Intelligent agent arrangement method and system

Abstract

The present invention relates to the field of model development technology, and discloses an intelligent agent orchestration method and system, the method comprising the following steps: extracting key content by receiving and processing information from project management tools and technical sources, and organizing it into a structured change request list; predicting the specific impact of each change request on the CI / CD process based on the structured change request list and existing configuration information; and outputting a set of reasonable configuration modification plans based on the prediction results. The present invention can automatically process each link in the CI / CD process, and can continuously optimize in a data-driven manner. At the same time, it can quickly respond to changes in demand and technology, minimize manual configuration changes, avoid introducing human errors, continuously optimize the CI / CD flow through data analysis and feedback loops, and use AI intelligent agents for self-adjustment and optimization to reduce dependence on human operators.

Description

Technical Field

[0001] The present invention relates to the field of model development, and more specifically, to an intelligent agent orchestration method and system. Background Art

[0002] With the development of artificial intelligence (AI) technology, especially the application of large models, it has become particularly important to build a CI / CD pipeline that can automatically integrate new AI modules and conduct comprehensive performance evaluation.

[0003] During the application of large models, rapidly changing requirements and technical environments require frequent adjustments to CI / CD pipeline configurations, which may increase complexity and introduce the risk of human error. Summary of the Invention

[0004] The present invention provides an intelligent agent arrangement method and system to solve technical problems in related technologies.

[0005] The present invention provides an agent orchestration method, comprising the following steps:

[0006] S100: Requirements and Technology Change Monitoring, which extracts key information from project management tools and technology sources by receiving and processing it, and organizes it into a structured change request list;

[0007] S200: Change impact assessment and planning: Based on a structured list of change requests and existing configuration information, predict the specific impact of each change request on the CI / CD process. Based on the prediction results, a set of reasonable configuration modification plans are output. Through cost-benefit analysis, the most appropriate plan is selected as the final implementation plan.

[0008] S300: Automated configuration adjustments. Based on the final implementation plan of S200, the new configuration is tested in a non-production environment and verified using a detailed set of deployment instructions. The output verification results show that the new configuration has passed all checks and is ready to be rolled out to the production environment.

[0009] S400: Continuous monitoring and feedback. Based on the new configuration shown in the verification results, it tracks the system performance under the new configuration in real time. Once an anomaly is detected, it immediately triggers a rollback mechanism to restore to the previous stable state. It also collects system operating status and outputs health status reports and abnormal event records.

[0010] S500: Data-driven continuous optimization. Based on health status reports and abnormal event records, the optimization strategy is generated through the model output, and the CI / CD process in S100-S400 is continuously optimized according to the optimization strategy.

[0011] Furthermore, S100 further includes the following steps:

[0012] S110, receiving project management tool information and technical information;

[0013] S120 uses natural language processing technology to parse messages from project management tools and extract key information;

[0014] S130, monitors technical information and extracts the latest release information of frameworks, libraries or tools;

[0015] S140, organizing the collected key information and release information into a structured change request list.

[0016] Furthermore, S200 further includes the following steps:

[0017] S210 , predicting the impact scope of each change request based on the structured change request list in S100 and existing configuration information;

[0018] S220, automatically generating a set of recommended configuration modification solutions based on the prediction results;

[0019] S230, compare the cost-effectiveness of multiple solutions and select the optimal solution as the final implementation plan.

[0020] Furthermore, S300 further includes the following steps:

[0021] S310, input the final implementation plan in S200, test the newly configured sandbox environment, and prove the feasibility of the new configuration settings after testing in a non-production environment;

[0022] S320, deployment instruction set, the instruction set refers to the detailed record of the specific steps for applying the new configuration to the production environment.

[0023] Furthermore, the specific steps applied to the production environment in S320 are as follows:

[0024] S321, generate detailed configuration change instructions: output detailed configuration change instructions through the final implementation plan including the introduction of new continuous integration tools and the update of existing test scripts;

[0025] S322, using declarative configuration management tools: configuration change instructions, Helm Charts configuration file path, Kubernetes cluster API key, output new tools and updated test scripts to successfully deploy to a non-production Kubernetes cluster;

[0026] S323, Verify the effect of the change: Based on the system performance indicators under the new configuration and user feedback, the verification results are output to show that the new configuration has passed all checks and is ready to be promoted to the production environment.

[0027] Furthermore, S400 further includes the following steps:

[0028] S410, input the verification result in S300 The corresponding new configuration and real-time monitoring data flow;

[0029] The real-time monitoring data stream includes information such as server load, error logs, and user behavior from the monitoring system;

[0030] S420 tracks the system performance after deployment in real time, focusing on the parts directly related to the change. If an anomaly is detected, a rollback mechanism is immediately triggered to restore the system to the previous stable state.

[0031] S430, collect the system's operating status under the new configuration, list all detected problems and their treatment status, and output a health status report and abnormal event record.

[0032] Furthermore, S200 also includes the following:

[0033] The calculation formula for change impact assessment is as follows:

[0034] ;

[0035] in, Refers to the specific impact of the predicted change on the existing system. is the first input feature vector containing detailed information about requirements and technical changes, are model parameters, where represents the transpose of the weight matrix, represents the weight matrix for change impact assessment, represents the bias term for the change impact assessment;

[0036] ;

[0037] in Indicates a requirement change request. Indicates technical change information. Represents contextual information;

[0038] The calculation formula for generating the recommended configuration modification solution is as follows:

[0039] ;

[0040] in, Represents a set of possible configuration modification scenarios and their expected effects, Represents the decision logic of a decision tree model, It is a historical data set that contains successful cases of similar changes in the past and the corresponding best practices. represents the decision path constructed based on input features and historical data, is based on the first input feature vector Feature extraction function;

[0041] ;

[0042] in Indicates a record of successfully implemented changes, Indicates best practice guidelines;

[0043] ;

[0044] in, represents the objective function, is the second input feature vector, containing detailed information about each recommended configuration modification solution, is the cost function, is the profit function, and are the weights of costs and benefits, respectively;

[0045] ;

[0046] in Indicates the specific configuration modification content. Represents contextual information;

[0047]

[0048] ;

[0049] in represents the implementation time cost of the i-th solution, including the time required for development, testing and deployment, represents the resource consumption cost of the i-th solution, represents the risk cost of the i-th option;

[0050] in represents the performance improvement of the i-th solution, represents the quality improvement of the i-th solution, represents the user experience improvement of the i-th solution;

[0051] The calculation formula for selecting the optimal solution is as follows:

[0052] ;

[0053] in represents the best solution, that is, the optimal solution.

[0054] Furthermore, the calculation of the optimization strategy generation model in S500 is shown as follows:

[0055] ;

[0056] in Indicates optimization strategy, which is a suggestion for improving the future CI / CD process. Represents a historical data set, including past success stories of similar changes and corresponding best practices. Indicates current feedback, including health status reports and abnormal event records, are the model parameters of the optimization strategy generation model, represents the reinforcement learning function;

[0057] ;

[0058] in Indicates health status report, Indicates abnormal event records.

[0059] The present invention also provides an agent orchestration system that performs one or more steps of the aforementioned agent orchestration method, including:

[0060] Information Collection and Analysis Module: Receives and processes information from project management tools and technical sources, extracts key content, and organizes it into a structured change request list;

[0061] Change Impact Assessment and Planning Module: Based on a structured list of change requests and existing configuration information, it predicts the specific impact of each change request on the CI / CD process. Based on the prediction results, it outputs a set of reasonable configuration modification plans. Through cost-benefit analysis, it selects the most appropriate plan as the final implementation plan.

[0062] Automated configuration adjustment module: Based on the final implementation plan, the new configuration is tested in a non-production environment and verified using a detailed set of deployment instructions. The output verification results show that the new configuration has passed all checks and is ready to be promoted to the production environment.

[0063] Continuous Monitoring and Feedback Module: Based on the new configuration displayed in the verification results, it tracks the system performance under the new configuration in real time. Once an anomaly is detected, it immediately triggers a rollback mechanism to restore the system to the previous stable state. It also collects system operation status and outputs health status reports and abnormal event records.

[0064] Data-driven continuous optimization module: Based on health status reports and abnormal event records, the module generates a model to output the optimization strategy, and continuously optimizes the CI / CD process according to the optimization strategy.

[0065] The present invention also proposes a storage medium storing non-transitory computer-readable instructions for executing one or more steps of the aforementioned method for sustainable communication optimization processing in a restricted connection environment.

[0066] The beneficial effects of the present invention are:

[0067] This invention can automate each link in the CI / CD process and continuously optimize it in a data-driven manner. At the same time, it can quickly respond to changes in demand and technology, minimize manual configuration changes, avoid the introduction of human errors, continuously optimize the CI / CD flow through data analysis and feedback loops, and use AI intelligent agents for self-adjustment and optimization, reducing dependence on human operators. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a flow chart of an intelligent agent orchestration method proposed by the present invention;

[0069] Figure 2 This is a structural block diagram of an intelligent agent orchestration system proposed by the present invention;

[0070] Figure 3 It is a process diagram of an intelligent agent orchestration method proposed in the present invention.

[0071] In the figure: 101, information collection and analysis module; 102, change impact assessment and planning module; 103, automated configuration adjustment module; 104, continuous monitoring and feedback module; 105, data-driven continuous optimization module. DETAILED DESCRIPTION

[0072] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed solely to enable those skilled in the art to better understand and implement the subject matter described herein, and that the functions and arrangements of the elements discussed may be varied without departing from the scope of this specification. Various examples may omit, substitute, or add various processes or components as needed. In addition, features described with respect to some examples may also be combined in other examples.

[0073] like Figure 1 As shown, an agent orchestration method includes the following steps:

[0074] S100: Requirements and Technology Change Monitoring, which extracts key information from project management tools and technology sources by receiving and processing it, and organizes it into a structured change request list;

[0075] S110, receiving project management tool information and technical information;

[0076] Project management tool information includes task updates and problem reports from project management tools (such as Jira, Trello, etc.);

[0077] Technical information includes new version release notes and best practice guides from technical blogs, community forums, and official documentation;

[0078] S120 uses natural language processing (NLP) technology to parse messages from project management tools and extract key information;

[0079] S130, monitors technical information and extracts the latest release information of frameworks, libraries or tools;

[0080] S140, organize the collected key information and release information into a structured change request list;

[0081] The structured change request list contains the identified requirements and technical changes and their priorities.

[0082] S200: Change impact assessment and planning: Based on a structured list of change requests and existing configuration information, predict the specific impact of each change request on the CI / CD process. Based on the prediction results, a set of reasonable configuration modification plans are output. Through cost-benefit analysis, the most appropriate plan is selected as the final implementation plan.

[0083] In one embodiment of the present invention, the specific steps are as follows:

[0084] S210 , predicting the impact scope of each change request based on the structured change request list in S100 and existing configuration information;

[0085] The existing configuration information includes the existing CI / CD configuration, the status of running services, etc.

[0086] In one embodiment of the present invention, the calculation formula for change impact assessment is as follows:

[0087] ;

[0088] in, is the first input feature vector containing detailed information about requirements and technical changes, are model parameters, including weights and biases, etc. represents the transpose of the weight matrix, Represents the weight matrix for change impact assessment, which is used to measure the importance of different input features. The bias term represents the change impact assessment, which adjusts the base level of the prediction results;

[0089] ;

[0090] in Indicates a request for a change in requirements (such as adding new features, modifying existing features, etc.), Indicates information about technical changes (such as the introduction of new frameworks, libraries, or tools), Represents contextual information (such as the current CI / CD configuration, the status of running services, etc.);

[0091] in Change Impact refers to the specific impact of the predicted change on the existing system, which can be expressed as a categorical label (such as "low impact", "medium impact", "high impact") or a continuous value (such as percentage of performance degradation);

[0092] S220, automatically generating a set of recommended configuration modification solutions based on the prediction results;

[0093] In one embodiment of the present invention, the calculation formula for generating the recommended configuration modification solution is as follows:

[0094] ;

[0095] in, (Recommended Configurations) represents a set of possible configuration modification schemes and their expected effects. The probability distribution can be used to represent the likelihood of each scheme being adopted. Represents the decision logic of a decision tree model, It is a historical data set that contains successful cases of similar changes in the past and the corresponding best practices. represents the decision path constructed based on input features and historical data, Based on the input feature vector Feature extraction function;

[0096] ;

[0097] in Indicates a record of successfully implemented changes, Indicates best practice guidelines.

[0098] S230, compare the cost-effectiveness of multiple solutions and select the optimal solution as the final implementation plan;

[0099] The cost-effectiveness calculation formula for comparing various options is as follows:

[0100] ;

[0101] in, (Objective Function) represents the objective function, is the second input feature vector, containing detailed information about each recommended configuration modification solution, is a cost function that measures the resource consumption required to implement the solution, is the benefit function, which measures the improvement or benefit brought about by the implementation of the solution. and are the weights of costs and benefits, respectively, reflecting their relative importance in the decision.

[0102] ;

[0103] in Indicates the specific configuration modification content. Represents contextual information (such as the current CI / CD configuration, the status of running services, etc.);

[0104]

[0105] ;

[0106] in represents the implementation time cost of the i-th solution, including the time required for development, testing and deployment, represents the resource consumption cost of the i-th solution, such as server resources, license fees, etc. Represents the risk cost of the i-th option, which evaluates the cost of potential risks and their response measures.

[0107] in Indicates the performance improvement of the i-th solution, such as shortened response time, increased throughput, etc. Indicates the quality improvement of the i-th solution, such as reduced error rate, improved stability, etc. Indicates the improvement in user experience of the i-th solution, such as an increase in user satisfaction score.

[0108] The calculation formula for selecting the optimal solution is as follows:

[0109] ;

[0110] in (Optimal Solution) represents the best solution, that is, the optimal solution. If the benefit is a positive contribution, a negative sign is used in the objective function;

[0111] S300: Automated configuration adjustments. Based on the final implementation plan of S200, the new configuration is tested in a non-production environment and verified using a detailed set of deployment instructions. The output verification results show that the new configuration has passed all checks and is ready to be rolled out to the production environment.

[0112] In one embodiment of the present invention, the following steps are specifically included:

[0113] S310, input the final implementation plan in S200, test the newly configured sandbox environment, and prove the feasibility of the new configuration settings after testing in a non-production environment;

[0114] In one embodiment of the present invention, the calculation formula for testing a newly configured sandbox environment is as follows:

[0115] ;

[0116] in, (Old Config) is the configuration in the existing production environment, including the existing CI / CD configuration, service status, etc. (Tested Config) indicates new configuration settings that have been verified in a non-production environment. Represents the process of applying a change or configuration to an existing system;

[0117] S320, deployment instruction set, which is a detailed record of the specific steps for applying the new configuration to the production environment;

[0118] In one embodiment of the present invention, the specific steps applied to the production environment are as follows:

[0119] S321, generate detailed configuration change instructions: Based on the final implementation plan, including the introduction of new continuous integration tools and the update of existing test scripts, output detailed configuration change instructions, such as the command for installing new tools, the specific location of the updated test scripts, etc.;

[0120] S322, using declarative configuration management tools: configuration change instructions, Helm Charts configuration file path, Kubernetes cluster API key, output new tools and updated test scripts to successfully deploy to a non-production Kubernetes cluster;

[0121] S323, Verify the effect of the change: Based on the system performance indicators under the new configuration (such as a 15% reduction in build time and a 10% increase in test coverage) and user feedback (such as no new error reports), the verification results are output to show that the new configuration has passed all checks and is ready to be promoted to the production environment;

[0122] The calculation formula for verifying the change effect is as follows:

[0123] ;

[0124] in Indicates monitoring indicators, including system performance indicators (such as CPU utilization, memory usage, etc.), user feedback (such as response time, error reports, etc.), Represents the validation model parameters, including threshold settings (such as the maximum allowable performance degradation percentage) and anomaly detection algorithms (such as statistical-based anomaly detection and machine learning models). (ValidationResult) indicates the validation result, including whether the validation is passed, the problems found and their severity. Represents a validation function.

[0125] S400: Continuous monitoring and feedback. Based on the new configuration shown in the verification results, it tracks the system performance under the new configuration in real time. Once an anomaly is detected, it immediately triggers a rollback mechanism to restore to the previous stable state. It also collects system operating status and outputs health status reports and abnormal event records.

[0126] In one embodiment of the present invention, the following steps are specifically included:

[0127] S410, input the verification result in S300 The corresponding new configuration and real-time monitoring data flow;

[0128] The real-time monitoring data stream includes information such as server load, error logs, and user behavior from the monitoring system;

[0129] S420 tracks the system performance after deployment in real time, focusing on the parts directly related to the change. If an anomaly is detected, a rollback mechanism is immediately triggered to restore the system to the previous stable state.

[0130] S430, collect the system's operating status under the new configuration, list all detected problems and their treatment status, and output a health status report and abnormal event record;

[0131] S500: Data-driven continuous optimization. Based on health status reports and abnormal event records, it generates models and outputs optimization strategies. It also continuously optimizes the CI / CD processes in S100-S400 based on the optimization strategies.

[0132] The optimization strategy refers to the improvement suggestions for the CI / CD process.

[0133] In one embodiment of the present invention, the calculation formula of the optimization strategy generation model is as follows:

[0134] ;

[0135] in Optimization Strategy (Optimization Strategy) represents an optimization strategy, which is an improvement suggestion for the future CI / CD process. It uses a rule set or action sequence to represent specific optimization measures. Represents a historical data set, including past success stories of similar changes and corresponding best practices. Indicates current feedback, including system feedback after the most recent change (such as health status reports and abnormal event records). are model parameters, including weights and biases, etc. represents the reinforcement learning function.

[0136] ;

[0137] in Indicates health status report, Indicates abnormal event records.

[0138] The steps for training the optimization strategy generation model include:

[0139] Step 1: Build the target network and generate the target network replication optimization strategy model;

[0140] Step 2: Randomly obtain an experience from the experience pool. The experience includes the state at time e , the action performed at time e , the state at time e+1 , perform the action Rewards received ;

[0141] The states at time e and e+1 represent the current feedback information at the e-th and e+1-th moments of the optimization strategy generation model input, respectively;

[0142] Then calculate the state The target network output corresponds to The probability value of ;

[0143] ;

[0144] express The maximum probability value output by the target network;

[0145] Step 3, according to and Calculate the loss value and update the optimization strategy generation model;

[0146] ;

[0147] express The output of the optimization strategy generation model corresponds to The probability value of

[0148] is the discount factor, a value between 0 and 1;

[0149] Step 4: Update the target network every fixed number of network updates so that its parameters are the same as those of the current optimization strategy generation model;

[0150] Step 5: The process is terminated until the number of collected experiences reaches the set threshold or the optimization strategy generation model converges.

[0151] In one embodiment of the present invention, the experience is derived from the record of historical action execution.

[0152] In one embodiment of the present invention, the reward The calculation formula is as follows:

[0153] ;

[0154] in, Indicates the current state at time e, which is determined by historical data and current feedback Joint decision, Indicates the action taken at time e, that is, the specific optimization strategy.

[0155] in Indicates performance improvement, ;

[0156] in Represents the new performance indicators after implementing the optimization strategy (such as build time, test coverage, etc.), Represents the old performance indicator before implementation.

[0157] in Indicates cost savings, ;

[0158] in Indicates the resource consumption cost after implementing the optimization strategy (such as server resources, license fees, etc.), Represents the resource consumption cost before implementation.

[0159] in Indicates a reduction in failure rate, ;

[0160] in represents the failure rate or number of abnormal events after the optimization strategy is implemented, Indicates the failure rate or number of abnormal events before implementation.

[0161] in It means time saving, ;

[0162] in represents the total deployment time after implementing the optimization strategy, Represents the total deployment time before implementation.

[0163] in 、 、 and They represent the weights corresponding to each parameter, reflecting their relative importance in decision making.

[0164] At least one embodiment of the present disclosure provides an agent orchestration system that performs one or more steps of the aforementioned agent orchestration method, including:

[0165] Information Collection and Parsing Module 101: Receives and processes information from project management tools and technologies, extracts key content, and organizes it into a structured change request list;

[0166] Change Impact Assessment and Planning Module 102: Based on the structured change request list and existing configuration information, the module predicts the specific impact of each change request on the CI / CD process. Based on the prediction results, it outputs a set of reasonable configuration modification plans. Through cost-benefit analysis, the module selects the most appropriate plan as the final implementation plan.

[0167] Automated configuration adjustment module 103: Based on the final implementation plan, the new configuration is tested in a non-production environment and verified using a detailed set of deployment instructions. The verification results are output showing that the new configuration has passed all checks and is ready to be rolled out to the production environment.

[0168] Continuous Monitoring and Feedback Module 104: Based on the new configuration displayed in the verification results, it tracks the system performance under the new configuration in real time. Once an abnormality is detected, it immediately triggers a rollback mechanism to restore the system to the previous stable state. It also collects system operating status and outputs health status reports and abnormal event records.

[0169] Data-driven continuous optimization module 105: Based on health status reports and abnormal event records, the optimization strategy is generated through the model output, and the CI / CD process is continuously optimized according to the optimization strategy.

[0170] At least one embodiment of the present disclosure provides a storage medium storing non-transitory computer-readable instructions for executing one or more steps in the aforementioned agent orchestration method.

[0171] The computer program may be stored / distributed on a suitable medium, such as an optical storage medium or solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. Any reference signs in the claims should not be construed as limiting the scope.

[0172] The above describes the embodiments of this embodiment, but this embodiment is not limited to the above specific implementation methods. The above specific implementation methods are merely illustrative and not restrictive. Ordinary technicians in this field can also make many forms based on the inspiration of this embodiment, all of which are protected by this embodiment.

Claims

1. An agent orchestration method, characterized in that: The following steps are involved: S100: Requirements and Technology Change Monitoring, which extracts key information from project management tools and technology sources by receiving and processing it, and organizes it into a structured change request list; S200: Change impact assessment and planning: Based on a structured list of change requests and existing configuration information, predict the specific impact of each change request on the CI / CD process. Based on the prediction results, a set of reasonable configuration modification plans are output. Through cost-benefit analysis, the most appropriate plan is selected as the final implementation plan. S300: Automated configuration adjustment. Based on the final implementation plan of S200, the new configuration is tested in a non-production environment and verified using a detailed set of deployment instructions. The output verification results show that the new configuration has passed all checks and is then rolled out to the production environment. S400: Continuous monitoring and feedback. Based on the new configuration shown in the verification results, it tracks the system performance under the new configuration in real time. Once an anomaly is detected, it immediately triggers a rollback mechanism to restore to the previous stable state. It also collects system operating status and outputs health status reports and abnormal event records. S5 00: Data-driven continuous optimization. Based on health status reports and abnormal event records, the optimization strategy is generated through the model output, and the CI / CD process in S100-S400 is continuously optimized according to the optimization strategy.

2. The agent arrangement method according to claim 1, characterized in that: S100 also includes the following steps: S110, receiving project management tool information and technical information; S120 uses natural language processing technology to parse messages from project management tools and extract key information; S130, monitors technical information and extracts the latest release information of frameworks, libraries or tools; S140, organizing the collected key information and release information into a structured change request list.

3. The agent arrangement method according to claim 2, characterized in that: S200 also includes the following steps: S210 , predicting the impact scope of each change request based on the structured change request list in S100 and existing configuration information; S220, automatically generating a set of recommended configuration modification solutions based on the prediction results; S230, compare the cost-effectiveness of multiple solutions and select the optimal solution as the final implementation plan.

4. The agent arrangement method according to claim 3, characterized in that: S300 also includes the following steps: S310, input the final implementation plan in S200, test the newly configured sandbox environment, and prove the feasibility of the new configuration settings after testing in a non-production environment; S320, deployment instruction set, the instruction set refers to the detailed record of the specific steps for applying the new configuration to the production environment.

5. The agent arrangement method according to claim 4, characterized in that: The specific steps for applying it to the production environment in S320 are as follows: S321, generate detailed configuration change instructions: output detailed configuration change instructions through the final implementation plan including the introduction of new continuous integration tools and the update of existing test scripts; S322, using declarative configuration management tools: configuration change instructions, Helm Charts configuration file path, Kubernetes cluster API key, output new tools and updated test scripts to successfully deploy to a non-production Kubernetes cluster; S323, Verify the effect of the change: Based on the system performance indicators under the new configuration and user feedback, the verification results are output to show that the new configuration has passed all checks and is ready to be promoted to the production environment.

6. The agent arrangement method according to claim 5, characterized in that: S400 also includes the following steps: S410, input the verification result in S300 The corresponding new configuration and real-time monitoring data flow; The real-time monitoring data stream includes server load, error logs, and user behavior information from the monitoring system; S420, real-time tracking of system performance after deployment, is directly related to the change. If an anomaly is detected, a rollback mechanism is immediately triggered to restore the system to the previous stable state. S430, collect the system's operating status under the new configuration, list all detected problems and their treatment status, and output a health status report and abnormal event record.

7. The agent arrangement method according to claim 6, characterized in that: The S200 also includes the following: The calculation formula for change impact assessment is as follows: ； in, Refers to the specific impact of the predicted change on the existing system. is the first input feature vector containing detailed information about requirements and technical changes, are model parameters, where represents the transpose of the weight matrix, represents the weight matrix for change impact assessment, represents the bias term for the change impact assessment; ； in Indicates a requirement change request. Indicates technical change information. Represents contextual information; The calculation formula for generating the recommended configuration modification solution is as follows: ； in, Represents a set of configuration modification plans and their expected effects, Represents the decision logic of a decision tree model, It is a historical data set that contains successful cases of similar changes in the past and the corresponding best practices. represents the decision path constructed based on input features and historical data, is based on the first input feature vector Feature extraction function; ； in Indicates a record of successfully implemented changes, Indicates best practice guidelines; ； in, represents the objective function, is the second input feature vector, containing detailed information about each recommended configuration modification solution, is the cost function, is the profit function, and are the weights of costs and benefits, respectively; ； in Indicates the specific configuration modification content. Represents contextual information; ； ； in represents the implementation time cost of the i-th solution, including the time required for development, testing and deployment, represents the resource consumption cost of the i-th solution, represents the risk cost of the i-th option; in represents the performance improvement of the i-th solution, represents the quality improvement of the i-th solution, represents the user experience improvement of the i-th solution; The calculation formula for selecting the optimal solution is as follows: ； in represents the best solution, that is, the optimal solution.

8. The agent arrangement method according to claim 7, characterized in that: The calculation of the optimization strategy generation model in S500 is as follows: ； in Indicates optimization strategy, which is a suggestion for improving the future CI / CD process. Represents a historical data set, including past success stories of similar changes and corresponding best practices. Indicates current feedback, including health status reports and abnormal event records, are the model parameters of the optimization strategy generation model, represents the reinforcement learning function; ； in Indicates health status report, Indicates abnormal event records.

9. An agent orchestration system, executing one or more steps of an agent orchestration method according to any one of claims 1 to 8, characterized in that: include: Information Collection and Analysis Module: Receives and processes information from project management tools and technical sources, extracts key content, and organizes it into a structured change request list; Change Impact Assessment and Planning Module: Based on a structured list of change requests and existing configuration information, it predicts the specific impact of each change request on the CI / CD process. Based on the prediction results, it outputs a set of reasonable configuration modification plans. Through cost-benefit analysis, it selects the most appropriate plan as the final implementation plan. Automated configuration adjustment module: Based on the final implementation plan, the new configuration is tested in a non-production environment and verified using a detailed set of deployment instructions. The output verification results show that the new configuration has passed all checks and is ready to be promoted to the production environment. Continuous Monitoring and Feedback Module: Based on the new configuration displayed in the verification results, it tracks the system performance under the new configuration in real time. Once an anomaly is detected, it immediately triggers a rollback mechanism to restore the system to the previous stable state. It also collects system operation status and outputs health status reports and abnormal event records. Data-driven continuous optimization module: Based on health status reports and abnormal event records, the module generates a model to output the optimization strategy, and continuously optimizes the CI / CD process according to the optimization strategy.

10. A storage medium, characterized in that: Non-transitory computer-readable instructions are stored for executing one or more steps in an agent orchestration method as described in any one of claims 1-8.

Images