Operation and maintenance operation script generation and execution method and device and storage medium

Abstract

The invention discloses an operation and maintenance operation script generation and execution method and device and a storage medium. The method comprises the steps of forming structured operation intention information based on a received voice operation and maintenance instruction, retrieving an operation and maintenance knowledge graph, generating an executable operation script in combination with a deep learning model, applying for the lowest operation authority required for executing the executable operation script, and executing the executable operation script according to the lowest operation authority. And performing pre-execution verification on the identified high-risk operation step in a sandbox environment, and running an executable operation script based on the lowest operation authority and the pre-execution verification. By forming structured operation intention information, manual conversion errors are solved; further, an executable operation script is dynamically generated based on the operation and maintenance knowledge graph and a deep learning model, and optimal practice is embedded; security is guaranteed through the lowest operation authority and sandbox environment pre-verification, and real-time decision and fault tolerance are carried out depending on an interface state, so that operation and maintenance operation standardization, experience accumulation, process auditing and human error risk elimination are realized.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus and storage medium for generating and executing operation and maintenance scripts. Background Technology

[0002] Currently, in the field of information system operations and maintenance (SEM), the transformation and execution of SEM requirements into specific operational instructions primarily relies on the personal knowledge and experience of SEM engineers. A typical workflow involves the engineer first understanding a natural language requirement such as "securely clean up historical data from before 2000 in a certain application system." Then, based on their understanding of system architecture, database operations, and security standards, the engineer manually translates this into a series of specific, executable operational instructions, such as "log in to the database," "perform a backup," "run a deletion script and check," and "log out of the database." These instructions may ultimately be implemented through manual operation of a graphical interface, execution of pre-written static scripts, or command-line tools. Summary of the Invention

[0003] To systematically reduce the risks of human error, improve the accuracy, completeness, and security of operation and maintenance operations, and achieve continuous accumulation and efficient reuse of operation and maintenance knowledge, this invention provides an operation and maintenance operation script generation and execution method, device, and storage medium. Based on the advantages of deep learning in complex semantic understanding and sequence generation, and the capabilities of knowledge graphs in structured storage and association of domain knowledge, combined with real-time environmental perception feedback provided by computer vision, an automated closed loop is constructed from natural language requirements to secure, reliable, and auditable execution.

[0004] As one aspect of the present invention, embodiments of the present invention provide a method for generating and executing operation and maintenance scripts, including:

[0005] The received voice operation and maintenance instructions are converted into text instructions, and the text instructions are semantically parsed to extract the operation and maintenance operation objects, operation actions and constraints to form structured operation intent information. Based on the operation intent information, retrieve the script templates and operation rules stored in the operation and maintenance knowledge graph, and generate an executable operation script by combining it with a deep learning model; Request the minimum operating permissions required to execute the executable operation script, and pre-execute and verify the high-risk operation steps identified by the script risk assessment model in a sandbox environment; Based on the minimum operating privileges and the pre-execution verification, the executable operation script is run, and the graphical user interface state related to the script execution is dynamically identified. The execution decision is made based on the semantic information of the identified interface state.

[0006] As another aspect of the present invention, embodiments of the present invention provide an operation and maintenance script generation and execution device, comprising:

[0007] The semantic understanding module is used to convert the received voice operation and maintenance instructions into text instructions, and to perform semantic parsing on the text instructions to extract the operation and maintenance operation objects, operation actions and constraints, and form structured operation intent information. The script generation module is used to retrieve script templates and operation rules stored in the operation and maintenance knowledge graph based on the operation intent information, and generate an executable operation script by combining it with a deep learning model. The security verification module is used to request the minimum operating permissions required to execute the executable operation script, and to pre-execute verification of high-risk operation steps identified by the script risk assessment model in a sandbox environment. The script execution module is used to run the executable operation script based on the minimum operation privileges and the pre-execution verification, and dynamically identify the graphical user interface state related to the script execution, and make operation decisions based on the semantic information of the identified interface state.

[0008] As another aspect of the present invention, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned operation and maintenance script generation and execution method.

[0009] As another aspect of the present invention, an embodiment of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the operation and maintenance script generation and execution method.

[0010] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following:

[0011] This invention provides a method for generating and executing operation and maintenance scripts. By converting received voice operation and maintenance instructions into text instructions and performing semantic parsing on these text instructions, the method extracts clear operation and maintenance objects, actions, and constraints, thereby forming structured operation intent information. This replaces manual translation and understanding, eliminating errors and omissions in instruction conversion caused by differences in personal experience or ambiguous expressions, ensuring that the original requirements are completely and accurately transformed into clear machine-processable intents. Furthermore, the system retrieves operation and maintenance knowledge graphs based on the operation intent information and generates executable operation scripts using a deep learning model, embedding best practices into the scripts. By requesting the minimum operation permissions required to execute the executable operation scripts and performing pre-execution verification in a sandbox environment, the system constrains the scope of operations from the permission and execution levels, preventing misoperations and information leaks. During operation, the system obtains the interface state of the graphical user interface and makes decisions or performs fault tolerance processing accordingly, ensuring the reliability of execution.

[0012] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0013] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0014] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0015] Figure 1 This is a flowchart of the operation and maintenance script generation and execution method provided in this embodiment of the invention;

[0016] Figure 2 This is a schematic diagram of the structure of the operation and maintenance script generation and execution device provided in this embodiment of the invention. Detailed Implementation

[0017] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0018] Current technologies relying on manual intervention have significant shortcomings. First, in the requirement transformation phase, they depend entirely on the individual engineer's knowledge and experience, which can easily lead to incorrect, inaccurate, or incomplete instructions due to insufficient knowledge or experience, such as omitting crucial risk mitigation steps (e.g., backups). Second, in the instruction implementation phase, direct manual operation or the use of static scripts makes it difficult to achieve strict minimum access control, increasing the risk of sensitive information (e.g., credentials) being leaked and the risk of human error is high. Furthermore, the knowledge generated throughout the operation and maintenance process (e.g., optimal operation sequences in specific scenarios, risk handling experience) is fragmented and individualized, making it difficult to accumulate effectively and pass on in a standardized manner. This also results in inefficient and difficult subsequent operation and maintenance audits due to the lack of structured and parsable execution records.

[0019] In view of the above problems, the inventors recognized that the core issue lies in the lack of an intelligent system capable of deeply understanding operational intentions and dynamically organizing and safely executing operations accordingly. Based on the advantages of deep learning in complex semantic understanding and sequence generation, and the capabilities of knowledge graphs in structured storage and association of domain knowledge, combined with real-time environmental awareness feedback provided by computer vision, the inventors propose a method for generating and executing operational operation scripts. This method aims to construct an automated closed loop from natural language requirements to secure, reliable, and auditable execution, thereby systematically reducing the risks of human error, improving the accuracy, completeness, and security of operational operations, and enabling the continuous accumulation and efficient reuse of operational knowledge.

[0020] It should be noted that the execution entity of each embodiment of this application can be a computing service system with data processing, network communication, and program execution functions, such as an electronic system capable of implementing the above functions, or a system for generating and executing operation and maintenance scripts. The following description uses an operation and maintenance script generation and execution system (hereinafter referred to as "the system") as an example to illustrate the following embodiments.

[0021] Example 1

[0022] Reference Figure 1 As shown, this embodiment of the invention provides a method for generating and executing operation and maintenance scripts, including:

[0023] S10: Convert the received voice operation and maintenance instructions into text instructions, and perform semantic parsing on the text instructions to extract the operation and maintenance operation objects, operation actions and constraints to form structured operation intent information; S20: Retrieve script templates and operation rules stored in the operation and maintenance knowledge graph based on the operation intent information, and generate an executable operation script by combining it with a deep learning model; S30: Request the minimum operating permissions required to execute the executable operation script, and pre-execute and verify the high-risk operation steps identified by the script risk assessment model in a sandbox environment; S40: Based on the minimum operating privileges and the pre-execution verification, run the executable operation script, dynamically identify the graphical user interface state related to the script execution, and make an operation decision based on the semantic information of the identified interface state.

[0024] Specifically, the system first extracts the operation objects, actions, and constraints using a specialized model trained on corpora from the operations and maintenance (O&M) domain, thus forming standardized, structured operation intent information. For example, for the instruction "Securely clean up data from before 2000 in a certain database," the system can accurately parse the operation object as "a certain database," the action as "clean up," and the constraint as "before 2000," and can also identify the additional requirements implied by the modifier "security." Then, based on the operation intent information, the system retrieves script templates and operation rules stored in the O&M knowledge graph, and combines this with a deep learning model to generate an executable operation script. This means that based on the relationships in the knowledge graph, the system fills in and adapts the constraints as parameters, and automatically inserts auxiliary operation steps such as backup and checks, thereby constructing a robust script logic sequence.

[0025] At the security execution level, the system requests the minimum operating permissions required to execute the executable script and pre-executes and verifies high-risk operation steps identified by the script risk assessment model in a sandbox environment before running the executable script. During this process, the system acquires the interface status of the relevant graphical user interface in real time and makes decisions or performs fault tolerance processing on the script execution based on the identified status (such as highlighting the "Confirm" button or displaying error warnings), thereby ensuring the reliability and adaptability of automated operations in complex graphical interface environments.

[0026] The decision-making process is based on the core intelligent response triggered by real-time recognition of the semantic state of the graphical user interface. Based on the recognized interface semantic information, the system autonomously selects and adjusts subsequent execution paths. For example, when the script reaches a critical step, if the deep learning model identifies a high-risk "Execute" or "Delete" button on the interface, the system can automatically pause and generate a semantic prompt requiring manual confirmation (such as "Detected 'Delete' operation, continue?"). Conversely, if a regular "Next" or "OK" button is detected, the system decides to automatically and safely click it to continue the process. It also includes fault tolerance: when the system detects an unexpected or erroneous interface state, it automatically initiates a preset recovery or correction process. For example, when an "Error" warning box pops up after an operation failure, the system automatically captures and records the error information and clicks the "Close" or "Cancel" button to restore the interface to a safe, retryable state; or when waiting for an element to appear, if the recognition timeout is caused by interface loading delay, the system automatically refreshes the page or rolls back the previous operation and then retryes execution.

[0027] In this embodiment of the invention, the step S10 described herein, which involves converting the received user-inputted voice operation and maintenance instructions into text instructions, may specifically include: processing the voice operation and maintenance instructions through a voice recognition component, filtering out non-critical spoken noise in the voice operation and maintenance instructions, and outputting accurate text instructions. The voice recognition component is trained based on a corpus of professional terminology and instruction sentence patterns in the operation and maintenance field.

[0028] Specifically, the core of this step lies in the use of a deeply optimized speech recognition component. This component is reinforced using a dedicated corpus containing professional terminology from the operations and maintenance field (such as "cluster," "load balancing," and "rollback") and typical command phrases (such as "restart [a service]" and "clean up historical data of [a database] from [a certain point in time]"). This training enables its acoustic and language models to more accurately recognize and process specific pronunciation combinations and grammatical structures in operations and maintenance scenarios.

[0029] When processing voice-based maintenance commands in real time, this component performs two key technical operations: First, it actively filters out non-critical spoken noise, such as interjections like "um" or "uh" used during thinking, unnecessary repetitions, or irrelevant narration, ensuring that the core command content is not disturbed; second, it prioritizes the identification and correction of high-frequency words and phrases in the maintenance field. For example, when pronunciation is unclear, it corrects potentially misheard words to the correct maintenance terminology based on the context.

[0030] In this embodiment of the invention, the step S10, which involves semantically parsing the text instruction to extract the operation and maintenance object, operation action, and constraints to form structured operation intent information, may specifically include steps S101-S102:

[0031] S101: Use a natural language processing model to parse the text instructions, identify the operation and maintenance domain entities in the text instructions as operation and maintenance operation objects and operation actions, and parse the time and condition parameters as constraints. The natural language processing model is a model trained and constructed based on historical operation and maintenance work orders and operation manual data. S102: Combine the operation and maintenance object, the operation action, and the constraint conditions into structured operation intent information.

[0032] Specifically, the training data for the natural language processing model consists of massive amounts of historical maintenance work orders and detailed operation manuals, enabling the model to internalize the specialized terminology system of the maintenance domain, common expressions of operational logic, and even implicit contextual conventions. When processing text instructions, the model first performs named entity recognition and semantic role labeling to accurately identify and classify entities in the maintenance domain. Among them, the target being operated on (such as "database A" or "service cluster B") is identified as the maintenance operation object, while the actions performed (such as "restart", "clean up", or "expand") are identified as operation actions.

[0033] Based on the identification of key entities, the model further analyzes the modifiers and limiting components in the instructions, specifically extracting time parameters (such as "before 2023", "first day of each month") and conditional parameters (such as "when disk usage is greater than 90%", "if backup is successful") as constraints. Finally, the system organizes and combines the above discretely extracted operation and maintenance objects, actions, and constraints according to a predefined structured framework (such as using JSON or a specific schema), thereby outputting a clear, structured operation intent information that can be seamlessly parsed and processed by a machine.

[0034] In this embodiment of the invention, step S20 may specifically include steps S201 to S204:

[0035] S201: Parse the operation intent information, extract the operation category and the operation object as search keywords, and use the search keywords as query conditions to traverse the nodes and edge relationships in the operation and maintenance knowledge graph to locate the relevant script templates and relevant operation rules;

[0036] S202: Based on the constraints in the operation intent information, select script templates and operation rules with a matching degree higher than the threshold from the relevant script templates and the relevant operation rules; S203: Using a deep learning model, the constraints are used as parameters to parameterize and adapt the script template, forming a parameterized script sequence;

[0037] S204: Dynamically combine the parameterized script sequence according to the operation rules, and insert auxiliary operation steps corresponding to the operation intent information to generate an executable operation script, wherein the auxiliary operation steps are obtained from the operation and maintenance knowledge graph and include at least one of backup, inspection or audit log generation.

[0038] Specifically, the system first extracts the core elements of the operation intent information—operation category (e.g., "data cleanup") and operation object (e.g., "database A")—as search keywords. Using these keywords as query conditions, the system traverses the nodes (representing entities, such as scripts, rules, and system components) and edges (representing relationships between entities, such as "belongs to," "depends on," and "contains") in the operation and maintenance knowledge graph. This allows for the precise identification of script template nodes and operation rule nodes associated with the current intent, completing the initial knowledge retrieval. Subsequently, the system introduces specific constraints from the operation intent information (e.g., "before 2000") as a fine-tuning filter. By calculating the semantic matching degree between the constraints and the context of the initially retrieved script templates and operation rules, the system retains only templates and rules with a matching degree higher than a preset threshold, ensuring a high degree of relevance between the selected knowledge elements and the current task.

[0039] The deep learning model receives the filtered script template and specific constraint parameters. The model not only populates the template variables with the parameters but also adapts and optimizes the template structure for context, outputting a parameter-ready script sequence. Finally, the system dynamically combines this script sequence according to the logic defined in the operation rules and retrieves and inserts corresponding auxiliary operation steps from the operation and maintenance knowledge graph (e.g., automatically inserting a backup command before performing deletion), thereby assembling and generating a complete and robust executable operation script.

[0040] In this embodiment of the invention, step S203 may specifically include steps S2031 to S2033:

[0041] S2031: Perform structured parsing on the constraints to generate a machine-processable parameter vector, and combine the operation intent information to generate a semantic embedding vector representing the context of this operation and maintenance.

[0042] S2032: Parse the script template into a script syntax tree containing variable placeholders, and input the parameter vector and the semantic embedding vector into a deep learning model;

[0043] S2033: Perform the following parameterization and adaptation steps using the deep learning model:

[0044] Based on the semantic embedding vector, the overall operation intent is understood, and the parameter vector is matched and filled into the corresponding variable placeholders in the script syntax tree to obtain the filled script syntax tree.

[0045] Based on the operation and maintenance knowledge graph, the filled script syntax tree is optimized and adjusted to obtain the adjusted script syntax tree;

[0046] The adjusted script syntax tree is serialized to output the final parameterized script sequence.

[0047] Specifically, first, the system performs structured parsing of the textual constraints, converting them into machine-processable parameter vectors. Simultaneously, the model generates a semantic embedding vector based on complete operational intent information. This vector encapsulates the deep semantics and context of the operation (e.g., the prudence mode implied in "safe cleanup"). Next, the system parses the textual script template into a structured intermediate representation, a script syntax tree, which explicitly identifies all placeholder variables that need to be filled. The parameter vector and semantic embedding vector are then fed together into the deep learning model.

[0048] The key parameterization and adaptation steps of the model execution consist of three stages: First, based on the semantic embedding vector, the overall operational intent is understood, and guided by this, the values ​​in the parameter vector are precisely matched and filled into the corresponding variable placeholders in the syntax tree, resulting in the filled script syntax tree. Then, the model performs logical optimization and adjustment on the tree based on the configuration associations and best practices stored in the operation and maintenance knowledge graph, such as enhancing fault tolerance logic, resulting in the adjusted script syntax tree. Finally, this optimized tree structure is serialized and restored to standard scripting language code, i.e., the final parameterized script sequence is output.

[0049] Furthermore, the step of optimizing and adjusting the padded script syntax tree based on the operation and maintenance knowledge graph to obtain an adjusted script syntax tree may specifically include: inputting the padded script syntax tree, the semantic embedding vector, and the associated system configuration snapshot extracted from the operation and maintenance knowledge graph into a script risk assessment model to predict potential abnormal paths and corresponding risk levels when the padded script syntax tree is executed in a real environment; for each potential abnormal path with a risk level exceeding a threshold, retrieving the corresponding risk mitigation strategy from the operation and maintenance knowledge graph; and inserting at least one protection node executing the risk mitigation strategy before or after the operation node related to the potential abnormal path in the padded script syntax tree to generate the adjusted script syntax tree.

[0050] Specifically, the input to the script risk assessment model is a comprehensive information body: including the populated script syntax tree (containing the operational logic with specific parameters), semantic embedding vectors (representing the context of operational intent), and snapshots of related system configurations extracted in real-time from the operations and maintenance knowledge graph (reflecting the real-time state of the target environment). By analyzing the relationships between these three elements, the model simulates and extrapolates various potential abnormal paths that may occur when the script executes in a real environment (e.g., failure branches due to insufficient resources, unavailability of dependent services, or excessively large parameter ranges), and assesses a risk level for each path. For each potential abnormal path whose risk level exceeds a threshold, the corresponding risk mitigation strategy is retrieved from the operations and maintenance knowledge graph. These strategies are pre-structured and stored solutions; for example, for the "insufficient disk space" risk, the strategy might be "insert disk space check and cleanup steps."

[0051] Finally, the system performs a substantial syntax tree reconstruction. In the padded script syntax tree, the relevant operation nodes leading to specific potentially abnormal paths are precisely located, and at least one protective node executing the risk mitigation strategy is inserted before or after that node. For example, a backup operation node is inserted before a high-risk deletion operation node. By implanting corresponding protective logic for all high-risk paths, a revised script syntax tree is ultimately generated, giving the script context-aware proactive defense capabilities, surpassing simple rule-based combinations of steps.

[0052] In this embodiment of the invention, step S40, which describes dynamically identifying the graphical user interface state related to the script execution and making a running decision based on the semantic information of the identified interface state, may specifically include steps S401 to S404:

[0053] S404: Capture an image of the relevant graphical user interface and compare the image with a preset element feature library to identify the interface state of the relevant graphical user interface, wherein the interface state includes the visual features and semantic state of specific buttons, tabs or warning boxes in the interface.

[0054] S402: When the interface state matches the expected state for the execution of the executable operation script, trigger the execution of the next operation instruction;

[0055] S403: When an error warning or an interface state that does not conform to the expected state is detected, execution is paused and a predefined exception handling process is triggered.

[0056] Specifically, by capturing images of relevant graphical user interfaces and comparing and recognizing them using a pre-built element feature library, the system can determine the current interface state in real time and accurately. This pre-built element feature library contains visual feature templates and semantic state definitions for specific buttons, tabs, or warning prompts in various operation and maintenance tool interfaces, such as "clickable," "disabled," and "highlighted warning." When the interface state matches the expected state for an executable operation script (for example, the expected "Next" button is clickable), the system will automatically trigger the execution of the next operation instruction, driving the script to continue running.

[0057] Conversely, when an error warning or an interface state inconsistent with the expected state is detected—for example, an unexpected "Connection Failure" dialog box pops up, or a target button becomes grayed out and unavailable—the system will immediately pause the current script execution and automatically trigger a predefined exception handling process. This process may include retrying the operation, recording the error state, notifying maintenance personnel, or executing a backup operation plan, achieving fault-tolerant processing based on real-time environmental feedback, thereby significantly improving the robustness and reliability of automated maintenance operations.

[0058] In practical applications, the implementation of this invention can be divided into two stages: the preparation stage (i.e., the template learning and knowledge base construction stage) and the usage stage (i.e., the instruction-driven and automated execution stage).

[0059] During the preparation phase, the system learns templates and constructs a knowledge graph under the supervision of operations engineers. Implementers first input standard template voice commands (VoiceCommandTemplate) using the system's voice input function. The system then uses its speech recognition module, enhanced with corpus data from the operations and maintenance domain, to convert the command into a text command (TextCommandTemplate) and display it, which is then verified by the implementers for accuracy. Next, the system calls a deep learning module trained on historical operations and maintenance data to transform the verified TextCommandTemplate into a reusable template operation script (OperateScriptTemplate), whose logic and steps are again reviewed and confirmed by the implementers. For non-scripted operations that must be completed through a graphical interface (e.g., clicking a specific button or entering information in a dialog box), the system uses its image recognition technology optimized for operations and maintenance interfaces to record the interaction process and generate a corresponding operation example template (OperateExampleTemplate), which also needs to be confirmed by the implementers. Finally, the system stores the associated TextCommandTemplate, OperateScriptTemplate, and OperateExampleTemplate to continuously accumulate and enrich its operation and maintenance knowledge base.

[0060] During the usage phase, the system leverages the established knowledge base to respond to new operational needs and execute automated operations. Implementers issue a new voice command (VoiceCommand) via voice input. The system converts this command into a text command (TextCommand) using its speech recognition module and displays it to the implementers for confirmation. Once confirmed, the system intelligently retrieves matching OperateScriptTemplate and OperateExampleTemplate from the knowledge base based on the TextCommand, combines and parameterizes them to generate a specific operation script instance (OperateScript). Implementers then conduct a final review of the generated OperateScript instance. After confirmation, they authorize the system to begin execution via voice command. The system then automatically completes the entire operational process—from permission request and sandbox verification to execution based on visual feedback and fault tolerance—according to the script.

[0061] It should be noted that the model used in this invention is constructed as follows:

[0062] The speech recognition model is obtained by collecting recordings of voice commands from maintenance personnel through the system and constructing a text corpus containing domain terminology. The basic ASR model is then fine-tuned and trained on this mixed data.

[0063] The natural language processing model uses historically accumulated operation and maintenance work orders, operation manuals and other texts as training sets. Through entity annotation and sequence annotation tasks, it trains a dedicated parsing model that can deeply understand the semantics of operation and maintenance.

[0064] The deep learning model for script generation and optimization uses structured script templates, parameterized examples, and corresponding operation contexts from knowledge graphs as training data. The model learns the generation and logical optimization rules from intent to script sequence through supervised learning.

[0065] The script risk assessment model is trained using configuration snapshots in the knowledge graph and manually annotated historical failure case data to learn to predict the potential risks of specific operations in specific environments.

[0066] The graphical interface recognition capability is achieved by systematically collecting and manually annotating screenshots of various mainstream operation and maintenance tools, and constructing a feature database that includes element positions, visual features, and semantic tags.

[0067] Example 2

[0068] Based on the same inventive concept, embodiments of the present invention also provide an operation and maintenance script generation and execution device, such as... Figure 2 As shown, it includes:

[0069] The semantic understanding module 10 is used to convert received voice operation and maintenance instructions into text instructions, and perform semantic parsing on the text instructions to extract operation and maintenance operation objects, operation actions, and constraints to form structured operation intent information; the script generation module 20 is used to retrieve script templates and operation rules stored in the operation and maintenance knowledge graph based on the operation intent information, and generate an executable operation script in combination with a deep learning model; the security verification module 30 is used to request the minimum operation permissions required to execute the executable operation script, and perform pre-execution verification on high-risk operation steps identified by the script risk assessment model in a sandbox environment; the script execution module 40 is used to run the executable operation script based on the minimum operation permissions and the pre-execution verification, and dynamically identify the graphical user interface state related to the script execution, and make operation decisions based on the semantic information of the identified interface state.

[0070] In this embodiment of the invention, the above-mentioned operation and maintenance script generation and execution device is a device corresponding to the operation and maintenance script generation and execution method provided in Embodiment 1. Its specific implementation can be referred to the detailed description of the operation and maintenance script generation and execution method in Embodiment 1. In this embodiment of the invention, it will not be repeated here.

[0071] Example 3

[0072] Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that executes the operation and maintenance script generation and execution method as described in Embodiment 1.

[0073] Example 4

[0074] Based on the same inventive concept, this embodiment of the invention also provides an operation and maintenance script generation and execution device, including the operation and maintenance script generation and execution method described in embodiment 1 above.

[0075] The principles by which the above-described apparatus, medium, related equipment, and system in the embodiments of the present invention solve the problem are similar to those of the aforementioned methods. Therefore, their implementation can refer to the implementation of the aforementioned methods, and repeated details will not be repeated.

[0076] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0077] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0078] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0079] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0080] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of equivalents of this invention, this invention also intends to include these modifications and variations.

Claims

1. An operation and maintenance operation script generation and execution method, characterized in that, include: The received voice operation and maintenance instructions are converted into text instructions, and the text instructions are semantically parsed to extract the operation and maintenance operation objects, operation actions and constraints to form structured operation intent information. Based on the operation intent information, retrieve the script templates and operation rules stored in the operation and maintenance knowledge graph, and generate an executable operation script by combining it with a deep learning model; Request the minimum operating permissions required to execute the executable operation script, and pre-execute and verify the high-risk operation steps identified by the script risk assessment model in a sandbox environment; Based on the minimum operating privileges and the pre-execution verification, the executable operation script is run, and the graphical user interface state related to the script execution is dynamically identified. The execution decision is made based on the semantic information of the identified interface state.

2. The operation and maintenance operation script generation and execution method according to claim 1, characterized in that, The step of retrieving script templates and operation rules stored in the operation and maintenance knowledge graph based on the operation intent information, and generating an executable operation script by combining it with a deep learning model, includes: The operation intent information is parsed, the operation category and the operation object are extracted as search keywords, and the search keywords are used as query conditions to traverse the nodes and edge relationships in the operation and maintenance knowledge graph to locate the relevant script templates and relevant operation rules. Based on the constraints in the operation intent information, script templates and operation rules with a matching degree higher than the threshold are selected from the relevant script templates and the relevant operation rules; By using a deep learning model, the constraints are used as parameters to parameterize and adapt the script template, forming a parameterized script sequence. The parameterized script sequence is dynamically combined according to the operation rules, and auxiliary operation steps corresponding to the operation intent information are inserted to generate an executable operation script. The auxiliary operation steps are obtained from the operation and maintenance knowledge graph and include at least one of backup, inspection or audit log generation.

3. The operation and maintenance operation script generation and execution method according to claim 2, characterized in that, The step of using a deep learning model to parameterize and adapt the constraints as parameters to the script template, forming a parameterized script sequence, includes: The constraints are structured and parsed to generate a machine-processable parameter vector, and the semantic embedding vector representing the context of this operation and maintenance operation is generated by combining the operation intent information. The script template is parsed into a script syntax tree containing variable placeholders, and the parameter vector and the semantic embedding vector are input into a deep learning model; The following parameterization and adaptation steps are performed using the deep learning model: Based on the semantic embedding vector, the overall operation intent is understood, and the parameter vector is matched and filled into the corresponding variable placeholders in the script syntax tree to obtain the filled script syntax tree. Based on the operation and maintenance knowledge graph, the filled script syntax tree is optimized and adjusted to obtain the adjusted script syntax tree; The adjusted script syntax tree is serialized to output the final parameterized script sequence.

4. The operation and maintenance operation script generation and execution method according to claim 3, characterized in that, The step of optimizing and adjusting the filled script syntax tree based on the operation and maintenance knowledge graph to obtain an adjusted script syntax tree includes: The filled script syntax tree, the semantic embedding vector, and the associated system configuration snapshot extracted from the operation and maintenance knowledge graph are input into the script risk assessment model to predict the potential abnormal paths and corresponding risk levels when the filled script syntax tree is executed in a real environment. For each potential abnormal path whose risk level exceeds the threshold, the corresponding risk mitigation strategy is retrieved from the operation and maintenance knowledge graph; In the padded script syntax tree, at least one protective node that executes the risk mitigation strategy is inserted before or after the operation node related to the potential abnormal path, thereby generating an adjusted script syntax tree.

5. The operation and maintenance operation script generation and execution method according to claim 1, characterized in that, The process of dynamically identifying the graphical user interface state related to the script execution and making operational decisions based on the semantic information of the identified interface state includes: Capture images of relevant graphical user interfaces and compare the images with a preset element feature library to identify the interface state of the relevant graphical user interface, wherein the interface state includes the visual features and semantic state of specific buttons, tabs or warning boxes in the interface; When the interface state matches the expected state for the execution of the executable operation script, the next operation instruction is triggered. When an error warning or an interface state that does not conform to the expected state is detected, execution is paused and a predefined exception handling process is triggered.

6. The operation and maintenance operation script generation and execution method according to claim 1, characterized in that, The step of converting received user-inputted voice operation and maintenance commands into text commands includes: The voice operation and maintenance instructions are processed by a voice recognition component, and non-critical spoken noise in the voice operation and maintenance instructions is filtered out to output accurate text instructions. The voice recognition component is trained based on a corpus of operation and maintenance professional terminology and instruction sentence patterns.

7. The method for generating and executing operation and maintenance scripts according to claim 1, characterized in that, The semantic parsing of the text instructions, extracting the operation and maintenance object, operation action, and constraints, to form structured operation intent information includes: The text instructions are parsed using a natural language processing model to identify the operation and maintenance domain entities in the text instructions as operation and maintenance operation objects and actions, and to parse the time and condition parameters as constraints. The natural language processing model is a model trained and constructed based on historical operation and maintenance work orders and operation manual data. The operation and maintenance operation objects, the operation actions, and the constraints are combined into structured operation intent information.

8. A device for generating and executing operation and maintenance scripts, characterized in that, include: The semantic understanding module is used to convert the received voice operation and maintenance instructions into text instructions, and to perform semantic parsing on the text instructions to extract the operation and maintenance operation objects, operation actions and constraints to form structured operation intent information. The script generation module is used to retrieve script templates and operation rules stored in the operation and maintenance knowledge graph based on the operation intent information, and generate an executable operation script by combining it with a deep learning model. The security verification module is used to request the minimum operating permissions required to execute the executable operation script, and to perform pre-execution verification of high-risk operation steps identified by the script risk assessment model in a sandbox environment. The script execution module is used to run the executable operation script based on the minimum operation privileges and the pre-execution verification, and dynamically identify the graphical user interface state related to the script execution, and make operation decisions based on the semantic information of the identified interface state.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the operation and maintenance script generation and execution method as described in any one of claims 1 to 7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the operation and maintenance script generation and execution method as described in any one of claims 1 to 7.

Images