Method and system for quickly creating operation and maintenance work order based on large model

Abstract

The invention discloses an operation and maintenance work order rapid creation method and system based on a large model, belongs to the technical field of large model application, and aims to solve the technical problems of how to reduce necessary input items of a work order, reduce input time, improve manual order creation efficiency and accelerate work order circulation speed. According to the technical scheme, the method comprises the following steps: interface packaging: taking work order parameters of problem description, business classification, problem level, work order source and work order title of a historical operation and maintenance work order as fine adjustment parameters of a large model, and packaging a large model reasoning interface through model pre-training; creating a work order: inputting work order problem description on a page, returning work order title, business classification, work order source and problem level necessary parameter information of work order creation through a large model reasoning interface, directly backfilling the information into a front-end work order creation page input box, and meanwhile, allowing work order creation personnel to adjust recommended work order parameters, and taking the created work order as a parameter for fine tuning of a subsequent large model.

Description

Technical Field

[0001] The present invention relates to the technical field of large model application, and in particular to a method and system for quickly creating an operation and maintenance work order based on a large model. Background Art

[0002] Operation and maintenance work order, referred to as work order, is a work order for handling operation and maintenance issues.

[0003] Currently, when filling out work orders manually, there is a lot of required information to be filled in, and there may be dependencies between the filled parameters, which makes the filling easy to make mistakes and the filling efficiency is low. In addition, due to the uneven professional quality of technical personnel, there is a strong reliance on experience in creating work orders, which is not friendly to novices.

[0004] Therefore, how to reduce the required input items for work orders, reduce input time, improve the efficiency of manual order creation, and speed up the turnover of work orders is a technical problem that needs to be solved urgently. Summary of the invention

[0005] The technical task of the present invention is to provide a method and system for quickly creating operation and maintenance work orders based on a large model to solve the problem of how to reduce the required input items of work orders, reduce input time, improve the efficiency of manual order creation, and speed up the circulation of work orders.

[0006] The technical task of the present invention is achieved in the following way: a method for quickly creating an operation and maintenance work order based on a large model, the method is specifically as follows:

[0007] Interface encapsulation: The problem description, business classification, problem level, work order source, and work order title of the historical operation and maintenance work order are used as the fine-tuning parameters of the large model. Through model pre-training, the large model reasoning interface is encapsulated;

[0008] Create a work order: You only need to enter the work order problem description on the page. The necessary parameter information for creating the work order, such as the work order title, business category, work order source, and problem level, will be returned through the big model inference interface and directly filled back into the input box on the front-end work order creation page. At the same time, the work order creator is allowed to adjust the recommended work order parameters and use the created work order as the parameters for subsequent big model fine-tuning.

[0009] As a preferred embodiment, the business classification adopts multi-level business classification, and there is a parent-child relationship between the business classification levels;

[0010] When the service classification is level 3, the large model reasoning interface is as follows:

[0011] Secondary business classification: Use historical work order data to train a multi-classification model for secondary business classification, and infer the secondary business classification ID based on the work order body content;

[0012] First-level business classification: Use the second-level business classification id classified by the multi-classification model of the second-level business classification and find the parent classification of the second-level business classification to obtain the first-level business classification;

[0013] Third-level business classification: read all third-level business classification IDs under the second-level business classification, and use the prompt method to determine the third-level business classification ID from the big model;

[0014] If the secondary business classification inference result is returned as null, all classifications will be returned as null directly.

[0015] Preferably, the work order title is generated by using the summary generation capability of the large model or keyword extraction; specifically, as follows:

[0016] (1) Description of the problem;

[0017] (2) Determine whether the number of words in the problem description is greater than the configured value:

[0018] ① If the word count of the problem description is less than the configured value, jump to step (7);

[0019] ② If the word count of the problem description is greater than the configured value, execute step (3);

[0020] (3) Generate summary information using the summary generation capability of the large model reasoning interface;

[0021] (4) Determine whether the number of characters in the summary information text is greater than the configured value:

[0022] ① If the number of characters in the summary information text is less than the configured value, jump to step (7);

[0023] ② If the number of characters in the summary information text is greater than the configured value, execute step (5);

[0024] (5) Use the keyword extraction capability of the large model reasoning interface to extract keywords from the generated summary information;

[0025] (6) Determine whether the number of extracted keywords is greater than the configured value:

[0026] ① If the number of extracted keywords is less than the configured value, jump to step (7);

[0027] ② If the number of words in the extracted keywords is greater than the configured value, select the first few keywords and execute step (7).

[0028] (7) Return the content as the work order title.

[0029] Preferably, the source of the work order is obtained through the trained source multi-classification model and inference based on the content of the work order body;

[0030] The problem level is obtained by inferring the trained problem-level multi-classification model based on the content of the work order.

[0031] Preferably, create a work order as follows:

[0032] (1) Enter the work order problem description on the front-end page;

[0033] (2) The backend calls the business classification interface of the large model and determines whether the returned classification is empty:

[0034] If the returned category id is not empty, determine whether the hierarchical relationship is logical:

[0035] If the hierarchical relationship is correct, the verification passes and the corresponding result is recorded;

[0036] If it does not return normally, the record will be empty;

[0037] If the returned result is not empty but the hierarchical relationship between categories is incorrect, the corresponding parent category is searched according to the secondary category, and only the classification results of the first two levels are recorded, and finally the results are temporarily stored in the variable;

[0038] (3) The backend calls the work order title interface, problem level interface, and work order source interface of the large model respectively and records the results;

[0039] (4) Obtain the pre-configured parameter rules and perform rule verification on the return value of the large model;

[0040] (5) The backend assembles the recorded results into JSON and returns it to the frontend. At the same time, the backend records the large model interface request and the returned result log, and the frontend displays it;

[0041] (6) The user determines whether the front-end echo is correct:

[0042] If it is incorrect, the front page supports manual modification. After the modification is completed, a work order is created. At this time, the user's order creation parameter log is recorded for subsequent fine-tuning of the large model parameters.

[0043] A system for quickly creating operation and maintenance work orders based on a large model, the system comprising:

[0044] Application service instance module, which handles ticket creation and logging;

[0045] Large model recommendation module, used to recommend work order parameters based on the problem description;

[0046] The information storage module is used to operate the database, save the log information, work order information and system configuration information, and add, delete, modify and query; the database is used to save various information of the system, including work orders, logs and system configuration tables.

[0047] Preferably, the large model recommendation submodule includes:

[0048] The business category recommendation submodule is used to recommend the corresponding business category based on the work order problem description;

[0049] Title recommendation submodule, used to generate work order titles;

[0050] The source recommendation submodule is used to determine the source of the work order;

[0051] The problem level recommendation submodule is used to determine the problem level of the work order.

[0052] Preferably, the business classification recommendation submodule adopts multi-level business classification, and there is a parent-child relationship between the business classification levels;

[0053] When the service classification is level 3, the large model reasoning interface is as follows:

[0054] Secondary business classification: Use historical work order data to train a multi-classification model for secondary business classification, and infer the secondary business classification ID based on the work order body content;

[0055] First-level business classification: Use the second-level business classification id classified by the multi-classification model of the second-level business classification and find the parent classification of the second-level business classification to obtain the first-level business classification;

[0056] Third-level business classification: read all third-level business classification IDs under the second-level business classification, and use the prompt method to determine the third-level business classification ID from the big model;

[0057] If the secondary business classification inference result is returned as null, all classifications will be returned as null directly.

[0058] Preferably, the title recommendation submodule generates a work order title through the summary generation capability or keyword extraction of the large model; the working process of the title recommendation submodule is as follows:

[0059] (1) Description of the problem;

[0060] (2) Determine whether the number of words in the problem description is greater than the configured value:

[0061] ① If the word count of the problem description is less than the configured value, jump to step (7);

[0062] ② If the word count of the problem description is greater than the configured value, execute step (3);

[0063] (3) Generate summary information using the summary generation capability of the large model reasoning interface;

[0064] (4) Determine whether the number of characters in the summary information text is greater than the configured value:

[0065] ① If the number of characters in the summary information text is less than the configured value, jump to step (7);

[0066] ② If the number of characters in the summary information text is greater than the configured value, execute step (5);

[0067] (5) Use the keyword extraction capability of the large model reasoning interface to extract keywords from the generated summary information;

[0068] (6) Determine whether the number of extracted keywords is greater than the configured value:

[0069] ① If the number of extracted keywords is less than the configured value, jump to step (7);

[0070] ② If the number of words in the extracted keywords is greater than the configured value, select the first few keywords and execute step (7).

[0071] (7) Return the content as the work order title;

[0072] The source recommendation submodule is obtained by inference based on the trained source multi-classification model and the content of the work order body;

[0073] The question-level recommendation submodule is obtained through the trained question-level multi-classification model and reasoning based on the content of the work order body.

[0074] Preferably, the application service instance module includes:

[0075] The rule configuration submodule is used to set and maintain the business rules for work order creation;

[0076] The operation and maintenance work order creation submodule is used to generate work orders based on user input and large model recommendations;

[0077] The logging submodule is used to record key information and user operations during the work order creation process and subsequent model fine-tuning;

[0078] The work order creation process is as follows:

[0079] (1) Enter the work order problem description on the front-end page;

[0080] (2) The backend calls the business classification interface of the large model and determines whether the returned classification is empty:

[0081] If the returned category id is not empty, determine whether the hierarchical relationship is logical:

[0082] If the hierarchical relationship is correct, the verification passes and the corresponding result is recorded;

[0083] If it does not return normally, the record will be empty;

[0084] If the returned result is not empty but the hierarchical relationship between categories is incorrect, the corresponding parent category is searched according to the secondary category, and only the classification results of the first two levels are recorded, and finally the results are temporarily stored in the variable;

[0085] (3) The backend calls the work order title interface, problem level interface, and work order source interface of the large model respectively and records the results;

[0086] (4) Obtain the pre-configured parameter rules and perform rule verification on the return value of the large model;

[0087] (5) The backend assembles the recorded results into JSON and returns it to the frontend. At the same time, the backend records the large model interface request and the returned result log, and the frontend displays it;

[0088] (6) The user determines whether the front-end echo is correct:

[0089] If it is incorrect, the front page supports manual modification. After the modification is completed, a work order is created. At this time, the user's order creation parameter log is recorded for subsequent fine-tuning of the large model parameters.

[0090] Among them, business classification: identifies the specific classification of work orders, examples: request work orders, event work orders, alarm work orders, etc., including multiple levels, with the concept of parent-child hierarchy.

[0091] Problem level: identifies the level of the work order, generally S1, S2, S3, etc. Each level corresponds to a different work order processing time, and the higher the level, the shorter the required processing time. The problem level is related to the business classification. Generally, only certain problem levels can be selected under a certain classification. For example, when creating a work order, the alarm classification can only be selected from the M1-M3 level. Of course, the level can be changed manually during the work order flow.

[0092] Work Order Source: Identifies the source of the work order, which generally includes conventional sources such as SMS, WeChat, and email, as well as automatic trigger sources such as monitoring alarms. The source is also related to the business category. For example, the work order source corresponding to the alarm category can only be alarms, not others.

[0093] The large model-based operation and maintenance work order rapid creation method and system of the present invention have the following advantages:

[0094] (I) Compared with the traditional manual ticket creation method, the present invention only needs to input the problem description to automatically generate the required parameters for creating a work order and fill them back on the page, which can greatly reduce the required input items and reduce the input time. The work order creator can quickly create the corresponding work order, speed up the work order circulation speed, and improve the efficiency of problem solving; at the same time, it can also reduce the error rate of manual ticket creation parameter filling, and the recommended parameters of the large model are trained based on historical work orders, which can effectively reduce the empirical dependence of creating work orders and is friendly to novices. At the same time, the present invention has good scalability and continuous learning capabilities;

[0095] (ii) The present invention can significantly reduce the number of required input items, reduce input time, effectively improve the efficiency of manual work order creation, speed up the work order circulation speed, and improve the efficiency of problem solving; at the same time, it can effectively reduce the empirical dependence on the creation of work orders and be friendly to novices. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] The present invention is further described below in conjunction with the accompanying drawings.

[0097] Attached Figure 1 Quickly create a system structure diagram for operation and maintenance work orders based on large models;

[0098] Attached Figure 2 A flowchart generated for the work order title;

[0099] Attached Figure 3 A flowchart for creating a work order. DETAILED DESCRIPTION

[0100] The large model-based operation and maintenance work order rapid creation method and system of the present invention are described in detail below with reference to the accompanying drawings and specific embodiments of the specification.

[0101] Embodiment 1:

[0102] This embodiment provides a method for quickly creating an operation and maintenance work order based on a large model, and the method is specifically as follows:

[0103] S1. Interface encapsulation: The problem description, business classification, problem level, work order source and work order title of the historical operation and maintenance work order are used as the fine-tuning parameters of the large model. Through model pre-training, the large model reasoning interface is encapsulated;

[0104] S2. Create a work order: You only need to enter the work order problem description on the page. The necessary parameter information for creating the work order, such as the work order title, business category, work order source, and problem level, will be returned through the big model inference interface and directly filled back into the input box on the front-end work order creation page. At the same time, the work order creator is allowed to adjust the recommended work order parameters and use the created work order as the parameters for subsequent big model fine-tuning.

[0105] The business classification in this embodiment adopts multi-level business classification, and there is a parent-child relationship between the business classification levels;

[0106] When the service classification is level 3, the large model reasoning interface is as follows:

[0107] Secondary business classification: Use historical work order data to train a multi-classification model for secondary business classification, and infer the secondary business classification ID based on the work order body content;

[0108] First-level business classification: Use the second-level business classification id classified by the multi-classification model of the second-level business classification and find the parent classification of the second-level business classification to obtain the first-level business classification;

[0109] Third-level business classification: read all third-level business classification IDs under the second-level business classification, and use the prompt method to determine the third-level business classification ID from the big model;

[0110] If the secondary business classification inference result is returned as null, all classifications will be returned as null directly.

[0111] As attached Figure 2 As shown, the work order title in this embodiment is generated by the summary generation capability of the large model or keyword extraction; the details are as follows:

[0112] (1) Description of the problem;

[0113] (2) Determine whether the number of words in the problem description is greater than the configured value:

[0114] ① If the word count of the problem description is less than the configured value, jump to step (7);

[0115] ② If the word count of the problem description is greater than the configured value, execute step (3);

[0116] (3) Generate summary information using the summary generation capability of the large model reasoning interface;

[0117] (4) Determine whether the number of characters in the summary information text is greater than the configured value:

[0118] ① If the number of characters in the summary information text is less than the configured value, jump to step (7);

[0119] ② If the number of characters in the summary information text is greater than the configured value, execute step (5);

[0120] (5) Use the keyword extraction capability of the large model reasoning interface to extract keywords from the generated summary information;

[0121] (6) Determine whether the number of extracted keywords is greater than the configured value:

[0122] ① If the number of extracted keywords is less than the configured value, jump to step (7);

[0123] ② If the number of words in the extracted keywords is greater than the configured value, select the first few keywords and execute step (7).

[0124] (7) Return the content as the work order title.

[0125] The source of the work order in this embodiment is obtained through a trained source multi-classification model and reasoning based on the content of the work order body.

[0126] The problem level in this embodiment is obtained by inferring the trained problem level multi-classification model based on the content of the work order body.

[0127] As attached Figure 3 As shown, in this embodiment, creating a work order is as follows:

[0128] (1) Enter the work order problem description on the front-end page;

[0129] (2) The backend calls the business classification interface of the large model and determines whether the returned classification is empty:

[0130] If the returned category id is not empty, determine whether the hierarchical relationship is logical:

[0131] If the hierarchical relationship is correct, the verification passes and the corresponding result is recorded;

[0132] If it does not return normally, the record will be empty;

[0133] If the returned result is not empty but the hierarchical relationship between categories is incorrect, the corresponding parent category is searched according to the secondary category, and only the classification results of the first two levels are recorded, and finally the results are temporarily stored in the variable;

[0134] (3) The backend calls the work order title interface, problem level interface, and work order source interface of the large model respectively and records the results;

[0135] (4) Obtain the pre-configured parameter rules and perform rule verification on the return value of the big model. The purpose is to accumulate some work order creation rules during the business operation process. The problem level and source of the work order may change during the processing, but generally will not change again after the work order is completed. Because the data relied on by the model training is historical work orders, the problem level and source parameters after the final completion are used for training, which may be inconsistent with what the actual user wants. Therefore, if you rely solely on the results returned by the big model, the results will be uncertain and the business will be affected, so it is necessary to verify the return value of the big model.

[0136] Common rules include: a category can only correspond to certain problem levels or sources. For example, the request category can only correspond to the request (R level); the event category corresponds to the event level S or M, but not the request level R; the source of the alarm ticket can only be alarm, and the source of other types of tickets does not need to be specified;

[0137] (5) The backend assembles the recorded results into JSON and returns it to the frontend. At the same time, the backend records the large model interface request and the returned result log, and the frontend displays it;

[0138] (6) The user determines whether the front-end echo is correct:

[0139] If it is incorrect, the front page supports manual modification. After the modification is completed, a work order is created. At this time, the user's order creation parameter log is recorded for subsequent fine-tuning of the large model parameters.

[0140] Embodiment 2:

[0141] As attached Figure 1 As shown, this embodiment provides a system for quickly creating an operation and maintenance work order based on a large model, and the system includes:

[0142] Application service instance module, which handles ticket creation and logging;

[0143] Large model recommendation module, used to recommend work order parameters based on the problem description;

[0144] The information storage module is used to operate the database, save the log information, work order information and system configuration information, and add, delete, modify and query; the database is used to save various information of the system, including work orders, logs and system configuration tables.

[0145] The large model recommendation submodule in this embodiment includes:

[0146] The business category recommendation submodule is used to recommend the corresponding business category based on the work order problem description;

[0147] Title recommendation submodule, used to generate work order titles;

[0148] The source recommendation submodule is used to determine the source of the work order;

[0149] The problem level recommendation submodule is used to determine the problem level of the work order.

[0150] The business classification recommendation submodule in this embodiment adopts multi-level business classification, and there is a parent-child relationship between the business classification levels;

[0151] When the service classification is level 3, the large model reasoning interface is as follows:

[0152] Secondary business classification: Use historical work order data to train a multi-classification model for secondary business classification, and infer the secondary business classification ID based on the work order body content;

[0153] First-level business classification: Use the second-level business classification id classified by the multi-classification model of the second-level business classification and find the parent classification of the second-level business classification to obtain the first-level business classification;

[0154] Third-level business classification: read all third-level business classification IDs under the second-level business classification, and use the prompt method to determine the third-level business classification ID from the big model;

[0155] If the secondary business classification inference result is returned as null, all classifications will be returned as null directly.

[0156] The title recommendation submodule in this embodiment generates a work order title through the summary generation capability or keyword extraction of the large model; Figure 2 As shown in the figure, the working process of the title recommendation submodule is as follows:

[0157] (1) Description of the problem;

[0158] (2) Determine whether the number of words in the problem description is greater than the configured value:

[0159] ① If the word count of the problem description is less than the configured value, jump to step (7);

[0160] ② If the word count of the problem description is greater than the configured value, execute step (3);

[0161] (3) Generate summary information using the summary generation capability of the large model reasoning interface;

[0162] (4) Determine whether the number of characters in the summary information text is greater than the configured value:

[0163] ① If the number of characters in the summary information text is less than the configured value, jump to step (7);

[0164] ② If the number of characters in the summary information text is greater than the configured value, execute step (5);

[0165] (5) Use the keyword extraction capability of the large model reasoning interface to extract keywords from the generated summary information;

[0166] (6) Determine whether the number of extracted keywords is greater than the configured value:

[0167] ① If the number of extracted keywords is less than the configured value, jump to step (7);

[0168] ② If the number of words in the extracted keywords is greater than the configured value, select the first few keywords and execute step (7).

[0169] (7) Return the content as the work order title.

[0170] The source recommendation submodule in this embodiment is obtained through the trained source multi-classification model and reasoning based on the content of the work order body.

[0171] The problem level recommendation submodule in this embodiment is obtained by inferring the trained problem level multi-classification model according to the content of the work order body.

[0172] The application service instance module in this embodiment includes:

[0173] The rule configuration submodule is used to set and maintain the business rules for work order creation;

[0174] The operation and maintenance work order creation submodule is used to generate work orders based on user input and large model recommendations;

[0175] The logging submodule is used to record key information and user operations during the work order creation process and subsequent model fine-tuning;

[0176] Among them, as attached Figure 3 As shown in the figure, the work order creation process is as follows:

[0177] (1) Enter the work order problem description on the front-end page;

[0178] (2) The backend calls the business classification interface of the large model and determines whether the returned classification is empty:

[0179] If the returned category id is not empty, determine whether the hierarchical relationship is logical:

[0180] If the hierarchical relationship is correct, the verification passes and the corresponding result is recorded;

[0181] If it does not return normally, the record will be empty;

[0182] If the returned result is not empty but the hierarchical relationship between categories is incorrect, the corresponding parent category is searched according to the secondary category, and only the classification results of the first two levels are recorded, and finally the results are temporarily stored in the variable;

[0183] (3) The backend calls the work order title interface, problem level interface, and work order source interface of the large model respectively and records the results;

[0184] (4) Obtain the pre-configured parameter rules and perform rule verification on the return value of the large model;

[0185] (5) The backend assembles the recorded results into JSON and returns it to the frontend. At the same time, the backend records the large model interface request and the returned result log, and the frontend displays it;

[0186] (6) The user determines whether the front-end echo is correct:

[0187] If it is incorrect, the front page supports manual modification. After the modification is completed, a work order is created. At this time, the user's order creation parameter log is recorded for subsequent fine-tuning of the large model parameters.

[0188] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for quickly creating an operation and maintenance work order based on a large model, characterized in that: The method is as follows: Interface encapsulation: The problem description, business classification, problem level, work order source, and work order title of the historical operation and maintenance work order are used as the fine-tuning parameters of the large model. Through model pre-training, the large model reasoning interface is encapsulated; Create a work order: Enter the work order problem description on the page, and return the necessary parameter information for creating the work order, such as the work order title, business category, work order source, and problem level, through the big model inference interface and directly fill it back into the input box of the front-end work order creation page. At the same time, the work order creator is allowed to adjust the recommended work order parameters and use the created work order as the parameters for subsequent big model fine-tuning.

2. The method for quickly creating an operation and maintenance work order based on a large model according to claim 1 is characterized in that: Business classification adopts multi-level business classification, and there is a parent-child relationship between business classification levels; When the service classification is level 3, the large model reasoning interface is as follows: Secondary business classification: Use historical work order data to train a multi-classification model for secondary business classification, and infer the secondary business classification ID based on the work order body content; First-level business classification: Use the second-level business classification id classified by the multi-classification model of the second-level business classification and find the parent classification of the second-level business classification to obtain the first-level business classification; Third-level business classification: read all third-level business classification IDs under the second-level business classification, and use the prompt method to determine the third-level business classification ID from the big model; If the secondary business classification inference result is returned as null, all classifications will be returned as null directly.

3. The method for quickly creating an operation and maintenance work order based on a large model according to claim 1 is characterized in that: The work order title is generated by the summary generation capability of the big model or keyword extraction; the details are as follows: (1) Description of the problem; (2) Determine whether the number of words in the problem description is greater than the configured value: ① If the word count of the problem description is less than the configured value, jump to step (7); ② If the word count of the problem description is greater than the configured value, execute step (3); (3) Generate summary information using the summary generation capability of the large model reasoning interface; (4) Determine whether the number of characters in the summary information text is greater than the configured value: ① If the number of characters in the summary information text is less than the configured value, jump to step (7); ② If the number of characters in the summary information text is greater than the configured value, execute step (5); (5) Use the keyword extraction capability of the large model reasoning interface to extract keywords from the generated summary information; (6) Determine whether the number of extracted keywords is greater than the configured value: ① If the number of extracted keywords is less than the configured value, jump to step (7); ② If the number of words in the extracted keywords is greater than the configured value, select the first few keywords and execute step (7). (7) Return the content as the work order title.

4. The method for quickly creating an operation and maintenance work order based on a large model according to claim 1 is characterized in that: The source of the work order is obtained through the trained source multi-classification model and inference based on the content of the work order body; The problem level is obtained by inferring the trained problem-level multi-classification model based on the content of the work order.

5. The method for quickly creating an operation and maintenance work order based on a large model according to any one of claims 1 to 4, characterized in that: To create a work order, follow these steps: (1) Enter the work order problem description on the front-end page; (2) The backend calls the business classification interface of the large model and determines whether the returned classification is empty: If the returned category id is not empty, determine whether the hierarchical relationship is logical: If the hierarchical relationship is correct, the verification passes and the corresponding result is recorded; If it does not return normally, the record will be empty; If the returned result is not empty but the hierarchical relationship between categories is incorrect, the corresponding parent category is searched according to the secondary category, and only the classification results of the first two levels are recorded, and finally the results are temporarily stored in the variable; (3) The backend calls the work order title interface, problem level interface, and work order source interface of the large model respectively and records the results; (4) Obtain the pre-configured parameter rules and perform rule verification on the return value of the large model; (5) The backend assembles the recorded results into JSON and returns it to the frontend. At the same time, the backend records the large model interface request and the returned result log, and the frontend displays it; (6) The user determines whether the front-end echo is correct: If it is incorrect, the front page supports manual modification. After the modification is completed, a work order is created. At this time, the user's order creation parameter log is recorded for subsequent fine-tuning of the large model parameters.

6. A system for quickly creating operation and maintenance work orders based on a large model, characterized in that: The system includes: Application service instance module, which handles ticket creation and logging; Large model recommendation module, used to recommend work order parameters based on the problem description; The information storage module is used to operate the database, save the log information, work order information and system configuration information, and add, delete, modify and query; the database is used to save various information of the system, including work orders, logs and system configuration tables.

7. The large model-based operation and maintenance work order rapid creation system according to claim 6 is characterized in that: The large model recommendation submodule includes: The business category recommendation submodule is used to recommend the corresponding business category based on the work order problem description; Title recommendation submodule, used to generate work order titles; The source recommendation submodule is used to determine the source of the work order; The problem level recommendation submodule is used to determine the problem level of the work order.

8. The large model-based operation and maintenance work order rapid creation system according to claim 6 is characterized in that: The business classification recommendation submodule adopts multi-level business classification, and there is a parent-child relationship between the business classification levels; When the service classification is level 3, the large model reasoning interface is as follows: Secondary business classification: Use historical work order data to train a multi-classification model for secondary business classification, and infer the secondary business classification ID based on the work order body content; First-level business classification: Use the second-level business classification id classified by the multi-classification model of the second-level business classification and find the parent classification of the second-level business classification to obtain the first-level business classification; Third-level business classification: read all third-level business classification IDs under the second-level business classification, and use the prompt method to determine the third-level business classification ID from the big model; If the secondary business classification inference result is returned as null, all classifications will be returned as null directly.

9. The large model-based operation and maintenance work order rapid creation system according to claim 6 is characterized in that: The title recommendation submodule generates a work order title through the summary generation capability or keyword extraction of the large model; the working process of the title recommendation submodule is as follows: (1) Description of the problem; (2) Determine whether the number of words in the problem description is greater than the configured value: ① If the word count of the problem description is less than the configured value, jump to step (7); ② If the word count of the problem description is greater than the configured value, execute step (3); (3) Generate summary information using the summary generation capability of the large model reasoning interface; (4) Determine whether the number of characters in the summary information text is greater than the configured value: ① If the number of characters in the summary information text is less than the configured value, jump to step (7); ② If the number of characters in the summary information text is greater than the configured value, execute step (5); (5) Use the keyword extraction capability of the large model reasoning interface to extract keywords from the generated summary information; (6) Determine whether the number of extracted keywords is greater than the configured value: ① If the number of extracted keywords is less than the configured value, jump to step (7); ② If the number of words in the extracted keywords is greater than the configured value, select the first few keywords and execute step (7). (7) Return the content as the work order title; The source recommendation submodule is obtained by inference based on the trained source multi-classification model and the content of the work order body; The question-level recommendation submodule is obtained through the trained question-level multi-classification model and reasoning based on the content of the work order body.

10. The large model-based operation and maintenance work order rapid creation system according to any one of claims 6 to 9, characterized in that: The application service instance module includes: The rule configuration submodule is used to set and maintain the business rules for work order creation; The operation and maintenance work order creation submodule is used to generate work orders based on user input and large model recommendations; The logging submodule is used to record key information and user operations during the work order creation process and subsequent model fine-tuning; The work order creation process is as follows: (1) Enter the work order problem description on the front-end page; (2) The backend calls the business classification interface of the large model and determines whether the returned classification is empty: If the returned category id is not empty, determine whether the hierarchical relationship is logical: If the hierarchical relationship is correct, the verification passes and the corresponding result is recorded; If it does not return normally, the record will be empty; If the returned result is not empty but the hierarchical relationship between categories is incorrect, the corresponding parent category is searched according to the secondary category, and only the classification results of the first two levels are recorded, and finally the results are temporarily stored in the variable; (3) The backend calls the work order title interface, problem level interface, and work order source interface of the large model respectively and records the results; (4) Obtain the pre-configured parameter rules and perform rule verification on the return value of the large model; (5) The backend assembles the recorded results into JSON and returns it to the frontend. At the same time, the backend records the large model interface request and the returned result log, and the frontend displays it; (6) The user determines whether the front-end echo is correct: If it is incorrect, the front page supports manual modification. After the modification is completed, a work order is created. At this time, the user's order creation parameter log is recorded for subsequent fine-tuning of the large model parameters.

Images