Object model capability arrangement method and system based on large model

Abstract

The invention discloses an object model capability arrangement method and system based on a large model. The method comprises the following steps: acquiring an arrangement text input by a user; analyzing the arrangement text based on the trained large model, and identifying keywords and intention information in the arrangement text; according to the keyword and the intention information, operation description and process information of a product model; and based on the operation description and the process information, generating a call chain corresponding to the object model so as to arrange a task execution sequence, optimize use of resources and execute corresponding tasks based on the call chain. According to the method, the semantic comprehension and generation capability of the large model is utilized, and the multi-dimensional information of the object model is combined, so that the intelligent object model function arrangement and optimization are realized, the operation complexity of object model arrangement is reduced, the efficiency is improved, and an efficient and personalized solution is provided for the fields of engineering design, scientific research, virtual reality and the like.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for orchestrating physical model capabilities based on a large model. Background Art

[0002] ‌Thing model orchestration‌ refers to the coordination of multiple thing models to complete complex tasks or achieve more efficient workflows. This usually involves the serial or parallel combination of thing models, so that the output of different thing models can be used as the input of other thing models, or the results of multiple thing models can be used together for decision-making. Object model orchestration plays an important role in building complex IoT systems and automated workflows. For example, by combining multiple thing models, the advantages of different models can be integrated to improve the accuracy of the overall system, as well as the automation of the workflow of multiple thing models, reducing manual intervention, improving processing efficiency, and realizing complex functions, thereby enhancing the reliability of the system.

[0003] Currently, object model orchestration is mainly used in the following scenarios: 1. Natural Language Processing (NLP): For example, combining speech recognition, named entity recognition, sentiment analysis, and text generation models to build intelligent dialogue systems; 2. Computer Vision: For example, combining object detection, image segmentation, and image classification models to build autonomous driving systems; 3. Recommendation system: Combines user behavior prediction, item similarity calculation and context analysis model to provide personalized recommendations; 4. Financial field: Combine fraud detection, risk assessment and credit scoring models to achieve a comprehensive financial risk control system.

[0004] Traditional physical model orchestration methods require clarifying the tasks to be completed and the specific physical models for each subtask, selecting appropriate physical models and training them, designing different types of physical models for each subtask, and designing data transmission methods between physical models to ensure consistency in data formats and interfaces between models. This results in traditional physical model orchestration methods requiring professional knowledge and complex programming skills, which makes users face problems of complex operations and low efficiency. Summary of the invention

[0005] The main purpose of the present invention is to provide a method and system for orchestrating physical model capabilities based on a large model, aiming to solve the problem that the physical model orchestration method in the prior art requires professional knowledge and complex programming skills, which makes users face the problems of complex operation and low efficiency.

[0006] To achieve the above object, the present invention provides a method for arranging physical model capabilities based on a large model, and the method for arranging physical model capabilities based on a large model comprises the following steps: Get the edited text entered by the user; Parsing the arrangement text based on the trained large model to identify keywords and intent information in the arrangement text; Generate an operation description and process information of a model according to the keywords and the intent information; Based on the operation description and the process information, a call chain corresponding to the object model is generated to arrange the task execution order, optimize the use of resources and execute the corresponding tasks based on the call chain.

[0007] Optionally, the step of obtaining the arrangement text input by the user includes: Integrate the trained large model into the orchestration system of the object model.

[0008] Optionally, the obtaining of the arrangement text input by the user specifically includes: The requirement information input by the user and containing the keywords and the intention information is obtained through a text input tool or a speech-to-text conversion tool to obtain the arrangement text.

[0009] Optionally, the parsing of the arrangement text based on the trained large model to identify keywords and intent information in the arrangement text specifically includes: The arrangement text is parsed based on the trained GPT-4 or BERT large model to identify keywords in the arrangement text that represent the user's physical system requirements for the object model, and the intent information is identified based on the semantics of the arrangement text.

[0010] Optionally, generating the operation description and process information of the object model according to the keyword and the intent information specifically includes: Determine the operation, control, state or condition of the object model according to the keyword, and determine the task or behavior to be completed by the object model according to the intention information; Based on the operation, control, state or condition of the physical model, the tasks or behaviors to be completed by the physical model are mapped to corresponding operation processes or tasks, and the operation description and process information of the physical model are generated.

[0011] Optionally, generating a call chain corresponding to the object model based on the operation description and the process information specifically includes: Defining a call chain structure corresponding to the object model; Parsing the operation description and process information corresponding to the object model; Based on the operation description and the parsed information of the process information, find the operation that is triggered first in the process, identify the trigger point information, and determine the dependency relationship between the operations; According to the dependency relationship between the trigger point information and the operations, a call chain corresponding to the object model is generated using a defined call chain structure.

[0012] Optionally, the method further includes: When an operation fails or the status changes, the call chain corresponding to the object model is dynamically adjusted through logs, status monitoring, and exception handling mechanisms.

[0013] In addition, to achieve the above-mentioned purpose, the present invention further provides a physical model capability orchestration system based on a large model, wherein the physical model capability orchestration system based on a large model includes: The editing text acquisition module is used to acquire the editing text input by the user; A keyword and intent parsing module, used to parse the arrangement text based on the trained large model, and identify the keywords and intent information in the arrangement text; An operation and process generation module, used to generate an operation description and process information of an object model according to the keywords and the intention information; A call chain generation module is used to generate a call chain corresponding to the object model based on the operation description and the process information, so as to arrange the task execution order, optimize the use of resources and execute the corresponding tasks based on the call chain.

[0014] In addition, to achieve the above-mentioned purpose, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a large-model-based physical model capability orchestration program stored in the memory and executable on the processor, wherein the large-model-based physical model capability orchestration program implements the steps of the large-model-based physical model capability orchestration method as described above when executed by the processor.

[0015] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a physical model capability orchestration program based on a large model, and when the physical model capability orchestration program based on a large model is executed by a processor, the steps of the physical model capability orchestration method based on a large model as described above are implemented.

[0016] In the present invention, by obtaining the arrangement text input by the user, the arrangement text can be parsed based on the trained large model to identify the keywords and intent information in the arrangement text; and the operation description and process information of the object model can be generated according to the keywords and the intent information; thereby, based on the operation description and the process information, a call chain corresponding to the object model is generated to arrange the task execution order, optimize the use of resources, and execute the corresponding tasks based on the call chain. The present invention utilizes the semantic understanding and generation capabilities of the large model, combined with the multi-dimensional information of the object model, to achieve intelligent object model function arrangement and optimization, reduce the operational complexity of the object model arrangement, improve efficiency, and provide efficient and personalized solutions for engineering design, scientific research, virtual reality and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of a preferred embodiment of the method for arranging physical model capabilities based on a large model of the present invention; Figure 2 It is a structural diagram of a preferred embodiment of the object model capability arrangement system based on a large model of the present invention; Figure 3 It is a structural diagram of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0019] At present, the traditional physical model orchestration method requires clarifying the tasks to be completed and the specific physical models of each subtask, selecting appropriate physical models and training them, designing different types of physical models for each subtask, and designing the data transmission method between physical models to ensure the consistency of data formats and interfaces between models. This results in the traditional physical model orchestration method requiring professional knowledge and complex programming skills, which makes users face the problems of complex operation and low efficiency.

[0020] In order to solve the problem that the physical model arrangement method in the prior art requires professional knowledge and complex programming skills, which makes users face the problem of complex operation and low efficiency, an embodiment of the present invention provides a physical model capability arrangement method based on a large model, which obtains the arrangement text input by the user; parses the arrangement text based on the trained large model to identify the keywords and intent information in the arrangement text; generates the operation description and process information of the physical model according to the keywords and the intent information; generates the call chain corresponding to the physical model based on the operation description and the process information, so as to arrange the task execution order, optimize the use of resources and execute the corresponding tasks based on the call chain. In the embodiment of the present invention, the semantic understanding and generation capabilities of the large model are utilized, combined with the multi-dimensional information of the physical model, to realize intelligent physical model function arrangement and optimization, reduce the operational complexity of physical model arrangement, improve efficiency, and provide efficient and personalized solutions for engineering design, scientific research, virtual reality and other fields.

[0021] The method for arranging physical model capabilities based on a large model according to a preferred embodiment of the present invention is as follows: Figure 1 As shown, the method for arranging physical model capabilities based on a large model includes the following steps: Step S10: Obtain the editing text input by the user.

[0022] It should be noted that in this application, the purpose is to allow users to generate and optimize the operation process and functional model of the physical system (i.e., the operating system of the physical model) by arranging text descriptions through advanced natural language processing technology. The system uses the semantic understanding and generation capabilities of the large model, combined with the multi-dimensional information of the physical model, to achieve intelligent physical model function arrangement and optimization, and provide efficient and personalized solutions for engineering design, scientific research, virtual reality and other fields. Therefore, it can be understood that in this application, through the powerful processing capabilities of the large model, users can directly input natural language descriptions and generate corresponding physical model operation processes, thereby simplifying the operation process and improving work efficiency.

[0023] Furthermore, in the present application, the step of obtaining the arrangement text input by the user includes: integrating the trained large model into the arrangement system of the object model.

[0024] Specifically, in this application, before implementing the object model capability orchestration, it is also necessary to integrate a well-trained large model into the object model orchestration system to understand the user's intention from the input natural language description; it can be understood that a large language model (for example, a GPT model, a BERT model, a T5 model, etc.) can be used to process natural language input and understand the user's intention; in this embodiment, a pre-trained GPT-4 model or a BERT model can be used to understand the user's intention from the input natural language description.

[0025] As an example, in this application, the trained large model can be integrated into the orchestration system of the object model according to the system architecture of the object model and the core tasks of user intent analysis (for example, querying device status, data or faults, setting device configuration or parameters, requesting to perform specific tasks, etc.); specifically, first, design a natural language understanding module. To integrate the large language model into the object model orchestration system, a natural language understanding (NLU) module is usually required, which is responsible for converting user input into machine-understandable operation instructions; then, construct an intent and action mapping layer. In order to convert natural language input into control instructions for the object model orchestration system, a mapping layer needs to be designed to match the recognized intent with the specific operation of the system (API, script, command, etc.). The intent and entity information provided by the large language model can be mapped to the corresponding operation instructions or device control interface; finally, construct a dialogue management layer and interactive test with the object model orchestration system. For more complex interactions (for example, multiple rounds of dialogue between users and the system), a dialogue management layer is required to track the dialogue context and maintain the state of the system. The large language model can infer the current state of the conversation based on the context; after the natural language intent is parsed and converted into control commands, the system needs to interact with the object model orchestration platform to further enhance the user experience.

[0026] In this application, by integrating a large language model into the object model orchestration system, the system's natural language understanding ability can be significantly improved, allowing users to interact with smart devices more naturally. By rationally designing modules such as intent recognition, action mapping, and dialogue management, combined with powerful language model processing capabilities, a more efficient and flexible operation method can be provided for industrial and IoT environments.

[0027] Furthermore, in the present application, the obtaining of the arrangement text input by the user specifically includes: obtaining the demand information input by the user including the keywords and the intent information through a text input tool or a speech-to-text conversion tool to obtain the arrangement text.

[0028] Specifically, in this application, the arrangement text input by the user can be obtained through a text input tool or a speech-to-text conversion tool, and the arrangement text refers to a way of describing the goals and steps to be achieved through natural language. In this context, the arrangement text allows the user to describe the requirements of his physical system, and the system generates corresponding instructions and operation processes by understanding these natural language descriptions.

[0029] The main function of the arrangement text described in this application is to bridge the user's intention and the execution function of the system, so that non-professional users can also control and optimize complex systems through simple language descriptions. The conventional form of the arrangement text is: usually including action instructions, system components, and specific parameters of the operation. Correspondingly, the arrangement text can be called an "Operation Description Generator".

[0030] Based on the above explanation about the arrangement text, as an example, the arrangement text in the present application contains keywords and intent information. For example, the arrangement text input by the user is: "Please initialize the starting sequence of the industrial robot and check its connection status.", where the keyword is "industrial robot" and the intent information is "initialization" and "check connection status"; or: "Design an automatic temperature control system to control the greenhouse environment", where the keyword is "automatic temperature control system" and the intent information is "control" and "greenhouse environment"; these arrangement texts can be obtained by the system directly in the user input interface, or the voice information input by the user can be obtained by the voice acquisition module, and then obtained by the voice-to-text conversion tool.

[0031] Step S20: parsing the arrangement text based on the trained large model to identify keywords and intent information in the arrangement text.

[0032] It can be understood that in this application, the semantic understanding ability of the GPT-4 or BERT large model is used to parse and analyze the arrangement text description input by the user (i.e., identify the keywords and intent information in the arrangement text), so as to identify key operation steps and requirements.

[0033] Furthermore, in the present application, the arrangement text is parsed based on the trained big model to identify keywords and intent information in the arrangement text, specifically including: parsing the arrangement text based on the trained GPT-4 or BERT big model, identifying keywords in the arrangement text of the user's physical system requirements for the object model, and identifying the intent information according to the semantics of the arrangement text.

[0034] Specifically, in this application, the use of pre-trained models such as GPT-4 or BERT can help understand and parse the semantics in the edited text; for example, the natural language understanding (NLU) module of the GPT-4 or BERT large model can be used to identify keywords in the edited text (e.g., the keyword "industrial robot"); for the intent in the object model edited text (e.g., obtaining sensor data, setting device status, etc.), a classification model can be trained or a large language model can be used directly to identify the intent.

[0035] As an example, the keywords extracted in this application can be specific terms input by the user, which are usually related to the operation, control, state or condition of the object model. For example: "switch", "adjust temperature", "set timer", "troubleshooting", etc. Intent information is the user's needs or action intentions, indicating the tasks or behaviors that the user wants the object model to complete. For example, the user's possible intentions are "control the switch state of the light bulb", "adjust the temperature of the air conditioner", "check the device status", etc.

[0036] Furthermore, in this application, if the choreographed text contains complex dependencies (e.g., interactions between devices, relationships between attributes, etc.), GPT-4 or BERT large models can help reason about these relationships. For example, through dependency parsing, it is possible to understand which inputs or conditions the execution of a service depends on.

[0037] For example: "Device A's temperature sensor reads a value, and if the temperature exceeds 30°C, device B starts the fan." It can be deduced from the model: Dependency: Device A (temperature sensor) reads temperature data, and device B (fan) takes action based on the temperature threshold.

[0038] In the actual formatted text parsing process, if the GPT-4 or BERT model cannot fully meet the needs of scenario analysis, the GPT-4 or BERT model can be fine-tuned. For example, a dataset of formatted text containing annotated object models can be used to perform task-specific training on large models to better understand and extract formatted text in a specific format.

[0039] In this application, the trained big model is used to effectively parse the object model arrangement text. The advantage of the big model is its powerful language understanding ability, which can identify and extract keywords, intentions, entities and other information in the text. Combining deep learning, natural language processing technology and structured data analysis, it can provide strong support for the automated arrangement and management of IoT devices.

[0040] Step S30: Generate an operation description and process information of the object model according to the keywords and the intention information.

[0041] It is understandable that in this application, based on the results of semantic understanding (i.e., the keywords and intent information in the arrangement text that are identified), the operation description and process of the model are generated; that is, based on the results of semantic understanding, the specific operation steps and description of the arrangement text are generated. These descriptions usually include specific actions such as starting the device, setting parameters, and checking the status to ensure that the physical device or system can operate according to the predetermined logic and sequence. After generating the operation description and process of the model, the system can generate the corresponding call chain based on the description to ensure that the logic and sequence of the operation are correct.

[0042] Furthermore, in the present application, the generation of the operation description and process information of the object model based on the keywords and the intention information specifically includes: determining the operation, control, state or condition of the object model based on the keywords, and determining the task or behavior to be completed by the object model based on the intention information; based on the operation, control, state or condition of the object model, mapping the task or behavior to be completed by the object model to the corresponding operation process or task, and generating the operation description and process information of the object model.

[0043] It is understandable that a physical model is an abstract representation of a real-world object, device, or system, and is often used in the Internet of Things (IoT), industrial control, smart devices, and other fields. Operational descriptions and process information are centered around the use, control, and maintenance of the object or device. The goals and functions of the physical model need to be clearly defined. For example, a physical model may be a smart home device (such as a smart light bulb, thermostat, door lock, etc.), or an industrial device (such as a robot, sensor, control system, etc. on a production line).

[0044] In this application, after extracting keywords and intent information through GPT-4 or BERT large models, the intent information can be mapped to specific operation processes or tasks. This requires the system to understand how each intent corresponds to the physical or virtual control function of the device. For example, when the user's intention is to "turn on the light bulb", the system needs to convert this intention into specific operation instructions (for example, sending a command through the smart home control platform to turn on the physical device). For complex operations, multiple steps may be involved in series. For example, "adjusting the air conditioner temperature" is not just setting the target temperature, but may also require additional operations such as adjusting the mode (heating / cooling), wind speed, etc.

[0045] Specifically, in this application, an operation description is generated based on the intent information. The operation description should clearly and concisely express the steps of how the user can complete the intent through the system. The operation description should conform to the user's language habits and be easy to understand and execute.

[0046] For example, Intent: Turn on or off a smart light bulb; Operation description: 1. Open the Smart Home APP. 2. Select the "Smart Light Bulb" device. 3. Click the "Switch" button and the light bulb will turn on.

[0047] Intent: Adjust the air conditioner temperature; Operation description: 1. Open the air conditioner control panel. 2. Select the "Temperature setting" option. 3. Use the slider to adjust to the desired temperature and click "Confirm" to save the setting.

[0048] Specifically, in this application, when generating process information, process information refers to the steps that describe how to perform a series of actions in a specific situation. This is usually expressed as a sequential process that describes how the object model interacts with the user and the state changes of the device.

[0049] For example: Process information: The process of adjusting the air conditioner temperature; 1. The user opens the air conditioner control interface. 2. The user selects "Temperature setting". 3. The user enters the target temperature (such as: 22°C). 4. The system receives the instruction and adjusts the air conditioner temperature to 22°C. 5. The system confirms the operation and displays the current temperature as 22°C.

[0050] Furthermore, in this application, for complex physical model operations, prompt words, conditions, and feedback mechanisms can also be added when generating the operation description and process information of the physical model. For example, if the user's operation cannot be successfully executed, the system should give an error prompt or feedback. For example: "The device is currently unable to connect to the network, please check the network settings." or "The temperature setting is out of the adjustable range, please reset it." etc.

[0051] In this application, by extracting keywords, identifying user intent information, and then mapping this information to the device's operation description and process, it can help users easily understand how to operate the object model and what steps should be performed in different scenarios. This process is crucial for user interaction design of IoT devices, smart homes, industrial control and other systems.

[0052] Step S40: Based on the operation description and the process information, generate a call chain corresponding to the object model, so as to arrange the task execution order, optimize the use of resources and execute the corresponding tasks based on the call chain.

[0053] It can be understood that in the present application, after generating the operation description and process of the object model, the object model system can generate a corresponding call chain according to the description to ensure that the logic and order of the operation are correct; wherein, the call chain mainly plays the following key roles: 1. Sequence guarantee: The call chain ensures that each step is executed according to the correct logic by defining the order of operations. For example, you cannot execute a motion command when the robot is not powered on.

[0054] 2. Dependency management: Some operations may depend on the results of previous steps. The call chain ensures that each step is executed after the necessary prerequisites are met by clarifying these dependencies.

[0055] 3. Error handling: During the operation, if a step fails, the call chain can perform error handling according to the set logic, such as retrying, skipping or terminating subsequent steps.

[0056] 4. Resource optimization: The call chain can optimize the use of resources. For example, by properly arranging the order of task execution, it can reduce waiting time and resource waste.

[0057] 5. Traceability: Provides a detailed execution path for operations to facilitate system problem diagnosis and system optimization.

[0058] Therefore, in this application, through the powerful processing capabilities of the large model, users can directly input natural language descriptions, generate corresponding physical model operation processes, and generate call chains corresponding to the physical models, so that the physical model system can use the call chains to more accurately and efficiently perform tasks corresponding to the edited text.

[0059] Furthermore, in the present application, the generation of a call chain corresponding to the physical model based on the operation description and the process information specifically includes: defining a call chain structure corresponding to the physical model; parsing the operation description and process information corresponding to the physical model; based on the parsed information of the operation description and the process information, finding the operation that is triggered first in the process, identifying the trigger point information, and determining the dependency relationship between the operations; and generating the call chain corresponding to the physical model using the defined call chain structure according to the trigger point information and the dependency relationship between the operations.

[0060] In this application, a call chain refers to a chain of call relationships between multiple operations or functional modules in a system when they are executed in sequence. For example, a call chain may involve triggering from one operation to the next operation, or one operation calling another operation and producing corresponding results.

[0061] The call chain of the object model can usually be constructed as follows: Node: Each operation or functional module (such as device operation, status check, etc.) can be regarded as a node.

[0062] Edges: The edges between nodes represent the calling relationship between operations. For example, operation A triggers operation B after completion, and operation C depends on the result of operation A, etc.

[0063] Therefore, when analyzing the object model operation and process information, the present application specifically includes the following steps: 1. Extract operations and attributes of the object model: First, extract all operations, attributes, states, events, and other information from the object model definition. This information may be obtained through APIs, data protocols (such as MQTT, RESTful interfaces, etc.).

[0064] 2. Analyze the operation description: By analyzing the operation description, understand the dependencies, trigger conditions, and execution order between operations. The operation description may contain conditional statements, event-driven execution processes, etc.

[0065] 3. Extract process information: Process information describes the logical relationship and execution order between operations. For example, in industrial automation, sensor data triggers the start of a device, and the state change of the device triggers the next step.

[0066] After parsing the object model operations and process information, when building a call chain, combine the operation description and process information to build the call chain according to the following steps: 1. Identify the trigger point: Find the first triggered operation in the process, which is usually the entry point of the system.

[0067] 2. Determine dependencies: Based on the operation description and process information, clarify the dependencies between operations. For example, operation B can only be executed after operation A is completed, and operation C may require the return value of operation A as input.

[0068] 3. Build links: Build call chains between operations based on trigger points and dependencies. For example: operation A triggers operation B; after operation B is completed, operation C is triggered; after operation C returns a result, operation D is triggered, etc.

[0069] Furthermore, in this application, after building the call chain, through the analysis of the call chain, it may be found that some operations can be executed in parallel, or some operations can be optimized and merged. Optimizing the execution order of the call chain helps to improve the performance of the system. Finally, it is necessary to ensure the correctness of the entire call chain, including the order of operations, whether the dependencies are reasonable, and whether the results meet expectations.

[0070] Furthermore, in the present application, the method also includes: when an operation fails or the state changes, dynamically adjusting the call chain corresponding to the object model through logs, state monitoring, and exception handling mechanisms.

[0071] Specifically, in this application, the physical model call chain may need to be dynamically adjusted according to external conditions in actual operation. For example, if an operation fails or the state changes, the call chain may need to be readjusted. The execution of the call chain can be dynamically adjusted through monitoring tools (such as logs, state monitoring, exception handling mechanisms, etc.).

[0072] Since the call chain of the object model needs to extract key content from the object model definition, operation description and process information, and build the call chain according to the dependencies and trigger conditions between operations, these call relationships can be converted into actual execution logic, optimized and verified through automation tools and code generation technology.

[0073] In the present invention, by obtaining the arrangement text input by the user, the arrangement text can be parsed based on the trained large model to identify the keywords and intent information in the arrangement text; and the operation description and process information of the object model can be generated according to the keywords and the intent information; thereby, based on the operation description and the process information, a call chain corresponding to the object model is generated to arrange the task execution order, optimize the use of resources, and execute the corresponding tasks based on the call chain. The present invention utilizes the semantic understanding and generation capabilities of the large model, combined with the multi-dimensional information of the object model, to achieve intelligent object model function arrangement and optimization, reduce the operational complexity of the object model arrangement, improve efficiency, and provide efficient and personalized solutions for engineering design, scientific research, virtual reality and other fields.

[0074] Furthermore, if Figure 2 As shown, based on the above-mentioned method for arranging physical model capabilities based on a large model, the present invention also provides a physical model capability arrangement system based on a large model, wherein the physical model capability arrangement system based on a large model includes: The arrangement text acquisition module 21 is used to acquire the arrangement text input by the user; A keyword and intention parsing module 22, used to parse the arrangement text based on the trained large model, and identify the keywords and intention information in the arrangement text; An operation and process generation module 23, used to generate an operation description and process information of a model according to the keywords and the intention information; The call chain generation module 24 is used to generate a call chain corresponding to the object model based on the operation description and the process information, so as to arrange the task execution order, optimize the use of resources and execute the corresponding tasks based on the call chain.

[0075] Furthermore, in the present application, the system further includes: The large model integration module is used to integrate the trained large model into the orchestration system of the object model.

[0076] The call chain adjustment module is used to dynamically adjust the call chain corresponding to the object model through logs, status monitoring, and exception handling mechanisms when an operation fails or the status changes.

[0077] Furthermore, in the present application, the arrangement text acquisition module 21 specifically includes: The demand information acquisition unit is used to acquire the demand information including the keywords and the intention information input by the user through a text input tool or a speech-to-text conversion tool to obtain the arrangement text.

[0078] Furthermore, in the present application, the keyword and intention parsing module 22 specifically includes: The large model recognition unit is used to parse the arrangement text based on the trained GPT-4 or BERT large model, identify keywords of the user's physical system requirements for the object model in the arrangement text, and recognize the intent information according to the semantics of the arrangement text.

[0079] Furthermore, in the present application, the operation and process generation module 23 specifically includes: an information determination unit, configured to determine an operation, control, state or condition of the object model according to the keyword, and to determine a task or behavior to be completed by the object model according to the intention information; A mapping and generation unit is used to map the tasks or behaviors to be completed by the physical model to the corresponding operation process or tasks based on the operation, control, state or condition of the physical model, and generate the operation description and process information of the physical model.

[0080] Furthermore, in the present application, the call chain generation module 24 specifically includes: A call chain definition unit, used to define a call chain structure corresponding to the object model; An operation and process parsing unit, used for parsing the operation description and process information corresponding to the object model; A dependency determination unit, configured to find out the first triggered operation in the process based on the operation description and the parsed information of the process information, identify the trigger point information, and determine the dependency relationship between the operations; The call chain generation subunit is used to generate a call chain corresponding to the object model using a defined call chain structure according to the trigger point information and the dependency relationship between the operations.

[0081] Furthermore, if Figure 3 As shown, based on the above-mentioned large-model-based object model capability orchestration method and system, the present invention also provides a terminal accordingly, and the terminal includes a processor 10, a memory 20 and a display 30. Figure 3 Only some components of the terminal are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0082] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Further, the memory 20 may also include both an internal storage unit of the terminal and an external storage device. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code of the installation terminal. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a large-model-based object model capability orchestration program 40 is stored on the memory 20, and the large-model-based object model capability orchestration program 40 can be executed by the processor 10, thereby realizing the large-model-based object model capability orchestration method in the present application.

[0083] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code or process data stored in the memory 20, such as executing the large model-based physical model capability orchestration method.

[0084] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 30 is used to display information on the terminal and to display a visual user interface. The components 10-30 of the terminal communicate with each other via a system bus.

[0085] In one embodiment, when the processor 10 executes the large model-based object model capability arrangement program 40 in the memory 20, the following steps are implemented: Parse user interface files and interface information; Generate the module code required by the application and the application code called by the interface according to the parsed file information, interface parameters and return value information; A test case for interface testing is generated according to the module code and the application code called by the interface.

[0086] Optionally, the parsing of the user interface file and the interface information specifically includes: Parsing the user interface file to obtain corresponding module information; Parsing the interface information in the user interface file; Obtain the number of parameters, parameter types, and return value type information in the user interface file.

[0087] Optionally, parsing the user interface file to obtain corresponding module information specifically includes: Create a plugin configuration file and create a file containing the N-API plugin code; Compile the plug-in configuration file and the file containing the N-API plug-in code to obtain a parsing plug-in; The parsing plug-in is run, and the user interface file is parsed based on the parsing plug-in to obtain corresponding module information.

[0088] Optionally, generating the module code required for the application and the application code for the interface call according to the parsed file information, interface parameters and return value information specifically includes: Define application interface call file template and interface call template; Generate the code for importing the module in the user interface file according to the module information corresponding to the user interface file, and obtain the module code required by the application; Generate application code for the interface call according to the interface parameters and the return value information.

[0089] Optionally, generating a code for importing a module in the user interface file according to the module information corresponding to the user interface file specifically includes: According to the application interface calling file template and the module information, compile and obtain the module corresponding to the user interface file based on the compilation configuration file; The module corresponding to the user interface file is imported, and the module corresponding to the user interface file is tested, and the code of the imported module in the user interface file is obtained based on the test result.

[0090] Optionally, generating the application code for the interface call according to the interface parameters and the return value information specifically includes: Define an interface function for receiving the parameter, and perform parameter checking on the parameter through the interface function to ensure that the parameter is a number; Generate application code for the interface call according to the parameter, the return value and the interface function.

[0091] Optionally, generating a test case for interface testing according to the module code and the application code called by the interface specifically includes: Configure the test framework and assertion library; In the test project directory, create a test file, and generate test cases for verifying the interface function according to the module code and the application code called by the interface; The test interface is tested based on the test case, and the corresponding test result is displayed.

[0092] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a large-model-based physical model capability orchestration program, and when the large-model-based physical model capability orchestration program is executed by a processor, the steps of the large-model-based physical model capability orchestration method as described above are implemented.

[0093] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or terminal including the element.

[0094] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the program can be stored in a computer-readable storage medium that can be read by a computer, and the program can include the processes of the above-mentioned method embodiments when executed. The computer-readable storage medium can be a memory, a disk, an optical disk, etc.

[0095] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A method for arranging physical model capabilities based on a large model, characterized in that: The method for arranging physical model capabilities based on a large model includes: Get the edited text entered by the user; Parsing the arrangement text based on the trained large model to identify keywords and intent information in the arrangement text; Generate an operation description and process information of a model according to the keywords and the intent information; Based on the operation description and the process information, a call chain corresponding to the object model is generated to arrange the task execution order, optimize the use of resources and execute the corresponding tasks based on the call chain.

2. The method for arranging physical model capabilities based on a large model according to claim 1, characterized in that: The step of obtaining the arrangement text input by the user includes: Integrate the trained large model into the orchestration system of the object model.

3. The method for arranging physical model capabilities based on a large model according to claim 1, characterized in that: The step of obtaining the arrangement text input by the user specifically includes: The requirement information input by the user and containing the keywords and the intention information is obtained through a text input tool or a speech-to-text conversion tool to obtain the arrangement text.

4. The method for arranging physical model capabilities based on a large model according to claim 1, characterized in that: The parsing of the arrangement text based on the trained large model to identify keywords and intent information in the arrangement text specifically includes: The arrangement text is parsed based on the trained GPT-4 or BERT large model to identify keywords in the arrangement text that represent the user's requirements for the physical system of the object model, and the intent information is identified based on the semantics of the arrangement text.

5. The method for arranging physical model capabilities based on a large model according to claim 1, characterized in that: The operation description and process information of the generated model according to the keywords and the intent information specifically include: Determine the operation, control, state or condition of the object model according to the keyword, and determine the task or behavior to be completed by the object model according to the intention information; Based on the operation, control, state or condition of the physical model, the tasks or behaviors to be completed by the physical model are mapped to corresponding operation processes or tasks, and the operation description and process information of the physical model are generated.

6. The method for arranging physical model capabilities based on a large model according to claim 1, characterized in that: The generating a call chain corresponding to the object model based on the operation description and the process information specifically includes: Defining a call chain structure corresponding to the object model; Parsing the operation description and process information corresponding to the object model; Based on the operation description and the parsed information of the process information, find the operation that is triggered first in the process, identify the trigger point information, and determine the dependency relationship between the operations; According to the dependency relationship between the trigger point information and the operations, a call chain corresponding to the object model is generated using a defined call chain structure.

7. The method for arranging physical model capabilities based on a large model according to claim 1, characterized in that: The method further comprises: When an operation fails or the status changes, the call chain corresponding to the object model is dynamically adjusted through logs, status monitoring, and exception handling mechanisms.

8. A system for orchestrating physical model capabilities based on a large model, characterized in that: The object model capability orchestration system based on the big model includes: The editing text acquisition module is used to acquire the editing text input by the user; A keyword and intent parsing module, used to parse the arrangement text based on the trained large model, and identify the keywords and intent information in the arrangement text; An operation and process generation module, used to generate an operation description and process information of an object model according to the keywords and the intention information; A call chain generation module is used to generate a call chain corresponding to the object model based on the operation description and the process information, so as to arrange the task execution order, optimize the use of resources and execute the corresponding tasks based on the call chain.

9. A terminal, characterized in that: The terminal includes: a memory, a processor, and a large-model-based physical model capability orchestration program stored in the memory and executable on the processor. When the large-model-based physical model capability orchestration program is executed by the processor, the steps of the large-model-based physical model capability orchestration method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a large-model-based physical model capability orchestration program, which, when executed by a processor, implements the steps of the large-model-based physical model capability orchestration method as described in any one of claims 1 to 7.

Images