Method, device and equipment for generating schema file in low-code platform

Abstract

The invention provides a method, a device and equipment for generating schema files in a low-code platform, and belongs to the technical field of artificial intelligence. The method comprises the following steps: in response to an operation instruction for a low-code application program, calling a large language model, and identifying the operation instruction to obtain at least one keyword; generating at least one feature vector based on the at least one keyword; based on the at least one feature vector, at least one description file is obtained from a platform meta-knowledge base, the platform meta-knowledge base is used for storing the corresponding relation between different feature vectors and the description file, and the description file is used for describing at least one attribute of the low-code application program; based on the at least one description file, a schema file of the low-code application program is generated, and the schema file is a structured description file corresponding to the low-code application program and is used for generating the low-code application program. According to the method and the device, the generation cost of the schema file can be reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a schema file generation method and device in a low-code platform and electronic equipment. BACKGROUND

[0002] A low-code platform is a software development tool or environment designed to simplify the process of developing applications and reduce the need for extensive coding. Through a low-code platform, developers can use visual interfaces, drag-and-drop components, and minimal programming to quickly build applications, which helps to accelerate the development cycle of applications, reduce technical barriers, and enable non-professional developers to participate in application development. A schema file is a structured description file corresponding to a low-code application developed by a low-code platform. With the development of artificial intelligence technology, LLM (Large Language Model) is applied in low-code platforms to generate schema files.

[0003] However, the main pricing unit of LLM is the number of tokens, and directly calling LLM to generate a schema file requires a large number of tokens, resulting in high generation costs of the schema file. SUMMARY

[0004] The embodiments of the present application provide a schema file generation method and device in a low-code platform and electronic equipment, which can reduce the generation cost of the schema file. The technical solution is as follows:

[0005] In a first aspect, a schema file generation method in a low-code platform is provided, and the method comprises:

[0006] In response to an operation instruction for a low-code application, a large language model is called to identify the operation instruction and obtain at least one keyword;

[0007] Based on the at least one keyword, at least one feature vector is generated;

[0008] Based on the at least one feature vector, at least one description file is obtained from a platform meta-knowledge base, the platform meta-knowledge base is used to store the corresponding relationship between different feature vectors and description files, and the description file is used to describe at least one attribute of the low-code application;

[0009] Based on the at least one description file, a schema file of the low-code application is generated, the schema file is a structured description file corresponding to the low-code application, and is used to generate the low-code application.

[0010] In a second aspect, a device for generating a schema file in a low-code platform is provided, the device comprising:

[0011] a recognition module, configured to call a large language model in response to an operation instruction for a low-code application, identify the operation instruction, and obtain at least one keyword;

[0012] A first generating module, configured to generate at least one feature vector based on the at least one keyword;

[0013] A first acquisition module is configured to acquire at least one description file from a platform meta-knowledge base based on the at least one feature vector, wherein the platform meta-knowledge base is configured to store a correspondence between different feature vectors and description files, and the description file is configured to describe at least one attribute of the low-code application;

[0014] The second generation module is used to generate a schema file of the low-code application based on the at least one description file. The schema file is a structured description file corresponding to the low-code application and is used to generate the low-code application.

[0015] In a third aspect, an electronic device is provided, comprising a processor and a memory; the memory stores at least one program code; the at least one program code is used to be called and executed by the processor to implement the method for generating a schema file in the low-code platform described in the first aspect.

[0016] In a fourth aspect, a computer-readable storage medium is provided, in which at least one computer program is stored. When the at least one computer program is executed by a processor, the method for generating a schema file in the low-code platform described in the first aspect can be implemented.

[0017] In the fifth aspect, a computer program product is provided, which includes a computer program, and when the computer program is executed by a processor, it can implement the method for generating a schema file in the low-code platform described in the first aspect.

[0018] The beneficial effects of the technical solution provided by the embodiments of the present application are:

[0019] An embodiment of the present application provides a platform meta-knowledge base, which stores the correspondence between different feature vectors and description files. Based on the platform meta-knowledge base, after receiving an operation instruction for a low-code application sent by a user, the large language model is called to identify at least one keyword from the operation instruction. Based on the feature vector generated by the at least one identified keyword, at least one description file can be queried by querying the platform meta-knowledge base. Each description file is used to describe at least one attribute of the low-code application, and then based on at least one description file, a complete schema file can be generated. Since there is no need for a large language model to directly generate a schema file, the number of tokens required for the large language model to generate the schema file is reduced, thereby reducing the cost of generating the schema file. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0021] Figure 1 This is a system architecture diagram of a low-code platform integrated with LLM provided in an embodiment of the present application;

[0022] Figure 2 This is a schematic diagram of an auxiliary application building scenario provided by an embodiment of the present application;

[0023] Figure 3 This is a schematic diagram of an auxiliary logic expansion scenario provided by an embodiment of the present application;

[0024] Figure 4 This is a schematic diagram of an auxiliary operation and maintenance scenario provided by an embodiment of the present application;

[0025] Figure 5 This is a schematic diagram of the principle of a method for generating a schema file in a low-code platform provided in an embodiment of the present application;

[0026] Figure 6 This is a block diagram of a process for generating a schema file in a low-code platform provided in an embodiment of the present application;

[0027] Figure 7 This is a flowchart of a method for generating a schema file in a low-code platform provided in an embodiment of the present application;

[0028] Figure 8 This is a schematic diagram of the structure of a platform meta-knowledge base provided in an embodiment of the present application;

[0029] Figure 9 This is a schematic diagram of the judgment logic in a cache provided by an embodiment of the present application;

[0030] Figure 10 This is a schematic diagram of a process for generating a schema file in a low-code platform provided in an embodiment of the present application;

[0031] Figure 11 This is a structural diagram of a device for generating a schema file in a low-code platform provided in an embodiment of the present application;

[0032] Figure 12 A structural block diagram of an electronic device provided by an exemplary embodiment of the present application is shown. DETAILED DESCRIPTION

[0033] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0034] It should be understood that the terms "each," "plurality," and "any" used in the embodiments of this application include two or more, "each" refers to each of the corresponding plurality, and "any" refers to any one of the corresponding plurality. For example, if a plurality of words includes 10 words, "each" refers to each of the 10 words, and "any" refers to any one of the 10 words.

[0035] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0036] Before executing the embodiments of the present application, the terms involved in the embodiments of the present application are first explained.

[0037] LLMs are deep learning models with relatively large parameters. Large models are deep learning-based models, specifically those that are "large" in nature. Specifically, large means more complex mathematical formulas and more parameters. In terms of data, each generation has seen an order of magnitude leap compared to the previous one, with significant increases in both the coverage and richness of the corpus. The current large language models' ability to learn in both zero-shot and small-shot scenarios primarily stems from extensive memorization of massive corpora during pre-training, followed by enhanced semantic encoding, long-range dependency modeling, and text generation capabilities, as well as the design of natural language task descriptions. In terms of training objectives, since the model is not explicitly guided to learn small-shot generalization, it performs poorly on language tasks such as niche corpora, logical comprehension, and mathematical problem-solving.

[0038] In natural language processing (NLP), a token refers to a sequence of related characters, such as a word or a punctuation mark. Since many NLP tasks (such as part-of-speech tagging, named entity recognition, and machine translation) require text decomposition, decomposing text into tokens can be considered a fundamental NLP task. In text processing, a token can be a word, number, punctuation mark, single letter, or any other element that can be used for text analysis. When decomposing text, token boundaries are typically determined based on spaces, punctuation marks, and other specific delimiters. Choosing the right tokenization method is crucial for NLP systems, as it directly impacts the accuracy and efficiency of other tasks. Since the number of tokens directly reflects the computing power consumed by large language models, large language models are currently primarily priced by token count.

[0039] Embedding converts natural language into machine language. This involves converting a user's input into a set of floating-point numbers (vectors). These vectors capture the model's understanding of the meaning and characteristics of the natural language. The similarity between two vectors can be measured by distance. The closer the two vectors are, the higher the similarity, and the corresponding natural language texts are also more similar.

[0040] A low-code application is an application built on a low-code platform. A low-code application may contain multiple forms, and each form consists of components.

[0041] Artificial intelligence (AI) refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve results. In other words, AI is a comprehensive technology within computer science that seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. AI also involves studying the design principles and implementation methods of various intelligent machines, enabling them to possess the capabilities of perception, reasoning, and decision-making.

[0042] With the research and advancement of artificial intelligence technology, it has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, autonomous driving, drones, robotics, smart healthcare, and smart customer service. It is believed that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role. The solutions provided in the embodiments of this application involve artificial intelligence technologies such as natural language processing, which are specifically explained through the subsequent embodiments. Natural language processing is an important field in the fields of computer science and artificial intelligence. It studies various theories and methods that enable effective communication between humans and computers using natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field involves natural language, that is, the language people use in daily life, and is closely related to the study of linguistics. Natural language processing technologies generally include text processing, semantic understanding, machine translation, robot question answering, knowledge graphs, and other technologies.

[0043] In recent years, low-code development platforms have gained increasing popularity due to their low barriers to entry and their ability to help non-professionals quickly develop applications. However, as applications become increasingly complex, the barriers to entry rise, leading to difficulties in scalability and opaque operations and maintenance. With the development of artificial intelligence (AI), LLM is being applied to low-code platforms to address these challenges, including difficulty in application development, scalability, and operations and maintenance.

[0044] Figure 1 A system architecture diagram of a low-code platform integrated with LLM is shown. Figure 1 The low-code platform integrated with LLM can achieve various functions, including assisting application construction, assisting logic expansion, and assisting platform operation and maintenance. The system consists of an access layer, an orchestration layer, an extension layer, and basic hardware. The access layer includes various interfaces, data message cards, cool apps, etc.; the orchestration layer includes agents, skills, prompts, indexes, memory, plug-ins, etc., as well as flow control, intent verification, and rights; the extension layer includes various small models and databases; and the basic hardware includes various processors.

[0045] Figure 2 A schematic diagram showing the auxiliary application building function of the low-code platform integrated with LLM, see Figure 2 , the auxiliary application building functions include building applications, building forms, building processes and content generation, etc. Among them, building applications include image-based applications (low-code applications generated based on images), Wensheng applications (low-code applications generated based on natural language) and application configuration automation (applications generated through configuration), etc. Building forms include Wensheng portals (portal pages generated based on natural language), Wensheng forms (forms generated based on natural language), Wensheng reports (reports generated based on natural language), etc. Building processes include Wensheng approval flows (approval processes generated based on natural language), Wensheng automation flows (automated processes generated based on natural language), etc. Content generation includes Wensheng texts (text generated based on natural language), Wensheng images (images generated based on natural language) and data filling, etc.

[0046] Figure 3 A schematic diagram showing the auxiliary logic extension function of the low-code platform integrated with LLM, see Figure 3 Auxiliary logic extension features include interconnection, extended code, formulas, and custom components. Interconnection includes Doc2 connectors and automatic data source configuration. Extended code includes automatic response code generation, glue code generation, and FaaS connection code generation. Formulas include Wensheng validation formulas, Wensheng linkage formulas, and Wensheng service formulas. Custom components include Wensheng custom models and Wensheng data source linkage.

[0047] Figure 4 A schematic diagram showing the auxiliary platform operation and maintenance functions of the low-code platform integrated with LLM, see Figure 4 Assisted platform operations and maintenance functions include customer service, user operations and maintenance, platform operations and maintenance, and conversation. Customer service includes intelligent Q&A and a personalized knowledge base. User operations and maintenance include problem diagnosis, performance prediction, optimization suggestions, and application status analysis. Platform operations and maintenance include water level prediction and intelligent flow control. Conversation includes NL2SQL and automatically generated analysis pages.

[0048] To implement each of these features, the low-code platform must generate a schema file, which is a structured description file for the low-code application. Currently, on low-code platforms, complete schema files are generated primarily by calling LLM. However, this requires a large number of tokens, resulting in high schema file generation costs.

[0049] In the low-code platform and LLM fusion scenario, in order to reduce the need to use a large number of tokens to call LLM to generate a complete schema file, resulting in a high cost of generating the schema file, the embodiment of the present application can pre-build a platform meta-knowledge base, which stores the correspondence between different feature vectors and description files. By obtaining the corresponding description file from the platform meta-knowledge base and then encapsulating the obtained description file, a schema file can be generated. In this process, LLM does not need to generate a complete schema file, but only needs to identify keywords from the user's operation instructions. Compared with generating a complete schema file, the number of tokens required is small and the cost is lower, so that a large language model can be used at low cost.

[0050] Figure 5 The schematic diagram of the principle of the method for generating a schema file in the low-code platform provided in the embodiment of the present application is shown in FIG. Figure 5 Users can input operation instructions into LLM. The operation instructions can be natural language. LLM recognizes the operation instructions and obtains multiple keywords. Then, it queries the platform meta-knowledge base to obtain the description file, and then calls the API (Application Programming Interface) to send the obtained description file to the low-code platform to generate a complete schema file.

[0051] See also Figure 6 In the embodiment of the present application, the low-code platform has multiple skills. For different skills, the generative artificial intelligence AIGC can generate different DSLs (Dynamic Script Language), so that multiple middle-layer DSLs can be obtained. Different middle-layer DSLs can correspond to different interfaces, and access the platform meta-knowledge base through different interfaces to obtain the corresponding description file, and then generate a complete schema file. Usually, the operation instruction sent by the user can be a piece of natural language. The user may not enter this natural language at one time, but multiple times when interacting with AIGC. In this way, AIGC needs to store the context DSL for interaction with the user, and then obtain the user's complete intention based on the stored context DSL, so as to generate the corresponding operation instruction. In addition, AIGC can provide a prompt template to the user during each interaction. Based on the provided prompt template, the user can select the prompt template that is the same as his intention, which can help AIGC understand the user's intention.

[0052] Currently, the cost of using LLM is very high, mainly due to the high computing power consumed, which is related to the cost of hardware, especially GPUs. In the scenario where a low-code platform is integrated with LLM, the embodiment of this application solves the problem of excessive tokens and low efficiency in intelligent integration scenarios by building a platform meta-knowledge base, ultimately achieving the low-cost use of large language models.

[0053] This application embodiment provides a method for generating a schema file in a low-code platform. Figure 7 , the method process provided in the embodiment of the present application includes:

[0054] 701. In response to an operation instruction for a low-code application, call a large language model to identify the operation instruction and obtain at least one keyword.

[0055] Operation instructions can be natural language, images, or other input from users on the low-code platform through interaction with the large language model. These instructions can include application operation instructions, form operation instructions, process operation instructions, configuration operation instructions, and question-and-answer instructions. Upon receiving an operation instruction for a low-code application from a user, the large language model can be invoked to identify the operation instruction and obtain at least one keyword.

[0056] Furthermore, the low-code platform in the embodiment of the present application can have multiple skills. For the skills possessed by the low-code platform, after the large language model identifies at least one keyword, it can also determine the relevant skills for each keyword, thereby facilitating the subsequent generation of different schema files based on different skills.

[0057] Further, considering that in the low-code application development scenario, the use scenario of the user is fixed, the operation instruction sent to the low-code platform can have certain similarity, such as operation on components, only adding, deleting, updating the name, updating other attributes, and the difference lies in the binding mapping relationship, that is, different applications, different forms, different component instances, etc. To reduce the number of calls to the large language model and reduce the use cost of the large language model, the embodiments of the present application will also obtain the keywords corresponding to the operation instruction after the execution of each operation instruction of the user, and analyze and cache the mapping relationship corresponding to the operation instruction. When receiving the operation instruction of the user for the low-code application, before calling the large language model, it can be queried first whether the operation instruction with a similarity higher than a preset threshold is stored in the cache. When the operation instruction with a similarity higher than the preset threshold is not stored in the cache, the large language model can be called to recognize the operation instruction and obtain at least one keyword operation. When the operation instruction with a similarity higher than the preset threshold is stored in the cache, at least one keyword corresponding to the operation instruction is obtained from the cache without calling the large language model.

[0058] In addition, after obtaining the analysis result from the cache and performing the subsequent schema file generation operation based on the analysis result, the accuracy and success rate of this time of obtaining data from the cache can also be analyzed, thereby better meeting the use requirements of the user.

[0059] Figure 8 A schematic diagram of a cache-based logical analysis process is shown, referring to Figure 8 The cache stores the mapping relationship of application operation instructions, form operation instructions, process operation instructions, configuration operation instructions, and question and answer instructions, and the feature vectors of different operation instructions. When receiving the operation instruction sent by the user, the similarity of the operation instruction sent by the user and the stored operation instruction is analyzed, the operation instruction with a similarity higher than a preset threshold is selected, and the subsequent operation is performed based on the keywords and mapping relationship corresponding to the operation instruction with a similarity higher than the preset threshold. Since the large language model does not need to be called, the use of token number when calling the large language model is reduced, and the use cost of the large language model is reduced.

[0060] 702. Based on the at least one keyword, at least one feature vector is generated.

[0061] In the embodiment of the present application, the low-code platform has multiple skills. After identifying at least one keyword, the large language model gives the skills associated with each keyword based on the text content of the at least one keyword, and then divides the at least one keyword into at least one keyword set based on the skills of the low-code platform. Each keyword set is related to a skill of the low-code platform, and then generates at least one corresponding feature vector based on each keyword set. When generating at least one corresponding feature vector based on each keyword set, a feature vector can be generated based on all the keywords included in the keyword set, and subsequent queries can be performed based on the feature vector; or a feature vector can be generated based on each keyword to obtain at least one feature vector, and subsequent queries can be performed based on at least one feature vector.

[0062] For example, Figure 10 In the example, the natural language input by the user is "Please help me build an application suitable for data analysis scenarios in the medical industry. The portal page includes components such as consultation lists, data cards, and rich text, and the theme color is preferably purple." The large language model is called for recognition, and keywords such as "build an application," "medical industry," "data analysis," "portal page," "consultation list," "data card," "rich text," and "purple" are obtained. These keywords can be divided into two keyword sets: one is a keyword set related to application building skills, including "build an application," "medical industry," and "data analysis," and the other is a keyword set related to page creation skills, including "portal page," "consultation list," "data card," "rich text," and "purple." For the keyword set related to application building skills, at least one feature vector can be generated, such as a feature vector based on the keyword "build an application," a feature vector based on the keyword "medical industry," and a feature vector based on the keyword "data analysis." For a set of keywords related to page creation skills, at least one feature vector can be generated, such as generating a feature vector based on the keyword "portal page", generating a feature vector based on the keyword "consultation list", generating a feature vector based on the keyword "data card", generating a feature vector based on the keyword "rich text", and generating a feature vector based on the keyword "purple".

[0063] 703. Based on the at least one feature vector, obtain at least one description file from the platform meta-knowledge base.

[0064] The platform meta-knowledge base is used to store the correspondence between different feature vectors and description files, which are used to describe at least one attribute of the low-code application. Figure 9, the platform meta-knowledge base includes the platform question and answer knowledge base and the platform system knowledge base. The platform question and answer knowledge base can use the product manuals, platform work order data, community questions and answers, and user feedback on the low-code platform as meta-knowledge, which is obtained by extracting the feature vectors (embedding) of these meta-knowledge and then storing them in the database. The platform system knowledge base includes description files such as templates, style descriptions, and layouts for different industries and scenarios. The industries involved in the platform system knowledge base include the service industry, retail industry, construction industry, medical industry, Internet industry, etc. Scenarios include procurement scenarios, inspection and management scenarios, data insight scenarios, business management scenarios, data analysis scenarios, etc. Templates include application templates, form templates, process templates, component templates, etc. To facilitate subsequent queries, the embodiment of the present application includes multiple sub-knowledge bases according to the skills of the low-code platform, and each sub-knowledge base corresponds to a skill of the low-code platform.

[0065] Specifically, based on at least one feature vector, obtaining at least one description file from the platform meta-knowledge base can include: based on at least one feature vector corresponding to each keyword set, obtaining at least one sub-description file corresponding to each keyword set by querying the sub-knowledge base corresponding to the skills related to each keyword set, and then encapsulating at least one sub-description file corresponding to each keyword set to obtain at least one description file.

[0066] For example, Figure 10Skill 1 is an application building skill, and Skill 2 is a page creation skill. Keyword sets related to application building skills include "building an application," "medical industry," and "data analysis," while key word sets related to page creation skills include "portal page," "consultation list," "data card," "rich text," and "purple." For Skill 1, based on the feature vector generated by the keyword "building an application," you can find a sub-knowledge base related to application building skills. Then, based on the feature vector generated by the keyword "medical industry," you can query description files related to the medical industry from this sub-knowledge base. Based on the feature vector generated by the keyword "data analysis," you can retrieve a sub-description file with the data analysis application scenario from the description files related to the medical industry. Then, you can encapsulate the obtained sub-description file to obtain a single description file. For skill 2, based on the feature vector generated by the keyword "portal page", you can find the sub-knowledge base related to the page creation skill, and then based on the feature vector generated by the keyword "consultation list", you can obtain the sub-description file corresponding to the consultation list from the sub-knowledge base, and then based on the feature vector generated by the keyword "data card", you can obtain the sub-description file corresponding to the data card from the sub-knowledge base, and then based on the feature vector generated by the keyword "rich text", you can obtain the sub-description file corresponding to the rich text, and then based on the feature vector generated by the keyword "purple", you can obtain the sub-description file corresponding to purple, and then encapsulate the obtained sub-description files to obtain a description file.

[0067] 704. Generate a schema file for the low-code application based on at least one description file.

[0068] Based on the at least one description file obtained, a schema file for the low-code application can be generated by encapsulating the description files. The schema file is a structured description file corresponding to the low-code application and is used to generate the low-code application.

[0069] Generally speaking, the schema file generated based on traditional AIGC is relatively large. For example, for a component page, the corresponding schema file also has hundreds of lines of code, so the generation process requires the use of a large number of tokens. The embodiment of the present application divides the schema file into a variable part, an enumeration part, and a default framework part, and simplifies common scenarios, removing uncommon and unchanging parts. In this way, when generating a schema file, generally only the description file corresponding to the variable part and the description file corresponding to the style and layout of the enumeration part are extracted. For the default framework part, it can be dynamically loaded from the platform meta-knowledge base. Since the acquired data is greatly reduced, the number of tokens used is reduced.

[0070] For example, the schema file generated by the embodiment of the present application has only a few lines of code, and its content is as follows:

[0071]

[0072]

[0073] While the schema file generated by the related art method includes hundreds of lines of code, it can be seen from the comparison that the schema file generated by the embodiment of the present application is smaller, and accordingly, the number of tokens consumed is smaller, and the computing power resource consumption is smaller.

[0074] In addition, the embodiment of the present application is based on the platform meta-knowledge base and the operation instructions sent by the user. LLM only needs to identify the operation instructions sent by the user, obtain the platform skills used, and then call the corresponding API interface for query to realize the entire function.

[0075] All of the above optional technical solutions can be combined in any way to form optional embodiments of the present application, and will not be described in detail here.

[0076] Please refer to Figure 11 , which shows a schematic diagram of the structure of a device for generating a schema file in a low-code platform provided in an embodiment of the present application. The device can be implemented by software, hardware, or a combination of both, and becomes all or part of an electronic device. The device includes:

[0077] The recognition module 1101 is configured to call a large language model in response to an operation instruction for a low-code application, identify the operation instruction, and obtain at least one keyword;

[0078] A first generating module 1102 is configured to generate at least one feature vector based on at least one keyword;

[0079] A first acquisition module 1103 is configured to acquire at least one description file from a platform meta-knowledge base based on at least one feature vector, where the platform meta-knowledge base is configured to store correspondences between different feature vectors and description files, where the description file is configured to describe at least one attribute of a low-code application;

[0080] The second generation module 1104 is used to generate a schema file of a low-code application based on at least one description file. The schema file is a structured description file corresponding to the low-code application and is used to generate the low-code application.

[0081] In another embodiment of the present application, the device further comprises:

[0082] A query module, configured to query, in response to the operation instruction, whether an operation instruction having a similarity with the operation instruction higher than a preset threshold is stored in the cache;

[0083] The recognition module is used to call the large language model to recognize the operation instruction and obtain at least one keyword when the cache does not store an operation instruction with a similarity higher than a preset threshold to the operation instruction.

[0084] In another embodiment of the present application, the device further comprises:

[0085] The second acquisition module is configured to acquire at least one keyword corresponding to the operation instruction from the cache when the cache stores an operation instruction whose similarity to the operation instruction is higher than a preset threshold.

[0086] In another embodiment of the present application, a first generation module is used to divide at least one keyword into at least one keyword set based on the text content of at least one keyword and the skills possessed by the low-code platform, each keyword set being related to a skill of the low-code platform; and based on each keyword set, generate at least one corresponding feature vector.

[0087] In another embodiment of the present application, the platform meta-knowledge base includes multiple sub-knowledge bases, each sub-knowledge base corresponds to a skill of the low-code platform, and the first acquisition module is used to obtain at least one sub-description file corresponding to each keyword set by querying the sub-knowledge base corresponding to the skill related to each keyword set based on at least one feature vector corresponding to each keyword set; and encapsulate the at least one sub-description file corresponding to each keyword set to obtain at least one description file.

[0088] An embodiment of the present application provides a platform meta-knowledge base, which stores the correspondence between different feature vectors and description files. Based on the platform meta-knowledge base, after receiving an operation instruction for a low-code application sent by a user, the large language model is called to identify at least one keyword from the operation instruction. Based on the feature vector generated by the at least one identified keyword, at least one description file can be queried by querying the platform meta-knowledge base. Each description file is used to describe at least one attribute of the low-code application, and then based on at least one description file, a complete schema file can be generated. Since there is no need for a large language model to directly generate a schema file, the number of tokens required for the large language model to generate the schema file is reduced, thereby reducing the cost of generating the schema file.

[0089] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0090] Figure 12FIG2 shows a block diagram of an electronic device 1200 according to an exemplary embodiment of the present application. Generally, the electronic device 1200 includes a processor 1201 and a memory 1202 .

[0091] The processor 1201 can be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 1201 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state; the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 1201 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 1201 may also include an artificial intelligence processor, which is used to process computing operations related to machine learning.

[0092] The memory 1202 may include one or more computer-readable storage media, which may be non-transitory computer-readable storage media, such as CD-ROM (Compact Disc Read-Only Memory), ROM, RAM (Random Access Memory), magnetic tape, floppy disk, and optical data storage device. The computer-readable storage medium stores at least one computer program, which, when executed, can implement a method for generating a schema file in a low-code platform.

[0093] Of course, the electronic device described above may also include other components, such as input / output interfaces and communication components. The input / output interface provides an interface between the processor and a peripheral interface module, which may be an output device, an input device, etc. The communication component is configured to facilitate wired or wireless communication between the electronic device and other devices.

[0094] Those skilled in the art will understand that Figure 12 The structure shown in the figure does not constitute a limitation on the electronic device 1200, and the electronic device 1200 may include more or fewer components than shown in the figure, or combine certain components, or adopt a different component arrangement.

[0095] An embodiment of the present application provides a computer-readable storage medium, which stores at least one computer program. When the at least one computer program is executed by a processor, it can implement the method for generating a schema file in the above-mentioned low-code platform.

[0096] An embodiment of the present application provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it can implement the method for generating a schema file in the low-code platform described above.

[0097] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for generating a schema file in a low-code platform, characterized in that: The method comprises: In response to an operation instruction for a low-code application, calling a large language model to identify the operation instruction and obtain at least one keyword; generating at least one feature vector based on the at least one keyword; Based on the at least one feature vector, obtaining at least one description file from a platform meta-knowledge base, wherein the platform meta-knowledge base is used to store a correspondence between different feature vectors and description files, and the description file is used to describe at least one attribute of the low-code application; Based on the at least one description file, a schema file of the low-code application is generated. The schema file is a structured description file corresponding to the low-code application and is used to generate the low-code application.

2. The method according to claim 1, characterized in that Before the step of responding to an operation instruction for a low-code application and calling a large language model to identify the operation instruction and obtain at least one keyword, the step further includes: In response to the operation instruction, querying whether an operation instruction having a similarity with the operation instruction higher than a preset threshold is stored in the cache; When the cache does not store an operation instruction whose similarity with the operation instruction is higher than a preset threshold, the large language model is called to identify the operation instruction to obtain at least one keyword operation.

3. The method according to claim 2, characterized in that The method further comprises: When an operation instruction having a similarity with the operation instruction higher than a preset threshold is stored in the cache, at least one keyword corresponding to the operation instruction is obtained from the cache.

4. The method according to claim 1, wherein The generating at least one feature vector based on the at least one keyword includes: Based on the text content of the at least one keyword and the skills possessed by the low-code platform, dividing the at least one keyword into at least one keyword set, each keyword set being related to a skill of the low-code platform; Based on each keyword set, at least one corresponding feature vector is generated.

5. The method according to claim 4, characterized in that The platform meta-knowledge base includes multiple sub-knowledge bases, each sub-knowledge base corresponds to a skill of the low-code platform, and obtaining at least one description file from the platform meta-knowledge base based on the at least one feature vector includes: Based on at least one feature vector corresponding to each keyword set, obtaining at least one sub-description file corresponding to each keyword set by querying a sub-knowledge base corresponding to skills related to each keyword set; At least one sub-description file corresponding to each keyword set is encapsulated to obtain the at least one description file.

6. A device for generating a schema file in a low-code platform, characterized in that: The device comprises: a recognition module, configured to call a large language model in response to an operation instruction for a low-code application, identify the operation instruction, and obtain at least one keyword; A first generating module, configured to generate at least one feature vector based on the at least one keyword; A first acquisition module is configured to acquire at least one description file from a platform meta-knowledge base based on the at least one feature vector, wherein the platform meta-knowledge base is configured to store a correspondence between different feature vectors and description files, and the description file is configured to describe at least one attribute of the low-code application; The second generation module is used to generate a schema file of the low-code application based on the at least one description file. The schema file is a structured description file corresponding to the low-code application and is used to generate the low-code application.

7. The device according to claim 6, characterized in that The device further comprises: A query module, configured to query, in response to the operation instruction, whether an operation instruction having a similarity with the operation instruction higher than a preset threshold is stored in the cache; The recognition module is configured to call a large language model to recognize the operation instruction and obtain at least one keyword when the cache does not store an operation instruction whose similarity with the operation instruction is higher than a preset threshold.

8. The device according to claim 7, characterized in that The device further comprises: The second acquisition module is configured to acquire at least one keyword corresponding to the operation instruction from the cache when the cache stores an operation instruction whose similarity to the operation instruction is higher than a preset threshold.

9. The device according to claim 6, characterized in that The first generation module is used to divide the at least one keyword into at least one keyword set based on the text content of the at least one keyword and the skills of the low-code platform, each keyword set is related to a skill of the low-code platform; based on each keyword set, generate at least one corresponding feature vector.

10. The device according to claim 9, characterized in that The platform meta-knowledge base includes multiple sub-knowledge bases, each sub-knowledge base corresponds to a skill of the low-code platform. The first acquisition module is used to obtain at least one sub-description file corresponding to each keyword set by querying the sub-knowledge base corresponding to the relevant skills of each keyword set based on at least one feature vector corresponding to each keyword set; encapsulate the at least one sub-description file corresponding to each keyword set to obtain the at least one description file.

11. An electronic device, characterized in that: It includes a processor and a memory; the memory stores at least one program code; the at least one program code is used to be called and executed by the processor to implement the method for generating a schema file in a low-code platform as described in any one of claims 1 to 5.

12. A computer-readable storage medium, characterized in that At least one computer program is stored in the computer-readable storage medium, and when the at least one computer program is executed by the processor, it can implement the method for generating a schema file in the low-code platform as described in any one of claims 1 to 5.

13. A computer program product, characterized in that The computer program product includes a computer program, which, when executed by a processor, can implement the method for generating a schema file in a low-code platform as described in any one of claims 1 to 5.

Images