Database operation and maintenance auxiliary method, device, equipment and medium based on large model

Abstract

The present application discloses a database operation and maintenance auxiliary method, device, equipment and medium based on a large model, which belongs to the field of large language models. The method includes: determining the user's initial demand portrait information based on the user's input, the mapping relationship between the preset operation and maintenance scenario and the input path, and the user intention recognition rule, the initial demand portrait information includes the target operation and maintenance scenario and the user intention information; obtaining supplementary data matching the initial demand portrait information from the information supplement database associated with database operation and maintenance, and generating the target prompt in combination with the prompt template corresponding to the initial demand portrait information; sending the target prompt to the large model corresponding to the initial demand portrait information, and generating reasoning answer information based on the output information fed back by the large model corresponding to the initial demand portrait information and pushing it to the user. According to the embodiment of the present application, the efficiency and quality of database operation and maintenance can be improved.

Description

Technical Field

[0001] The present application relates to the field of large language models, and in particular to a database operation and maintenance auxiliary method, device, equipment and medium based on large models. Background Art

[0002] Database operations and maintenance ensure the stable, efficient, and secure operation of database systems. This requires manual effort and empirical analysis by operations personnel to address issues such as failures, data recovery, and performance bottlenecks. However, with the continuous advancement of database technology, the complexity of database systems continues to grow. Manual analysis and processing during database operations and maintenance is lengthy and prone to errors. This presents a significant challenge for operations personnel with limited understanding of database deployment and operational status, further reducing their efficiency and quality. Summary of the Invention

[0003] The embodiments of the present application provide a database operation and maintenance auxiliary method, device, equipment and medium based on a large model, which can improve the efficiency and quality of database operation and maintenance.

[0004] In the first aspect, an embodiment of the present application provides a database operation and maintenance assistance method based on a big model, including: determining the user's initial demand portrait information based on the user's input, the mapping relationship between the preset operation and maintenance scenarios and the input path, and the user intention recognition rules, the initial demand portrait information includes the target operation and maintenance scenario and user intention information; obtaining supplementary data matching the initial demand portrait information from an information supplement database associated with database operation and maintenance, and generating a target prompt in combination with the prompt template corresponding to the initial demand portrait information; sending the target prompt to the big model corresponding to the initial demand portrait information, and generating inference answer information based on the output information fed back by the big model corresponding to the initial demand portrait information and pushing it to the user.

[0005] In the second aspect, an embodiment of the present application provides a database operation and maintenance assistance device based on a big model, including: a user demand determination module, which is used to determine the user's initial demand portrait information based on the user's input, the mapping relationship between the preset operation and maintenance scenarios and the input path, and the user intention recognition rules, and the initial demand portrait information includes the target operation and maintenance scenario and user intention information; a prompt generation module, which is used to obtain supplementary data matching the initial demand portrait information from an information supplement database associated with database operation and maintenance, and generate a target prompt in combination with the prompt template corresponding to the initial demand portrait information; an information push module, which is used to send the target prompt to the big model corresponding to the initial demand portrait information, and generate reasoning answer information based on the output information fed back by the big model corresponding to the initial demand portrait information and push it to the user.

[0006] In a third aspect, an embodiment of the present application provides a database operation and maintenance assistance device based on a large model, comprising: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the database operation and maintenance assistance method based on a large model of the first aspect is implemented.

[0007] In the fourth aspect, an embodiment of the present application provides a database operation and maintenance assistance system based on a big model, including: a database cloud platform, used to execute the database operation and maintenance assistance method based on a big model of the first aspect; a big model subsystem, which is communicated with the database cloud platform, used to receive target prompts sent by the database cloud platform, and obtain output information based on the target prompts and feed it back to the database cloud platform.

[0008] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are executed by a processor, the large model-based database operation and maintenance assistance method of the first aspect is implemented.

[0009] In a sixth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the large model-based database operation and maintenance assistance method of the first aspect.

[0010] The embodiments of the present application provide a database operation and maintenance assistance method, apparatus, device and medium based on a large model, which can determine the initial demand profile information that can characterize the user's needs based on the user's input, the mapping relationship between the operation and maintenance scenario and the input path, and the user intention recognition rules, and obtain supplementary data that matches the user's needs from the information supplement database. The supplementary data can characterize the user's needs, and the supplementary data is filled into the prompt template corresponding to the initial demand profile to generate a target prompt. The target prompt can gather professional information on database operation and maintenance, and reflect more complete user needs with structured, comprehensive and detailed data. The target prompt is input into the large model, which can use the reasoning ability and natural language processing ability of the large model to feedback the services required by the user and assist the user in database operation and maintenance. Even if the user has a limited understanding of the database deployment and operating status, the database operation and maintenance can be smoothly and accurately implemented with assistance, thereby improving the efficiency and quality of database operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0012] Figure 1 A flowchart of a database operation and maintenance assistance method based on a large model provided in one embodiment of the present application;

[0013] Figure 2 A flowchart of an example of extracting database table names provided in an embodiment of the present application;

[0014] Figure 3 A schematic diagram of an example of adjusting the weight parameters of a large model provided in an embodiment of the present application;

[0015] Figure 4 A schematic diagram of an example of a database operation and maintenance assistance system based on a large model provided in an embodiment of the present application;

[0016] Figure 5 A flowchart of an example of an operation and maintenance scenario provided in an embodiment of the present application;

[0017] Figure 6 A flowchart of an example of obtaining user needs provided in an embodiment of the present application;

[0018] Figure 7 A flowchart of an example of an assembly target prompt word provided in an embodiment of the present application;

[0019] Figure 8 A flowchart of an example of question-and-answer feedback provided in an embodiment of the present application;

[0020] Figure 9 A schematic diagram of the structure of a database operation and maintenance auxiliary device based on a large model provided in one embodiment of the present application;

[0021] Figure 10 A schematic diagram of the structure of a database operation and maintenance auxiliary device based on a large model provided in one embodiment of the present application;

[0022] Figure 11 A schematic diagram of the structure of a database operation and maintenance assistance system based on a large model provided in one embodiment of the present application. DETAILED DESCRIPTION

[0023] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is merely to provide a better understanding of the present application by illustrating examples of the present application. It should be noted that the acquisition, storage, use, processing, etc. of information and data in the embodiments of the present application are authorized by the user or relevant agencies and comply with the relevant provisions of national laws and regulations.

[0024] Database operations and maintenance ensure the stable, efficient, and secure operation of database systems. This requires manual effort and experience-based analysis by operations personnel to address issues such as failures, data recovery, and performance bottlenecks. With the rise of cloud computing, database cloud platforms have emerged. These platforms offer database management services across multiple cloud architectures and database types. They support containerized deployment and enable full lifecycle management of database instances, from creation, monitoring, migration, backup, recovery, and destruction. This improves efficiency and resource utilization while reducing maintenance costs. However, with the continuous advancement of database technology, the complexity of database systems continues to increase. For example, database cloud platforms can manage dozens of database types and hundreds of thousands of instances. This wide variety of databases, the sheer number of instances, and the constant updating of database versions increase the complexity of database operations and maintenance. In this context, manual analysis and processing during database operations and maintenance is lengthy and prone to errors. This is particularly challenging for operations personnel with limited understanding of database deployment and operational status, further reducing efficiency and accuracy.

[0025] The present application provides a database operation and maintenance assistance method, device, equipment, system, medium and program product based on a big model, which can integrate the big model of operation and maintenance scenarios into the database cloud platform, and use the reasoning ability and natural language processing ability of the big model to solve the technical problems in database operation and maintenance, provide multi-scenario and multi-dimensional technical support for database operation and maintenance, assist database operation and maintenance, reduce the difficulty of database operation and maintenance, and enable operation and maintenance personnel with limited understanding of database deployment and operation status to complete database operation and maintenance efficiently and with high quality, thereby improving the efficiency and quality of database operation and maintenance.

[0026] It should be noted that the large model in the embodiment of the present application is the abbreviation of the large language model, which will not be described in detail in the following text. The following describes the database operation and maintenance auxiliary method, device, equipment, system and medium based on the large model provided by the present application.

[0027] The present application provides a database operation and maintenance assistance method based on a big model, which can be used in database operation and maintenance scenarios. The database in the database operation and maintenance scenario can be uniformly managed by a database cloud platform. The database operation and maintenance assistance method based on a big model can be specifically executed by the database cloud platform. The database cloud platform can be implemented as a database operation and maintenance assistance device, equipment, system, etc. based on a big model, which is not limited here. Figure 1 A flowchart of a database operation and maintenance assistance method based on a large model provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the database operation and maintenance auxiliary method based on the large model includes steps S101 to S103.

[0028] In step S101, the user's initial demand profile information is determined based on the user's input, the mapping relationship between the preset operation and maintenance scenarios and the input path, and the user intention recognition rules.

[0029] The operation and maintenance scenario represents the specific scenario type of database operation and maintenance, and can be pre-set based on one or more of the information such as the architecture of the database cloud platform, historical data of various database operations in different historical periods, and historical operation and maintenance knowledge information of various database operations. The historical operation and maintenance knowledge information of database operation and maintenance may include but is not limited to database operation and maintenance manuals, database product guides, database troubleshooting guides, expert experience summaries, etc. An operation and maintenance knowledge base can be constructed based on the historical operation and maintenance knowledge information of database operation and maintenance. For example, the historical operation and maintenance knowledge information of database operation and maintenance can be processed by data extraction, vectorization, etc. to obtain data with a unified structure, thereby constructing an operation and maintenance knowledge base. The operation and maintenance knowledge base can also provide information support for generating target prompts in subsequent steps in the embodiment of this application. Statistics and summarization of problems in the database operation and maintenance process can be performed based on the architecture of the database cloud platform, historical data of various database operations in different historical periods, and historical operation and maintenance knowledge information of various database operations, thereby obtaining a variety of operation and maintenance scenarios. The types and number of operation and maintenance scenarios are not limited here and can be set according to specific needs. For example, O&M scenarios may include, but are not limited to, explanation scenarios, optimization scenarios, and error correction scenarios. These three categories can be further categorized, such as knowledge question-and-answer scenarios, Structured Query Language (SQL) translation scenarios, parameter explanation scenarios, permission explanation scenarios, SQL optimization scenarios, configuration optimization scenarios, slow log optimization scenarios, configuration verification scenarios, and fault diagnosis and processing scenarios. Different O&M scenarios may correspond to different big models, with the corresponding big models specialized for that O&M scenario. The big models corresponding to these O&M scenarios can be trained on the original big model based on historical database data and historical O&M knowledge for that scenario, and then the weight parameters of the big model can be adjusted. For example, the weight parameters of the original big model can be adjusted using supervised learning or other methods based on the historical database data and historical O&M knowledge for that scenario, resulting in a big model that is specific to that scenario. The big model for this scenario can grasp the domain-specific terminology, operational procedures, and best practices for that scenario. Using the corresponding large model for processing in the operation and maintenance scenario can further improve the quality of database operation and maintenance.

[0030] The user's input can specifically realize the interaction between the user and the human-computer interaction interface. The user's input may include but is not limited to the user's page selection, function area selection, button selection, icon selection, input box text input, file upload and other input operations on the human-computer interaction interface. If the user's input includes file upload, the uploaded file format is not limited here. For example, the file can be in text format, picture format, video format, voice format, etc., which is not limited here. If the user's input includes multimodal data such as text format, picture format, video format, voice format, etc., the multimodal data can be identified, analyzed and processed separately. If an abnormality occurs in the identification, a prompt message can be fed back to the user to guide the user to re-enter. The human-computer interaction interface can pre-integrate the operation and maintenance scenarios with the architecture of the database cloud platform, and combine the existing management function distribution of the database cloud platform to realize the location of the functions corresponding to each operation and maintenance scenario on the human-computer interaction interface, so that users can easily enter the use state and reduce the difficulty of users in performing database operation and maintenance. For example, buttons, icons, input boxes and other controls corresponding to the knowledge question and answer scenario can be set on the homepage of the cloud platform database; buttons, icons, input boxes and other controls corresponding to the parameter explanation scenario can be set on the parameter management page of the cloud platform database; buttons, icons, input boxes and other controls corresponding to the configuration verification scenario can be set on the configuration management page of the cloud platform database; buttons, icons, input boxes and other controls corresponding to the fault diagnosis and processing scenario can be set on the alarm management page of the cloud platform database. Users can reflect their own needs through input on the human-computer interaction interface. The user's input can include a single input or a series of continuous inputs. The user's input can form an input path, which is the path formed by the interactive information corresponding to the user's input, and can explicitly or implicitly reflect the user's needs. The mapping relationship between the operation and maintenance scenario and the input path can be preset, so that when the input path is known, the operation and maintenance scenario indicated by the input path can be obtained.

[0031] The initial demand portrait information can characterize the user's demand for database operation and maintenance, and can be determined based on the user's input, preset mapping relationships, and user intent recognition rules. The initial demand portrait information may include target operation and maintenance scenarios and user intent information, but is not limited to this. In some examples, the initial demand portrait information may also include basic information of the database operated by the user and other information that can be expanded based on the basic information. The database operated by the user may specifically be an instance database. The basic information may include but is not limited to information that can characterize the instance database, such as database type, database architecture, and database deployment container information. Classification, labeling, and other processing can be performed based on the basic information to obtain feature labels of the instance database, so that the database cloud platform can more accurately determine the user's intent. Other information that can be expanded based on the basic information may include but is not limited to database query patterns, database operation and maintenance goals, database workload, and other information that can serve as feature labels for the database.

[0032] The target operation and maintenance scenario is the operation and maintenance scenario of the database operation and maintenance required by the user this time, which can be determined based on the user's input and the mapping relationship. Through the mapping relationship, the user's current operation and maintenance scenario can be quickly located, so that information and assistance that is more suitable for the user's current operation and maintenance scenario can be provided in the subsequent database operation and maintenance assistance process. User intent information represents the user's intention for database operation and maintenance currently required, that is, it represents what kind of database operation and maintenance the user wants to perform and what kind of assistance he hopes to get. User intent information can be determined based on user input and user intent identification rules. User intent identification rules are rules for identifying user intent, and user intent information can be determined by matching user input with intention identification rules. User intent identification rules may include but are not limited to one or more of keywords, phrases, regular expressions, logical combinations, etc. Logical combinations may include multiple rules combined using logical operators such as AND and OR. The user's intention can be accurately inferred through user intent identification rules, thereby effectively guiding the process in the database operation and maintenance assistance process, improving interaction efficiency, database operation and maintenance quality, and user experience.

[0033] In step S102, supplementary data matching the initial demand portrait information is obtained from an information supplement database associated with database operation and maintenance, and a target prompt is generated in combination with a prompt template corresponding to the initial demand portrait information.

[0034] The information supplement database can provide supplementary information for database operation and maintenance assistance based on the large model, so as to alleviate the "hallucination" problem that may exist in the large model when the large model is used in the future. The "hallucination" problem of the large model refers to the errors, inaccuracies or content that is inconsistent with reality that may appear when the large model generates answers. In some examples, the information supplement database may include an operation and maintenance knowledge base and / or an instance database. The specific content of the operation and maintenance knowledge base can refer to the relevant descriptions in the above embodiments and will not be repeated here. The instance database is a database that stores instance data, and the objects of database operation and maintenance in the embodiments of the present application include the instance database. For example, the instance database corresponding to the transaction business may store instance data such as order data, payer data, and payee data; the instance database corresponding to the user management business may store instance data such as user identification and user identity information. Data related to the initial demand portrait information can be queried in the information supplement database, and data related to the initial demand portrait information can be used as supplementary data.

[0035] Different specific prompt templates can be set according to different user needs. The prompt templates may include but are not limited to reflection iteration templates, chain thinking templates, etc. If the demand represented by the initial demand portrait has a corresponding specific prompt template, then the specific prompt template can be determined as the prompt template corresponding to the initial demand portrait information. A general prompt template can also be preset. If the demand represented by the initial demand portrait information does not have a corresponding specific prompt template, then the general prompt template is determined as the prompt template corresponding to the initial demand portrait information. Supplementary data can be filled into the prompt template and assembled into a target prompt. The target prompt can be regarded as more comprehensive and detailed demand portrait information. The supplementary data that matches the initial demand portrait information in the information supplement database can characterize user needs. Generating target prompts through supplementary data can further constrain the thinking boundary of the large model, reduce the reasoning delay of the large model and improve the reasoning accuracy of the large model.

[0036] In step S103, a target prompt is sent to the large model corresponding to the initial demand portrait information, and based on the output information fed back by the large model corresponding to the initial demand portrait information, inference answer information is generated and pushed to the user.

[0037] Multiple large models can be pre-trained. When a large model is needed to assist in database operation and maintenance, the one that best meets the user's needs is selected from the multiple large models for use. The initial demand profile information may include the target operation and maintenance scenario, basic information about the instance database, and other information expanded based on this basic information. The target operation and maintenance scenario, basic information about the instance database, and other information expanded based on this basic information can be combined to jointly determine the large model corresponding to the initial demand profile information. For example, based on the target operation and maintenance scenario, database type, database query mode, database operation and maintenance goals, database workload, etc., a large model that meets the user's needs can be selected from multiple pre-trained large models in a data-driven manner to process the target prompt. If no large model that meets the user's needs is found among the multiple pre-trained large models, a general large model is selected to process the target prompt. In some examples, the large model can receive the target prompt through an application programming interface (API). Correspondingly, after obtaining the target prompt, a request body and token that conform to the API format of the big model can be constructed. After encapsulating the request body and token, a Hypertext Transfer Protocol (HTTP) call is initiated to transmit the encapsulated information to the big model so that the big model can process the target prompt.

[0038] In some examples, the output information fed back by the large model can be directly pushed to the user as inference answer information. In other examples, the user's needs represented by the initial demand profile information may include other needs besides large model processing, such as data query and performance monitoring. These other needs can create other tasks for execution, and the execution result information can be obtained. The output information fed back by the large model and the execution result information can be integrated to obtain inference answer information and pushed to the user. Inference answer information includes information provided to the user to meet the user's needs. In some examples, the Server-Sent Events (EES) mechanism can be used, and the front end of the database cloud platform can actively and asynchronously push inference answer information to the user to display the inference answer information in real time.

[0039] In an embodiment of the present application, the initial demand profile information that can characterize the user's needs can be determined based on the user's input, the mapping relationship between the operation and maintenance scenario and the input path, and the user intention recognition rules. Supplementary data that matches the user's needs is obtained from the information supplement database. The supplementary data can characterize the user's needs. The supplementary data is filled into the prompt template corresponding to the initial demand profile to generate a target prompt. The target prompt can gather professional information on database operation and maintenance, and reflect more complete user needs with structured, comprehensive and detailed data. The target prompt is input into the large model, which can use the reasoning ability and natural language processing ability of the large model to feedback the services required by the user and assist the user in database operation and maintenance. Even if the user has a limited understanding of the database deployment and operating status, the database operation and maintenance can be smoothly and accurately implemented with assistance, thereby improving the efficiency and quality of database operation and maintenance.

[0040] In some embodiments, the above step S101 can be specifically refined as follows: obtaining the operation and maintenance scenario corresponding to the input path formed by the user's input in the mapping relationship as the target operation and maintenance scenario; obtaining user intention information based on the matching result of the user's input and the user intention recognition rule; determining the target instance database based on the user's input, and obtaining basic information of the target instance database; obtaining initial demand portrait information based on the target operation and maintenance scenario, user intention information and basic information.

[0041] In the mapping relationship, the operation and maintenance scenario corresponding to the input path formed by the user's input is the target operation and maintenance scenario. For example, the user clicks the parameter management label on the homepage of the cloud platform database, and the cloud platform database jumps to the parameter management page. The user clicks the button representing parameter A1 and the button representing the use of the large model on the parameter management page. The input path formed by the user's input can be parameter management label → parameter A1 button + large model button. In the mapping relationship, the operation and maintenance scenario corresponding to the input path of parameter management label → parameter A1 button + large model button is the parameter interpretation scenario, and the current target operation and maintenance scenario can be determined to be the parameter interpretation scenario. In the scenario where the user's input is text input or file upload, if the user's input is text input, the input text can be directly obtained. If the user's input is file upload, the uploaded file can be processed by data extraction, conversion, etc. to obtain text, and the text can be analyzed to obtain the input path. For example, if a user enters "I want to query the database connection problem, "MySQL database cannot be connected remotely, error code is XXXX" in the input box, the input path formed by the user's input can be database connection → cannot connect remotely → error code XXXX. In the mapping relationship, the operation and maintenance scenario corresponding to the input path of database connection → cannot connect remotely → error code XXXX is the fault diagnosis and processing scenario.

[0042] The number of user intent recognition rules can be multiple, and one user intent recognition rule can correspond to one user intent, that is, one user intent recognition rule can correspond to one user intent information. According to the user input, the information to be matched with the user intent recognition rule can be obtained, and the information to be matched can be matched with multiple user intent recognition rules one by one. If the match is successful, the user intent information corresponding to the successfully matched user intent recognition rule is determined as the user intent information in the initial demand profile information. In some examples, word segmenters such as jieba and hanlp can be used to extract vocabulary and phrase matching information from the user input, and the vocabulary and phrase matching information can be matched with the user intent recognition rule. If the user intent recognition rule includes keywords, the user input can be matched with the user intent recognition rule by detecting whether the information to be matched contains keywords. If the user intent recognition rule includes regular expressions, the user input can be matched with the user intent recognition rule by matching the information to be matched with the regular expressions. If the user intent recognition rule includes logical combinations, the user input can be matched with the user intent recognition rule by matching the information to be matched with the logical combinations. For example, the matching of the user input with the user intent recognition rule can be expressed using the following formula (1):

[0043] (1)

[0044] in, The user intention is determined based on the user input and the user intention recognition rules, which can be represented by user intention information; 、 、…、 n preset user intentions; 、 、…、 n user intention recognition rules corresponding to the n preset user intentions; For user input; Indicates matching; Indicates unknown user intent; Indicates except 、 、…、 Other than the situation.

[0045] In the case where the information to be matched successfully matches multiple user intent identification rules, the matching result of the information to be matched and the multiple user intent identification rules can not only indicate whether the match is successful, but also indicate the degree of match. The user intent information corresponding to the user intent identification rule with the highest degree of match indicated by the matching result can be determined as the user intent information in the initial demand profile information. The matching result that can indicate the degree of match can be obtained based on the priority of the information to be matched and each user intent identification rule or the weight coefficient of the matching of the information to be matched and each user intent identification rule, etc., which is not limited here.

[0046] The user input may indicate one or more instance databases, and the target instance database includes the instance database indicated by the user. The specific content of the basic information of the instance database can be found in the relevant description of the above embodiment and will not be repeated here. The initial demand profile information may also include the basic information of the target instance database and other information expanded based on this basic information. The specific content of this other information can be found in the relevant description of the above embodiment and will not be repeated here.

[0047] The database cloud platform can include a front-end and a back-end. The front-end can determine the user's initial demand profile information. After determining the initial demand profile information, the front-end can transmit the initial demand profile information to the back-end. The back-end can detect anomalies in the transmission of the initial demand profile information and automatically trigger retransmission if an anomaly is found.

[0048] In some embodiments, after obtaining the initial demand portrait information, a task set can be generated based on the initial demand portrait information, and the tasks in the task set can be assigned to the corresponding task queue for processing. The task set includes at least one task. The classification of tasks can be set in advance, and different task queues can be set for different types of tasks. Tasks belonging to the same type can be assigned to corresponding task processing chains, so that tasks are assigned to corresponding task queues and asynchronously called in sequence to achieve multi-scenario, high-concurrency database operation and maintenance task processing. The task set may include large model tasks and other tasks such as database access and performance monitoring. The task queue to which the large model task belongs can be monitored. When the task set includes the large model task, supplementary data can be obtained from the information supplement database, and the target prompt can be generated in combination with the prompt template corresponding to the initial demand portrait information to send the target prompt to the large model.

[0049] In some examples, the database cloud platform can also regularly query the initiation status of each task. If a task is not confirmed or confirmed abnormally, the task processing chain needs to resend the task and record the exception and resend records in the log to facilitate subsequent tracking and investigation of the event.

[0050] In some embodiments, the information supplement database includes an operation and maintenance knowledge base and / or an instance database. In the application embodiment, the operation and maintenance database is basically an instance database. The specific content of the operation and maintenance knowledge base and / or the instance database can be found in the relevant description above and will not be repeated here. The above step S102 can be specifically refined as follows: searching the operation and maintenance knowledge base for knowledge information that matches the initial demand portrait information as supplementary data, and / or calling the interface to obtain instance data that matches the initial demand portrait information from the instance database corresponding to the initial demand portrait information as supplementary data; filling the supplementary data into the corresponding position in the prompt template to generate the target prompt.

[0051] The initial demand profile information may include the target operation and maintenance scenario, user intent information, basic information about the target instance database, and other information expanded from this basic information. Keywords can be extracted from the initial demand profile information to retrieve operation and maintenance knowledge information related to the user's needs from the operation and maintenance knowledge base. In some examples, a first search can be conducted in the operation and maintenance knowledge base using a first keyword extracted from the initial demand profile information. First search result data that meets the quality standard can be identified as supplementary data. A second search can be conducted in the operation and maintenance database using the second keyword based on the first keyword, and second search result data that meets the quality standard can be identified as supplementary data. The first keyword is a keyword extracted from the initial demand profile, and the first search result data includes operation and maintenance knowledge information retrieved from the operation and maintenance knowledge base using the first keyword. After obtaining the first search result data, the first search result data can be quality-checked to verify its accuracy, completeness, and relevance to the initial demand profile information. First search result data that meets the quality standard can be screened from the first search result data. If any first search result data that meets the quality standard exists, the first search result data that meets the quality standard can be identified as supplementary data. The second keyword can be obtained by performing synonym expansion, context expansion, or other methods on the first keyword. The second search result data includes operation and maintenance knowledge information retrieved from the operation and maintenance knowledge base using the second keyword. The second search result data is quality-checked, and second search result data that meets quality standards is selected from the second search result data as supplementary data.

[0052] In some examples, instance data corresponding to the target O&M scenario can be obtained from the target instance database corresponding to the initial demand profile information as supplementary data. For example, if the target O&M scenario includes slow log optimization, database table structure information can be obtained from the target instance database; if the target O&M scenario includes parameter optimization, historical alarm information and current resource usage can be obtained from the target instance database.

[0053] In some examples, the instance data in the supplementary data includes table structure information, and the initial demand profile information also includes an SQL statement selected by the user's input. The SQL statement in the initial demand profile information can be parsed to generate a syntax tree (Abstract Syntax Tree, AST) for the SQL statement. Specifically, the JSqlParser library in Java can be used to parse the SQL statement into a syntax tree. The nodes in the syntax tree are traversed to obtain a target node containing a database table name, and the database table name in the target node is obtained. Based on the obtained database table name, the tables in the instance database are traversed to obtain the table structure information as supplementary data.

[0054] For example, Figure 2 This is a flowchart of an example of extracting database table names provided in an embodiment of the present application, such as Figure 2 As shown, after obtaining the syntax tree, the process of extracting the database table name may include steps a1 to a8.

[0055] In step a1, the SQL statement is parsed into a Statement object. Specifically, a parsing tool such as CCJSqlParserUtil can be used to parse the SQL statement into a Statement object.

[0056] In step a2, a Select object is obtained by filtering from the Statement object.

[0057] In step a3, if the Select object is a simple join-table query, a simple join-table query is performed. This simple join-table query is also known as non-nested subquery table name resolution. Specifically, the From node, Expression node, and Select node in the syntax tree are parsed sequentially. The Expression node can contain a Where node, a Having node, and a Function node. The Select node can contain a function field, a character field, and a Case field.

[0058] In step a4, if the Select object belongs to a nested subquery, each subquery is traversed and the simple join table query of step a3 is performed on each subquery.

[0059] In step a5, determine whether the From node, Where node, Having node, and Function node contain nested subqueries. If so, proceed to step a3; if not, proceed to step a6.

[0060] In step a6, the table name in the node is obtained.

[0061] In step a7, obtain the table name in the Select node.

[0062] In step a8, a table name set is obtained based on the table name integration. The table names obtained in step a6 and step a7 can be added to the table name set and duplicates can be removed. Subsequently, the tables in the table name set can be traversed through the API to obtain table structure information.

[0063] A syntax tree is a structure that accurately parses SQL statements. By breaking down SQL statements into individual nodes within the syntax tree, the syntactic location of table names can be clearly identified. The syntax tree can parse complex SQL statements, such as nested subqueries, multi-table joins, and union queries, layer by layer, ensuring accurate and comprehensive table name extraction and preventing omissions. Syntax tree-based database table name extraction is also more efficient.

[0064] In some examples, the SQL statement can be converted into multiple subtasks that can be processed in parallel, thereby improving the efficiency of obtaining table structure information. Specifically, the SQL statement can be split into multiple subtasks according to the query logic; multiple subtasks are processed in parallel, and the table structure information corresponding to each subtask is obtained in the instance database as supplementary data. The query logic may include but is not limited to multi-table logic, subquery logic, filter condition logic, etc. For example, if the SQL statement involves multi-table connection, such as the SQL statement includes a JOIN statement, the query of each table can be treated as an independent subtask; the independent subqueries in the SQL statement can be broken down into separate statements, and the separate statements can be used as subtasks; if the part of the SQL statement involving conditions, such as the WHERE condition, includes multiple logical branches in the form of OR or complex combinations, it can be split according to the conditions, and the split parts can be used as subtasks. Parallel processing of these obtained subtasks can quickly obtain the corresponding table structure information from the instance database, thereby improving the efficiency of obtaining table structure information as supplementary data.

[0065] In some embodiments, the large model processing target prompt needs to obtain relevant information from the instance database. This information may include sensitive information such as user identity information and transaction information. To ensure the security of sensitive information, the sensitive information may be desensitized, and different desensitization processes may be applied based on the sensitivity of the sensitive information. Specifically, if the information required by the large model processing target prompt corresponding to the initial demand profile information to obtain from the instance database includes level-one sensitive information, the level-one sensitive information may be replaced with secure desensitized information to obtain desensitized data. If the information required by the large model processing target prompt corresponding to the initial demand profile information to obtain from the instance database includes level-two sensitive information, the level-two sensitive information may be processed using substitution masking and / or masking with placeholders to obtain desensitized data. The desensitized data is then transmitted to the large model corresponding to the initial demand profile information as feedback output information. Level-one sensitive information has a higher sensitivity than level-two sensitive information. For example, level-one sensitive information may include, but is not limited to, user identity information, bank card numbers, order numbers, and other information, while level-two sensitive information may include, but is not limited to, email addresses, merchant numbers, and addresses. The higher the sensitivity of sensitive information, the stricter the desensitization measures.

[0066] For Level 1 sensitive information, the entire Level 1 sensitive information can be replaced, for example, with security-sensitive information such as a random value or hash value. The desensitized data includes security-sensitive information. Security-sensitive information will not leak Level 1 sensitive information. In other words, if a user obtains security-sensitive information, they cannot restore the Level 1 sensitive information.

[0067] For secondary sensitive information, replacement desensitization and / or mask desensitization are used. The desensitized data includes the information processed by replacement desensitization and / or mask desensitization. Mask desensitization can be expressed by the following formula (2):

[0068] (2)

[0069] in, To mask the desensitized information; For original information; The mask rule can be used to define the part to be retained after desensitization and the part to be replaced after desensitization. It is a masking function, which is used to desensitize the original information according to the masking rules. The masking rules may include rules for masking and desensitizing sensitive information such as phone numbers, email addresses, user identification numbers, bank card numbers, merchant numbers, addresses, etc. For example, the masking rule corresponding to the phone number may be to retain the first 3 digits and the last 4 digits of the phone number, with * in the middle; the masking rule corresponding to the email address may be to retain the first character and the domain name, with * in the middle; the masking rule corresponding to the user identification number may be to retain the first 6 digits and the last 4 digits, with * in the middle; the masking rule corresponding to the bank card number may be to retain the first 6 digits and the last 4 digits, with * in the middle; the masking rule corresponding to the merchant number may be to retain the first 6 digits and the last 4 digits, with * in the middle; the masking rule corresponding to the address may be to retain the first 6 digits and the last 4 digits, with * in the middle.

[0070] The database cloud platform can parse information that needs to be transmitted to the large model in real time. This information can be transmitted to the large model by inputting streaming JSON data. While desensitizing sensitive information, the database cloud platform also performs format conversion and monitors the parsing and conversion processes for anomalies. If anomalies occur, local exception handling is performed to avoid impacting database operations and maintenance.

[0071] By desensitizing sensitive information, the sensitive parts of the sensitive information can be effectively shielded while avoiding affecting the real-time interaction between users and large models.

[0072] In some embodiments, if the user is dissatisfied with the inference answer information, the user can be guided to re-enter the information, and the content of the inference answer information can be used in the re-generation process of the target prompt to improve the degree of compliance of the inference answer information with the user's needs. Specifically, if the received user feedback information indicates that the inference answer information does not meet the user's needs, the user can be again interacted with to determine new initial demand profile information; new supplementary data matching the new initial demand profile information can be obtained from the information supplementation database; at least part of the information in the inference answer information and / or at least part of the information in the target prompt corresponding to the inference answer information can be used as supplementary data; based on the supplementary data, combined with the prompt template corresponding to the new initial demand profile information, a new target prompt can be generated for processing by the large model.

[0073] The input interaction with the user again to determine the new initial demand profile information and the content of obtaining the supplementary data from the information supplement database are basically the same as the content of determining the initial demand profile and obtaining the supplementary data from the information supplement database above, and will not be repeated again. The input interaction with the user again may include the user's evaluation mark of each information in the reasoning answer information and the evaluation mark of the content in the target prompt. The evaluation mark may include a mark that meets the requirements and a mark that does not meet the requirements. The reasoning answer information with the mark that meets the requirements and / or the reasoning answer information with the mark that meets the requirements can be used as positive supplementary data, and the reasoning answer information with the mark that meets the requirements and / or the reasoning answer information with the mark that does not meet the requirements can be used as negative supplementary data. The positive supplementary data and the negative supplementary data are both supplementary data. The positive supplementary data instructs the large model to provide answers according to the positive supplementary data, and the negative supplementary data instructs the large model to avoid the negative supplementary data, enriching the contextual information from both positive and negative directions, thereby further improving the accuracy of the large model's answers.

[0074] In some embodiments, the weight parameters of the large model include incremental weight parameters and original weight parameters of the original large model. The incremental weight parameters are obtained by training the original large model based on the operation and maintenance knowledge information in the operation and maintenance knowledge base. The weight parameters, original weight parameters, and incremental weight parameters here can all be implemented as parameter matrices. The weight parameters of the large model can be specifically implemented as the core weight parameter matrix that has the greatest impact on the model behavior in the large model. The large model can be adjusted according to user feedback to make the large model more in line with user needs and the output more accurate. Specifically, after receiving user feedback information, the original weight parameters can be frozen and the incremental weight parameters can be decomposed into two low-rank matrices; the received user feedback information is used as a reward signal to adjust the low-rank matrix; the updated weight parameters of the large model are obtained based on the original weight parameters and the new incremental weight parameters obtained based on the adjusted low-rank matrix.

[0075] Freezing the original weight parameters means not adjusting the original weight parameters. In the embodiment of the present application, the incremental weight parameters are adjusted, and the low-rank adaptation (LoRA) technology is used to decompose the incremental weight parameters into two low-rank matrices. The rank of the low-rank matrix is ​​smaller than the rank of the incremental weight parameters. By adjusting the low-rank matrix, the weight parameters of the large model are updated, which accelerates the convergence of the large model and further reduces the computing and storage costs. For example, Figure 3 A schematic diagram of an example of adjusting the weight parameters of a large model provided in an embodiment of the present application, such as Figure 3 As shown, the adjustment method of the weight parameters of the large model can refer to the forward propagation of LoRA technology. The forward propagation of LoRA technology can be shown as follows (3):

[0076] (3)

[0077] in, The features before the weight parameter matrix of the large model can be realized as d-dimensional vector features; is the weight parameter of the large model; is the original weight parameter; is the incremental weight parameter; and are the two low-rank matrices decomposed, ,Right now It can be realized as a d×r low-rank matrix, the initial Can be an all-zero matrix, ,Right now It can be realized as a low-rank matrix of r×d, the initial It can be a Gaussian random distribution matrix, d is much smaller than r; It is the feature after passing through the front and middle parameter matrix of the large model.

[0078] The ranks of the two low-rank matrices obtained by decomposing the incremental weight parameters can determine the number and expressiveness of the incremental weight parameters. The lower the rank of the low-rank matrix means that fewer trainable parameters are required, but it may limit the adaptability of the trained large model to complex tasks; the higher the rank of the low-rank matrix, the more trainable parameters are required, which can make the trained large model better fit the specific task, but it will increase resource consumption, such as increasing computing cost and storage requirements. In the embodiment of the present application, a suitable rank can be selected in the process of training the large model to balance the model capability and the number of trainable parameters. Specifically, a rank range can be set based on the resource consumption information of the original large model for training, and the rank in the rank range is negatively correlated with the resource consumption represented by the resource consumption information; multiple ranks are selected within the rank range as candidate ranks of the low-rank matrix, and the weight parameters of the large model are updated based on the low-rank matrices of different candidate ranks; the performance parameters of the large model after the weight parameters corresponding to the low-rank matrices of different candidate ranks are obtained by using the verification sample set, and the candidate rank corresponding to the large model with the highest performance represented by the performance parameters is determined as the rank of the low-rank matrix.

[0079] Resource consumption information may include, but is not limited to, information such as the task complexity of model training, the data scale required for model training, and the hardware resources for model training. Based on the resource consumption information, a rank range with resource consumption within an acceptable range can be selected. For example, the rank range can be [4, 16], that is, the minimum value of the rank range is 4 and the maximum value is 16. Multiple candidate ranks are selected within the rank range. For each candidate rank, the incremental weight parameter is decomposed into a low-rank matrix with the rank of the candidate rank, and the large model is trained. The trained large model corresponding to each candidate rank is verified using a validation sample set to obtain the performance parameters of the large model. The performance parameters characterize the performance of the large model. The performance parameters may include, but are not limited to, loss value, accuracy, and other parameters. The candidate rank corresponding to the large model with the highest performance is determined as the rank of the low-rank matrix, and the large model with the highest performance is put into use to generate reasoning answer information based on the output information fed back by the large model. In this way, a low-rank matrix with a more appropriate rank can be automatically selected during the training of the large model, thereby improving the reasoning accuracy of the large model.

[0080] By introducing small, low-rank weight matrices into the decisive layer of the large model, adjusting the large model through user feedback information, and using a small number of incremental weight coefficients to adapt to database operation and maintenance scenarios, the convergence speed of the large model is accelerated and the computing and storage costs of large model training are reduced.

[0081] In the embodiment of the present application, the database operation and maintenance assistance method based on a large model can be executed by a database operation and maintenance assistance system based on a large model. The database operation and maintenance assistance method based on a large model will be described below in conjunction with the database operation and maintenance assistance system based on a large model. Figure 4 A schematic diagram of an example of a database operation and maintenance auxiliary system based on a large model provided in an embodiment of the present application is shown as follows: Figure 4 As shown, the database operation and maintenance assistance system based on the large model may include a scene setting unit 21, a user demand acquisition unit 22, a prompt word assembly unit 23 and a question and answer feedback unit 24.

[0082] The scenario setting unit 21 is used to construct and manage operation and maintenance scenarios, set up a human-computer interaction interface, provide an effective operation and maintenance scenario framework, and lay the foundation and basis for subsequent database operation and maintenance assistance. Figure 5 A flowchart of an example of an operation and maintenance scenario provided in an embodiment of the present application is shown as follows: Figure 5 As shown, the operation and maintenance scenario construction process may include steps b1 to b7.

[0083] In step b1, an operation and maintenance knowledge base is constructed. The construction of the operation and maintenance knowledge base is described above and will not be repeated here. For ease of storage, historical operation and maintenance knowledge information can be vectorized and stored in the operation and maintenance knowledge base in vector form. The operation and maintenance knowledge information in the operation and maintenance knowledge base needs to be updated promptly.

[0084] In step b2, the large model is trained and adjusted. Based on the open-source original large model, the weight coefficients of the original large model can be adjusted using methods such as supervised learning, allowing the trained and adjusted large model to capture the operational knowledge. Based on operational scenarios and user needs, different large models can be trained and adjusted using operational knowledge corresponding to different scenarios and user needs. For example, large models can be trained and adjusted for scenarios and needs such as programming assistance, event analysis, resource orchestration, and long text generation. Multiple large models can be integrated into the database cloud platform for easy access during database operational support.

[0085] In step b3, the operation and maintenance scenarios are divided. Data of different types of databases and operation and maintenance data of different historical periods can be obtained, and statistical summary can be used to obtain multiple operation and maintenance scenarios. The classification of operation and maintenance scenarios can be found in Figure 5 The relevant descriptions in the above embodiments will not be repeated here.

[0086] In step b4, a prompt template is set. Prompt templates can be customized for different operation and maintenance scenarios. Each prompt template is customized with personalized vocabulary, grammatical structure, and optimized prompt strategies. This allows the prompt template to optimize the output of the large model, making it more accurate and targeted, thereby improving the inference efficiency and quality of the large model.

[0087] In step b5, functional structure integration is performed. The implementation of the O&M scenario depends on the database cloud platform. By combining the database cloud platform architecture with the various O&M scenarios obtained in step b3, the specific structure of the database cloud platform can be selected to deploy the O&M scenario functions, achieving seamless integration and effective integration between the O&M scenario and the database cloud platform.

[0088] In step b6, a human-computer interaction interface is set up. The front-end human-computer interaction interface of the database cloud platform is set up for operation and maintenance scenarios. This ensures that a concise, friendly, and efficient human-computer interaction interface is provided while ensuring accurate transmission of necessary information. Elements such as icons, buttons, and input boxes can be used to simplify user operations, reduce user operation difficulty, and improve user experience.

[0089] In step b7, manage operation permissions. You can use a role-based access control (RBAC) strategy to assign permissions for different operation and maintenance scenarios to different user roles. This ensures that users can use functions in the operation and maintenance scenarios that match their permissions when performing database operation and maintenance, improving security and operational accuracy.

[0090] The specific contents of the above steps b1 to b7 can be found in the relevant descriptions in the above embodiments and will not be repeated here.

[0091] The user demand acquisition unit 22 can understand the user's operation and maintenance needs through the objective trigger points pre-set by the scenario setting unit 21 and the user's principal behavior represented by the received user input, and assist the database cloud platform to efficiently transmit user needs to the big model to obtain more accurate answers. Figure 6 This is a flowchart of an example of obtaining user needs provided in an embodiment of the present application, such as Figure 6 As shown, the process of obtaining user needs may include steps c1 to c13.

[0092] In step c1, user demand acquisition is triggered. The user logs in to the database cloud platform and performs operations according to the requirements, triggering the process of acquiring user requirements.

[0093] In step c2, authentication is performed to determine whether the authentication is successful. The user can be determined to have the operation authority according to the authority control policy set by the scene setting unit 21, thereby determining whether the authentication is successful. If the authentication fails, step c3 is executed; if the authentication is successful, step c4 is executed.

[0094] In step c3, a prompt message is sent to prompt the user to apply for permission upgrade or retry.

[0095] In step c4, interactive input is performed. The interactive input can be guided by the user through the human-computer interaction interface preset by the scene setting unit 21. The human-computer interaction interface can support the input of multimodal data, such as text, image, and voice, and the operation and maintenance scene is configured with customized guidance.

[0096] In step c5, it is determined whether the format of the data input by the user complies with the specification. This determination can be achieved by verifying the format of the data input by the user. If it complies with the specification, step c6 is executed; if it does not, the process returns to step c4.

[0097] In step c6, a preliminary demand profile is constructed. The database cloud platform's front-end automatically adds basic information about the target instance database and categorizes and labels the basic information to assign data tags to the target instance database. The database cloud platform also uses these tags to assist in identifying user needs and subsequent processing.

[0098] In step c7, the preliminary demand profile information is transmitted. The front-end of the database cloud platform can transmit the preliminary demand profile information to the back-end of the database cloud platform via the HTTP protocol.

[0099] In step c8, it is detected whether the transmission is successful. If the transmission is successful, step c9 is executed; if the transmission is unsuccessful, the process returns to step c7, i.e., an automatic retransmission is initiated.

[0100] In step c9, multimodal data is identified and converted. The database cloud platform's backend can analyze the transmitted multimodal data to identify auxiliary scenarios and user needs.

[0101] In step c10, it is determined whether the recognition is normal. If so, step c11 is executed; if not, the process returns to step c4, where the user is guided to re-initiate interactive input, and the reason for the abnormal recognition is fed back to the user.

[0102] In step c11, tasks are distributed. A task set is generated based on the initial demand profile information. Based on the target operation and maintenance scenario, feature tags, database type, user intent information, etc. in the initial demand profile information, tasks in the task set are assigned to the corresponding task processing chain and stored in the corresponding task queue, awaiting asynchronous invocation, thus achieving multi-scenario, high-concurrency operation and maintenance auxiliary access.

[0103] In step c12, the task status is polled. The database cloud platform regularly queries the task initiation status and ensures smooth task execution through a consumption confirmation retry mechanism.

[0104] In step c13, it is determined whether the task is successfully initiated. If so, the process ends; if not, the process returns to step c11.

[0105] The specific contents of the above steps c1 to c13 can be found in the relevant descriptions in the above embodiments, which will not be repeated here.

[0106] The prompt word assembly unit 23 can execute the distributed large model tasks through the multi-threaded consumption task queue, call the API interface to obtain professional context information in the information supplement database, and assemble prompt words in combination with the prompt word template that matches the user needs. The assembled target prompt words can form a structured, comprehensive and detailed user demand portrait. Figure 7 This is a flowchart of an example of assembling target prompt words provided in an embodiment of the present application, such as Figure 7 As shown, the process of assembling the target prompt word may include steps d1 to d16.

[0107] In step d1, the task queue is monitored.

[0108] In step d2, determine whether there are any pending tasks. A pending task is a task that is waiting to be processed. If there are pending tasks, proceed to step d3; if not, proceed to step d1.

[0109] In step d3, pending tasks are executed in batches. Tasks waiting to be consumed are queued and processed in batches according to first-in, first-out order. The thread pool size can also be dynamically adjusted, controlling the number of tasks processed in a batch by controlling the threads. Once a task is processed, a consumption confirmation flag is immediately provided, indicating that processing has begun.

[0110] In step d4, the supplementary data in the operation and maintenance knowledge base is injected. The supplementary data can be retrieved from the operation and maintenance knowledge base through the retrieval enhancement generation technology.

[0111] In step d5, it is determined whether the search enhancement generation is abnormal. If so, step d6 is executed; if not, step d7 is executed.

[0112] In step d6, exception handling is performed, and relevant information may be recorded in a log after the exception handling is performed.

[0113] In step d7, the quality of the supplementary information is determined to be up to standard. This can be determined by determining whether the searched data meets the quality standard. If so, proceed to step d8. If not, return to step d4 and perform a second search.

[0114] In step d8, the instance database is queried for supplementary data.

[0115] In step d9, it is determined whether the instance database query is abnormal. If an abnormality is found, step d6 is executed; if not, step d10 is executed.

[0116] In step d10, it is determined whether the instance database query is comprehensive. If so, step d11 is executed; if not, the process returns to step d8 and performs a secondary query.

[0117] In step d11, a prompt template is searched and a corresponding prompt template can be selected based on the template adaptation rule engine according to user needs.

[0118] In step d12, it is determined whether the prompt template query is abnormal. If so, step d6 is executed; if not, step d13 is executed.

[0119] In step d13, it is determined whether a corresponding prompt template is found. If so, step d14 is executed; if not, the process returns to step d11 and a universal prompt template is used.

[0120] In step d14, the target prompt is assembled based on the supplementary data and the prompt template.

[0121] In step d15, it is determined whether the prompt assembly is abnormal. If so, step d6 is executed; if not, step d16 is executed.

[0122] In step d16, the operation and maintenance knowledge base is dynamically updated. For newly emerged database operation and maintenance issues, the operation and maintenance knowledge information corresponding to the database operation and maintenance issues can be pushed to the operation and maintenance knowledge base during the prompt assembly process to update the operation and maintenance knowledge base so that the supplementary data used to assemble the prompt is the latest data.

[0123] The specific contents of the above steps d1 to d16 can be found in the relevant descriptions in the above embodiments, which will not be repeated here.

[0124] The question-answer feedback unit 24 can obtain the output of the large model and present the inference answer information to the user in real time. It can also guide secondary input interaction and large model updates based on user feedback. Through frequent communication between users and the database cloud platform, it can also promote the accumulation of knowledge and skill improvement of the database cloud platform. Figure 8 This is a flowchart of an example of question-and-answer feedback provided in an embodiment of the present application, such as Figure 8 As shown, the question-answer feedback process may include steps e1 to e14.

[0125] In step e1, the corresponding big model is called. If multiple big models exist, the big model corresponding to the initial demand profile information can be called. The corresponding big model can be selected from the multiple big models in a data-driven manner based on characteristics such as database type, database query mode, database operation and maintenance goals, and workload.

[0126] In step e2, it is determined whether the corresponding large model is found. If so, step e3 is executed; if not, the process returns to step e1.

[0127] In step e3, a request body is constructed for encapsulation and calling. In order to decouple the database cloud platform from the big model, the flexibility and versatility of database operation and maintenance are improved, and the database cloud platform can quickly access multiple types of big models. The database cloud platform can store big model information of various types of big models. The big model information may include model alias, model real name, operation and maintenance knowledge base and other information. The big model information can be stored in an SQL table. The model alias can be matched by the front end of the database cloud platform through data drive. The model real name and the model alias are mapped one-to-one. The model real name can determine the encapsulation format of the request body, i.e., Request, and the parsing type field can determine the parsing format of the response body, i.e., Response. The relay identifiers such as the model alias, operation and maintenance knowledge base, and question and answer information of the big model can be combined to control the call of the operation and maintenance knowledge base and / or the instance database and obtain supplementary data.

[0128] In step e4, the information is dynamically parsed, desensitized, and converted. The database cloud platform can parse the JSON data in the input stream in real time, correcting, desensitizing, and converting the format of sensitive information.

[0129] In step e5, it is determined whether the parsing and conversion are abnormal. If there is an abnormality, the process returns to step e4 and a local abnormality processing can be skipped to avoid affecting the overall process. If there is no abnormality, the process proceeds to step e6.

[0130] In step e6, the stream information is pushed and displayed. The database cloud platform can push the inference answer information to the user in the form of stream information, and the inference answer information is displayed in real time on the front end of the database cloud platform.

[0131] In step e7, it is determined whether the push is abnormal. If so, the process returns to step e6, where local exception handling can be skipped to avoid affecting the overall process. If not, the process proceeds to step e8.

[0132] In step e8, user feedback information is received.

[0133] In step e9, it is determined whether the user has adopted the inference answer information. The database cloud platform can determine whether the user has adopted the inference answer information based on the user feedback information. If the user has adopted the inference answer information, step e10 is executed; if not, step e14 is executed.

[0134] In step e10, the database operation and maintenance auxiliary process information is stored. The question-and-answer process between the user and the database cloud platform, user feedback information, processing steps, and reasoning answer information can be stored in a complete historical record in the corresponding database.

[0135] In step e11, it is determined whether the historical records exceed the storage limit. If so, the process returns to step e10 and performs content truncation to ensure the integrity and validity of the data storage. If not, the process proceeds to step e12.

[0136] In step e12, a summary report is generated. The database cloud platform can regularly summarize information about the database operation and maintenance auxiliary process and generate a summary report. The summary report can include question and answer information, frequently asked questions, user feedback, optimization results, etc., thus supporting database operation and maintenance analysis. The information in the summary report can also be used by the data source for further in-depth pre-training of large models.

[0137] In step e13, customized user recommendations are generated. Based on the user's historical interaction data with database operation and maintenance assistance, personalized operation and maintenance recommendations are generated for the user. Operation and maintenance learning resources can also be pushed to the user, and learning plans can be implemented for the user, thereby improving the user's database operation and maintenance skills and resolving operation and maintenance bottlenecks.

[0138] In step e14, the user demand acquisition unit 22 is called again. Calling the user demand acquisition unit 22 again triggers a new interaction cycle. In the new interaction cycle, the initial demand profile information, target prompt, and inference answer information from the previous interaction cycle are also filled into the prompt template to generate a new target prompt, enriching the context information of the target prompt and optimizing the response of the large model.

[0139] The specific contents of the above steps e1 to e14 can be found in the relevant descriptions in the above embodiments, which will not be repeated here.

[0140] The present application also provides a database operation and maintenance auxiliary device based on a large model. Figure 9 A schematic diagram of the structure of a database operation and maintenance auxiliary device based on a large model provided in an embodiment of the present application is shown in FIG. Figure 9 As shown, the database operation and maintenance auxiliary device 200 based on the large model may include a user demand determination module 201, a prompt generation module 202 and an information push module 203.

[0141] The user demand determination module 201 can be used to determine the user's initial demand portrait information based on the user's input, the mapping relationship between the preset operation and maintenance scenario and the input path, and the user intention recognition rules. The initial demand portrait information includes the target operation and maintenance scenario and user intention information.

[0142] The prompt generation module 202 can be used to obtain supplementary data matching the initial demand profile information from an information supplement database associated with database operation and maintenance, and generate a target prompt in combination with a prompt template corresponding to the initial demand profile information.

[0143] The information push module 203 can be used to send target prompts to the large model corresponding to the initial demand portrait information, and generate inference answer information based on the output information fed back by the large model corresponding to the initial demand portrait information and push it to the user.

[0144] In some embodiments, the user demand determination module 201 can be specifically used to: obtain the operation and maintenance scenario corresponding to the input path formed by the user's input in the mapping relationship as the target operation and maintenance scenario; obtain user intention information based on the matching results of the user's input and the user intention recognition rules; determine the target instance database based on the user's input, and obtain basic information of the target instance database; obtain initial demand portrait information based on the target operation and maintenance scenario, user intention information and basic information.

[0145] In some embodiments, the user demand determination module 201 may also be used to generate a task set based on the initial demand profile information, and assign the tasks in the task set to corresponding task queues for processing.

[0146] The prompt generation module 202 can be specifically used to: when the task set includes a large model task, obtain supplementary data from the information supplement database, and generate a target prompt in combination with the prompt template corresponding to the initial demand portrait information.

[0147] In some embodiments, the information supplement database includes an operations and maintenance knowledge base and / or an instance database. The prompt generation module 202 may be specifically configured to: retrieve knowledge information matching the initial demand profile information from the operations and maintenance knowledge base as supplementary data; and / or, invoke an interface to retrieve instance data matching the initial demand profile information from the instance database corresponding to the initial demand profile information as supplementary data; and fill the supplementary data into the corresponding position in the prompt template to generate the target prompt.

[0148] In some examples, the prompt generation module 202 can be specifically used to: use the first keyword extracted from the initial demand portrait information to perform a search in the operation and maintenance knowledge base, and determine the first search result data that meets the quality standard conditions as supplementary data; expand based on the first keyword to obtain the second keyword, use the second keyword to perform a second search in the operation and maintenance database, and determine the second search result data that meets the quality standard conditions as supplementary data.

[0149] In some examples, the instance data in the supplementary data includes table structure information, and the initial demand profile information also includes an SQL statement selected by the user. The prompt generation module 202 can be specifically configured to: parse the SQL statement in the initial demand profile information to generate a syntax tree for the SQL statement; traverse the nodes in the syntax tree to filter out a target node containing a database table name, and obtain the database table name from the target node; and traverse the tables in the instance database based on the obtained database table name to obtain the table structure information as the supplementary data.

[0150] In some examples, the instance data in the supplementary data includes table structure information, and the initial demand profile information also includes the SQL statement selected by the user. The prompt generation module 202 can be specifically configured to: split the SQL statement into multiple subtasks according to the query logic; process the multiple subtasks in parallel; and obtain the table structure information corresponding to each subtask from the instance database as supplementary data.

[0151] In some embodiments, the information push module 203 can also be used for: when the information that needs to be obtained from the instance database for the large model processing target prompt corresponding to the initial demand portrait information includes first-level sensitive information, the first-level sensitive information is replaced with secure desensitized information to obtain desensitized data; when the information that needs to be obtained from the instance database for the large model processing target prompt corresponding to the initial demand portrait information includes second-level sensitive information, the second-level sensitive information is processed using a replacement desensitization method and / or a mask desensitization method to obtain desensitized data; and the desensitized data is transmitted to the large model corresponding to the initial demand portrait information as feedback output information.

[0152] In some embodiments, the user demand determination module 201 may also be used to: when the received user feedback information indicates that the inference answer information does not meet the user demand, again interact with the user to determine new initial demand profile information.

[0153] The prompt generation module 202 can also be used to: obtain new supplementary data that matches the new initial demand portrait information from the information supplement database; use at least part of the information in the reasoning answer information and / or at least part of the information in the target prompt corresponding to the reasoning answer information as supplementary data; based on the supplementary data, combined with the prompt template corresponding to the new initial demand portrait information, generate a new target prompt for large model processing.

[0154] In some embodiments, the weight parameters of the large model include incremental weight parameters and original weight parameters of the original large model, and the incremental weight parameters are obtained by training the original large model based on the operation and maintenance knowledge information in the operation and maintenance knowledge base. The database operation and maintenance assistance device 200 based on the large model may also include a large model adjustment module. The large model adjustment module can be used to: freeze the original weight parameters, decompose the incremental weight parameters into two low-rank matrices, and the dimension of the low-rank matrix is ​​smaller than the dimension of the incremental weight parameters; use the received user feedback information as a reward signal to adjust the low-rank matrix; obtain the updated weight parameters of the large model based on the original weight parameters and the new incremental weight parameters obtained based on the adjusted low-rank matrix.

[0155] In some examples, the large model adjustment module can also be used to: set a rank range based on the resource consumption information of the original large model for training, and the rank in the rank range is negatively correlated with the resource consumption represented by the resource consumption information; select multiple ranks within the rank range as candidate ranks of the low-rank matrix, and update the weight parameters of the large model based on the low-rank matrices of different candidate ranks; use the verification sample set to obtain the performance parameters of the large model after the weight parameters corresponding to the low-rank matrices of different candidate ranks are updated, and determine the candidate rank corresponding to the large model with the highest performance represented by the performance parameters as the rank of the low-rank matrix.

[0156] It should be noted that the large model-based database operation and maintenance assistance device 200 is a device corresponding to the above-mentioned large model-based database operation and maintenance assistance method. All implementation methods in the above-mentioned method embodiments are applicable to the embodiments of the device and can achieve the same technical effects.

[0157] This application also provides a database operation and maintenance auxiliary device based on a large model. Figure 10 A schematic diagram of the structure of a database operation and maintenance auxiliary device based on a large model provided in an embodiment of the present application is shown in FIG. Figure 10 As shown, the database operation and maintenance auxiliary device 300 based on the large model includes a memory 301, a processor 302 and a computer program stored in the memory 301 and executable on the processor 302.

[0158] In some examples, the processor 302 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0159] The memory 301 may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical, or other physical / tangible memory storage device. Therefore, typically, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the database operation and maintenance assistance method based on a large model according to an embodiment of the present application.

[0160] The processor 302 runs a computer program corresponding to the executable program code by reading the executable program code stored in the memory 301, so as to implement the database operation and maintenance auxiliary method based on the large model in the above embodiment.

[0161] In some examples, the database operation and maintenance auxiliary device 300 based on the large model may further include a communication interface 303 and a bus 304. Figure 10 As shown, the memory 301 , the processor 302 , and the communication interface 303 are connected via a bus 304 and communicate with each other.

[0162] The communication interface 303 is mainly used to implement communication between the modules, devices, units and / or equipment in the embodiment of the present application. Input devices and / or output devices can also be connected through the communication interface 303.

[0163] The bus 304 includes hardware, software, or both, and couples the components of the large-model-based database operation and maintenance auxiliary device 300 to each other. By way of example, and not limitation, the bus 304 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-E) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, the bus 304 may include one or more buses. Although embodiments herein describe and illustrate a particular bus, this application contemplates any suitable bus or interconnect.

[0164] This application also provides a database operation and maintenance assistance system based on a large model. Figure 11 A schematic diagram of the structure of a database operation and maintenance auxiliary system based on a large model provided in an embodiment of the present application is shown in FIG. Figure 11 As shown, the database operation and maintenance auxiliary system based on the big model includes a database cloud platform 41 and a big model subsystem 42.

[0165] The database cloud platform 41 can host and manage multiple instance databases. The number and types of instance databases in the database cloud platform 41 are not limited. The database cloud platform can be used to implement the large model-based database operation and maintenance assistance method described above, achieving the same technical effects. To avoid repetition, the details are not described here.

[0166] The large model subsystem 42 provides large models. These large models may include original large models and large models trained using the operation and maintenance knowledge base. The large model subsystem 42 can provide a variety of large models to address various operation and maintenance scenarios. The number and types of large models provided by the large model subsystem 42 are not limited. The large model subsystem 42 is in communication with the database cloud platform 41 and is configured to receive target prompts from the database cloud platform 41 and generate output information based on the target prompts and feed it back to the database cloud platform 41.

[0167] The specific contents of the database cloud platform 41 and the large model subsystem 42 can be found in the relevant descriptions in the above embodiments, which will not be repeated here.

[0168] The present application also provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are executed by a processor, the large model-based database operation and maintenance auxiliary method in the above-mentioned embodiment can be implemented, and the same technical effect can be achieved. To avoid repetition, the above-mentioned computer-readable storage medium may include a non-transitory computer-readable storage medium, such as a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc., which is not limited here.

[0169] The present application also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the large model-based database operation and maintenance assistance method in the above-mentioned embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0170] It should be understood that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. For device embodiments, equipment embodiments, system embodiments, computer-readable storage medium embodiments, and computer program product embodiments, the relevant parts can be referred to the description part of the method embodiment. This application is not limited to the specific steps and structures described above and shown in the figures. Those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of this application. In addition, for the sake of brevity, a detailed description of known method technologies is omitted here.

[0171] Aspects of the present application have been described above with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each block in the flowcharts and / or block diagrams, as well as combinations of blocks in the flowcharts 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, a special-purpose computer, or other programmable data processing device to produce a machine such that execution of these instructions by the processor of the computer or other programmable data processing device enables the implementation of the functions / actions specified in one or more blocks in the flowcharts and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It should also be understood that each block in the block diagrams and / or flowcharts, as well as combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware that performs the specified functions or actions, or by a combination of dedicated hardware and computer instructions.

[0172] Those skilled in the art should understand that the above embodiments are illustrative rather than restrictive. Different technical features appearing in different embodiments can be combined to achieve beneficial effects. Based on a study of the drawings, the specification and the claims, those skilled in the art should be able to understand and implement other variations of the disclosed embodiments. In the claims, the term "comprising" does not exclude other devices or steps; the quantifier "one" does not exclude a plurality; the terms "first" and "second" are used to identify names rather than to indicate any specific order. Any figure marks in the claims should not be understood as limiting the scope of protection. The functions of multiple parts appearing in the claims can be implemented by a separate hardware or software module. The fact that certain technical features appear in different dependent claims does not mean that these technical features cannot be combined to achieve beneficial effects.

Claims

1. A database operation and maintenance auxiliary method based on a large model, characterized in that: include: Determine the user's initial demand profile information based on the user's input, the mapping relationship between the preset operation and maintenance scenario and the input path, and the user intention recognition rule, wherein the initial demand profile information includes the target operation and maintenance scenario and the user intention information; Acquire supplementary data matching the initial demand profile information from an information supplement database associated with database operation and maintenance, and generate a target prompt in combination with a prompt template corresponding to the initial demand profile information; Sending the target prompt to the large model corresponding to the initial demand profile information, and generating inference answer information based on the output information fed back by the large model corresponding to the initial demand profile information and pushing it to the user; The determining of the user's initial demand profile information based on the user's input, the mapping relationship between the preset operation and maintenance scenario and the input path, and the user intention recognition rule includes: Acquire, from the mapping relationship, an operation and maintenance scenario corresponding to an input path formed by the user's input as the target operation and maintenance scenario; Obtaining the user intent information according to a matching result between the user input and the user intent recognition rule; Determine a target instance database according to the user's input, and obtain basic information of the target instance database; The initial demand profile information is obtained based on the target operation and maintenance scenario, the user intention information and the basic information.

2. The method according to claim 1, characterized in that Also includes: Generate a task set based on the initial demand profile information, and assign the tasks in the task set to corresponding task queues for processing; The acquiring of supplementary data matching the initial demand profile information from an information supplement database associated with database operation and maintenance, and generating a target prompt in combination with a prompt template corresponding to the initial demand profile information, includes: In the case where the task set includes a large model task, the supplementary data is obtained from the information supplement database, and the target prompt is generated in combination with the prompt template corresponding to the initial demand portrait information.

3. The method according to claim 1, characterized in that The information supplement database includes an operation and maintenance knowledge base and / or an instance database; The acquiring of supplementary data matching the initial demand profile information from an information supplement database associated with database operation and maintenance, and generating a target prompt in combination with a prompt template corresponding to the initial demand profile information, includes: Retrieving knowledge information matching the initial demand profile information from an operation and maintenance knowledge base as the supplementary data, and / or calling an interface to obtain instance data matching the initial demand profile information from an instance database corresponding to the initial demand profile information as the supplementary data; Fill the supplementary data into the corresponding position in the prompt template to generate the target prompt.

4. The method according to claim 3, characterized in that The retrieving the knowledge information matching the initial demand profile information from the operation and maintenance knowledge base as the supplementary data includes: Performing a search in the operation and maintenance knowledge base using a first keyword extracted from the initial demand profile information, and determining first search result data that meets the quality standard condition as the supplementary data; The first keyword is expanded to obtain a second keyword, and the second keyword is used to perform a secondary search in the operation and maintenance knowledge base, and the second search result data that meets the quality standard condition is determined as the supplementary data.

5. The method according to claim 3, characterized in that The instance data in the supplementary data includes table structure information, and the initial demand profile information also includes a structured query language SQL statement selected by the user's input; The calling interface obtains instance data matching the initial demand portrait information from an instance database corresponding to the initial demand portrait information as the supplementary data, including: Parsing the SQL statements in the initial demand profile information to generate a syntax tree of the SQL statements; Traversing the nodes in the syntax tree, screening out a target node containing a database table name, and obtaining the database table name in the target node; According to the acquired database table name, the tables in the instance database are traversed to obtain table structure information as the supplementary data.

6. The method according to claim 3, characterized in that The instance data in the supplementary data includes table structure information, and the initial demand profile information also includes a structured query language SQL statement selected by the user's input; The calling interface obtains instance data matching the initial demand portrait information from an instance database corresponding to the initial demand portrait information as the supplementary data, including: Split the SQL statement into multiple subtasks according to the query logic; The multiple subtasks are processed in parallel, and table structure information corresponding to each subtask is obtained in the instance database as the supplementary data.

7. The method according to claim 1, characterized in that Also includes: When the information required to be obtained from the instance database for processing the target prompt by the large model corresponding to the initial demand profile information includes first-level sensitive information, the first-level sensitive information is replaced with secure desensitized information to obtain desensitized data; When the information required to be obtained from the instance database for processing the target prompt by the large model corresponding to the initial demand profile information includes secondary sensitive information, the secondary sensitive information is processed using a replacement desensitization method and / or a mask desensitization method to obtain desensitized data; The desensitized data is transmitted to the large model corresponding to the initial demand portrait information as the output information of feedback.

8. The method according to claim 1, characterized in that Also includes: When the received user feedback information indicates that the inference answer information does not meet the user's needs, performing input interaction with the user again to determine new initial demand profile information; Acquire new supplementary data matching the new initial demand profile information from the information supplement database; using at least part of the inference answer information and / or at least part of the target prompt corresponding to the inference answer information as the supplementary data; Based on the supplementary data and in combination with the prompt template corresponding to the new initial demand portrait information, a new target prompt is generated for processing by the large model.

9. The method according to claim 1, characterized in that The weight parameters of the large model include incremental weight parameters and original weight parameters of the original large model, wherein the incremental weight parameters are obtained by training the original large model based on the operation and maintenance knowledge information in the operation and maintenance knowledge base; The method further comprises: Freezing the original weight parameters, and decomposing the incremental weight parameters into two low-rank matrices, where the dimensions of the low-rank matrices are smaller than the dimensions of the incremental weight parameters; Using the received user feedback information as a reward signal to adjust the low-rank matrix; The updated weight parameters of the large model are obtained according to the original weight parameters and the new incremental weight parameters obtained based on the adjusted low-rank matrix.

10. The method according to claim 9, characterized in that Also includes: Setting a rank range based on resource consumption information for training the original large model, wherein the ranks in the rank range are negatively correlated with the resource consumption represented by the resource consumption information; Selecting multiple ranks within the rank range as candidate ranks of the low-rank matrix, and updating weight parameters of the large model based on the low-rank matrices of different candidate ranks respectively; Using the verification sample set, the performance parameters of the large model after the weight parameters corresponding to the low-rank matrix of different candidate ranks are obtained, and the candidate rank corresponding to the large model with the highest performance represented by the performance parameters is determined as the rank of the low-rank matrix.

11. A database operation and maintenance auxiliary device based on a large model, characterized in that: include: A user demand determination module is used to determine the user's initial demand profile information based on the user's input, the mapping relationship between the preset operation and maintenance scenario and the input path, and the user intention recognition rule. The initial demand profile information includes the target operation and maintenance scenario and the user intention information; A prompt generation module is used to obtain supplementary data matching the initial demand profile information from an information supplement database associated with database operation and maintenance, and generate a target prompt based on a prompt template corresponding to the initial demand profile information; The information push module is used to send the target prompt to the large model corresponding to the initial demand portrait information, and generate the inference answer information based on the output information fed back by the large model corresponding to the initial demand portrait information and push it to the user. The user demand determination module is specifically used to: obtain the operation and maintenance scenario corresponding to the input path formed by the user's input in the mapping relationship as the target operation and maintenance scenario; obtain the user intention information based on the matching result of the user's input and the user intention recognition rule; determine the target instance database based on the user's input and obtain the basic information of the target instance database; obtain the initial demand portrait information based on the target operation and maintenance scenario, the user intention information and the basic information.

12. A database operation and maintenance auxiliary device based on a large model, characterized in that: include: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, the large model-based database operation and maintenance assistance method according to any one of claims 1 to 10 is implemented.

13. A database operation and maintenance auxiliary system based on a large model, characterized in that: include: A database cloud platform, configured to execute the database operation and maintenance assistance method based on a large model as claimed in any one of claims 1 to 10; The large model subsystem is communicatively connected to the database cloud platform, and is used to receive the target prompt sent by the database cloud platform, and obtain output information based on the target prompt and feed it back to the database cloud platform.

14. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the large model-based database operation and maintenance assistance method according to any one of claims 1 to 10.

15. A computer program product, characterized in that It includes a computer program, which, when executed by a processor, implements the database operation and maintenance assistance method based on a large model as described in any one of claims 1 to 10.

Images