Defect processing method and apparatus, and storage medium

Abstract

Embodiments of the present application disclose a defect processing method and device and a storage medium. The method comprises: in response to input multi-modal information of a target defect, performing fusion analysis and processing on the multi-modal information by using a defect management model to obtain problem description information and a problem type of the target defect; querying a target historical defect matching the problem description information of the target defect from a defect knowledge base; obtaining a judgment result of whether the target defect meets a requirement according to a similarity between description information of each requirement in a requirement knowledge base and the problem description information of the target defect; evaluating a target repair success rate of the target defect according to the problem type of the target defect, a code context complexity associated with the target defect, and description information of the target historical defect; and transferring the target defect to a corresponding defect processing node for processing according to the target repair success rate and / or the judgment result to obtain a processing result of the target defect.

Description

Technical Field

[0001] This application relates to, but is not limited to, the field of vehicle networking technology, and in particular to a defect handling method, device, and storage medium. Background Technology

[0002] Current defect management systems primarily rely on manual operation to complete the entire process from defect (BUG) entry to repair and verification. The defect management process includes: first, testers conduct thorough testing of the software development project to identify existing defects; then, the discovered defects are submitted through the management system, initiating the processing flow. Once a project defect is submitted, it is assigned to the relevant developers for analysis, repair, and initial self-testing. Afterward, the defect is returned to the testers for final retesting to confirm whether the defect has been resolved. However, because the above defect management process requires manual intervention at each stage, the defect processing cycle is lengthy and inefficient, making it difficult to meet the high-efficiency collaboration requirements of agile development. Summary of the Invention

[0003] One objective of this invention is to provide at least one defect handling method, apparatus, and storage medium. Its advantages are that, in response to the multimodal information of the target defect, it automatically performs fusion analysis and processing using a defect management model, quickly obtains problem description information and problem type, avoids the tedious process of manually collecting, organizing, and analyzing information, saves a lot of time, and realizes intelligent analysis of multimodal bugs.

[0004] Another objective of this invention is to provide at least one defect handling method, apparatus, and storage medium. Its advantage lies in the fact that, since the defect knowledge base contains detailed descriptions of multiple historical defects, such as problem description information, problem type, cause analysis information, repair information, and historical repair success rate, historical defects that match the target defect can be quickly queried from the defect knowledge base. When handling the target defect, the cause analysis and repair information of the matching historical defects can be referenced, which facilitates developers to quickly locate the cause of the problem and formulate a repair plan, greatly shortening the defect repair time.

[0005] Another objective of this invention is to provide at least one defect handling method, apparatus, and storage medium, the advantage of which is that, based on the similarity between the description information of each requirement in the requirement knowledge base and the problem description information of the target defect, it determines whether the target defect meets the requirements, avoids misjudging the requirements as defects, or omitting defects that do not meet the requirements, and improves the accuracy of defect handling.

[0006] Another objective of this invention is to provide at least one defect handling method, apparatus, and storage medium. Its advantage lies in assessing the probability of successfully fixing a target defect based on its problem type, associated code context complexity, and description information of historical defects. Based on the target probability and / or the assessment result, the target defect is routed to the corresponding defect handling node for processing. This facilitates the rational allocation of development resources. Defects with different fixation probabilities and / or those that do not meet requirements can be routed to different processing nodes (such as human and / or AI programmers), thereby achieving intelligent and dynamic defect routing. Simultaneously, integrating AI capabilities into the defect management process enables human-machine collaboration, improving overall processing efficiency and optimizing the entire defect management workflow.

[0007] To achieve the above objectives, the technical solution of this application embodiment is implemented as follows:

[0008] In a first aspect, embodiments of this application provide a defect handling method, the method comprising:

[0009] In response to the multimodal information of the target defect, the defect management model is used to perform fusion analysis and processing on the multimodal information to obtain the problem description information and problem type of the target defect;

[0010] The defect knowledge base is used to search for target historical defects that match the problem description information of the target defect. The defect knowledge base includes multiple historical defects, and the description information of each historical defect includes at least one of the following: problem description information, problem type, cause analysis information, repair information, and repair success rate.

[0011] Based on the similarity between the description information of each requirement in the requirement knowledge base and the problem description information of the target defect, the judgment result of whether the target defect meets the requirements is obtained.

[0012] The success rate of target fixes is evaluated based on the problem type of the target defect, the code context complexity associated with the target defect, and the description information of the target historical defects.

[0013] Based on the target repair success rate and / or judgment results, the target defect is transferred to the corresponding defect processing node for processing, and the processing result of the target defect is obtained.

[0014] Secondly, embodiments of this application provide a defect processing apparatus, the apparatus comprising:

[0015] The response module is used to respond to the multimodal information of the target defect input;

[0016] The processing module is used to fuse and analyze multimodal information using a defect management model to obtain problem description information of the target defect;

[0017] The processing module is also used to query the target historical defects from the defect knowledge base that match the problem description information of the target defect; the defect knowledge base includes multiple historical defects, and the description information of each historical defect includes at least one of the following: problem description information of the historical defect, problem type, cause analysis information, repair information, and historical repair success rate;

[0018] The processing module is also used to determine whether the target defect meets the requirements based on the similarity between the description information of each requirement in the requirement knowledge base and the problem description information of the target defect.

[0019] The processing module is also used to assess the probability of successfully fixing the target defect based on the problem type of the target defect, the code context complexity associated with the target defect, and the description information of the target historical defects.

[0020] The processing module is also used to transfer the target defect to the defect processing node for processing based on the target probability and / or judgment result, so as to obtain the processing result of the target defect.

[0021] Thirdly, embodiments of this application provide a computer device including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements some or all of the steps in the method described in the first aspect above.

[0022] Fourthly, embodiments of this application provide a storage medium storing one or more computer programs, which can be executed by one or more processors to implement some or all of the steps in the method described in the first aspect above.

[0023] Fifthly, embodiments of this application provide a computer program product, including a computer program or instructions, which, when executed by a processor, implement some or all of the steps in the method described in the first aspect above.

[0024] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of this application. Attached Figure Description

[0025] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the technical solutions of this application.

[0026] Figure 1 A flowchart illustrating an optional defect handling method provided in this application embodiment. Figure 1 ;

[0027] Figure 2 A flowchart illustrating an optional defect handling method provided in this application embodiment. Figure 2 ;

[0028] Figure 3 A flowchart illustrating an optional defect handling method provided in this application embodiment. Figure 3 ;

[0029] Figure 4 A flowchart illustrating an optional defect handling method provided in this application embodiment. Figure 4 ;

[0030] Figure 5 A flowchart illustrating an optional defect handling method provided in this application embodiment. Figure 5 ;

[0031] Figure 6 A schematic diagram of the architecture of an optional defect handling method provided in an embodiment of this application;

[0032] Figure 7 A flowchart illustrating an optional defect handling method provided in this application embodiment. Figure 6 ;

[0033] Figure 8 A schematic diagram of an optional defect handling device provided in an embodiment of this application;

[0034] Figure 9 This is a schematic diagram of the hardware entity of a computer device provided in an embodiment of this application. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application are further described in detail below with reference to the accompanying drawings and embodiments. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0036] In the following description, references to "some embodiments" refer to a subset of all possible embodiments. It is understood that "some embodiments" may be the same or different subsets of all possible embodiments and may be combined with each other without conflict. The terms "first / second / third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first / second / third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used herein is for descriptive purposes only and is not intended to limit the scope of this application.

[0038] The bug management process in related technologies includes the following steps: First, testers conduct thorough testing of the software development project to identify any defects. Then, the discovered defects are submitted through the aforementioned management system, initiating the processing flow. Once a project defect is submitted, it is assigned to the relevant developers for analysis, repair, and initial self-testing. Afterward, the defect is returned to the testers for final retesting to confirm whether the defect has been resolved.

[0039] It's important to note that traditional bug management processes suffer from several critical pain points: First, the entire process, from bug submission and workflow to fix verification, relies on manual operation, resulting in lengthy processing times and low efficiency. Second, repeated communication and confirmation between testers and developers regarding submitted bugs further consumes significant time. Third, when handling assigned bugs, developers not only need to spend time understanding the bug description but also invest effort in reproducing the problem before they can begin fixing it, leading to difficulties in problem analysis and localization, and low efficiency. Fourth, many bugs in software development projects share common characteristics. Developers often encounter bugs that have already been solved by others or that they themselves have handled but forgotten due to the passage of time, forcing them to reinvest effort in bug analysis, resulting in low reuse rates of historical bug experience. Fifth, a large amount of repetitive and simple bug fixing work consumes developers' valuable time, leading to high labor costs and hindering development efficiency. Sixth, for newcomers lacking experience, studying past bugs is necessary for quick learning, but traditional processes struggle to provide an efficient learning path.

[0040] To address one or more of the aforementioned issues, embodiments of this application provide a defect handling method. This method responds to the multimodal information of a target defect, automatically performs fusion analysis using a defect management model to quickly obtain problem description information and problem type, queries a defect knowledge base for target historical defects that match the problem description information of the target defect, and determines whether the target defect meets the requirements based on the similarity between the description information of each requirement in the requirement knowledge base and the problem description information of the target defect. Based on the problem type of the target defect, the code context complexity associated with the target defect, and the description information of the target historical defects, the method assesses the target probability of successful repair of the target defect. Based on the target probability and / or the determination result, the target defect is transferred to the corresponding defect handling node for processing to obtain the processing result of the target defect. In this way, historical defects that match the target defect can be quickly retrieved from the defect knowledge base. When dealing with the target defect, the cause analysis and repair information of the matching historical defects can be referenced, which helps developers quickly locate the cause of the problem and formulate a repair plan, greatly shortening the defect repair time. At the same time, defects with different repair probabilities and / or defects that do not meet the requirements can be transferred to different processing nodes (such as human and / or AI programmers), thereby realizing intelligent dynamic routing of defects. Furthermore, the integration of artificial intelligence capabilities into the defect management process enables human-machine collaboration, improves overall processing efficiency, and optimizes the entire defect management process.

[0041] This method can be executed by the processor of a computer device. The computer device can refer to a server, laptop, tablet, desktop computer, smart TV, set-top box, mobile device (such as a mobile phone, portable video player, personal digital assistant, dedicated messaging device, portable gaming device), or any other device with data processing capabilities.

[0042] Reference Figure 1 , Figure 1 This is a schematic diagram illustrating the implementation flow of a defect handling method provided in an embodiment of this application. This method can be executed by a computer device. Here, it will be combined with... Figure 1 The steps shown are explained below.

[0043] Step 101: In response to the input multimodal information of the target defect, the multimodal information is fused and analyzed using the defect management model to obtain the problem description information of the target defect.

[0044] In this application embodiment, the target defect can be a specific problem or anomaly in software, system or product that needs to be identified, analyzed and fixed in a software testing scenario. Defects are also known as bugs or faults.

[0045] In this embodiment, modal information can be different types of information about the target defect. These different types of information have different data sources, including but not limited to text descriptions, images, videos, log information, and sensor data. Specifically, the modal information corresponding to the text description type is the text description information itself, which can be user-input feedback information. For example, if the navigation logo is stuck while connected to Wi-Fi, the text description information can provide the defect background and details. The modal information corresponding to the image type is image information, which can be a photo taken by the user regarding the target defect. The modal information corresponding to the video type is video information, which can be a video taken by the user during the entire process of the operation that caused the target defect. Image and video information can visually present the appearance and scene of the defect. Log information can be the system's record of the target defect's operational data and abnormal situations; that is, log information can record operational data and abnormal situations.

[0046] In this embodiment of the application, the multimodal information of the target defect includes two or more different modalities of information such as text description information, image information, video information, and log information.

[0047] In this embodiment, the defect management model can be a large language model used to process and analyze defect-related information. For example, the defect management model can be an intelligent model built based on machine learning algorithms such as convolutional neural networks and recurrent neural networks in deep learning, or a rule-based model designed based on traditional statistical methods and expert experience. The defect management model is used to fuse, analyze, and reason about information from different input modalities to mine more accurate problem description information. Of course, the defect management model can also output the problem type of the target defect and the operational steps to obtain the target defect.

[0048] In this embodiment of the application, the problem description information may be a clear and comprehensive description of the defect problem obtained by analyzing and processing at least one modal information of the target defect. The problem description information may also include, but is not limited to, the defect version.

[0049] In this application embodiment, the problem type can be classified by interface, logic type, underlying layer and system, etc., to obtain interface type, logic type, underlying layer type and system type, etc. Of course, the problem type can also be classified by function, performance, security and compatibility, etc., to obtain including but not limited to functional defects, performance defects, security vulnerabilities and compatibility problems.

[0050] In this embodiment, after receiving multiple modal information of the target defect, the computer device invokes the defect management model. The defect management module uses specific algorithms and rules to perform fusion analysis on the multimodal information. By integrating the characteristics of different modal information, it uncovers the key features and patterns of the target defect hidden behind the information, thereby obtaining a more comprehensive problem description of the target defect. The problem description information is then classified to determine the problem type to which it belongs. Finally, the problem description information and problem type of the target defect are output. In this way, the complementarity of multimodal information can be fully utilized to more comprehensively and accurately understand the characteristics and essence of the target defect, and obtain more accurate problem description information and problem type to which the problem description information belongs.

[0051] In some embodiments, if the multimodal information includes two or more of text description information, image information, video information, and log information, before using the defect management model to perform fusion analysis on the multimodal information to obtain the problem description information of the target defect, the multimodal information can be processed as follows: the text description information is segmented to obtain the first information after segmentation; the images in the image information and video information are image recognized to obtain the second information after image recognition; and the log information is parsed to obtain the third information after parsing. The first information, the second information, and the third information are then fused and analyzed to obtain the problem description information of the target defect.

[0052] In this embodiment, the computer device performs word segmentation processing on the text description information of the target defect, such as using Natural Language Processing (NLP) word segmentation technology, to obtain the segmented structured text, i.e., the first information. For example, the text description information of the target defect is "interface is lagging and cannot be clicked", which is decomposed into ["interface", "lagging", "cannot be clicked"]. The defect management module performs image recognition on the image or video frame of the target defect, such as target detection and abnormal region segmentation, to extract visual features or labels, i.e., the second information, such as identifying "screen flickering" or "UI element misalignment". The defect management module also performs parsing processing on the system logs of the target defect, such as error logs and performance logs, such as regular expression matching and time series analysis, to obtain key log events or error codes (the third information), such as parsing "memory usage exceeds the threshold".

[0053] Furthermore, the defect management model is used to perform one or more of the following processes on the first, second, and third information: semantic association, spatiotemporal alignment, and conflict resolution, to obtain the problem description information of the target defect. Here, for example, semantic association can be to associate "stuttering" in the text with "Central Processing Unit (CPU) load 100%" in the log; spatiotemporal alignment can be to resolve conflicts if screen flickering occurs at a certain point in the video, and the log records a GPU driver error; conflict resolution can resolve conflicts when different modal conclusions contradict each other, such as the text description information being "crash" but the log record showing no abnormalities, by adjusting the weights or confidence levels, thereby obtaining the comprehensive problem description information of the target defect, such as "the system failed to render the interface due to GPU driver timeout (log error code 0x8003) (black screen detected in the video frame), and the 'stuttering' phenomenon reported by the user is related to this."

[0054] In this way, by integrating multimodal information such as text, images, videos, and logs, the limitations of a single information source can be overcome, avoiding misjudgments and omissions due to insufficient information. At the same time, the fusion of multimodal information makes the analysis perspective richer and more comprehensive, with different modal information complementing and verifying each other, thereby obtaining more accurate problem description information.

[0055] Step 102: Query the target historical defects from the defect knowledge base that match the problem description information of the target defect.

[0056] The defect knowledge base includes multiple historical defects, and the description information of each historical defect includes at least one of the following: problem description information, problem type, cause analysis information, repair information, and repair success rate.

[0057] In this embodiment of the application, the defect knowledge base can be a database used to store and manage historical defect information. The defect knowledge base contains detailed records of each historical defect, such as problem description, problem type, cause analysis information, repair information and repair success rate, etc., to assist in the analysis and resolution of new defects.

[0058] In this embodiment of the application, the target historical defect may be a past defect record in the defect knowledge base that matches or is similar to the target defect. The descriptive information of the target historical defect can be used to provide reference analysis and repair solutions for the resolution of the target defect.

[0059] In this embodiment, the descriptive information of the target historical defect includes, but is not limited to, the problem description information, problem type, root cause analysis information, repair information, and repair success rate of the historical defect. Specifically, the problem description information can be a detailed description of the historical defect; the problem type can be the type of problem to which the historical defect belongs; the root cause analysis information can be an analysis of the root cause of the historical defect, such as code logic errors, resource leaks, configuration errors, etc.; the repair information can be solutions to the historical defect, such as code modification, configuration adjustment, version rollback, etc.; and the repair success rate can be the effective proportion of cases in which the repair solution has been successfully repaired in similar historical cases.

[0060] In this embodiment, the computer device compares the problem description information of the current target defect with the description information of each historical defect in the defect knowledge base to find a matching target historical defect and obtain its description information. Alternatively, keywords can be extracted from the problem description information of the target defect, and the target historical defect matching the keywords can be retrieved from the defect knowledge base. In this way, information from target historical defects, including problem type, cause analysis, repair information, and historical repair success rate, can be used to provide reference and guidance for handling the current target defect. This facilitates developers in quickly locating the cause of the problem and developing a repair plan, significantly shortening the defect repair time, improving the efficiency and accuracy of defect handling, and avoiding duplication of effort and errors. Furthermore, through this method, the defect knowledge base can continuously accumulate and utilize historical knowledge, achieving continuous improvement and optimization of defect management.

[0061] Step 103: Based on the similarity between the description information of each requirement in the requirement knowledge base and the problem description information of the target defect, obtain the judgment result of whether the target defect meets the requirements.

[0062] In this embodiment of the application, the requirement knowledge base can be a database used to store and manage system requirements. The requirement knowledge base contains detailed description information for each requirement, such as requirement details, functional points, priority, acceptance criteria, inputs and outputs, constraints, and other types of information, which are used to guide development and testing.

[0063] In this embodiment of the application, the judgment result includes the result that the target defect meets the design requirements and the result that the target defect does not meet the design requirements.

[0064] In this embodiment, the computer device searches a requirements knowledge base for requirements that match the problem description information of the target defect. Then, it can utilize natural language processing, machine learning methods such as text embedding, cosine similarity, edit distance algorithms, or Jaccard coefficients, or a large language model, to calculate the similarity between the problem description information of the target defect and the description information of each matching requirement. Further, based on the calculated similarity and combined with a pre-set similarity threshold or evaluation rule, it is determined whether the target defect meets a certain requirement. For example, if the calculated cosine similarity is greater than or equal to the similarity threshold, it indicates that the problem description information of the target defect and the description information of the matching requirement are semantically similar, thus obtaining a judgment result that the target defect meets the requirement design; if the calculated cosine similarity is less than the similarity threshold, it is considered that the two are significantly different, thus obtaining a judgment result that the target defect does not meet the requirement design. In this way, similarity can accurately determine whether defects affect the fulfillment of requirements, avoiding the bias that may occur by relying solely on subjective judgment. Furthermore, by assessing the degree of impact of defects on requirements, an objective and quantitative method is provided, thereby more accurately grasping whether a product or system meets expected requirements. This avoids misjudging situations that meet requirements as defects, or omitting defects that do not meet requirements, thus improving the accuracy of defect handling.

[0065] Step 104: Based on the problem type of the target defect, the code context complexity associated with the target defect, and the description information of the target historical defects, assess the success rate of the target defect's remediation.

[0066] In this embodiment, the code context complexity associated with the target defect can be understood as the complexity of the code region where the target defect resides when the target defect is associated with code. Code context complexity includes, but is not limited to, the structural complexity, logical complexity, and code size. Structural complexity includes, but is not limited to, nesting levels and module call relationships; logical complexity includes, but is not limited to, the complexity of conditional statements and loop statements; and code size can be the code length. It should be noted that the more complex the code context, the higher the code context complexity, and the more difficult it is to understand and fix the defect.

[0067] In this embodiment, the target repair success rate represents the probability of successfully repairing a target defect when the artificial intelligence (AI) processing node is used. Here, the target repair success rate can be obtained by evaluating the target defect based on the problem type of the target defect, the code context complexity associated with the target defect, and the description information of the target's historical defects.

[0068] Step 105: Based on the target repair success rate and / or judgment result, transfer the target defect to the corresponding defect processing node for processing to obtain the processing result of the target defect.

[0069] In this embodiment of the application, based on the target repair success rate and / or judgment result, the target defect is transferred to the corresponding defect processing node for processing to obtain the processing result of the target defect. This can include the following methods:

[0070] The first approach is to determine if the judgment result indicates that the target defect meets the requirements. In this case, the target defect can be directly transferred to the test processing node. Since the current processing node may also be the test processing node, a prompt message can be generated to remind the tester to close the processing information of the target defect. Based on the prompt message, the processing result of the target defect can be obtained.

[0071] The second approach is to transfer the target defect to the AI ​​processing node for processing if the target repair success rate meets the probability condition corresponding to the problem type of the target defect, thereby obtaining the processing result of the target defect.

[0072] The third approach is to transfer the target defect to the development processing node for processing if the judgment result indicates that the target defect does not meet the requirements and the success rate of target repair does not meet the probability conditions corresponding to the problem type of the target defect.

[0073] In this way, defects with different repair probabilities and / or defects that do not meet the requirements can be transferred to different processing nodes, thereby realizing intelligent dynamic diversion of defects; at the same time, the capabilities of artificial intelligence are integrated into the defect management process to achieve human-machine collaboration, improve overall processing efficiency, and optimize the entire defect management process.

[0074] In some embodiments, step 102, which involves querying a target historical defect from a defect knowledge base that matches the problem description information of the target defect, is combined with... Figure 2 Please provide an explanation.

[0075] Step 201: Based on the similarity between the description information of each historical defect in the defect knowledge base and the problem description information of the target defect, obtain the matching degree between each historical defect and the target defect.

[0076] Step 202: Filter out historical defects from the defect knowledge base that have a matching degree that reaches the preset matching degree threshold, and identify them as target historical defects.

[0077] In this embodiment, the computer device can utilize natural language processing or machine learning methods such as text embedding, cosine similarity, edit distance algorithm, or Jaccard coefficient to calculate the similarity between the description information of each historical defect in the defect knowledge base and the problem description information of the target defect, thereby determining the matching degree between each historical defect and the target defect. Further, based on the obtained matching degree and combined with a preset matching degree threshold, historical defects are filtered to obtain historical defects with a matching degree greater than or equal to the preset matching degree threshold. These historical defects with a matching degree greater than or equal to the preset matching degree threshold are considered as target historical defects that have reference significance for the target defect.

[0078] Thus, when encountering a target defect, by quickly calculating the matching degree with historical defects, the most relevant target historical defects can be screened from a large number of historical defects. Since the screened target historical defects and the target defect have a high degree of similarity in problem description, the processing methods of the target historical defects can be referenced, eliminating the need to start from scratch in analysis and exploration, thereby greatly shortening the defect processing time and improving work efficiency. At the same time, the root cause and solution of the target defect can be more accurately determined, reducing the situation of repair failure or repeated repair due to incorrect judgment, and improving the accuracy of defect processing.

[0079] In some embodiments, step 102, which involves querying a target historical defect from a defect knowledge base that matches the problem description information of the target defect, is combined with... Figure 3 Please provide an explanation.

[0080] Step 301: Query multiple first historical defects corresponding to the problem type of the target defect from the defect knowledge base.

[0081] In this embodiment of the application, the first historical defect may be a historical defect in the defect knowledge base that corresponds to the problem type of the target defect.

[0082] Understandably, the defect knowledge base stores a large number of historical defects of different problem types. After obtaining the problem type of the target defect, the defect knowledge base can be queried using the problem type as an index to quickly locate multiple historical defects that are the same or similar to the target defect in terms of problem type, thereby obtaining multiple first historical defects.

[0083] In this way, the initial screening of problem types significantly reduces the number of historical defects that require further analysis, narrows the search space, and improves the efficiency of information retrieval.

[0084] Step 302: Based on the matching degree between the description information of each first historical defect and the problem description information of the target defect, select the first historical defects whose matching degree reaches the preset matching degree threshold from multiple first historical defects, and determine them as the target historical defects.

[0085] Understandably, after obtaining multiple first historical defects corresponding to the problem type of the target defect, the similarity between the description information of each first historical defect and the problem description information of the target defect can be calculated to determine the matching degree between each first historical defect and the target defect. Furthermore, based on the obtained matching degree and a preset matching degree threshold, the multiple first historical defects are filtered to obtain those with a matching degree greater than or equal to the preset matching degree threshold. These first historical defects with a matching degree greater than or equal to the preset matching degree threshold are considered as target historical defects that are of reference value to the target defect. It should be noted that the higher the matching degree, the greater the probability that the defects are the same. Thus, by combining matching degree filtering, the most valuable target historical defects can be quickly located. Developers or relevant personnel can quickly obtain valuable information without having to search through a large number of irrelevant historical records, thereby saving time, speeding up defect processing, and improving overall work efficiency.

[0086] As described above, the embodiments of this application employ a dual screening mechanism based on problem type and description information matching degree to ensure the similarity between the target historical defect and the target defect in multiple key dimensions. Processing these target historical defects allows for a more accurate determination of the cause of the target defect, its potential impact, and appropriate solutions. This reduces the risk of misjudgment due to referencing irrelevant or low-similarity historical cases, thereby improving the accuracy of defect processing.

[0087] In some embodiments, step 104, which assesses the success rate of target remediation based on the problem type of the target defect, the code context complexity associated with the target defect, and the descriptive information of the target historical defects, is combined with... Figure 4 Please provide an explanation.

[0088] Step 401: Determine the model capability threshold of the defect management model corresponding to the problem type of the target defect.

[0089] Different problem types correspond to different model capability thresholds.

[0090] In this embodiment, the model capability threshold is used to characterize the current capability threshold of the large language model when processing defects of different problem types. It should be noted that different problem types have different capability requirements for the large language model due to differences in complexity, scope of impact, and processing difficulty; therefore, different problem types correspond to different model capability thresholds. For example, for security vulnerability issues, which generally require manual processing, the model capability threshold can be set relatively low; while for relatively simple user interface (UI) issues, such as simple syntax errors or text errors, the model capability threshold may be set relatively high.

[0091] Step 402: Obtain the code context complexity associated with the target defect, the historical repair success rate of the target historical defect, and the model capability threshold corresponding to the problem type of the target defect, and their respective weights.

[0092] In this embodiment, weights are assigned to three dimensions of data: the code context complexity associated with the target defect, the historical repair success rate of the target historical defect, and the model capability threshold corresponding to the problem type of the target defect. These weights reflect their relative importance in the evaluation process. The weights can be determined based on historical experience, expert knowledge, or automatically learned through machine learning algorithms. For example, if code context complexity is considered to have the greatest impact on the repair success rate, it can be assigned a higher weight.

[0093] Understandably, after obtaining the problem description information of the target defect, the computer device can call the code associated with the target defect based on the problem description information. By analyzing the code associated with the target defect, the context complexity of the code associated with the target defect can be obtained. Furthermore, the weights corresponding to the code context complexity associated with the target defect, the historical repair success rate of the target historical defect, and the model capability threshold corresponding to the problem type of the target defect can be obtained.

[0094] Step 403: Based on multiple weights, the code context complexity associated with the target defect, the historical repair success rate of the target historical defect, and the model capability threshold corresponding to the problem type of the target defect, a reinforcement learning algorithm is used to evaluate the success probability of repairing the target defect and obtain the target repair success rate.

[0095] In this embodiment of the application, reinforcement learning algorithms are used to evaluate the probability of successful repair when the AI ​​processing node automatically repairs the target defect. Reinforcement learning algorithms include policy gradient algorithm, Q-Leaning, DQN, PPO and / or A3C.

[0096] Understandably, after obtaining the weights corresponding to the code context complexity associated with the target defect, the historical repair success rate of the target historical defect, and the model capability threshold corresponding to the problem type of the target defect, the computer device uses a reinforcement learning algorithm to evaluate the probability of successfully repairing the target defect by combining multiple weights, the code context complexity associated with the target defect, the historical repair success rate of the target historical defect, and the model capability threshold corresponding to the problem type of the target defect. This yields the probability of successful repair when the AI ​​processing node automatically repairs the target defect.

[0097] Thus, by comprehensively considering multiple key dimensions of data, such as code context complexity, historical repair success rate, and model capability threshold, and assigning corresponding weights to each dimension, it is possible to comprehensively and meticulously reflect the various conditions affecting the success of target defect repair. Reinforcement learning algorithms can effectively integrate and analyze this multi-dimensional data, avoiding the one-sidedness of relying on a single factor for evaluation, thereby significantly improving the accuracy of assessing the probability of target defect repair success through AI processing nodes. It should be noted that referring to the historical repair success rate of the target defect allows for full utilization of past experience in handling similar defects. If the historical repair success rate for similar defects is high, the expectation of successful repair of the current target defect will be correspondingly increased; conversely, if the historical repair success rate is low, the difficulty and risk of repair will be assessed more cautiously.

[0098] Furthermore, the importance of each key dimension of data may vary depending on the type of defect or the project scenario. Reinforcement learning algorithms can adaptively adjust the weights of each factor based on the constantly changing environment and data. For example, in projects with extremely high code quality requirements, the weight of code context complexity may automatically increase during the learning process; while in scenarios that emphasize the accumulation of historical experience, the weight of historical fix success rate may be increased. This ability to adaptively adjust weights allows the evaluation model to better adapt to various complex and changing realities, improving the flexibility and effectiveness of the evaluation.

[0099] In some embodiments, step 105, based on the target repair success rate, transfers the target defect to the corresponding defect processing node for processing to obtain the processing result of the target defect. Figure 5 Please provide an explanation.

[0100] Step 501: Based on the description information of the target's historical defects, generate the target defect analysis results.

[0101] In this embodiment, after identifying a target historical defect, the computer device analyzes and processes the acquired descriptive information of the target historical defect. Through sorting, comparing, and summarizing the descriptive information of the target historical defect, and combining it with the characteristics of the current target defect and known information, a target analysis result regarding the target defect is derived. It should be noted that the target analysis result may cover multiple aspects, such as analogies and inferences based on the cause analysis of historical defects to speculate on the causes of the target defect; predictions of the possible impact range of the target defect based on its impact; and suggestions on the repair direction and measurement of the target defect based on its repair information. Thus, by referring to the descriptive information of the target historical defect, target analysis results are quickly generated, reducing manual analysis time.

[0102] In one example, if a target historical defect was caused by a performance issue due to improper optimization of the database query statement and has been successfully fixed, then for a target defect with a similar performance issue, the target analysis results may indicate that the problem may also lie in the database query part, and suggest checking the relevant query statements and referring to the previous optimization methods for fixing.

[0103] Step 502: If the target repair success rate meets the probability condition corresponding to the problem type of the target defect, the target defect is transferred to the AI ​​processing node.

[0104] Among them, the defect processing node includes the AI ​​processing node.

[0105] In this embodiment, the probability condition for the target repair success rate to satisfy the problem type of the target defect includes: the target repair success rate is greater than or equal to the probability threshold corresponding to the problem type of the target defect, such as 95%. For example, the problem type of the target defect can be a UI type, and the target repair success rate of the target defect is 90%. The target repair success rate of 90% is equal to the probability threshold corresponding to the UI type. In this case, it indicates that the target repair success rate satisfies the probability condition corresponding to the problem type of the target defect. Furthermore, the target defect can be transferred to the AI ​​processing node.

[0106] In this way, once the conditions are met, the code is automatically generated at the AI ​​processing node, which is faster than writing code manually. Automated testing can also provide rapid feedback, greatly shortening the defect handling cycle.

[0107] Step 503: Through AI processing nodes, based on the description information of the target defect and the target analysis results, call the code context associated with the target defect to generate the target repair code for the target defect;

[0108] Step 504: Perform automated tests on the target fix code and generate a test report on the target defect;

[0109] The test report includes at least one of the following: test results, location of the fix code, and parameters of the fix code.

[0110] Step 505: If the test results indicate that the target repair code has been successfully tested, submit the target repair code and update the status of the target defect from the abnormal state to the normal state.

[0111] The processing results include the status of the target fix code and the target defect.

[0112] In this embodiment, after the target defect is transferred to the AI ​​processing node, the AI ​​processing node, based on the description information of the target defect and the target analysis results, calls the code context associated with the target defect and automatically generates target repair code for the target defect. Further, the AI ​​processing node performs automated testing on the target repair code and generates a test report for the target defect. This test report includes the test results, the location of the repair code, and one or more parameters of the repair code. Then, if the test results indicate that the target repair code has been successfully tested, the target repair code is submitted to the code repository, and the status of the target defect is updated from an abnormal state to a normal state, thereby completing the status update of the target defect.

[0113] In this way, target analysis results are generated based on historical defect information, and past experience is used to judge the causes and impacts of defects. The AI ​​processing node generates repair code based on the target analysis results and related information, reducing human judgment errors and coding errors, and improving repair accuracy. At the same time, by introducing AI processing nodes to participate in defect repair, intelligent operation from analysis to repair is realized, reducing reliance on human experience and skills, improving the level of intelligence, reducing manual workload, improving resource utilization efficiency, and reducing enterprise operating costs.

[0114] In some embodiments, step 105 transfers the target defect to the corresponding defect processing node for processing based on the target repair success rate and the judgment result, and obtains the processing result of the target defect. This process is described below.

[0115] If the target probability does not meet the probability conditions corresponding to the problem type of the target defect, and the judgment result does not meet the requirements, the target defect will be transferred to the development processing node. The development processing node will then repair and automate the target defect based on the description information of the target defect and the target analysis results. After the test is completed, the target defect will be associated in the code commit record to obtain the processing result of the target defect.

[0116] In this embodiment, when the target repair success rate is less than the probability threshold corresponding to the problem type of the target defect, it is determined that the target repair success rate does not meet the probability condition corresponding to the problem type of the target defect, and the judgment result does not meet the requirements. At this time, the target defect can be transferred to the development processing node. The development processing node is usually managed by developers or related technical personnel who possess professional programming and technical capabilities and are able to conduct in-depth analysis and repair of complex or difficult-to-handle defects.

[0117] Then, upon receiving the target defect, the development processing node formulates a corresponding remediation strategy based on the existing problem description information and the generated target analysis results, and repairs the context code associated with the target defect to obtain the target repair code. After the repair is completed, automated testing is performed on the target repair code to verify whether the repair was successful.

[0118] Furthermore, after automated testing concludes, if the test results indicate that the target fix code has been successfully tested, the target fix code is submitted to the code repository, and the target defect is associated with the code commit record. This facilitates subsequent tracing and querying of the defect fixing process. In this way, the final result of the target defect processing is obtained, which may include the fixed code, test reports, and updates to the defect status.

[0119] In this way, defects with low fix success rates and those that do not meet requirements are routed to the development stage, ensuring that these complex or critical defects receive focused attention and handling from professionals. Developers, leveraging their expertise and experience, can analyze the target defects more deeply based on the provided target defect analysis results, improving the success rate of defect fixes. Simultaneously, this reduces communication costs and processing delays, contributing to improved team collaboration efficiency.

[0120] In some embodiments, step 105 transfers the target defect to the corresponding defect processing node for processing based on the judgment result, and obtains the processing result of the target defect, which is explained through the following process.

[0121] If the judgment result indicates that the target defect meets the requirements, the target defect is transferred to the test processing node; the defect processing node includes the test processing node; in response to the test processing node's closing operation on the target defect, the processing result of the target defect is obtained.

[0122] In this embodiment, if the judgment result indicates that the target defect meets the requirements, the target defect can be transferred to the test processing node. The test processing node is a link in the entire defect handling process, specifically responsible for performing relevant test operations on the defect to further confirm its status and impact. If the target defect meets the requirements, the test processing node performs a closure operation on the target defect, obtaining the processing result. Of course, if the judgment result indicates that the target defect does not meet the requirements, such as if the description information of the target defect contradicts the definition of the requirement description information, a requirement conflict prompt message and a comparison report can be generated for testers to analyze and process.

[0123] In this way, by determining whether the target defect meets the requirements, defects that meet the requirements are directly transferred to the test processing node, and the processing result is obtained after the test processing node is closed. This avoids the incorrect allocation of these defects to more complex processing links (such as development processing nodes), thereby rationally allocating processing resources and improving the efficiency of the entire defect handling process.

[0124] In some embodiments, step 105 transfers the target defect to the corresponding defect processing node for processing based on the target repair success rate and / or judgment result. After obtaining the processing result of the target defect, the following process can also be executed.

[0125] In response to the retesting of the target defect, if the retesting results indicate that the target defect has been successfully tested, the problem description information, problem type, target analysis results, target repair code, and target repair success rate of the target defect are added to the defect knowledge base.

[0126] In this embodiment, in response to the retesting of the target defect and the submission of target repair code, a retesting result is obtained. If the retesting result indicates successful testing, the record relationship of the target defect can be established. Furthermore, the problem description information, problem type, target analysis result, and target repair code of the target defect can be added to the defect knowledge base as repair information and target repair success rate, thereby expanding the defect knowledge base. Simultaneously, using the data from the expanded defect knowledge base, the defect management model and the programmer model used by the AI ​​processing node are continuously trained and optimized, improving their problem analysis and code generation capabilities.

[0127] Below, refer to Figure 6 and Figure 7 As shown, the defect handling method provided in the embodiments of this application is illustrated in an feasible scenario.

[0128] Step 701: Testers submit bugs.

[0129] Here, refer to Figure 6As shown, for user interaction, testers enter the description of the bug (corresponding to the multimodal information mentioned above) through the bug entry interface 601. The bug description can be multimodal description information, such as text description, image, video, and log information.

[0130] Step 702, BUG preprocessing.

[0131] Here, continue to refer to Figure 6 The description of the bug is preprocessed by the preprocessing module 602, such as text segmentation, image recognition, and log parsing.

[0132] Step 703: AI analysis, requirement matching, and retrieval of the BUG knowledge base.

[0133] Here, continue to refer to Figure 6 For the AI ​​core, the preprocessed data enters the AI ​​analysis module 603 (corresponding to the aforementioned defect management module). The AI ​​analysis module 603 interacts with the requirement knowledge base 604 and the BUG knowledge base 605 (corresponding to the aforementioned defect database). Specifically, the AI ​​analysis module 603 performs fusion processing on the BUG descriptions to obtain detailed BUG description information. Semantic understanding is then performed on this detailed BUG description information, and it is matched with the description information of each requirement in the requirement knowledge base 604 to obtain a requirement matching result. If the requirement matching result indicates that the BUG meets the requirements, step 704 is executed; otherwise, step 705 is executed. Simultaneously, the historical BUG knowledge base 605 is searched to find similar historical defects and their solutions, as well as the AI ​​analysis results for the BUGs.

[0134] Step 704: Directly disable the bug.

[0135] Here, if the bug description matches the requirements, the tester will be prompted to close the bug. If the bug description contradicts the requirements definition, a "requirement conflict" message will be automatically displayed, and a comparison report (consistency check) will be generated.

[0136] Step 705: Assess the probability of AI repair.

[0137] Here, we construct an evaluation index system, which includes, but is not limited to, the type of bug (UI interface, logic, performance, etc.), the complexity of the code context associated with the bug, the success rate of fixing historical bugs similar to the bug, and the current capability threshold of the AI ​​model.

[0138] Here, continue to refer to Figure 6 If the requirement matching result indicates that the bug does not meet the requirements, reinforcement learning is used to dynamically adjust the weights based on factors such as the problem type, historical data, and AI model capabilities to evaluate the success probability of AI automatically fixing the bug.

[0139] Here, after obtaining the success probability of AI automatically fixing the bug, the bug is divided according to the success probability. If the success probability is less than the probability threshold, step 706 is executed; if the success probability is greater than the probability threshold, step 710 is executed.

[0140] Step 706: Manual analysis and repair.

[0141] Here, continue to refer to Figure 6 For complex bugs that require manual fixing (607), the AI ​​analysis results (possible causes, similar bugs, and fix suggestions) can be provided to the backend developers (corresponding to the development processing nodes mentioned above). The developers can then analyze and fix the bugs by referring to the AI ​​analysis results, and the fixed code can be automated for testing (611).

[0142] In one feasible scenario, a tester submits a defect causing an intermittent crash, along with logs and video. The AI ​​system analyzes the logs, searches the bug knowledge base, and provides possible causes and links to similar historical bugs to help developers quickly locate the problem.

[0143] Step 707: Submit the code and mark the bug.

[0144] Here, continue to refer to Figure 6 After developers complete automated testing of the bug, if the self-test result is that the test passes, they submit code 612 and associate it with the bug number.

[0145] Step 708: Testers conduct retesting.

[0146] Here, the tester retests the submitted code (which can be submitted by the AI ​​programmer or the developer) 613. If the retest fails, the process returns to step 706. If the retest passes, the process proceeds to step 709.

[0147] Step 709: Close the bug and update the bug knowledge base.

[0148] Here, continue to refer to Figure 6 Once the testers pass the retest, the bug and related information (description, solution, code, etc.) will be automatically updated to bug knowledge base 605. Using the data in the bug knowledge base, the large model and AI programmer model will be continuously trained and optimized to improve their problem analysis and code generation capabilities. At the same time, bug 614 will be closed.

[0149] Step 710: AI programmer fixes.

[0150] Here, continue to refer to Figure 6For simple bugs, AI programmers can fix them (608). In this case, the AI ​​analysis results can be provided to the backend AI programmers (corresponding to the AI ​​processing node mentioned above). Based on the AI ​​analysis results, such as the bug description and the associated code context, the AI ​​programmers automatically generate fix code. The AI ​​programmers then perform automated testing on the fix code and generate a self-test report (609).

[0151] In one feasible scenario, a tester submits a defect indicating a UI text display error, along with a screenshot. The defect management model identifies the problem type, searches the defect knowledge base to find similar issues, assesses a high probability of fix, and directly forwards it to an AI processing node, such as an AI programmer. The AI ​​programmer automatically modifies the code, generates a self-test report, and submits it. After verification by the tester, the defect is closed.

[0152] Step 711: The AI ​​automatically tests and generates a report, then submits the code.

[0153] Here, continue to refer to Figure 6 The AI ​​programmer performs automated testing on the repair code and generates a self-test report. If the AI ​​programmer's self-test results indicate that the test is successful, the AI ​​programmer submits code 610 and repeats steps 708, 706, 707, and / or step 709.

[0154] As described above, this application's embodiments utilize the semantic understanding and knowledge reasoning capabilities of large-scale models to achieve intelligent analysis, requirement matching, and knowledge base retrieval of multimodal bugs. Thus, leveraging AI's knowledge reasoning and historical experience reduces human error and improves the quality of bug fixes. AI programmers automatically generate code and execute tests, automating the fixing of some bugs, reducing manual processing, accelerating bug flow, shortening the bug fixing cycle, improving efficiency, and reducing costs. It also reduces manual fixing of repetitive bugs, lowering labor costs. Based on factors such as problem type and historical data, it assesses the success probability of AI-automated bug fixing, achieving intelligent dynamic routing, reducing communication costs, and improving collaboration efficiency between testing and development. Integrating AI capabilities into the bug management process enables human-machine collaboration, improving overall efficiency. The system automatically extracts information from the bug handling process, automatically builds and updates the bug knowledge base, achieving knowledge accumulation and reuse, facilitating team members' learning and experience reuse. Newcomers can quickly get started and begin development.

[0155] Reference Figure 8 , Figure 8 This is a schematic diagram of an optional defect handling device provided in an embodiment of this application. The defect handling device 8 includes:

[0156] Response module 801 is used to respond to the multimodal information of the input target defect;

[0157] Processing module 802 is used to perform fusion analysis on multimodal information using a defect management model to obtain problem description information of the target defect;

[0158] The processing module 802 is also used to query the target historical defect from the defect knowledge base that matches the problem description information of the target defect; the defect knowledge base includes multiple historical defects, and the description information of each historical defect includes at least one of the following: problem description information of the historical defect, problem type, cause analysis information, repair information and historical repair success rate;

[0159] The processing module 802 is also used to obtain a judgment result on whether the target defect meets the requirements based on the similarity between the description information of each requirement in the requirement knowledge base and the problem description information of the target defect.

[0160] The processing module 802 is also used to assess the probability of successfully fixing the target defect based on the problem type of the target defect, the code context complexity associated with the target defect, and the description information of the target historical defects.

[0161] The processing module 802 is also used to transfer the target defect to the defect processing node for processing based on the target probability and / or judgment result, so as to obtain the processing result of the target defect.

[0162] This application provides a computer device including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements some or all of the steps in the above-described method.

[0163] This application provides a storage medium that stores one or more computer programs, which can be executed by one or more processors to implement some or all of the steps in the above-described method. The storage medium can be transient or non-transient.

[0164] This application provides a computer program including computer-readable code, wherein when the computer-readable code is executed in a computer device, a processor in the computer device performs some or all of the steps in the above-described method.

[0165] This application provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program. When the computer program is read and executed by a computer, it implements some or all of the steps in the above-described method. This computer program product can be implemented specifically through hardware, software, or a combination thereof. In some embodiments, the computer program product is specifically embodied as a computer storage medium; in other embodiments, the computer program product is specifically embodied as a software product, such as a software development kit (SDK), etc.

[0166] It should be noted that the descriptions of the various embodiments above tend to emphasize the differences between them, while their similarities or commonalities can be referred to interchangeably. The descriptions of the above embodiments of the device, storage medium, computer program, and computer program product are similar to the descriptions of the above method embodiments and have similar beneficial effects. For technical details not disclosed in the embodiments of the device, storage medium, computer program, and computer program product of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0167] Figure 9 This application provides a hardware entity diagram of a computer device as an embodiment of the present application, such as... Figure 9 As shown, the hardware entity of the computer device 9 includes a processor 901 and a memory 902, wherein the memory 902 stores a computer program that can run on the processor 901, and the processor 901 executes the program to perform the following steps.

[0168] In response to the multimodal information of the target defect, the defect management model is used to perform fusion analysis and processing on the multimodal information to obtain the problem description information and problem type of the target defect;

[0169] The defect knowledge base is used to search for target historical defects that match the problem description information of the target defect. The defect knowledge base includes multiple historical defects, and the description information of each historical defect includes at least one of the following: problem description information, problem type, cause analysis information, repair information, and repair success rate.

[0170] Based on the similarity between the description information of each requirement in the requirement knowledge base and the problem description information of the target defect, the judgment result of whether the target defect meets the requirements is obtained.

[0171] The success rate of target fixes is evaluated based on the problem type of the target defect, the code context complexity associated with the target defect, and the description information of the target historical defects.

[0172] Based on the target repair success rate and / or judgment results, the target defect is transferred to the corresponding defect processing node for processing, and the processing result of the target defect is obtained.

[0173] The memory 902 stores computer programs that can run on the processor. The memory 902 is configured to store instructions and applications that can be executed by the processor 901. It can also cache data to be processed or already processed by the processor 901 and various modules in the computer device 9 (e.g., image data, audio data, voice communication data and video communication data). It can be implemented by flash memory or random access memory (RAM).

[0174] When the processor 901 executes the program, it implements the steps of the method in any of the above embodiments. The processor 901 generally controls the overall operation of the computer device 9.

[0175] This application provides a computer storage medium that stores one or more programs, which can be executed by one or more processors to implement the steps of the methods described in any of the above embodiments.

[0176] It should be noted that the descriptions of the storage medium and device embodiments above are similar to the descriptions of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the storage medium and device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0177] The aforementioned processor can be at least one of the following: Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), Central Processing Unit (CPU), Controller, Microcontroller, and Microprocessor. It is understood that other electronic devices can also implement the functions of the aforementioned processor, and this application does not specifically limit the specific implementation.

[0178] The aforementioned computer storage media / memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM), etc.; or it can be various terminals that include one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0179] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above steps / processes do not imply a sequential order of execution; the execution order of each step / process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above embodiments of this application are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0180] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0181] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0182] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0183] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0184] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0185] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence or the part that contributes to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an in-vehicle terminal (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, magnetic disks, or optical disks.

[0186] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A defect handling method, characterized in that, The method includes: In response to the input multimodal information of the target defect, the defect management model is used to perform fusion analysis on the multimodal information to obtain the problem description information and problem type of the target defect; The system queries a defect knowledge base to find a target historical defect that matches the problem description information of the target defect. The defect knowledge base includes multiple historical defects, and the description information of each historical defect includes at least one of the following: the problem description information, problem type, cause analysis information, repair information, and repair success rate of the historical defect. Based on the similarity between the description information of each requirement in the requirement knowledge base and the problem description information of the target defect, a judgment result is obtained on whether the target defect meets the requirements. The success rate of target remediation is evaluated based on the problem type of the target defect, the code context complexity associated with the target defect, and the description information of the target historical defects. Based on the target repair success rate and / or the judgment result, the target defect is transferred to the corresponding defect processing node for processing to obtain the processing result of the target defect.

2. The method according to claim 1, characterized in that, The step of querying the target historical defect from the defect knowledge base that matches the problem description information of the target defect includes: The matching degree between each historical defect and the target defect is obtained based on the similarity between the description information of each historical defect in the defect knowledge base and the problem description information of the target defect. Historical defects that match a preset matching threshold are selected from the defect knowledge base and identified as the target historical defects.

3. The method according to claim 1, characterized in that, The step of querying the target historical defect from the defect knowledge base that matches the problem description information of the target defect includes: Retrieve multiple first historical defects from the defect knowledge base that correspond to the problem type of the target defect; Based on the matching degree between the description information of each first historical defect and the problem description information of the target defect, the first historical defect whose matching degree reaches a preset matching degree threshold is selected from the plurality of first historical defects and determined as the target historical defect.

4. The method according to any one of claims 1 to 3, characterized in that, The step of evaluating the target repair success rate of the target defect based on the problem type of the target defect, the code context complexity associated with the target defect, and the description information of the target historical defects includes: Determine the model capability threshold of the defect management model corresponding to the problem type of the target defect; wherein, different problem types correspond to different model capability thresholds; Obtain the code context complexity associated with the target defect, the historical repair success rate of the target historical defect, and the model capability threshold corresponding to the problem type of the target defect, and their respective weights; Based on multiple weights, the code context complexity associated with the target defect, the historical repair success rate of the target historical defect, and the model capability threshold corresponding to the problem type of the target defect, a reinforcement learning algorithm is used to evaluate the success probability of repairing the target defect, thereby obtaining the target repair success rate.

5. The method according to any one of claims 1 to 3, characterized in that, The step of transferring the target defect to the corresponding defect processing node for processing based on the target repair success rate, and obtaining the processing result of the target defect, includes: Based on the description information of the target's historical defects, the target analysis results of the target defects are generated; If the target repair success rate meets the probability condition corresponding to the problem type of the target defect, the target defect is transferred to the artificial intelligence processing node, and the defect processing node includes the artificial intelligence processing node. Through the artificial intelligence processing node, based on the description information of the target defect and the target analysis results, the code context associated with the target defect is invoked to generate the target repair code for the target defect; Automated testing is performed on the target fix code to generate a test report of the target defect; the test report includes at least one of the following: test results, location of the fix code, and fix code parameters; If the test result indicates that the target fix code has been successfully tested, the target fix code is submitted, and the status of the target defect is updated from an abnormal state to a normal state; wherein, the processing result includes the status of the target fix code and the target defect.

6. The method according to claim 5, characterized in that, The step of transferring the target defect to the corresponding defect processing node for processing based on the target probability and the judgment result, to obtain the processing result of the target defect, includes: If the target probability does not meet the probability condition corresponding to the problem type of the target defect, and the judgment result does not meet the requirements, the target defect is transferred to the development processing node so that the development processing node can repair and automate the target defect based on the description information of the target defect and the target analysis result. After the test is completed, the target defect is associated in the code commit record to obtain the processing result of the target defect.

7. The method according to claim 5, characterized in that, The step of transferring the target defect to the corresponding defect processing node for processing based on the judgment result, and obtaining the processing result of the target defect, includes: If the judgment result indicates that the target defect meets the requirements, the defect will be transferred to the test processing node; the defect processing node includes the test processing node. In response to the test processing node's shutdown operation on the target defect, the processing result of the target defect is obtained.

8. The method according to claim 5 or 6, characterized in that, The method further includes: In response to the retesting of the target defect, if the retesting result of the target defect indicates that the test was successful, the problem description information, the problem type, the target analysis result, the target repair code, and the target repair success rate of the target defect are added to the defect knowledge base.

9. A defect processing device, characterized in that, The device includes: The response module is used to respond to the multimodal information of the target defect input; The processing module is used to perform fusion analysis on the multimodal information using a defect management model to obtain the problem description information of the target defect; The processing module is further configured to query a target historical defect from a defect knowledge base that matches the problem description information of the target defect; the defect knowledge base includes multiple historical defects, and the description information of each historical defect includes at least one of the following: the problem description information, problem type, cause analysis information, repair information, and historical repair success rate of the historical defect; The processing module is also used to obtain a judgment result on whether the target defect meets the requirements based on the similarity between the description information of each requirement in the requirement knowledge base and the problem description information of the target defect. The processing module is further configured to evaluate the probability of successfully fixing the target defect based on the problem type of the target defect, the code context complexity associated with the target defect, and the description information of the target historical defect. The processing module is further configured to transfer the target defect to a defect processing node for processing based on the target probability and / or the judgment result, so as to obtain the processing result of the target defect.

10. A storage medium, characterized in that, The storage medium stores one or more computer programs, which can be executed by one or more processors to implement the defect handling method as described in any one of claims 1 to 8.

Images