage customer information, including a customer list, customer details page, the ability to add new customers, and the ability to search by customer name."

[0032] In step S102, intent parsing and the generation of an atomic application skeleton are performed. Specifically, after receiving a user request, the AI ​​engine 20 initiates an intent parsing process. This process first identifies the core entity "Customer" and related operations such as "Manage," "List," "Details," "Add," and "Search." Based on these parsing results, the AI ​​engine 20 automatically generates an atomic application skeleton. It should be noted that this application skeleton is not a simple interface prototype, but a structured data collection, the contents of which may include, but are not limited to: 1. Data Model: Define a data structure named "Customer," containing fields such as "Customer Name," "Contact Number," "Email," "Company Address," and "Customer Status," and specify the data type for each field, such as text, number, date, etc. 2. Application Interface: Generate a set of application interface definitions conforming to the RESTful style, such as GET / api / customers for retrieving the customer list, POST / api / customers for creating a new customer, GET / api / customers / {id} for retrieving detailed information of a specific customer, and GET / api / customers?name={keyword} for implementing the search function. 3. Basic Business Logic: Generate preliminary business processes and business rules. For example, define the business process for "Add New Customer," from opening the form to saving the data; define basic business rules, such as the "Contact Number" field must be 11 digits and the "Email Address" field must conform to a standard format. 4. Interface Layout: Generate basic interface layout suggestions for functions such as "Customer List," "Customer Details," and "Add Customer," for example, using table components for the list page and form components for the details page and the add page.

[0033] Subsequently, in step S103, component mapping is performed and the components are presented on the visualization canvas. After receiving the atomic application skeleton generated by the AI ​​engine 20, the interface engine 40 performs a smart mapping operation. Based on the definitions in the skeleton, it automatically selects and instantiates the corresponding components from the atomic component library 42. For example, the definition of the customer list is mapped to a table component, and the definition of customer details is mapped to a read-only form component, which are then presented on the visualization canvas 41 to form a preliminary application interface that is visible and interactive to the user.

[0034] In step S104, the user can perform visual re-editing. This step is crucial for transforming the AI-generated generic prototype into a highly personalized application. User 10 can deeply modify and enhance the application entirely through visual operations on the visualization canvas 41, without writing any code. Specifically, the operations that User 10 can perform include: 1. Interface and style adjustments: The user can directly drag the column headers of the table component on the canvas to adjust the order and width of each column in the "Customer List," or modify the font, color, border, and other styles in the right-hand property panel. 2. Function enhancement: If the user finds the customer details page information insufficiently intuitive, they can drag a "chart component" from the atomic component library 42 to the canvas area of ​​the details page. The configuration panel of the chart component will then pop up, allowing the user to bind its data source to the customer's consumption record interface, select "bar chart" as the chart type, set the X-axis to "month," and the Y-axis to "consumption amount," thereby dynamically generating a consumption analysis chart for each customer. 3. Custom Component Integration: Assume the enterprise developers have pre-developed a "map component" to display customer locations on a map and uploaded it to the user's extended component library. The sales manager can find this component in the component library, drag and drop it onto the customer details page, and bind the component's "address" input attribute to the "company address" field in the current customer data model in its property panel. 4. Business Process Modification: If the user wants the system to automatically send a welcome email after successfully adding a new customer, they can select the "Save" button in the "Add Customer" form, right-click, and select "Edit Business Process." At this time, the interface engine 40 will load and display the editing interface of the visual process engine component 31. The original AI-generated process only contains three nodes: "Start," "Save Data," and "End." The user can drag a "Send Email" node from the node library, place it between "Save Data" and "End," and connect it to the process line. Then, the user clicks the "Send Email" node, and in the pop-up configuration window, sets the recipient address to the current new customer's "email address" field and edits the email's title and content template. This operation intuitively modifies the application's backend business process. 5. Application Interface Modification and Integration: In order to enrich customer profiles, users may want to query customers' credit scores from external credit reporting systems. Figure 3This illustrates the signaling interaction process between entities in this embodiment. Users can drag and drop an instance of a visual interface aggregation engine component 33 from the atomic-level component library 42 and configure it. In the configuration interface, the user enters the URL of the external credit scoring system API 60, selects "POST" as the request method, adds the API key required for authentication in the request header, and maps the customer's ID number to the parameters of the external API in the request body. Next, the credit score field returned by the API is bound to a newly added "Credit Score" display component on the customer details page. Thus, when the user views the customer details, the deployed application 50 will automatically call the external system API 60 and display the obtained data in real time. 6. AI Model Parameter Configuration: If the user wants to add a "Smart Summary" function to the customer details page to automatically generate a summary based on the customer's follow-up records, a button and a text display box can be dragged in. Then, an AI call function is associated with the button's click event. In the attribute configuration area of ​​the interface engine 40, the user can directly configure the parameters of the large language model used in this call. For example, a user can select a local model fine-tuned from enterprise private data from a drop-down list and design a prompt in the "Prompt Word" input box: "Please play the role of a senior sales consultant and generate a customer profile summary of no more than 200 words based on the following customer follow-up records: {records}, focusing on analyzing their potential needs and follow-up suggestions." Here, {records} is a dynamic variable linked to the customer's follow-up record data. Furthermore, users can adjust parameters such as the "Temperature" parameter (e.g., setting it to 0.7 to increase the creativity of the summary) and the "Maximum Output Length" via a slider or input box.

[0035] In step S105, the platform provides a real-time preview function. User 10 can click the "Preview" button at any time, and the system will immediately display the actual running effect of the application in a new window or in the preview mode of the current canvas. The user can interact with the application in the preview to test all functions such as adding customers, viewing details, and triggering email sending to verify whether the modifications meet expectations. If problems are found, the user can return to the editing interface (S104) to continue making modifications, thus forming an efficient "edit-preview" iterative loop.

[0036] In step S106, one-click deployment is performed. Once user 10 is completely satisfied with the application, they simply click the "One-Click Deployment" button. The platform backend will automatically complete all the work of code building, packaging, containerization, and deployment to the target server environment, ultimately generating a standalone, deployed application 50.

[0037] In this embodiment, users can build and deploy a CRM system integrated with AI capabilities. Compared to traditional methods that require weeks or even months of development time, the method described in this embodiment significantly improves development efficiency and verifies the beneficial effects of this application in lowering technical barriers and achieving high-quality personalized development.

[0038] Example 2

[0039] This embodiment takes the development of an internal expense reimbursement system as an example, aiming to illustrate how the platform provided in this application utilizes the visual process engine component 31 and the rule engine component 32 to build and edit applications with complex business logic.

[0040] Its development process can be followed in the same way. Figure 2 The steps are shown below. First, an administrative manager (User 10) inputs a natural language request into the platform: "Create an expense reimbursement application where employees can submit expense reports, which then need to be approved by their direct manager. If the amount exceeds 5,000 yuan, it also needs to be approved by the director. After approval, the finance department will make the payment."

[0041] After performing the intent parsing and generating the atomic application skeleton (S102), the AI ​​engine 20 can generate a preliminary application skeleton, which includes a "reimbursement form" data model (fields such as: claimant, reason, amount, invoice image, etc.), a reimbursement submission form page for employees to fill out, and an approval list page for managers to view. The basic business process it generates may only be a simple linear process of "submission-approval-end".

[0042] Next, in the step of visual re-editing by the user (S104), the administrative manager will further customize this basic process and rules.

[0043] First, there is the visual arrangement of business processes. The administrative manager locates the business process associated with the "Submit" button on the visualization canvas 41 and enters the editing interface of the process engine component 31. This interface is presented as a graphical flowchart editor.

[0044] Figure 4This is a schematic diagram of the expense reimbursement process constructed in this embodiment. The administrative manager constructs this process using the following drag-and-drop operations: 1. First, for the initial process generated by AI, the administrative manager can delete the common "Approval" node. 2. Then, from the node library on the left, drag a decision node "Determine if the amount is greater than 5000 yuan D201" onto the canvas and connect it after the "Employee submits expense report S201" node. 3. Click on the decision node D201, and set its judgment condition in the attribute panel on the right. Specifically, select the "Amount" field in the "Expense Report" data model, select ">" as the operator, and enter "5000" as the value. 4. Next, drag two "Manual Task" type nodes from the node library onto the canvas. Name the first node "Manager Approval S202" and configure its handler as "the submitter's direct superior"; name the second node "Director Approval S203" and configure its handler as "Director". 5. Subsequently, draw a process line from the "Yes" branch of decision node D201 (i.e., amount > 5000 yuan) and connect it to the "Director Approval S203" node; simultaneously, draw another process line from the "No" branch of decision node D201 and connect it to the "Manager Approval S202" node. 6. To eliminate the need for repeated manager approval after director approval, the output of the "Director Approval S203" node can be directly connected to subsequent steps. 7. Finally, drag in a "Manual Task" node, name it "Financial Payment S204," and configure its handler as the "Finance" role. Connect the approval outputs of both the "Manager Approval S202" and "Director Approval S203" nodes to the "Financial Payment S204" node.

[0045] Through the above purely graphical operations, an administrative manager can modify a simple linear process into a complex business process that includes conditional branches and multi-level approvals. The logic of this process is similar to... Figure 4 Consistent with the above.

[0046] Correspondingly, users can also configure business rules visually. If the administrative manager wants to add some automated processing rules to improve efficiency, they can open the editing interface of the visual rule engine component 32. This interface provides a rule builder in a natural language style similar to "if...then...". As an optional implementation, users can create the following rules: 1. Rule 1 (Auto Approval): Construct the following logic through drop-down menus and input boxes: "If the reason for the expense report includes intra-city transportation and the hourly portion of the expense report's occurrence time is later than 10 PM, then set the approval status of the expense report to Auto Approval." This rule can be executed after the expense report is submitted but before entering the manual approval process. For regular small-amount expense reports that meet the conditions, manual approval can be skipped directly, and the process can proceed to the financial payment stage. 2. Rule 2 (Budget Verification): Create another rule: "If the expense type of the expense report is equal to entertainment expenses and the amount of the expense report is greater than 5% of the remaining entertainment expenses in the department's budget, then set the approval status of the expense report to Rejection, and set the rejection reason to Entertainment Expenses Exceed Budget." This rule enables linked verification with budget data, allowing non-compliant applications to be intercepted early in the process.

[0047] This embodiment shows that the platform provided in this application, through the visual process engine component 31 and rule engine component 32, enables business personnel without programming skills to design, build and modify complex business logic and automation rules, thereby quickly developing internal applications that meet the actual management needs of enterprises.

[0048] Example 3

[0049] This embodiment takes the construction of an intelligent question-answering robot based on an enterprise's private knowledge base as an example, and focuses on how the platform provided in this application can use the visual interface aggregation engine component 33 and the AI ​​large model parameter configuration function to quickly develop and customize AI native applications.

[0050] The development process begins when a product operations staff member (User 10) inputs a requirement: "I need a Q&A robot that can be embedded in our product website and answer customers' pre-sales inquiries based on the product documents I upload."

[0051] After executing step S102, AI Engine 20 can generate a basic dialogue interface application skeleton containing a chat input box, a message display area, and a send button, and reserve a backend logic call point for obtaining answers.

[0052] In the visual re-editing step (S104), product operations personnel need to complete two core configurations: connecting to the private knowledge base and fine-tuning the AI ​​model.

[0053] The first core configuration involves connecting to the data source via the visual API aggregation engine component 33. In the application's backend logic editing view, operations personnel locate the AI-reserved "Knowledge Retrieval" section and drag in an instance of the API aggregation engine component 33. The configuration process is as follows: 1. In the component's property panel, configure the "API Address" to the retrieval API address of the company's internal knowledge base system, such as https: / / internalkb.mycompany.com / api / v1 / search. 2. Set the "Request Method" to "GET". 3. In the "Request Parameters" configuration area, add a parameter named "query" and dynamically bind its value to the content entered by the user in the "Chat Input Box". 4. Necessary authentication information can also be added to the "Request Headers," such as an Authorization header with a value of an API access token. 5. In the "Response Processing" section, specify that the content under the array path "data.hits" in the JSON data returned by the API is the required knowledge fragment, and name the output of this path "retrieved-context".

[0054] With the above form-based configuration, operators can connect to external data interfaces and extract data, providing the necessary contextual information for subsequent AI responses, and the entire process requires no coding.

[0055] The second core configuration involves the visual configuration of the large language model that drives the robot's answers. The operations staff selects the AI ​​module responsible for generating the final answer, and its property panel displays a wealth of configurable items: 1. Model Selection: The panel provides a "Model" dropdown list, listing all available large models integrated into the platform. The operations staff can choose a general model or a "MyProduct-Expert-V2" model that the company has fine-tuned using proprietary product documentation and deployed on a local server. Choosing the latter significantly improves the professionalism and accuracy of the answers. 2. Prompt Word Engineering: The panel provides a multi-line text box for editing "System Prompt Words." The operations staff can enter the following instructions: "You are a professional customer service representative for XX product. Your task is to answer user questions in a friendly, concise, and accurate manner based solely on the background information provided below. If the background information does not contain relevant information, politely reply, 'Sorry, I cannot provide accurate information about this question at the moment. I suggest you contact our human customer service.' Fabricating information is absolutely prohibited. Background information: {retrieved_context}. User question: {user_query}." Here, `{retrieved_context}` is dynamically bound to the output of the previous interface aggregation engine component, while `{user_query}` is bound to the user's chat input. 3. Parameter Fine-tuning: The panel also provides a series of sliders and input boxes for adjusting the model's generation behavior. "Temperature": To make the answers more stable and closer to the original knowledge base text, operators can lower the temperature slider to 0.1. "Maximum Output Tokens": To avoid overly long answers, this value can be set to 300. "Repetition Penalty": This value can be set to 1.1 to reduce the likelihood of the model generating repetitive statements.

[0056] After completing these two core configurations, operations staff can use the real-time preview (S105) function to test the chatbot's dialogue, verifying the accuracy and style of its responses. If unsatisfied, they can return at any time (S104) to adjust prompts or model parameters until the desired effect is achieved. Finally, through one-click deployment (S106), this customized intelligent question-answering chatbot is successfully published on the company's official website.

[0057] It is understood that this embodiment clearly demonstrates that the solution of this application greatly simplifies the development process of AI-native applications. Non-technical users can easily complete the integration of external data and the deep customization of AI model behavior through a visual approach, greatly improving the development flexibility and final quality of AI applications.

[0058] Example 4

[0059] This embodiment takes the creation of an IoT device monitoring dashboard as an example to highlight how the openness and scalability achieved by the platform through the "user extended component library" can meet the deep customization needs of specific industries or scenarios.

[0060] In this scenario, two roles are involved: one is a developer with front-end development capabilities, and the other is a factory workshop manager (user 10) who needs to use the monitoring dashboard.

[0061] First, to meet the needs of factory monitoring, developers can pre-develop a professional "real-time dashboard" component. This component can be written using a common front-end framework (such as Vue.js or React). Its function is to receive a data source URL, periodically poll the URL to obtain values, and display the values ​​in real time in the form of a dynamic dashboard. The component also defines its own configuration items, such as "data source URL", "refresh frequency (seconds)", "range", and "alarm threshold". After development, the component can be packaged and uploaded to the "User Extended Component Library" category in the atomic-level component library 42 of this application platform.

[0062] Subsequently, the factory shop floor manager (User 10), a user without a technical background, can perform the following operations. Their requirement is: "Create a dashboard for real-time monitoring of temperature and humidity in Shop 1."

[0063] After the user logs into the platform and performs the steps of receiving natural language requirements (S101) and generating a preliminary skeleton (S102), the AI ​​engine 20 generates a blank Kanban page containing a title and grid layout, which is then displayed on the visualization canvas 41 (S103).

[0064] In the visualization re-editing step (S104), the factory manager begins building their monitoring dashboard: 1. In the atomic-level component library 42 on the left side of the interface engine 40, expand the "User Extended Component Library" category and find the "Real-time Dashboard" component previously uploaded by the developers. 2. Use the mouse to drag this "Real-time Dashboard" component into the grid layout of the canvas on the right. To monitor temperature and humidity simultaneously, drag twice to create two dashboard instances side-by-side on the canvas. 3. Specifically, the user can first select the first dashboard component. At this time, the configuration items defined by the custom component will be displayed in the property panel on the right side of the interface engine 40. In the "Data Source URL" input box, enter the API address pointing to the workshop temperature sensor according to the device documentation: http: / / 192.168.1.101 / api / sensor / temperature. In the "Refresh Frequency (seconds)" input box, enter "5", indicating that the data is updated every 5 seconds. In "Measuring Range", enter "0-100", with the unit being "℃". In the "Alarm Threshold" field, enter "60", indicating that the dashboard should issue a red alarm when the temperature exceeds 60 degrees Celsius. 4. Next, select the second dashboard component and configure it in the same way, only setting the "Data Source URL" to the humidity sensor's API address: http: / / 192.168.1.102 / api / sensor / humidity, and setting the measurement range to "0-100", with the unit being "%RH".

[0065] After configuration, the factory manager clicks "Real-time Preview" (S105), and two dynamic dashboards immediately appear on the screen, displaying real-time temperature and humidity readings in the workshop, with the data refreshed every 5 seconds. Once satisfied with the results, the manager clicks "One-Click Deployment" (S106), and this professional IoT monitoring dashboard is officially deployed to the factory's internal network for easy access.

[0066] This embodiment demonstrates the openness and scalability of the platform presented in this application. By allowing professional developers to contribute and extend functional components, the platform can provide professional technical capabilities to end-users in a simple and user-friendly manner, thereby efficiently meeting the specific application needs of different vertical fields and achieving deep integration of technology and business.

[0067] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An application development platform that integrates AI generation and visual orchestration, characterized in that, include: The AI ​​engine is used to receive users' natural language application development requirements and generate an atomic application skeleton that includes data models, application interfaces, and basic business logic consisting of business processes and business rules. A UI engine is provided to offer a visual canvas for presenting the atomic application skeleton and to allow users to edit the application skeleton through visual operations. The visual editing includes modifying at least one of the business processes, business rules, and application interfaces of the application skeleton.

2. The platform according to claim 1, characterized in that, The interface engine includes a visual process engine component, and the modification of the business process is accomplished by operating the process engine component on the visual canvas.

3. The platform according to claim 1, characterized in that, The interface engine includes a visual rules engine component, and the modification of business rules is accomplished by operating the rules engine component on the visual canvas.

4. The platform according to claim 1, characterized in that, The interface engine includes a visual interface aggregation engine component, and the modification of the application interface is accomplished by operating the interface aggregation engine component on the visual canvas.

5. The platform according to claim 1, characterized in that, The visual editing also includes configuring the parameters of the large language model used by the AI ​​engine.

6. An application development method integrating AI generation and visual choreography, using an application development platform integrating AI generation and visual choreography as described in any one of claims 1-5, characterized in that, include: The intent parsing and generation steps involve receiving the user's natural language application development requirements through an AI engine and generating an atomic application skeleton that includes a data model, application interface, and basic business logic consisting of business processes and business rules. The visual editing process provides a visual canvas on which the atomized application skeleton is presented, and responds to the user's visual operations to edit the application skeleton. The visual editing step includes modifying at least one of the business processes, business rules, and application interfaces of the application skeleton.

7. The method according to claim 6, characterized in that, The modification of the business process is accomplished by responding to the user's visual process editing operation on the visual canvas.

8. The method according to claim 6, characterized in that, The modification of the business rules is accomplished by responding to the user's visual rule editing operation on the visual canvas.

9. The method according to claim 6, characterized in that, The modification of the application interface is accomplished by responding to the visualization interface aggregation operation performed by the user on the visualization canvas.

10. The method according to claim 6, characterized in that, The visualization editing step also includes configuring the parameters of the large language model used by the AI ​​engine that performs the intent parsing and generation steps.

Images