Fault problem analysis method and device based on large language model, equipment and medium

Abstract

The embodiment of the invention discloses a fault problem analysis method and device based on a large language model, equipment and a medium, and the method comprises the steps: obtaining error data, and carrying out the arrangement, and obtaining the error arrangement data; accessing the error arrangement data to an agent platform, and setting a plurality of cloud functions in the agent platform; the different cloud functions are used for acquiring target error arrangement data required by different analysis functions from the error arrangement data; creating an intelligent agent according to the fault analysis workflow arrangement requirement, the plurality of cloud functions and a pre-constructed error code knowledge base; and acquiring a fault analysis instruction input by a user, and calling the large language model, the cloud function and the error code knowledge base by the intelligent agent according to the fault analysis instruction to obtain a fault problem analysis result. According to the method, the big language model is combined with retrieval enhancement generation, the diagnosis capability of the big language model on intelligent analysis of fault problems can be effectively enhanced, and the accuracy and analysis efficiency of fault analysis are effectively improved.

Description

Technical Field

[0001] Embodiments of the present application relate to the field of Internet technologies, and in particular, to a method, apparatus, device, and medium for analyzing fault problems based on large language models. Background Art

[0002] With the development of Internet technologies, software systems have been widely applied in various fields. Inevitably, various problems will occur during the operation of the system. When a fault problem occurs in the system, it is necessary to analyze log data, etc. to determine the characteristics of the fault problem. Existing automatic log analysis solutions often rely on simple keyword matching and threshold alarms, and it is difficult to handle diverse log data and high-dimensional abnormal patterns in complex systems. Although existing automatic log analysis solutions can help operation and maintenance personnel identify fault problems to a certain extent, operation and maintenance personnel still need to further manually analyze log data to judge the concentration, dispersion, etc. of errors, so as to determine the characteristics of the fault problem. It can be seen that how to accurately and efficiently analyze fault problems has become an urgent problem to be solved. Summary of the Invention

[0003] In view of the above problems, the present application proposes a method, apparatus, device, and medium for analyzing fault problems based on large language models, which are used to solve the following problems: existing fault analysis methods require a large amount of labor costs and have low analysis efficiency.

[0004] According to one aspect of the embodiments of the present application, a method for analyzing fault problems based on large language models is provided, including:

[0005] Obtain the collected error data, and organize the error data to obtain organized error data;

[0006] Connect the organized error data to an agent platform, and set multiple cloud functions in the agent platform; different cloud functions are used to obtain target organized error data required for different analysis functions from the organized error data;

[0007] Create an agent according to the fault analysis workflow orchestration requirements, multiple cloud functions, and a pre-constructed error code knowledge base;

[0008] Obtain a fault analysis instruction input by a user, and the agent calls a large language model, cloud functions, and the error code knowledge base according to the fault analysis instruction to obtain a fault problem analysis result.

[0009] Further, obtaining the collected error data and organizing the error data to obtain organized error data further includes:

[0010] Obtain the error data collected by the stability platform; wherein, the error data includes error information, error codes, and environmental information when the error occurs.

[0011] Clean the error data, group and count the cleaned error data, and form error collation data; among them, the error collation data contains error correlation information in multiple dimensions.

[0012] Further, the error correlation information in multiple dimensions at least includes: scenario information, error information, error code, error occurrence date, container instance at the time of error occurrence, available zone at the time of error occurrence, operating environment at the time of error occurrence, application service information at the time of error occurrence, number of instance errors, number of available zone errors, and number of operating environment errors.

[0013] Further, creating an agent according to the fault analysis workflow orchestration requirements, multiple cloud functions, and a pre-constructed error code knowledge base further includes:

[0014] Create multiple routing branches of the agent through a visual agent creation tool according to the fault analysis workflow orchestration requirements; different routing branches are used to provide different analysis functions;

[0015] Create corresponding working nodes for each routing branch and configure the cloud function corresponding to the routing branch for the working nodes;

[0016] Create an error code parsing process for the agent and configure the error code knowledge base for the error code parsing process.

[0017] Further, the construction process of the error code knowledge base includes:

[0018] Obtain an error code description file; among them, the error code description file records error codes and error description information corresponding to the error codes;

[0019] Use a knowledge base tool to vectorize the error codes in the error code description file, and create an error code knowledge base based on the error description information corresponding to the error codes in the error code description file.

[0020] Further, the agent calls a large language model, cloud functions, and an error code knowledge base according to a fault analysis instruction, and the obtained fault problem analysis result further includes:

[0021] The agent calls a large language model to identify the intent of the fault analysis instruction, obtains the user analysis intent, and determines the target routing branch that matches the user analysis intent;

[0022] Execute the target routing branch, call the cloud function corresponding to the target routing branch, and obtain the target error collation data required for the corresponding analysis function from the error collation data through the cloud function;

[0023] Execute the error code parsing process, and determine the error description information corresponding to the target error code in the target error collation data according to the error code knowledge base;

[0024] Provide the target error collation data, the error description information corresponding to the target error code, and the preset system prompt information to the large language model for analysis and reasoning to obtain the analysis result of the fault problem.

[0025] Further, executing the error code parsing process and determining the error description information corresponding to the target error code in the target error collation data according to the error code knowledge base further includes:

[0026] Execute the error code parsing process, and extract the target error code included in the target error collation data;

[0027] Query the error code knowledge base to obtain the error description information corresponding to the target error code.

[0028] According to another aspect of the embodiments of the present application, a fault problem analysis device based on a large language model is provided, including:

[0029] A collation module, adapted to obtain the collected error data and collate the error data to obtain error collation data;

[0030] A setting module, adapted to connect the error collation data to the intelligent agent platform and set multiple cloud functions in the intelligent agent platform; different cloud functions are used to obtain the target error collation data required for different analysis functions from the error collation data;

[0031] A creation module, adapted to create an intelligent agent according to the fault analysis workflow orchestration requirements, multiple cloud functions, and the pre-constructed error code knowledge base;

[0032] An analysis module, adapted to obtain the fault analysis instruction input by the user, and the intelligent agent calls the large language model, cloud functions, and error code knowledge base according to the fault analysis instruction to obtain the analysis result of the fault problem.

[0033] Further, the collation module is further adapted to:

[0034] Obtain the error data collected by the stability platform; wherein, the error data includes error information, error codes, and environmental information when the error occurs;

[0035] Perform data cleaning on the error data, and perform grouped statistics on the cleaned error data to form error collation data; wherein, the error collation data contains error correlation information in multiple dimensions.

[0036] Further, the error correlation information in multiple dimensions at least includes: scenario information, error information, error code, error occurrence date, container instance at the time of error, availability zone at the time of error, operating environment at the time of error, application service information at the time of error, number of instance errors, number of availability zone errors, and number of operating environment errors.

[0037] Further, the creation module is further adapted to:

[0038] Create multiple routing branches of the intelligent agent according to the fault analysis workflow orchestration requirements through the visual intelligent agent creation tool; different routing branches are used to provide different analysis functions;

[0039] Create corresponding worker nodes for each routing branch and configure cloud functions corresponding to the routing branch for the worker nodes;

[0040] Create an error code parsing process for the intelligent agent and configure an error code knowledge base for the error code parsing process.

[0041] Further, the device further includes: a knowledge base construction module; the knowledge base construction module is adapted to:

[0042] Obtain an error code description file; wherein, the error code description file records error codes and error description information corresponding to the error codes;

[0043] Use a knowledge base tool to vectorize the error codes in the error code description file, and create an error code knowledge base based on the error description information corresponding to the error codes in the error code description file.

[0044] Further, the analysis module is further adapted to:

[0045] The intelligent agent calls a large language model to recognize the intention of the fault analysis instruction, obtain the user analysis intention, and determine the target routing branch that matches the user analysis intention;

[0046] Execute the target routing branch, call the cloud function corresponding to the target routing branch, and obtain the target error collation data required for the corresponding analysis function from the error collation data through the cloud function;

[0047] Execute the error code parsing process, and determine the error description information corresponding to the target error code in the target error collation data according to the error code knowledge base;

[0048] Provide the target error collation data, the error description information corresponding to the target error code, and the preset system prompt information to the large language model for analysis and reasoning to obtain the analysis result of the fault problem.

[0049] Further, the analysis module is further adapted to:

[0050] Execute the error code parsing process to extract the target error code contained in the target error collation data;

[0051] Query the error code knowledge base to obtain the error description information corresponding to the target error code.

[0052] According to another aspect of the embodiments of the present application, there is provided a computing device, including: a processor, a memory, a communication interface, and a communication bus. The processor, the memory, and the communication interface complete communication with each other through the communication bus;

[0053] The memory is used to store at least one executable instruction, and the executable instruction causes the processor to execute the operations corresponding to the above-mentioned fault problem analysis method based on the large language model.

[0054] According to still another aspect of the embodiments of the present application, there is provided a computer storage medium, in which at least one executable instruction is stored, and the executable instruction causes the processor to execute the operations corresponding to the above-mentioned fault problem analysis method based on the large language model.

[0055] According to yet another aspect of the embodiments of the present application, there is provided a computer program product, including at least one executable instruction, and the executable instruction causes the processor to execute the operations corresponding to the above-mentioned fault problem analysis method based on the large language model.

[0056] According to the fault problem analysis method, device, equipment and medium based on the large language model provided by the embodiments of the present application, by collating the error data, error collation data is obtained, and the preprocessing of the error data is realized, making it more convenient for analysis; the error collation data is connected to the intelligent agent platform, and multiple cloud functions are set in the intelligent agent platform. Through different cloud functions, the ability to quickly assemble the error collation data required for different analysis functions can be completed, and the intelligent extraction and assembly of the error collation data are conveniently realized, effectively improving the data processing efficiency; according to the fault analysis workflow orchestration requirements, multiple cloud functions, and the pre-constructed error code knowledge base, an intelligent agent is created; when the user needs to perform specific fault problem intelligent analysis, a fault analysis instruction can be input, and the intelligent agent calls the large language model, cloud functions, and error code knowledge base according to the fault analysis instruction, and conveniently obtains the fault problem analysis result. This solution combines the large language model with retrieval augmented generation by arranging the error code knowledge base into the workflow of the intelligent agent, which can effectively enhance the diagnostic ability of the large language model for intelligent analysis of fault problems, effectively improve the accuracy of fault analysis, greatly improve the fault analysis efficiency, reduce manual participation and time costs, and realize accurate and efficient intelligent analysis of fault problems.

[0057] The above description is only an overview of the technical solution of the embodiments of the present application. In order to be able to understand the technical means of the embodiments of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the embodiments of the present application more obvious and understandable, the following specifically describes the specific implementation manners of the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the embodiments of the present application. And throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0059] Figure 1 shows a schematic flow chart of a fault problem analysis method based on a large language model according to an embodiment of the present application;

[0060] Figure 2a shows a schematic flow chart of a fault problem analysis method based on a large language model according to another embodiment of the present application;

[0061] Figure 2b shows a schematic diagram of an error code knowledge base;

[0062] Figure 2c shows a schematic diagram of a cloud function management list;

[0063] Figure 2d shows a schematic diagram of the creation of an agent;

[0064] Figure 2e shows a schematic diagram of a preset system prompt message;

[0065] Figure 2f shows a schematic diagram of a fault problem analysis result;

[0066] Figure 3 shows a block diagram of a fault problem analysis device based on a large language model according to an embodiment of the present application;

[0067] Figure 4 shows a schematic diagram of the structure of a computing device according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0069] First, the noun terms related to one or more embodiments of this application are explained.

[0070] Large Language Model (LLM): Refers to an artificial intelligence language model with a large scale, a huge amount of training data, and numerous parameters, such as OpenAI GPT, BERT, etc. Such models are designed to understand and generate natural language, can better train to analyze, reason, and output key information of text, and can handle various tasks.

[0071] Agent: Refers to an agent that can perceive the environment and take actions to achieve specific goals, and can be used to solve a certain type of intelligent problem.

[0072] Service Level Objective (SLO): A key indicator for measuring the reliability of a system and the quality of service, which defines the performance goals that a service should achieve within a specific time window, and usually refers to the availability standard for measuring a certain service, API, etc.

[0073] SLO breach: A term in operations and SRE (Site Reliability Engineering), which means that the actual performance of the system breaks through the threshold specified by the SLO, resulting in the failure to meet the service reliability standard.

[0074] Retrieval-Augmented Generation (RAG): A hybrid artificial intelligence technology that combines information retrieval and text generation. It enhances the output quality of the generation model by retrieving relevant information from external knowledge sources, and can be used to enhance the scenario accuracy of the content generated by the LLM.

[0075] Figure 1 The flowchart of a fault problem analysis method based on a large language model according to an embodiment of the present application is shown, as Figure 1 shown, the method includes the following steps:

[0076] Step S101, obtain the collected error data, and sort out the error data to obtain the error sorted data.

[0077] In the embodiments of the present application, error data generated during the operation of each application service of the system can be collected in real time through a stability platform, etc. Among them, the error data includes error information, error codes, and environmental information when the error occurs, etc. The environmental information can specifically include information such as availability zones, operating environments, and container instances. Usually, the error information is a string of characters such as numbers.

[0078] For the convenience of analysis, the collected error data can be first processed such as cleaned and denoised, and then sorted such as grouped and statistically analyzed, so as to organize the error data into error organized data with a basic detailed data structure. The error organized data can contain error correlation information in multiple dimensions. Specifically, the multiple dimensions can include: scenario dimension, error information dimension, error code dimension, error occurrence date dimension, environment dimension when the error occurs, etc. Among them, setting the error occurrence date dimension can be used for regular trend analysis, so as to obtain the error organized data within a certain time period based on the error occurrence date dimension.

[0079] Step S102: Connect the error organized data to the agent platform and set multiple cloud functions in the agent platform.

[0080] Among them, the agent platform is a comprehensive system that supports the development, deployment, and management of agents. It can support the reasoning, decision-making, and execution of agents, and can perform life cycle management on agents (such as creation, deployment, update, and retirement, etc.), version control, and permission management. It also supports connecting to multiple large language models. The embodiments of the present application apply the agent platform to the field of fault problem analysis and provide multiple analysis functions. It can automatically and intelligently assemble the required error organized data according to the user's fault analysis instructions by the agent, call the large language model to reason and analyze the error organized data, and obtain the fault problem analysis result. Specifically, the multiple analysis functions can include tracking analysis function, scenario historical error analysis function, scenario analysis function connected according to the business ID, resource analysis function affected by the error, SLO breach analysis function, error analysis function for each connected scenario, error analysis function for each connected business, etc.

[0081] In step S102, the error organized data can be encapsulated into an API and connected to the agent platform for the agent platform to use, such as querying, assembling, etc. And multiple cloud functions also need to be set in the agent platform. Among them, there is a corresponding relationship between the cloud functions and the analysis functions. Different cloud functions are used to obtain the target error organized data required for different analysis functions from the error organized data. That is to say, the cloud functions are used to provide the data assembly ability required for their corresponding analysis functions.

[0082] Step S103: Create an agent according to the fault analysis workflow orchestration requirements, multiple cloud functions, and a pre-constructed error code knowledge base.

[0083] Considering that in a system, the specific type of error is usually represented in the form of an error code, and it is difficult for a large language model to directly recognize the meaning of the error code. To solve the above problems, in the embodiments of the present application, a technology combining a large language model with retrieval-augmented generation is proposed to enhance the diagnostic ability of intelligent analysis of fault problems. Among them, an error code knowledge base is pre-constructed, and the corresponding relationships between error codes, vectorized error codes, and error description information are recorded in the error code knowledge base. By introducing the error code knowledge base, relevant information of the error code (such as vectorized error codes and error description information) can be retrieved and provided to the large language model conveniently, realizing retrieval-augmented generation, which can effectively enhance the diagnostic ability of the large language model for intelligent analysis of fault problems and improve the accuracy of fault analysis.

[0084] To create an intelligent agent, it is also necessary to obtain the fault analysis workflow orchestration requirements. The fault analysis workflow orchestration requirements refer to the design requirements of the visualization process set by the operation and maintenance personnel for the workflow of the intelligent agent's intelligent analysis of fault problems. Specifically, in the intelligent agent platform, the workflow of the intelligent agent's intelligent analysis of fault problems can be orchestrated through a visualization intelligent agent creation tool according to the fault analysis workflow orchestration requirements, multiple cloud functions, and the error code knowledge base, thus completing the convenient creation of the intelligent agent.

[0085] Step S104: Obtain the fault analysis instruction input by the user, and the intelligent agent calls the large language model, cloud function, and error code knowledge base according to the fault analysis instruction to obtain the fault problem analysis result.

[0086] When the user wants to conduct specific intelligent analysis of fault problems based on the error data generated by the system, a fault analysis instruction can be input. The fault analysis instruction is a task-based operation command input by the user, such as "analyze the error trend of the historical error data generated in the manuscript review submission scenario and give analysis and troubleshooting suggestions", etc. Obtain the fault analysis instruction, and the intelligent agent perceives the user's analysis intention and takes actions according to the fault analysis instruction, automatically calling the large language model, cloud function, and error code knowledge base to obtain the fault problem analysis result.

[0087] According to the fault problem analysis method based on the large language model provided by the embodiments of the present application, by sorting out the error data, error sorting data is obtained, realizing the preprocessing of the error data and making it more convenient for analysis; the error sorting data is connected to the agent platform, and multiple cloud functions are set in the agent platform. Through different cloud functions, the ability to quickly assemble the error sorting data required for different analysis functions can be completed, and the intelligent extraction and assembly of the error sorting data are conveniently realized, effectively improving the data processing efficiency; according to the fault analysis workflow orchestration requirements, multiple cloud functions, and the pre-constructed error code knowledge base, an agent is created; when the user needs to perform specific intelligent analysis of fault problems, a fault analysis instruction can be input, and the agent calls the large language model, cloud functions, and error code knowledge base according to the fault analysis instruction, and conveniently obtains the fault problem analysis result. By orchestrating the error code knowledge base into the workflow of the agent, this solution realizes the combination of the large language model and retrieval augmented generation, can effectively enhance the diagnostic ability of the large language model for intelligent analysis of fault problems, effectively improves the accuracy of fault analysis, and greatly improves the fault analysis efficiency, reduces manual participation and time costs, and realizes accurate and efficient intelligent analysis of fault problems.

[0088] Figure 2a FIG. shows a schematic flowchart of a fault problem analysis method based on a large language model according to another embodiment of the present application, as Figure 2a shown, the method includes the following steps:

[0089] Step S201, obtain the error data collected by the stability platform, perform data cleaning on the error data, and perform grouped statistics on the cleaned error data to form error sorting data.

[0090] The stability platform is a technical system built to ensure the continuous and reliable operation of the system and services. The stability platform can collect the error data generated by each application service of the system in real time during operation. The error data includes error information, error codes, and environmental information when the error occurs, etc.

[0091] After obtaining the error data, the error data can be cleaned and sorted. Among them, the cleaning process can include noise reduction based on fingerprints, filtering out low-level error data, etc. According to the error information, error codes, error occurrence dates, container instances, availability zones, operating environments, and application services, etc., grouped statistics are performed on the cleaned error data to form error sorting data with a basic detailed data structure. The error sorting data contains error correlation information in multiple dimensions.

[0092] Specifically, the error correlation information in multiple dimensions may at least include: scenario information, error information, error code, error occurrence date, container instance at the time of error, availability zone at the time of error, operating environment at the time of error, application service information at the time of error, instance error count, availability zone error count, and operating environment error count, etc.

[0093] For example, the data structure of a piece of error collation data may be:

[0094] Scenario name: Manuscript callback

[0095] Error information: *

[0096] Error code: 177016

[0097] Error count: 5

[0098] Error occurrence date: 20250320

[0099] Application service name: Name1

[0100] Running POD instance of the application service: Instance1

[0101] Availability zone: sh001

[0102] Operating environment: prod (production environment)

[0103] Instance error count: 5

[0104] Availability zone error count: 5

[0105] Operating environment error count: 5

[0106] Among them, the error information being * means that it is obtained after denoising the original digital error information and other forms based on fingerprints.

[0107] Step S202: Obtain the error code description file, use the knowledge base tool to vectorize the error codes in the error code description file, and create an error code knowledge base based on the error description information corresponding to the error codes in the error code description file.

[0108] Among them, the error code description file records error codes and the error description information corresponding to the error codes. The error description information is information used to describe the specific type of error. Error codes and the error description information corresponding to each error code can be collected, and after cleaning and de-duplicating the error codes, an error code description file is formed. The error code description file can specifically be in csv format or excel format, etc.

[0109] For example, the error code description file records that: the error code is 21051, and the corresponding error description information is that the manuscript description length is zero; the error code is 21052, and the corresponding error description information is that the manuscript description length is too long and has exceeded the limit; the error code is 21053, and the corresponding error description information is that the manuscript description type does not exist or does not match; the error code is 21054, and the corresponding error description information is that the manuscript description type does not match the corresponding partition type; the error code is 21055, and the corresponding error description information is that the manuscript description type does not match the corresponding creation type.

[0110] After obtaining the error code description file, use the knowledge base tool to create a knowledge base for the error code description file to obtain an error code knowledge base. Specifically, vectorize the error codes in the error code description file, and create an error code knowledge base based on the error description information corresponding to the error codes in the error code description file.

[0111] Figure 2b Fig. shows a schematic diagram of the error code knowledge base, such as Figure 2b As shown, the error code knowledge base contains multiple error code records, which are stored using the error code as the vector field; for each error code record, an edit control is also set, so that users can edit the error code, error description information, etc. in the error code record by triggering the edit control. As Figure 2b As shown, in the error code record corresponding to ID1, the error code is 10001, and the error description information is that this type does not support submitting manuscripts; in the error code record corresponding to ID2, the error code is 10002, and the error description information is that this manuscript already exists; in the error code record corresponding to ID3, the error code is 10003, and the error description information is that this manuscript does not exist.

[0112] Step S203, connect the error sorting data to the intelligent agent platform, and set multiple cloud functions in the intelligent agent platform.

[0113] The error sorting data can be encapsulated into an API and connected to the intelligent agent platform for the intelligent agent platform to assemble and use. Moreover, multiple cloud functions need to be set in the intelligent agent platform. Among them, there is a corresponding relationship between the cloud functions and the analysis functions, and different cloud functions are used to obtain the target error sorting data required for different analysis functions from the error sorting data.

[0114] Figure 2c Fig. shows a schematic diagram of the cloud function management list, such as Figure 2cAs shown, there is a cloud function named "Tracking Analysis" whose cloud function description information is to analyze a certain recall; a cloud function named "Scenario Historical Error Analysis" whose cloud function description information is to analyze scenario historical errors; a cloud function named "Basic Information Query_Scenario_Business ID" whose cloud function description information is to query the list of accessed scenarios according to the business ID. For each cloud function, there are also management controls, information editing controls, and deletion controls. Users can manage whether the cloud function is put on the shelf by triggering the management control, edit the name, cloud function description information, etc. of the cloud function by triggering the information editing control, and delete the cloud function by triggering the deletion control.

[0115] Step S204, create an agent according to the fault analysis workflow orchestration requirements, multiple cloud functions, and a pre-built error code knowledge base.

[0116] Among them, the fault analysis workflow orchestration requirements refer to the design requirements of the visual process set by the operation and maintenance personnel for the workflow of the agent's intelligent analysis of fault problems, and specifically record the design requirements of each work node in each routing branch of the agent. In the embodiment of the present application, the creation of the agent can be conveniently completed through a visual agent creation tool, and the visual agent creation tool is specifically a low-code workflow orchestration tool.

[0117] Specifically, through the visual agent creation tool, multiple routing branches of the agent are created according to the fault analysis workflow orchestration requirements. Among them, there is a corresponding relationship between the routing branches and the analysis functions. Different routing branches are used to provide different analysis functions, and there is also a corresponding relationship between the cloud functions and the analysis functions. Then, there is also a corresponding relationship between the routing branches and the cloud functions.

[0118] Figure 2d Shows a schematic diagram of the creation of an agent, as Figure 2d shown, the name of the agent can be set, a specific version of the large language model can be selected as the dialogue model, additional parameters can be configured, and multiple routing branches can be set, such as the tracking analysis branch, the scenario historical error analysis branch, the analysis branch of the accessed scenarios according to the business ID, the resource analysis branch affected by the error, the SLO break line analysis branch, the error analysis branch of each accessed scenario, the error analysis branch of each accessed business, etc. Specifically, it can be analyzed based on scenario analysis (such as the error sorting data of the current day and historical data), or based on the recall ID (such as the business ID, etc.).

[0119] Next, corresponding worker nodes are created for each routing branch, and the cloud function corresponding to the routing branch is configured for the worker nodes. Moreover, an error code parsing process for the agent needs to be created, and an error code knowledge base is configured for the error code parsing process, so as to orchestrate the error code knowledge base into the agent's workflow. Before providing the data to the large language model for the final summary, the error description information of the error code is recalled and provided to the large language model, realizing retrieval-augmented generation, which can effectively enhance the diagnostic ability of the large language model for intelligent analysis of fault problems and improve the accuracy of fault analysis.

[0120] Step S205: Obtain the fault analysis instruction input by the user. The agent calls the large language model, the cloud function, and the error code knowledge base according to the fault analysis instruction to obtain the fault problem analysis result.

[0121] When the user wants to perform specific intelligent analysis of fault problems based on the error data generated by the system, the user can input a specific fault analysis instruction. Obtain the fault analysis instruction, and the agent perceives the user's analysis intention and takes actions according to the fault analysis instruction, calls the large language model, the cloud function, and the error code knowledge base to obtain the fault problem analysis result.

[0122] Specifically, the agent calls the large language model to perform intention recognition on the fault analysis instruction to obtain the user's analysis intention, and determines the target routing branch that matches the user's analysis intention; executes the target routing branch, calls the cloud function corresponding to the target routing branch, and obtains the target error sorting data required for the corresponding analysis function from the error sorting data through the cloud function; executes the error code parsing process, and determines the error description information corresponding to the target error code in the target error sorting data according to the error code knowledge base. For example, execute the error code parsing process, extract the target error code included in the target error sorting data, and then query the error code knowledge base to obtain the error description information corresponding to the target error code; then provide the target error sorting data, the error description information corresponding to the target error code, and the preset system prompt information (i.e., System Prompt) to the large language model for analysis and reasoning to obtain the fault problem analysis result. Among them, those skilled in the art can set the specific content of the preset system prompt information according to actual needs, and no specific limitation is made here. In an alternative embodiment, the specific content of the preset system prompt information can be as Figure 2e shown.

[0123] For example, when the user analysis intention obtained by the intelligent agent calling the large language model for intention recognition of the fault analysis instruction is the intention of "intelligent analysis recall ID", and the determined matching target routing branch is the "tracking analysis branch", the working node will route and navigate to the "tracking analysis branch", and then call the pre-configured cloud function corresponding to this routing branch through the large language model to obtain the target error collation data required in the current analysis process from the error collation data accessed by the intelligent agent platform; before providing the target error collation data to the large language model for the final summary, it is also necessary to execute the error code parsing process to recall the error description information according to the error code. Specifically, execute the error code parsing process, extract the target error code included in the target error collation data, and obtain the error description information corresponding to the target error code by querying the error code knowledge base; then provide the target error collation data and the error description information corresponding to the target error code to the large language model for the final inference, and give the expected preset system prompt information.

[0124] Figure 2f shows a schematic diagram of the analysis result of a fault problem, such as Figure 2f shown, the analysis result of the fault problem can include basic information, error trend analysis (including error concentration, error dispersion, etc.), impact data, analysis and troubleshooting suggestions, etc. It can perform accurate intelligent analysis of fault problems from multiple dimensions and effectively improve the fault intelligent diagnosis ability.

[0125] According to the fault problem analysis method based on the large language model provided by the embodiments of the present application, by cleaning and sorting the error data, error sorting data with a basic detailed data structure is formed, making it more convenient for analysis; the error sorting data is connected to the intelligent agent platform, and multiple cloud functions are set in the intelligent agent platform. Through different cloud functions, the intelligent extraction and assembly of the error sorting data required for different analysis functions are conveniently realized, effectively improving the data processing efficiency; through the visual intelligent agent creation tool, the creation of the intelligent agent can be conveniently completed, effectively improving the intelligent agent creation efficiency. The created intelligent agent can provide multiple analysis functions; when the user needs to perform specific intelligent analysis of the fault problem, the unified intelligent agent can adaptively call the large language model, cloud function, and error code knowledge base according to the fault analysis instruction input by the user, perform intelligent diagnostic analysis on the error sorting data, and give a multi-dimensional fault problem analysis result including the trend of the error, the characteristics of the error, the data affected by the error, etc.; this solution introduces the intelligent reasoning ability of the large language model into the field of fault analysis, and arranges the error code knowledge base into the workflow of the intelligent agent, and can conveniently recall and provide the error description information of the error code to the large language model, realizing the combination of the large language model and retrieval augmented generation, effectively enhancing the diagnostic ability of the large language model for intelligent analysis of fault problems, effectively improving the accuracy of fault analysis, greatly improving the fault analysis efficiency, effectively reducing the manual participation of the operation and maintenance personnel, and realizing accurate and efficient intelligent analysis of fault problems.

[0126] Figure 3 The structural block diagram of a fault problem analysis device based on the large language model according to an embodiment of the present application is shown, as Figure 3 shown, the device includes: a sorting module 310, a setting module 320, a creating module 330, and an analyzing module 340.

[0127] The sorting module 310 is adapted to: acquire the collected error data, and sort the error data to obtain error sorting data.

[0128] The setting module 320 is adapted to: connect the error sorting data to the intelligent agent platform, and set multiple cloud functions in the intelligent agent platform. Different cloud functions are used to obtain the target error sorting data required for different analysis functions from the error sorting data.

[0129] The creating module 330 is adapted to: create an intelligent agent according to the fault analysis workflow arrangement requirements, multiple cloud functions, and the pre-constructed error code knowledge base.

[0130] The analyzing module 340 is adapted to: acquire the fault analysis instruction input by the user, and the intelligent agent calls the large language model, cloud function, and error code knowledge base according to the fault analysis instruction to obtain the fault problem analysis result.

[0131] Furthermore, the sorting module 310 is further adapted to: obtain error data collected by the stability platform; wherein the error data includes error information, error codes, and environmental information at the time of error occurrence; perform data cleaning on the error data, and perform grouped statistics on the cleaned error data to form error sorting data; wherein the error sorting data contains error correlation information in multiple dimensions.

[0132] Furthermore, the error correlation information in multiple dimensions at least includes: scenario information, error information, error codes, error occurrence date, container instance at the time of error occurrence, availability zone at the time of error occurrence, operating environment at the time of error occurrence, application service information at the time of error occurrence, instance error count, availability zone error count, and operating environment error count.

[0133] Furthermore, the creation module 330 is further adapted to: create multiple routing branches of the intelligent agent according to the requirements of the fault analysis workflow orchestration through the visual intelligent agent creation tool; different routing branches are used to provide different analysis functions; create corresponding working nodes for each routing branch and configure the cloud function corresponding to the routing branch for the working node; create an error code parsing process for the intelligent agent and configure an error code knowledge base for the error code parsing process.

[0134] Furthermore, the device further includes: a knowledge base construction module 350; the knowledge base construction module 350 is adapted to: obtain an error code description file; wherein the error code description file records error codes and error description information corresponding to the error codes; use the knowledge base tool to perform vectorization processing on the error codes in the error code description file, and create an error code knowledge base based on the error description information corresponding to the error codes in the error code description file.

[0135] Furthermore, the analysis module 340 is further adapted to: have the intelligent agent call the large language model to perform intent recognition on the fault analysis instruction to obtain the user analysis intent, and determine the target routing branch that matches the user analysis intent; execute the target routing branch, call the cloud function corresponding to the target routing branch, and obtain the target error sorting data required for the corresponding analysis function from the error sorting data through the cloud function; execute the error code parsing process, and determine the error description information corresponding to the target error code in the target error sorting data according to the error code knowledge base; provide the target error sorting data, the error description information corresponding to the target error code, and the preset system prompt information to the large language model for analysis and reasoning to obtain the fault problem analysis result.

[0136] Furthermore, the analysis module 340 is further adapted to: execute the error code parsing process, extract the target error code included in the target error sorting data; query the error code knowledge base to obtain the error description information corresponding to the target error code.

[0137] The descriptions of the above modules refer to the corresponding descriptions in the method embodiments and will not be elaborated here.

[0138] According to the fault problem analysis device based on a large language model provided by an embodiment of the present application, by cleaning and organizing error data, error organized data with a basic detailed data structure is formed, making it more convenient for analysis; the error organized data is connected to an intelligent agent platform, and multiple cloud functions are set in the intelligent agent platform. Through different cloud functions, intelligent extraction and assembly of the error organized data required for different analysis functions are conveniently achieved, effectively improving the data processing efficiency; through a visual intelligent agent creation tool, the creation of an intelligent agent can be conveniently completed, effectively improving the intelligent agent creation efficiency, and the created intelligent agent can provide multiple analysis functions; when a user needs to perform specific intelligent analysis of a fault problem, a unified intelligent agent can adaptively call a large language model, cloud functions, and an error code knowledge base according to the fault analysis instruction input by the user, perform intelligent diagnostic analysis on the error organized data, and give a multi-dimensional fault problem analysis result including error trends, error characteristics, data affected by errors, etc.; this solution introduces the intelligent reasoning ability of the large language model into the field of fault analysis, and arranges the error code knowledge base into the workflow of the intelligent agent, enabling the error description information of the error code to be conveniently recalled and provided to the large language model, realizing the combination of the large language model and retrieval augmented generation, effectively enhancing the diagnostic ability of the large language model for intelligent analysis of fault problems, effectively improving the accuracy of fault analysis, greatly improving the fault analysis efficiency, effectively reducing the manual participation of operation and maintenance personnel, and realizing accurate and efficient intelligent analysis of fault problems.

[0139] An embodiment of the present application provides a non-volatile computer storage medium, and the computer storage medium stores at least one executable instruction or computer program, and the executable instruction or computer program can enable a processor to execute the operations corresponding to the method for analyzing a fault problem based on a large language model in any of the above method embodiments.

[0140] An embodiment of the present application provides a computer program product, and the computer program product includes at least one executable instruction or computer program, and the executable instruction or computer program can enable a processor to execute the operations corresponding to the method for analyzing a fault problem based on a large language model in any of the above method embodiments.

[0141] Figure 4 The structural schematic diagram of a computing device according to an embodiment of the present application is shown, and the specific implementation of the computing device is not limited in the specific embodiments of the present application.

[0142] As Figure 4As shown, the computing device may include: a processor 402, a communications interface 404, a memory 406, and a communication bus 408.

[0143] Among them: The processor 402, the communications interface 404, and the memory 406 communicate with each other through the communication bus 408. The communications interface 404 is used to communicate with network elements of other devices such as clients or other servers. The processor 402 is used to execute the program 410, and specifically can execute the relevant steps in the above-mentioned method embodiments for analyzing fault problems of the computing device based on the large language model.

[0144] Specifically, the program 410 may include program code, and the program code includes computer operation instructions.

[0145] The processor 402 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application. One or more processors included in the computing device may be of the same type of processor, such as one or more CPUs; or may be of different types of processors, such as one or more CPUs and one or more ASICs.

[0146] The memory 406 is used to store the program 410. The memory 406 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

[0147] The program 410 is specifically used to cause the processor 402 to execute the method for analyzing fault problems based on the large language model in any of the above method embodiments. For the specific implementation of each step in the program 410, reference may be made to the corresponding steps and units in the above-mentioned embodiments of the method for analyzing fault problems based on the large language model, which will not be elaborated here. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding process descriptions in the foregoing method embodiments, which will not be repeated here.

[0148] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. A variety of general-purpose systems may also be used in conjunction with the teachings provided herein. The structure required to construct such systems will be apparent from the above description. Additionally, the embodiments of the present application are not directed to any particular programming language. It should be understood that the content of the embodiments of the present application described herein can be implemented using a variety of programming languages, and the descriptions of specific languages above are for the purpose of disclosing the best mode of the embodiments of the present application.

[0149] In the specification provided herein, a number of specific details are set forth. However, it is understood that the embodiments of the present application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0150] Similarly, it should be understood that, in order to streamline the present disclosure and assist in understanding one or more of the various inventive aspects, in the description of the exemplary embodiments of the present application above, the various features of the embodiments of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. However, the disclosed method should not be construed as reflecting an intention that the claimed embodiments of the present application require more features than are expressly recited in each claim. Rather, as reflected in the following claims, the inventive aspects lie in less than all the features of the single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate embodiment of the present application.

[0151] Those skilled in the art will appreciate that the modules in the devices in the embodiments can be adaptively changed and disposed in one or more devices different from those of the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and in addition, they can be divided into multiple sub-modules or sub-units or sub-components. Except that at least some of such features and / or processes or units are mutually exclusive, any combination can be adopted for all the features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all the processes or units of any method or device so disclosed. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) can be replaced by an alternative feature providing the same, equivalent, or similar purpose.

[0152] In addition, those skilled in the art can understand that although some embodiments described herein include certain features included in other embodiments rather than other features, the combination of features of different embodiments means that it is within the scope of the embodiments of the present application and forms different embodiments. For example, in the following claims, any one of the claimed embodiments can be used in any combination.

[0153] Each component embodiment of the embodiments of the present application can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of the present application. The embodiments of the present application can also be implemented as a device or apparatus program (for example, a computer program and a computer program product) for executing part or all of the methods described herein. Such a program for implementing the embodiments of the present application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0154] It should be noted that the above embodiments illustrate the embodiments of the present application rather than limit the embodiments of the present application, and those skilled in the art can design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in the claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The embodiments of the present application can be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In the unit claims listing several devices, several of these devices can be embodied by the same item of hardware. The use of the words first, second, and third, etc. does not denote any order. These words can be interpreted as names.

Claims

1. A method for analyzing fault problems based on a large language model, comprising: Obtaining the collected error data, and sorting out the error data to obtain error sorted data; Connecting the error sorted data to an agent platform, and setting multiple cloud functions in the agent platform; different cloud functions are used to obtain the target error sorted data required for different analysis functions from the error sorted data; Creating an agent according to the fault analysis workflow orchestration requirements, multiple cloud functions, and a pre-constructed error code knowledge base; Obtaining a fault analysis instruction input by a user, and the agent calls a large language model, cloud functions, and the error code knowledge base according to the fault analysis instruction to obtain a fault problem analysis result.

2. The method according to claim 1, wherein the obtaining the collected error data, and sorting out the error data to obtain error sorted data further comprises: Obtaining the error data collected by the stability platform; wherein the error data includes error information, error codes, and environment information when the error occurs; Performing data cleaning on the error data, and performing grouped statistics on the cleaned error data to form the error sorted data; wherein the error sorted data contains error correlation information in multiple dimensions.

3. The method according to claim 2, wherein the error correlation information of the multiple dimensions at least includes: Scenario information, error information, error codes, error occurrence date, container instance when the error occurs, availability zone when the error occurs, running environment when the error occurs, application service information when the error occurs, instance error count, availability zone error count, and running environment error count.

4. The method according to any one of claims 1-3, wherein the creating an agent according to the fault analysis workflow orchestration requirements, multiple cloud functions, and a pre-constructed error code knowledge base further comprises: Creating multiple routing branches of the agent according to the fault analysis workflow orchestration requirements through a visual agent creation tool; Different routing branches are used to provide different analysis functions; Creating corresponding working nodes for each routing branch and configuring the cloud function corresponding to the routing branch for the working node; Creating an error code parsing process of the agent and configuring the error code knowledge base for the error code parsing process.

5. The method according to any one of claims 1-4, wherein the construction process of the error code knowledge base comprises: Obtaining an error code description file; wherein the error code description file records error codes and error description information corresponding to the error codes; Using a knowledge base tool to perform vectorization processing on the error codes in the error code description file, and creating the error code knowledge base according to the error description information corresponding to the error codes in the error code description file.

6. The method according to any one of claims 1-5, wherein the agent calling a large language model, cloud functions, and the error code knowledge base according to the fault analysis instruction to obtain a fault problem analysis result further comprises: The agent calls the large language model to perform intent recognition on the fault analysis instruction to obtain a user analysis intent, and determines a target routing branch that matches the user analysis intent; Execute the target routing branch, call the cloud function corresponding to the target routing branch, and obtain the target error sorting data required for the corresponding analysis function from the error sorting data through the cloud function; Execute the error code parsing process, and determine the error description information corresponding to the target error code in the target error sorting data according to the error code knowledge base; Provide the target error sorting data, the error description information corresponding to the target error code, and the preset system prompt information to the large language model for analysis and reasoning to obtain the analysis result of the fault problem.

7. The method according to claim 6, wherein the step of executing the error code parsing process and determining the error description information corresponding to the target error code in the target error sorting data according to the error code knowledge base further comprises: Execute the error code parsing process, and extract the target error code included in the target error sorting data; Query the error code knowledge base to obtain the error description information corresponding to the target error code.

8. A fault problem analysis device based on a large language model, comprising: A sorting module, adapted to obtain the collected error data and sort the error data to obtain error sorting data; A setting module, adapted to connect the error sorting data to an intelligent agent platform and set a plurality of cloud functions in the intelligent agent platform; different cloud functions are used to obtain the target error sorting data required for different analysis functions from the error sorting data; A creation module, adapted to create an intelligent agent according to the requirements of the fault analysis workflow arrangement, a plurality of cloud functions, and a pre-constructed error code knowledge base; An analysis module, adapted to obtain a fault analysis instruction input by a user, and the intelligent agent calls a large language model, a cloud function, and the error code knowledge base according to the fault analysis instruction to obtain an analysis result of the fault problem.

9. A computing device, comprising: A processor, a memory, a communication interface, and a communication bus, through which the processor, the memory, and the communication interface complete communication with each other; The memory is used to store at least one executable instruction, and the executable instruction causes the processor to execute the operations corresponding to the method for analyzing fault problems based on a large language model according to any one of claims 1-7.

10. A computer storage medium, in which at least one executable instruction is stored, and the executable instruction causes a processor to execute the operations corresponding to the method for analyzing fault problems based on a large language model according to any one of claims 1-7.

11. A computer program product, including at least one executable instruction, and the executable instruction causes a processor to execute the operations corresponding to the method for analyzing fault problems based on a large language model according to any one of claims 1-7.

Images