SaaS application low-code construction method and system fused with large model instruction analysis

Abstract

The embodiment of the invention provides a SaaS application low-code construction method and system fused with large model instruction analysis, and the method provides a one-stop solution for rapid construction of SaaS applications by combining a low-code development tool and a large model instruction analysis technology. Moreover, by integrating the dependency management capability of the component diagram and the natural language processing capability of the large model, the low-code platform can quickly convert the natural language demand of the user into an operable application system. According to the embodiment of the invention, by combining a low-code development tool and a large model instruction analysis technology, the technical threshold and time cost of application development can be effectively reduced.

Claims

1. A low-code construction method for SaaS applications integrating large model instruction parsing, characterized in that: The method is applied to a large model, and the method comprises: Receiving natural language text input by a user, and converting the natural language text into a demand vector; the dimension of the demand vector is aligned with the dimension of the feature vector of the low-code platform component; Performing semantic matching on the demand vector and the feature vector by calculating similarity, and selecting a first component from the low-code platform component library to form a first component combination according to the semantic matching result; Sending the first component combination to the low-code platform, so that the low-code platform parses the dependency relationship between the first components in the first component combination through a component graph to generate a dependency chain; the component graph is constructed by the low-code platform with the components in the component library as nodes and the dependency relationship between components as edges; the node contains the feature vector; Receive the dependency chain sent by the low-code platform, and adjust the order of the first component combination according to the similarity and the dependency chain to obtain a second component combination; Sending the second component combination to the low-code platform, so that the low-code platform generates the front-end code and back-end logic code of each second component in the second component combination through a template engine, and generates a front-end page according to the front-end code; Receive the front-end code and back-end logic code of the second component sent by the low-code platform, and associate the front-end code and back-end logic code of the second component; connect the back-end logic code of the second component to the database, and build a mapping relationship between the second component and the database.

2. The low-code construction method for SaaS applications integrating large model instruction parsing according to claim 1 is characterized in that: The step of receiving a natural language text input by a user and converting the natural language text into a demand vector comprises: Perform word segmentation on the natural language text to obtain a Token sequence; Map the token sequence after word segmentation into a high-dimensional space to form a token vector through word embedding; The Token vectors are combined into a context vector capable of representing the demand semantics of the entire natural language text through the Transformer architecture, and the context vector is the demand vector.

3. The low-code construction method for SaaS applications integrating large model instruction parsing according to claim 1 is characterized in that: The step of selecting a first component from the component library to form a first component combination according to the semantic matching result also includes: generating a new component when there is no component matching the semantic matching result in the component library; and the method for generating the new component includes: Extracting key information from the natural language text and analyzing user needs based on the key information; the key information includes function information, attribute information and dependency relationship information; Determine the type of the new component according to user needs, convert the function information into corresponding HTML elements, add necessary elements according to the attribute information, reserve placeholders for interaction logic and data connection in the dependency information, and generate a basic HTML structure; Design the style and interaction logic of the new component, and connect the new component to the database; Adding the new component to the component library and updating the component diagram; The step of selecting a first component from the component library to form a first component combination according to the semantic matching result also includes: Extract key information from the natural language text, and compare the key information with the learned database to check whether the user requirements meet the standards; If not, the user is prompted and given operation options; the operation options include: adding a field related to the key information to the database, building a system without the key information, and building a system after replacing the key information; The method also includes: comparing the key information with the learned database, and providing fields related to the key information for the user to select.

4. The low-code construction method for SaaS applications integrating large model instruction parsing according to claim 1 is characterized in that: The step of sorting the first component combination according to the similarity and the dependency chain to obtain the second component combination comprises: Using a semantic distance measurement method to calculate the similarity between the demand vector and each of the feature vectors, to obtain a similarity score; Defining conflict detection rules and conflict resolution strategies, and detecting and resolving dependency conflicts for each of the first components according to the conflict detection rules and the conflict resolution strategies; the conflict resolution strategies include priority rules and user preferences; The first components are comprehensively scored according to the similarity and the dependency chain, and the first components are sorted according to the comprehensive score; the comprehensive score includes a weighted score and a dependency score.

5. The low-code construction method for SaaS applications integrating large model instruction parsing according to claim 1 is characterized in that: Associating the front-end code and the back-end logic code of the second component; The steps of connecting the backend logic code of the second component to a database and constructing a mapping relationship between the second component and the database include: Constructing each of the second components and their dependencies into a combination graph; According to the mapping relationship of the combination graph, the front-end code of the second component is associated with the back-end logic code through an interface binding algorithm; Construct a mapping relationship between the second component and the database, and describe the data flow and processing logic through a data flow diagram.

6. The low-code construction method for SaaS applications integrating large model instruction parsing according to claim 1 is characterized in that: The method also includes optimizing the large model, and the optimization method of the large model includes: receiving feedback text from a user, and converting the feedback text into a feedback vector; the dimension of the feedback vector is aligned with the dimension of the demand vector; Using a fusion algorithm to fuse the feedback vector with the demand vector to generate a new training sample; The new training samples are input into the large model, and the parameters of the large model are updated using an incremental learning algorithm.

7. A low-code construction device for SaaS applications integrating large model instruction parsing, characterized in that: The device comprises: A text conversion module, for receiving natural language text input by a user, and converting the natural language text into a demand vector; the dimension of the demand vector is aligned with the dimension of the feature vector of the low-code platform component; A semantic matching module, used to perform semantic matching on the demand vector and the feature vector by calculating similarity, and select a first component from the low-code platform component library to form a first component combination according to the semantic matching result; A first sending module is used to send the first component combination to a low-code platform, so that the low-code platform parses the dependency relationship between the first components in the first component combination through a component graph to generate a dependency chain; the component graph is constructed by the low-code platform with the components in the component library as nodes and the dependency relationship between components as edges; the node contains the feature vector; A component acquisition module, configured to receive the dependency chain sent by the low-code platform, and adjust the order of the first component combination according to the similarity and the dependency chain to obtain a second component combination; A second sending module is used to send the second component combination to the low-code platform, so that the low-code platform generates the front-end code and back-end logic code of each second component in the second component combination through a template engine, and generates a front-end page according to the front-end code; An application building module is used to receive the front-end code and back-end logic code of the second component sent by the low-code platform, and to associate the front-end code and back-end logic code of the second component; to connect the back-end logic code of the second component to the database, and to build a mapping relationship between the second component and the database.

8. A SaaS application low-code construction system integrating large model instruction parsing, characterized in that: The construction system includes a low-code platform, a large model, and a database: The low-code platform includes a component library composed of multiple components; the component library is constructed as a component graph with components as nodes and dependencies between components as edges; the nodes contain the feature vectors; the component graph is indexed by the dimensions of the feature vectors; the low-code platform is used to receive the first component combination obtained by the large model, parse the dependency relationship between the first components in the first component combination through the component graph, generate a dependency chain, and send it to the large model; receive the second component combination obtained by the large model, generate the front-end code and back-end logic code of each second component in the second component combination through the template engine, and send it to the large model; generate a front-end page according to the front-end code; The large model is used to receive natural language text input by a user and convert the natural language text into a demand vector; the dimension of the demand vector is aligned with the dimension of the feature vector; the demand vector and the feature vector are semantically matched by calculating the similarity, and a first component is selected from the low-code platform component library according to the semantic matching result to form a first component combination; the dependency chain sent by the low-code platform is received, and the first component combination is adjusted and sorted according to the similarity and the dependency chain to obtain a second component combination; the front-end code and back-end logic code of the second component sent by the low-code platform are received, and the front-end code and back-end logic code of the second component are associated; the back-end logic code of the second component is connected to a database, and a mapping relationship between the second component and the database is constructed.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, it implements the steps of the low-code construction method of SaaS application integrating large model instruction parsing as described in any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the low-code construction method for SaaS applications integrating large model instruction parsing as described in any one of claims 1 to 6 are implemented.

Images