System operation and maintenance processing method and device based on MCP tool, equipment and medium

Abstract

The application relates to a system operation and maintenance processing method and device based on an MCP tool, equipment and a medium. The method comprises the following steps: obtaining historical fault data and current operation data of an operation and maintenance system, and constructing a state evaluation prompt word according to the historical fault data and the current operation data; inputting the state evaluation prompt word into a preset large model to obtain a system state score, a risk level and a current abnormal type of the operation and maintenance system; constructing a tool selection prompt word according to the system state score, the risk level, the current abnormal type and an MCP tool list; inputting the tool selection prompt word into the preset large model to obtain a candidate tool set; determining and running a target tool in the candidate tool set through a preset strategy to obtain a running result; and weighting and fusing the running result of the target tool according to a tool weight to obtain a system operation and maintenance strategy. The problem that it is difficult to dynamically generate the most suitable operation and maintenance processing strategy according to the real-time state of an operation and maintenance system is solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a system operation and maintenance processing method, apparatus, equipment and medium based on the MCP (Model Context Protocol) tool. Background Technology

[0002] Operations and maintenance (O&M) systems are crucial for ensuring the stable operation of IT infrastructure, application services, and cloud platforms. Currently, O&M systems primarily operate in the following modes: First, the fixed-tool mode, where a single O&M tool (e.g., monitoring-only or alarm-only) is statically configured to handle all or specific types of faults, lacking flexibility; second, the static configuration mode, where tool selection and invocation rely on pre-written configuration files, requiring a system restart for changes to take effect, resulting in slow response times; and third, the traditional rule-based engine mode, which selects tools through complex manual rules or scripts, making rule writing and maintenance difficult, and the decision-making process hard to interpret. Furthermore, O&M tools themselves lack unified management standards; expanding with new tools typically requires modifying the core system code, and dynamic registration and hot reloading are not possible, leading to chaotic tool ecosystem management and high expansion costs.

[0003] Because the current operation and maintenance process lacks the ability to intelligently perceive and make real-time decisions on complex and dynamically changing operation and maintenance scenarios, and also lacks an effective mechanism for standardizing and loosely coupling the integration of diverse operation and maintenance tools, it is difficult to dynamically generate the most suitable operation and maintenance handling strategy based on the real-time status of the operation and maintenance system.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This application provides a system operation and maintenance processing method, apparatus, equipment and medium based on MCP tools to solve the above-mentioned technical problem of difficulty in dynamically generating the most suitable operation and maintenance processing strategy according to the real-time status of the operation and maintenance system.

[0006] According to one aspect of the embodiments of this application, this application provides a system operation and maintenance processing method based on MCP tools, including: upon receiving an operation and maintenance request, acquiring historical fault data and current operating data of the operation and maintenance system, and constructing status assessment prompts based on the historical fault data and current operating data; calling a preset large model and inputting the status assessment prompts into the preset large model to obtain the system status score, risk level, and current anomaly type of the operation and maintenance system obtained after status assessment by the preset large model; acquiring an MCP tool list and constructing tool selection prompts based on the system status score, risk level, current anomaly type, and the MCP tool list; inputting the tool selection prompts into the preset large model to obtain a candidate toolset selected by the preset large model; determining and running target tools in the candidate toolset using a preset strategy to obtain running results, wherein the preset strategy is to first select target tools based on the tool prediction effect, and then calculate the tool weight of each target tool based on the tool performance; and weighting and fusing the running results of the target tools according to the tool weights to obtain a system operation and maintenance strategy.

[0007] Optionally, the historical fault data and current running data of the operation and maintenance system are obtained, including: attempting to read the current running data and historical fault data of the operation and maintenance system from the target cache; if the attempt to read from the target cache fails, attempting to read the current running data and historical fault data from the target database; if the attempt to read from the target database fails, then generating historical fault data and current running data using preset default values.

[0008] Optionally, the status assessment prompt is input into a preset large model to obtain the system status score, risk level, and current anomaly type of the operation and maintenance system obtained after the status assessment is performed by the preset large model. This includes: sending the status assessment prompt as the first input parameter to the first application interface of the preset large model, wherein the status assessment prompt includes an instruction to return data in a target format; receiving the first response data returned by the first application interface, and extracting the response data system status score, risk level, and current anomaly type from the first response data.

[0009] Optionally, the tool selection prompt is input into a preset large model to obtain a candidate toolset selected by the preset large model. This includes: sending the tool selection prompt as a second input parameter to a second application interface of the preset large model, and receiving second response data returned by the second application interface, wherein the tool selection prompt includes an instruction to return data in a target format; obtaining a tool identifier list by parsing the second response data, wherein the tool identifier list includes at least one tool identifier recommended by the preset large model after analysis based on system status score, risk level, current anomaly type, and functional descriptions of each tool in the MCP tool list; matching and verifying the tool identifier list with available tools in the operation and maintenance system, and integrating the MCP tools corresponding to the target tool identifiers that pass the matching and verification into a candidate toolset.

[0010] Optionally, target tools are determined and run in the candidate toolset using a preset strategy to obtain the running results, including: obtaining the historical performance of each candidate tool in the candidate toolset, the state matching degree between the candidate tool and the system state score, and the load matching degree between the candidate tool and the current resource load of the operation and maintenance system, wherein the historical performance represents the historical comprehensive performance of the candidate tool; weighting the historical performance, state matching degree, and load matching degree to obtain the predicted performance of the candidate tool; comparing the predicted performance of the tool with the performance threshold, and selecting the target tool whose predicted performance is greater than the performance threshold from the candidate tools; obtaining the recent feedback performance and time decay coefficient of the target tool from the MCP tool library, and calculating the initial weight using the historical performance, recent feedback performance, and time decay coefficient, wherein the recent feedback performance is the feedback performance of the target tool in the target rounds before the current round; and normalizing the initial weight of each target tool to obtain the tool weight of each target tool.

[0011] Optionally, the system operation and maintenance strategy is obtained by weighted fusion of the operation results of the target tools based on the tool weights, including: extracting operation and maintenance actions from the operation results of each target tool, wherein the operation results include at least one operation and maintenance action and the initial priority of the operation and maintenance action; for each operation and maintenance action, establishing a correspondence between the operation and maintenance action, the initial priority, and the action weight, wherein the action weight is the tool weight of the target tool that generated the operation and maintenance action; multiplying the initial priority corresponding to the operation and maintenance action by the action weight to obtain the fusion priority; for the same operation and maintenance action in different operation results, adding the fusion priorities of the same operation and maintenance action to obtain the target priority; arranging each operation and maintenance action in descending order according to the target priority, and integrating the operation and maintenance actions ranked before the target sequence number to form the system operation and maintenance strategy.

[0012] Optionally, after obtaining the system operation and maintenance strategy, for the current target tool, the method further includes calculating the current round feedback effect of the current target tool in the following manner: executing the system operation and maintenance strategy and recording the target operation and maintenance actions in the system operation and maintenance strategy, wherein the target operation and maintenance actions are the actions in the running results output by the current target tool; obtaining the processing result score of each target operation and maintenance action; calculating the contribution ratio of the current target tool to the target operation and maintenance action during fusion, and calculating the contribution effect value of the current target tool to the target operation and maintenance action based on the processing result score and the contribution ratio; adding the contribution effect values ​​of each target operation and maintenance action to obtain the current round feedback effect of the current target tool, and updating the recent feedback effect of the current target tool in the MCP tool library with the current round feedback effect.

[0013] According to another aspect of the embodiments of this application, this application provides a system operation and maintenance processing device based on MCP tools, including: an acquisition module, used to acquire historical fault data and current operating data of the operation and maintenance system upon receiving an operation and maintenance request, and construct status assessment prompt words based on the historical fault data and current operating data; an invocation module, used to invoke a preset large model and input the status assessment prompt words into the preset large model to obtain the system status score, risk level, and current anomaly type of the operation and maintenance system obtained after status assessment by the preset large model; a construction module, used to acquire an MCP tool list and construct tool selection prompt words based on the system status score, risk level, current anomaly type, and MCP tool list; an input module, used to input the tool selection prompt words into the preset large model to obtain a candidate toolset selected by the preset large model; a determination module, used to determine and run target tools in the candidate toolset according to a preset strategy to obtain running results, wherein the preset strategy is to first select target tools based on the tool prediction effect, and then calculate the tool weight of each target tool based on the tool performance; and a fusion module, used to perform weighted fusion of the running results of the target tools according to the tool weights to obtain a system operation and maintenance strategy.

[0014] According to another aspect of the embodiments of this application, this application provides an electronic device, including a memory, a processor, a communication interface and a communication bus. The memory stores a computer program that can run on the processor. The memory and the processor communicate with each other through the communication bus and the communication interface. When the processor executes the computer program, it implements the steps of the above method.

[0015] According to another aspect of the embodiments of this application, this application also provides a computer-readable medium having processor-executable non-volatile program code that causes the processor to perform the above-described method.

[0016] Compared with related technologies, the technical solutions provided in this application have the following advantages: This application provides a system operation and maintenance method based on MCP tools, including: upon receiving an operation and maintenance request, acquiring historical fault data and current operating data of the operation and maintenance system, and constructing status assessment prompts based on the historical fault data and current operating data; calling a preset large model and inputting the status assessment prompts into the preset large model to obtain the system status score, risk level, and current anomaly type of the operation and maintenance system obtained after status assessment by the preset large model; acquiring an MCP tool list and constructing tool selection prompts based on the system status score, risk level, current anomaly type, and the MCP tool list; inputting the tool selection prompts into the preset large model to obtain a candidate tool set selected by the preset large model; determining and running target tools in the candidate tool set through a preset strategy to obtain the running results, wherein the preset strategy is to first select target tools based on the tool prediction effect, and then calculate the tool weight of each target tool based on the tool performance; and weighting and fusing the running results of the target tools according to the tool weights to obtain the system operation and maintenance strategy. By calling a pre-set large model to assess the system status, and then using the pre-set large model to intelligently select tools from the MCP tool list based on the assessment results, the tool operation results are optimized and integrated by sampling a multi-step progressive strategy based on prediction effect and tool performance, thereby obtaining the best operation and maintenance strategy. This solves the problem of difficulty in dynamically generating the most suitable operation and maintenance handling strategy according to the real-time status of the operation and maintenance system. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an optional system operation and maintenance method based on an MCP tool, according to an embodiment of this application. Figure 2 This is a flowchart illustrating an optional system parameter acquisition degradation process according to an embodiment of this application. Figure 3 This is a schematic diagram illustrating the execution flow of an optional preset strategy provided according to an embodiment of this application; Figure 4 This is a schematic diagram of an optional weighted fusion mechanism provided according to an embodiment of this application; Figure 5This is a schematic diagram illustrating an optional closed-loop feedback and optimization method according to an embodiment of this application. Figure 6 This is a block diagram of an optional system operation and maintenance processing device based on an MCP tool, according to an embodiment of this application. Figure 7 This is a schematic diagram of an optional electronic device structure provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] In the following description, the use of suffixes such as "module," "part," or "unit" to denote elements is solely for the purpose of illustration and has no specific meaning in itself. Therefore, "module" and "part" may be used interchangeably.

[0022] Operations and maintenance (O&M) systems are crucial for ensuring the stable operation of IT infrastructure, application services, and cloud platforms. Currently, O&M systems primarily operate in the following modes: First, the fixed-tool mode, where a single O&M tool (e.g., monitoring-only or alarm-only) is statically configured to handle all or specific types of faults, lacking flexibility; second, the static configuration mode, where tool selection and invocation rely on pre-written configuration files, requiring a system restart for changes to take effect, resulting in slow response times; and third, the traditional rule-based engine mode, which selects tools through complex manual rules or scripts, making rule writing and maintenance difficult, and the decision-making process hard to interpret. Furthermore, O&M tools themselves lack unified management standards; expanding with new tools typically requires modifying the core system code, and dynamic registration and hot reloading are not possible, leading to chaotic tool ecosystem management and high expansion costs.

[0023] Because the current operation and maintenance process lacks the ability to intelligently perceive and make real-time decisions on complex and dynamically changing operation and maintenance scenarios, and also lacks an effective mechanism for standardizing and loosely coupling the integration of diverse operation and maintenance tools, it is difficult to dynamically generate the most suitable operation and maintenance handling strategy based on the real-time status of the operation and maintenance system.

[0024] To address the problems mentioned in the background art, according to one aspect of an embodiment of this application, a system operation and maintenance method based on the MCP tool is provided, such as... Figure 1 As shown, it includes: Step 101: Upon receiving an operation and maintenance request, obtain historical fault data and current operating data of the operation and maintenance system, and construct status assessment prompts based on the historical fault data and current operating data; Step 102: Call the preset large model and input the status assessment prompt words into the preset large model to obtain the system status score, risk level and current anomaly type of the operation and maintenance system after the status assessment is performed by the preset large model; Step 103: Obtain the MCP tool list and select prompt words based on the system status score, risk level, current anomaly type, and the tool used to build the MCP tool list; Step 104: Input the tool selection prompts into the preset large model to obtain the candidate tool set selected by the preset large model; Step 105: Determine and run the target tool in the candidate tool set through a preset strategy to obtain the running results. The preset strategy is to first select the target tool based on the prediction effect of the tool, and then calculate the tool weight of each target tool based on the tool performance. Step 106: The running results of the target tool are weighted and fused according to the tool weight to obtain the system operation and maintenance strategy.

[0025] The embodiments provided in this application can be used in a variety of operation and maintenance scenarios, including information technology infrastructure operation and maintenance, application service operation and maintenance, and cloud platform operation and maintenance.

[0026] This application provides a Large Language Model (LLM) as a pre-defined large model, which can also be a GPT or Claude model. The intelligent understanding and reasoning capabilities of the pre-defined large model are used to replace traditional decision-making models that rely on fixed rules or human experience.

[0027] Upon receiving an operation and maintenance request (e.g., a fault alarm), the system retrieves historical fault data and current operational data, and constructs status assessment prompts based on this data. A pre-defined large-scale model is then invoked to process these prompts, performing intelligent analysis and evaluation of the system, and outputting a quantified system status score, risk level, and specific current anomaly type.

[0028] Obtain a standardized list of MCP operation and maintenance tools, then combine system status score, risk level, current anomaly type, and tool list information to construct tool selection prompts. Next, call the preset large model to process the tool selection prompts again, and the preset large model will intelligently recommend an initial set of candidate tools.

[0029] This application introduces MCP as a standardized tool execution framework, enabling various operation and maintenance tools to be dynamically managed and invoked through a unified interface, achieving plug-and-play functionality without modifying the core system code.

[0030] The candidate toolset is optimized using a pre-defined strategy. The core of this strategy is to first filter and then weight the tools: first, target tools are selected based on the predictive performance of each candidate tool; then, the tool weights of each target tool are calculated and normalized based on its performance data.

[0031] By predicting the effects, we ensure the tool's suitability for the current scenario. Then, by using dynamic weights calculated based on historical and recent performance, we ensure that better tools have a higher weight in the final decision, thereby driving the entire system to continuously evolve towards better decisions.

[0032] Execute the selected target tools and collect the results generated by each tool (i.e., its recommended maintenance actions). These results are then weighted and merged according to the tool weights of each target tool to generate a final, unified-priority system maintenance strategy.

[0033] As an optional embodiment, obtaining historical fault data and current running data of the operation and maintenance system includes: attempting to read the current running data and historical fault data of the operation and maintenance system from the target cache; if the attempt to read from the target cache fails, attempting to read the current running data and historical fault data from the target database; if the attempt to read from the target database fails, generating historical fault data and current running data using preset default values.

[0034] Current operational data includes system monitoring data, fault information, and resource load.

[0035] First, we attempt to read the required current operating data and historical fault data of the operation and maintenance system from the target cache. The target cache in this application can be a Redis cache, a high-performance in-memory database suitable for storing and quickly accessing frequently accessed data.

[0036] If reading data from the Redis cache fails, the system will fall back to attempting to read the same data from the target database. The target database in this application is MySQL, which can serve as the persistent primary data store. Prioritizing reading data from the high-performance Redis cache significantly reduces data input / output latency, thereby accelerating the startup speed of the entire operation and maintenance process, especially when handling high-frequency or real-time-critical operation and maintenance requests.

[0037] The reason for failing to read data from the Redis cache may be that the data does not exist or the cache service is malfunctioning.

[0038] If reading from the MySQL database also fails, the default values ​​will be used to generate or replace the required historical failure data and current running data, ensuring that subsequent processes can obtain basic data and continue running even if the data source is temporarily unavailable, rather than being completely interrupted.

[0039] Figure 2 The flowchart illustrates the degradation process for obtaining system parameters provided in this application. First, it attempts to read data from the Redis cache (priority 1). If successful, the process jumps directly to the bottom, calling the monitoring service API (Application Programming Interface) to obtain real-time monitoring data (i.e., currently running data). If Redis cache retrieval fails, it attempts to read system status data and historical fault data from the MySQL database (priority 2). If MySQL database retrieval succeeds, it also proceeds to calling the monitoring service API. If MySQL database retrieval also fails, the default degradation strategy (priority 3) is executed, and then the monitoring service API is called again to obtain real-time monitoring data. Ultimately, all paths converge on the step of calling the monitoring service API to obtain real-time monitoring data, and the parameters are returned, completing the entire data acquisition process.

[0040] When acquiring data, this application first accesses the fastest, but not necessarily persistent, Redis cache. If that fails, it accesses the slower but more reliable main database, MySQL. If both are unavailable, a built-in, guaranteed-usability default value is used as a last resort. This strategy maximizes system availability and responsiveness while preventing service interruption due to the unavailability of a single data source. Even in the extreme case where both the cache and database are unavailable, the system can still generate basic data input by using preset default values, enabling subsequent critical steps such as large-scale model state evaluation to be performed. Although this may be based on incompletely accurate data, it maintains the integrity of the processing chain, buying time for system recovery or manual intervention.

[0041] As an optional embodiment, the status assessment prompt is input into a preset large model to obtain the system status score, risk level, and current anomaly type of the operation and maintenance system obtained after the status assessment by the preset large model. This includes: sending the status assessment prompt as a first input parameter to the first application interface of the preset large model, wherein the status assessment prompt includes an instruction to return data in a target format; receiving the first response data returned by the first application interface, and extracting the response data system status score, risk level, and current anomaly type from the first response data.

[0042] The completed state evaluation prompt is sent as the first input parameter to the first application interface of the preset large model. This first application interface is the application programming interface provided by the preset large model for handling such requests. The state evaluation prompt contains an explicit instruction requiring the preset large model to return its evaluation result data in the target format.

[0043] For example, the status assessment prompt could be: "Assess system status:" System monitoring data: {monitoring_data} Fault information: {fault_info} Resource load: {resource_load} Return JSON: {\"status_score\":0.0-1.0, \"risk_level\":\"high / medium / low\", \"reason\":\"reason\"}".

[0044] The target format involved in this application is JSON (JavaScript Object Notation).

[0045] Receive the first response data returned by the first application interface. The first response data is structured data in JSON format. Extract the value of the status_score field as the system status score, extract the value of the risk_level field as the risk level, and extract the value of the fault_type field as the current exception type. The current exception type can be a fault type such as "CPU overload", "memory leak", "service unavailable", or "network packet loss".

[0046] By using standardized API calls and structured data formats (such as JSON) to constrain the ambiguity that may arise from natural language interaction, the evaluation results are guaranteed to be in a stable format and have clear content, which greatly reduces the complexity of interface integration and subsequent maintenance costs.

[0047] As an optional embodiment, inputting tool selection prompts into a preset large model to obtain a candidate toolset selected by the preset large model includes: sending the tool selection prompts as a second input parameter to a second application interface of the preset large model, and receiving second response data returned by the second application interface, wherein the tool selection prompts include an instruction to return data in a target format; obtaining a tool identifier list by parsing the second response data, wherein the tool identifier list includes at least one tool identifier recommended by the preset large model after analysis based on system status score, risk level, current anomaly type, and functional descriptions of each tool in the MCP tool list; matching and verifying the tool identifier list with available tools in the operation and maintenance system, and integrating the MCP tools corresponding to the target tool identifiers that pass the matching and verification into a candidate toolset.

[0048] The constructed tool selection prompt is sent as the second input parameter to the second application interface of the pre-defined large model. The core content of the tool selection prompt includes the system status score, risk level, current anomaly type, and functional descriptions of each tool in the MCP tool list. Similar to the status assessment step described above, the tool selection prompt also includes instructions requiring the large model to return recommended results in the target format.

[0049] For example, the tool selection prompt could be: "Select maintenance tools based on system status:" System status: {system_status} Fault type: {fault_type} Resource load: {resource_load} List of available tools: {available_tools} Please select 3-4 of the most suitable tools.

[0050] The system receives the second response data returned by the second application interface and parses this data to obtain a list of tool identifiers recommended by the large model. The list of tool identifiers contains tool identifiers for one or more tools recommended by the preset large model based on its understanding of the current system state, risks, and anomaly types, as well as its analysis of the functional descriptions of each MCP tool.

[0051] In the embodiments of this application, not all tools recommended by the preset large model are directly adopted. Instead, the parsed list of tool identifiers needs to be matched and verified against the currently registered and available MCP tools in the operation and maintenance system. Only those tool identifiers that exist in the actual available tool library will be retained. The MCP tools corresponding to these verified tool identifiers are ultimately integrated into a candidate toolset.

[0052] This embodiment combines semantic-based preliminary tool selection from a large model with existence verification based on registration status, which together improves the accuracy and reliability of the final generated candidate toolset, reduces the probability of needing to backtrack and reselect due to tool unavailability, and improves the efficiency of overall operation and maintenance.

[0053] As an optional embodiment, a target tool is determined and run in a candidate toolset using a preset strategy to obtain the running results. This includes: acquiring the historical performance of each candidate tool in the candidate toolset, the state matching degree between the candidate tool and the system state score, and the load matching degree between the candidate tool and the current resource load of the operation and maintenance system, wherein the historical performance represents the historical comprehensive performance of the candidate tool; performing a weighted calculation on the historical performance, state matching degree, and load matching degree to obtain the predicted performance of the candidate tool; comparing the predicted performance of the tool with the performance threshold, and selecting the target tool whose predicted performance is greater than the performance threshold from the candidate tools; acquiring the recent feedback performance and time decay coefficient of the target tool from the MCP tool library, and calculating the initial weight using the historical performance, recent feedback performance, and time decay coefficient, wherein the recent feedback performance is the feedback performance of the target tool in the target rounds before the current round; and normalizing the initial weight of each target tool to obtain the tool weight of each target tool.

[0054] First, obtain three key parameters for each candidate tool in the candidate toolset: historical performance, which represents the tool's overall historical performance data over all or a longer period; state matching degree, which represents the degree to which the tool's function matches the current system state score (which has been evaluated by a pre-set large model); and load matching degree, which represents the degree to which the tool adapts to the current resource load of the operation and maintenance system.

[0055] For each candidate tool, the historical performance, state matching degree, and load matching degree are weighted and calculated to obtain the tool prediction performance for each candidate tool. The specific calculation formula can be: Tool Prediction Performance = Historical Performance First weight + state matching degree Second weight + load matching degree The third weight, of which the first, second, and third weights can be adjusted according to the actual situation.

[0056] The prediction effect of each tool is compared with the preset effect threshold. Only candidate tools whose prediction effect is greater than the threshold are selected as the target tools for the next stage. This completes the first round of screening based on effect prediction. The effect threshold can be set according to the actual situation.

[0057] By using weighted calculations to integrate multiple evaluation dimensions into a single predictive performance index, a rapid and quantitative assessment of the tool's fit with the current scenario is achieved, and the tool is initially screened.

[0058] For the selected target tools, recent feedback results and time decay coefficients are obtained from the MCP tool library. Recent feedback results represent the performance score calculated based on the actual processing results of the target tool in the target rounds (e.g., 1) before the current round (i.e., this fault handling), which reflects the latest dynamic performance of the tool.

[0059] The time decay coefficient is a preset coefficient used to decay historical data over time in weight calculations, making recent performance have a greater impact.

[0060] Using the obtained historical effects, recent feedback effects, and time decay coefficient, an initial weight is calculated for each target tool. The specific calculation formula is: Initial Weight = (Historical Effects) / (Recent Feedback Effects) Fifth weighting + recent feedback results Fifth weight) The time decay coefficient, in which the fourth and fifth weights can be adjusted according to the actual situation.

[0061] By introducing recent feedback effects and a time decay factor, the weights of tools are no longer static but can be dynamically adjusted based on their latest practical performance. Tools that perform well have increased influence in subsequent decisions, while those that perform poorly have decreased influence, thus enabling the entire system to continuously improve and optimize itself based on historical experience.

[0062] Since the initial weight values ​​of each target tool have different dimensions, the initial weights of all target tools are normalized so that the sum of all tool weights is 1, and finally the tool weight of each target tool is obtained.

[0063] Figure 3 This diagram illustrates the execution flow of the preset strategy provided in this application. As shown, the first stage involves selecting basic tools. The system calculates the predictive tool's effectiveness based on three dimensions: history, state, and load. The formula is: history (corresponding to historical effectiveness) multiplied by 0.5, state (corresponding to state matching degree) multiplied by 0.3, and load (corresponding to load matching degree) multiplied by 0.2. Next, a judgment phase occurs. If the calculated predictive effectiveness of a tool exceeds a set threshold, the tool is added to the basic tool list; otherwise, it is excluded. Tools that pass the selection process enter the second stage of weight adjustment. The system calculates the weight using the formula: history (effectiveness) multiplied by 0.6, feedback (corresponding to recent feedback effectiveness) multiplied by 0.4, and then normalizes the calculated weights. Finally, the third stage, result fusion and preparation, begins. The system executes each MCP tool and outputs the optimized final result.

[0064] As an optional implementation, the system operation and maintenance strategy is obtained by weighted fusion of the operation results of the target tools based on tool weights. This includes: extracting operation and maintenance actions from the operation results of each target tool, wherein the operation results include at least one operation and maintenance action and the initial priority of the operation and maintenance action; establishing a correspondence between the operation and maintenance action, the initial priority, and the action weight for each operation and maintenance action, wherein the action weight is the tool weight of the target tool that generated the operation and maintenance action; multiplying the initial priority corresponding to the operation and maintenance action by the action weight to obtain the fusion priority; adding the fusion priorities of the same operation and maintenance action in different operation results to obtain the target priority; and arranging each operation and maintenance action in descending order according to the target priority, and integrating the operation and maintenance actions that precede the target priority to form the system operation and maintenance strategy.

[0065] Process the execution results generated by each target tool one by one. The execution results include at least one specific operation and maintenance action (e.g., "restart server A", "clean up the / var / log directory"), as well as an initial priority assigned to each operation and maintenance action, which represents the numerical value of the action's relative importance within its source tool.

[0066] For each extracted operation and maintenance action, a clear mapping relationship is established, recording the operation and maintenance action, its initial priority, and its corresponding action weight. The operation and maintenance action weight here directly adopts the tool weight of the target tool that generated this running result.

[0067] For each operational action, its initial priority is multiplied by its corresponding action weight to obtain a fused priority. This calculation is equivalent to passing the confidence level at the tool level to each specific action it suggests.

[0068] Since different tools may propose the same or similar actions (for example, both tool A and tool B suggest "increase caching"), it is necessary to identify these same operational actions, and then add up all the fusion priority values ​​from different tools for the same action to finally obtain a target priority for the action. This achieves a weighted aggregation of consensus on the suggestions of different tools for the same processing method.

[0069] All operations and maintenance actions are sorted in descending order of target priority (from high to low). Finally, the operations and maintenance actions listed before the target number are selected and integrated to form the final system operation and maintenance strategy. The system operation and maintenance strategy is an executable sequence of operations and maintenance actions sorted by comprehensive priority, and the target number is a pre-set positive integer.

[0070] Figure 4The diagram illustrates the weighted fusion mechanism provided in this application. First, three parallel input nodes represent tool result 1 and its weight W1, tool result 2 and its weight W2, and tool result N and its weight WN, respectively. These three inputs converge at the central weighted fusion node for processing. Next, the fusion priority is calculated using the formula that priority equals the sum of the products of the weights and the priorities. Then, the processing schemes are sorted. Finally, the process outputs Top-N processing schemes, selecting the top N optimal processing schemes, where N is a pre-defined positive integer one less than the target index.

[0071] This embodiment achieves a multi-source decision fusion mechanism based on efficiency and consensus by passing tool weights to each specific operation and maintenance action it outputs and by accumulating priority weights for the same actions proposed by different tools. The generated system operation and maintenance strategy not only incorporates the collective wisdom of multiple intelligent tools, but also effectively distinguishes the value of suggestions through quantified weights and priority ranking, thereby significantly improving the robustness of the final operation and maintenance decision.

[0072] As an optional embodiment, after obtaining the system operation and maintenance strategy, for the current target tool, the method further includes calculating the current round feedback effect of the current target tool in the following manner: executing the system operation and maintenance strategy and recording the target operation and maintenance actions in the system operation and maintenance strategy, wherein the target operation and maintenance actions are the actions in the running results output by the current target tool; obtaining the processing result score of each target operation and maintenance action; calculating the contribution ratio of the current target tool to the target operation and maintenance action during fusion, and calculating the contribution effect value of the current target tool to the target operation and maintenance action based on the processing result score and the contribution ratio; adding the contribution effect values ​​of each target operation and maintenance action to obtain the current round feedback effect of the current target tool, and updating the recent feedback effect of the current target tool in the MCP tool library with the current round feedback effect.

[0073] The system operation and maintenance strategy generated by weighted fusion is executed, and all operation and maintenance actions adopted in the strategy are recorded. The actions contributed by the current target tool, which is calculating its feedback effect, are called target operation and maintenance actions.

[0074] Obtain the processing result score after each target operation and maintenance action is executed. The processing result score is a quantitative indicator used to measure the execution effect of the action (e.g., successfully solve the corresponding problem, partially alleviate it, be ineffective or have a negative impact).

[0075] For each target operation and maintenance action, the contribution ratio of the current target tool to that action in the previous fusion steps is first calculated. The contribution ratio can be determined based on the ratio of the tool weight of the current target tool when generating the action to the sum of the weights of all tools that generated the same action. Then, the processing result score of the target operation and maintenance action is multiplied by its contribution ratio to obtain the contribution effect value of the current target tool to this target operation and maintenance action. The contribution effect value reflects the effect responsibility that the tool should share in this round of processing for each specific action it proposed and adopted.

[0076] By using contribution proportions, it avoids simply averaging or attributing the effectiveness of all actions to a single tool. Instead, it fairly distributes the success or failure of an action according to the proportion of each tool's contribution to the decision-making process. This ensures the fairness and accuracy of feedback, with tools that are highly effective, have high weight, and whose recommendations are adopted receiving more significant feedback.

[0077] The summation of the contribution values ​​calculated by the current target tool for all target operation and maintenance actions yields the overall feedback effect of the current target tool in this round of processing. This feedback effect is then updated in the MCP tool library to record the latest performance data of the current target tool in previous target rounds.

[0078] The calculated feedback effect of this round is updated to the MCP tool library, so that the key performance indicator of the tool's recent feedback effect can dynamically and realistically reflect its latest performance. This provides the latest basis for calculating the tool weight when handling the next fault, thereby driving the system to continuously learn from experience and achieve self-optimization.

[0079] Figure 5 The closed-loop feedback and optimization diagram provided for this application is shown in the figure. First, the processing plan is executed; then, the system enters the stage of collecting and processing feedback; next, the system calculates the processing effect based on four dimensions: success rate, processing time, recovery status, and resource consumption. The values ​​of these four dimensions are pre-quantified, and the calculation formula is: success rate multiplied by 0.4, processing time multiplied by 0.3, recovery status multiplied by 0.2, and resource consumption multiplied by 0.1; then, the system enters the effect judgment stage, and branches processing according to the calculated effect value: if the effect is greater than 0.7, the weight is increased, specifically by multiplying the weight by 1.1; if the effect is less than 0.5, the weight is decreased, specifically by multiplying the weight by 0.9; if the effect is between 0.5 and 0.7, the weight remains unchanged; after all branch processing is completed, the system normalizes the weights, then updates the tool library, marking the completion of closed-loop optimization, and finally returns to the previous tool selection stage, forming a continuous optimization cycle.

[0080] The system operation and maintenance processing method based on MCP tools provided in this application is applied to the overall architecture of an intelligent operation and maintenance system based on a large model. It mainly consists of five parts: user interface layer, parameter acquisition layer, core processing layer, tool library layer, and feedback optimization layer. After receiving the operation and maintenance request, the user interface layer enters the parameter acquisition layer. The parameter acquisition layer contains a parameter acquisition module, which attempts to retrieve data from the Redis cache and MySQL database in turn. If both fail, it calls the monitoring service API to obtain real-time monitoring data. The acquired data then enters the middle core processing layer. First, the large model status assessment module assesses the system status, and then the large model tool selection module enters the large model tool selection module. This module contains three rounds of optimization: the first round selects basic tools, the second round adjusts weights, and the third round fuses the results. The processed data flows to the bottom tool library layer, which contains the MCP tool library management, specifically listing MCP tool 1 - CPU monitoring tool, MCP tool 2 - memory cleanup tool, MCP tool 3 - disk cleanup tool, MCP tool 4 - service restart tool, and MCP tool 5 - load balancing tool. After these tools are executed, the results are aggregated to the result fusion module, which finally outputs the processing solution. The feedback optimization layer is used to collect feedback after the processing solution is output and enter it into the effect evaluation module. Based on the evaluation results, the tool library is updated to complete the closed-loop optimization and ensure that the system can continuously learn and improve.

[0081] This application provides a system operation and maintenance method based on MCP tools, including: upon receiving an operation and maintenance request, acquiring historical fault data and current operating data of the operation and maintenance system, and constructing status assessment prompts based on the historical fault data and current operating data; calling a preset large model and inputting the status assessment prompts into the preset large model to obtain the system status score, risk level, and current anomaly type of the operation and maintenance system obtained after status assessment by the preset large model; acquiring an MCP tool list and constructing tool selection prompts based on the system status score, risk level, current anomaly type, and the MCP tool list; inputting the tool selection prompts into the preset large model to obtain a candidate tool set selected by the preset large model; determining and running target tools in the candidate tool set through a preset strategy to obtain the running results, wherein the preset strategy is to first select target tools based on the tool prediction effect, and then calculate the tool weight of each target tool based on the tool performance; and weighting and fusing the running results of the target tools according to the tool weights to obtain the system operation and maintenance strategy. By calling a pre-set large model to assess the system status, and then using the pre-set large model to intelligently select tools from the MCP tool list based on the assessment results, the tool operation results are optimized and integrated by sampling a multi-step progressive strategy based on prediction effect and tool performance, thereby obtaining the best operation and maintenance strategy. This solves the problem of difficulty in dynamically generating the most suitable operation and maintenance handling strategy according to the real-time status of the operation and maintenance system.

[0082] According to another aspect of the embodiments of this application, this application provides a system operation and maintenance processing device based on the MCP tool, such as... Figure 6 As shown, it includes: The acquisition module 601 is used to acquire historical fault data and current operating data of the operation and maintenance system when a maintenance request is received, and to construct status assessment prompt words based on the historical fault data and current operating data; Module 602 is used to call the preset large model and input the status assessment prompt words into the preset large model to obtain the system status score, risk level and current anomaly type of the operation and maintenance system after the status assessment is performed by the preset large model. Module 603 is used to obtain the list of MCP tools and select prompt words based on the system status score, risk level, current anomaly type, and the list of MCP tools. Input module 604 is used to input tool selection prompts into a preset large model to obtain a set of candidate tools selected by the preset large model; The determination module 605 is used to determine and run target tools in the candidate toolset through a preset strategy to obtain the running results. The preset strategy is to first select target tools based on the prediction effect of the tools, and then calculate the tool weight of each target tool based on the tool performance. The fusion module 606 is used to weight and fuse the running results of the target tool according to the tool weight to obtain the system operation and maintenance strategy.

[0083] It should be noted that the acquisition module 601 in this embodiment can be used to execute step 101 in this application embodiment, the calling module 602 in this embodiment can be used to execute step 102 in this application embodiment, the construction module 603 in this embodiment can be used to execute step 103 in this application embodiment, the input module 604 in this embodiment can be used to execute step 104 in this application embodiment, the determination module 605 in this embodiment can be used to execute step 105 in this application embodiment, and the fusion module 606 in this embodiment can be used to execute step 106 in this application embodiment.

[0084] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments.

[0085] According to another aspect of the embodiments of this application, this application provides an electronic device, such as... Figure 7As shown, the device includes a memory 701, a processor 703, a communication interface 705, and a communication bus 707. The memory 701 stores a computer program that can run on the processor 703. The memory 701 and the processor 703 communicate through the communication interface 705 and the communication bus 707. When the processor 703 executes the computer program, it implements the steps of the above method.

[0086] The memory and processor in the aforementioned electronic devices communicate with each other via a communication bus and a communication interface. The communication bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into an address bus, a data bus, a control bus, etc.

[0087] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0088] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0089] According to another aspect of the embodiments of this application, a computer-readable medium having processor-executable non-volatile program code is also provided.

[0090] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments, and will not be repeated here.

[0091] In specific implementation, the embodiments of this application can be referred to the above embodiments and have corresponding technical effects.

[0092] It is understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described herein, or combinations thereof.

[0093] For software implementation, the techniques described herein can be implemented by units that perform the functions described herein. The software code can be stored in memory and executed by a processor. The memory can be implemented in the processor or external to the processor.

[0094] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0095] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0096] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0097] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0098] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0099] If the aforementioned function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, or the part that contributes to the prior art, or a part of the technical solution, 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 a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps 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 USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks. It should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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. In the absence of further restrictions, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0100] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A system operation and maintenance method based on MCP tools, characterized in that, include: Upon receiving an operation and maintenance request, the system obtains historical fault data and current operating data of the operation and maintenance system, and constructs status assessment prompts based on the historical fault data and the current operating data. Call the preset large model and input the status assessment prompt words into the preset large model to obtain the system status score, risk level and current anomaly type of the operation and maintenance system after the status assessment is performed by the preset large model; Obtain the MCP tool list, and select prompt words based on the system status score, the risk level, the current anomaly type, and the MCP tool list construction tool; Input the tool selection prompts into the preset large model to obtain a set of candidate tools selected by the preset large model; The target tool is determined and run in the candidate tool set by a preset strategy to obtain the running result. The preset strategy is to first select the target tool based on the prediction effect of the tool, and then calculate the tool weight of each target tool based on the tool performance. The system operation and maintenance strategy is obtained by weighting and fusing the running results of the target tool according to the tool weight.

2. The method according to claim 1, characterized in that, The acquisition of historical fault data and current operational data of the operation and maintenance system includes: Attempt to read the current operating data and historical fault data of the operation and maintenance system from the target cache; If an attempt to read from the target cache fails, an attempt is made to read the current running data and the historical fault data from the target database. If an attempt to read from the target database fails, the historical fault data and the current running data are generated using preset default values.

3. The method according to claim 1, characterized in that, The step of inputting the status assessment prompts into the preset large model to obtain the system status score, risk level, and current anomaly type of the operation and maintenance system obtained after status assessment by the preset large model includes: The status assessment prompt is sent as the first input parameter to the first application interface of the preset large model, wherein the status assessment prompt includes an instruction to return data in the target format; Receive the first response data returned by the first application interface, and extract the response data system status score, the risk level, and the current anomaly type from the first response data.

4. The method according to claim 1, characterized in that, The step of inputting the tool selection prompts into the preset large model to obtain a candidate tool set selected by the preset large model includes: The tool selection prompt is sent as a second input parameter to the second application interface of the preset large model, and the second response data returned by the second application interface is received, wherein the tool selection prompt includes an instruction to return data in the target format; The tool identifier list is obtained by parsing the second response data. The tool identifier list includes at least one tool identifier recommended by the preset big model after analysis based on the system status score, the risk level, the current anomaly type and the functional description of each tool in the MCP tool list. The tool identifier list is matched and verified with the available tools in the operation and maintenance system, and the MCP tools corresponding to the target tool identifiers that pass the matching and verification are integrated into the candidate tool set.

5. The method according to claim 1, characterized in that, The process of determining and running the target tool from the candidate toolset using a preset strategy to obtain the running results includes: The historical performance of each candidate tool in the candidate toolset is obtained, as well as the state matching degree between the candidate tool and the system state score and the load matching degree between the candidate tool and the current resource load of the operation and maintenance system. The historical performance represents the overall historical performance of the candidate tool. The predicted effect of the candidate tool is obtained by weighting the historical effect, the state matching degree and the load matching degree. The predicted effect of the tool is compared with an effect threshold, and the target tool whose predicted effect is greater than the effect threshold is selected from the candidate tools; The recent feedback effect and time decay coefficient of the target tool are obtained from the MCP tool library, and the initial weight is calculated using the historical effect, the recent feedback effect and the time decay coefficient, wherein the recent feedback effect is the feedback effect of the target tool in the target rounds before the current round; The initial weights of each target tool are normalized to obtain the tool weights of each target tool.

6. The method according to claim 1, characterized in that, The step of weighting and fusing the running results of the target tool according to the tool weight to obtain the system operation and maintenance strategy includes: Extract the operation and maintenance actions from the running results of each of the target tools, wherein the running results include at least one operation and maintenance action and the initial priority of the operation and maintenance action; For each of the operation and maintenance actions, a correspondence is established between the operation and maintenance action, the initial priority, and the action weight, wherein the action weight is the tool weight of the target tool that generates the operation result to which the operation and maintenance action belongs; The initial priority corresponding to the operation and maintenance action is multiplied by the action weight to obtain the fused priority; For the same operation and maintenance action in different operation results, the fusion priority of the same operation and maintenance action is added together to obtain the target priority; The operation and maintenance actions are arranged in descending order according to the target priority, and the operation and maintenance actions that are ranked before the target number are integrated to form the system operation and maintenance strategy.

7. The method according to claim 6, characterized in that, After obtaining the system operation and maintenance strategy, for the current target tool, the method further includes calculating the current round feedback effect of the current target tool in the following manner: Execute the system operation and maintenance strategy and record the target operation and maintenance actions in the system operation and maintenance strategy, wherein the target operation and maintenance actions are the actions in the operation results output by the current target tool; Obtain a score for the processing result of each of the target maintenance actions; Calculate the contribution ratio of the current target tool to the target operation and maintenance action during the fusion process, and calculate the contribution effect value of the current target tool to the target operation and maintenance action based on the processing result score and the contribution ratio; The contribution effect values ​​of each of the target operation and maintenance actions are summed to obtain the current round feedback effect of the current target tool, and the current round feedback effect is used to update the recent feedback effect of the current target tool in the MCP tool library.

8. A system operation and maintenance processing device based on MCP tools, characterized in that, include: The acquisition module is used to acquire historical fault data and current operating data of the operation and maintenance system when a maintenance request is received, and to construct status assessment prompt words based on the historical fault data and the current operating data. The calling module is used to call the preset large model and input the status assessment prompt words into the preset large model to obtain the system status score, risk level and current anomaly type of the operation and maintenance system after the status assessment is performed by the preset large model. The construction module is used to obtain the MCP tool list and select prompt words based on the system status score, the risk level, the current anomaly type, and the MCP tool list construction tool. The input module is used to input the tool selection prompts into the preset large model to obtain the candidate tool set selected by the preset large model; The determination module is used to determine and run target tools in the candidate tool set according to a preset strategy, and obtain the running results. The preset strategy is to first select target tools based on the prediction effect of the tools, and then calculate the tool weight of each target tool based on the tool performance. The fusion module is used to perform weighted fusion of the running results of the target tool according to the tool weight to obtain the system operation and maintenance strategy.

9. An electronic device comprising a memory, a processor, a communication interface, and a communication bus, wherein the memory stores a computer program executable on the processor, and the memory and the processor communicate via the communication bus and the communication interface, characterized in that... When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable medium having processor-executable non-volatile program code, characterized in that, The program code causes the processor to execute the method of any one of claims 1 to 7.

Images