Transaction system operation and maintenance troubleshooting method and system based on agent, and electronic equipment

Abstract

The invention relates to a transaction system operation and maintenance troubleshooting method based on an agent, and the method comprises the following steps: responding to an input operation and maintenance event alarm and / or an operation and maintenance knowledge question, enabling a coordination agent to judge whether the input is related to operation and maintenance, if yes, calling a plan agent, and if not, calling a local large language model to generate reply information for the input; in response to the calling plan agent, generating operation and maintenance fault check plan process information based on a preset system operation and maintenance knowledge base and a preset prompt word, and calling an execution management agent; and in response to calling of the execution management agent, calling a subordinate execution agent to execute the operation and maintenance fault check plan process based on the operation and maintenance fault check plan process information, and outputting an execution result after execution is completed, thereby realizing operation and maintenance fault troubleshooting of the transaction system. Compared with the prior art, the method has the advantages of fault scene self-adaption, cross-scene troubleshooting, low false positive rate and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer technology, in particular to a transaction system operation and maintenance fault diagnosis method and system based on an agent, and an electronic device. BACKGROUND

[0002] As a core infrastructure in the financial field, the safe operation of the transaction system requires high real-time and accuracy in fault diagnosis and disposal. However, the current mainstream fault diagnosis technology mainly relies on rule engines or traditional AI models, such as the isolated forest algorithm based on machine learning, the LSTM anomaly detection model of deep learning, and the rule matching of expert systems. These methods have achieved automatic detection to some extent, but generally have the following significant problems:

[0003] (1) Algorithm effectiveness lags behind: The rule base relies on manual updates and cannot dynamically adapt to changes in transaction system patterns, resulting in detection delays and low effectiveness. At the same time, the training data of traditional models is often static and difficult to cope with real-time evolving system behavior, resulting in a high false negative rate.

[0004] (2) Poor flexibility: Single AI models are usually designed for specific system scenarios or fault types. If the fault conditions exceed the preset range, the fault diagnosis mechanism cannot be effectively triggered. In addition, these models often ignore context associations, resulting in inadequate handling of complex dependency relationships.

[0005] (3) Insufficient resource consumption and scalability: Traditional methods require a large amount of computing resources when processing massive transaction data, and the model training period is long, making it difficult to efficiently expand to distributed or cloud-based transaction environments. At the same time, the high false positive rate easily leads to alert fatigue, increasing the workload of operation and maintenance personnel.

[0006] (4) Lack of reflection and iteration mechanism: Existing technologies are mostly one-way detection, lacking automatic reflection and optimization of the diagnosis process, making it difficult to completely eliminate repeated faults, and in high-frequency trading scenarios, the response time is long, which may cause economic losses.

[0007] Chinese patent application publication No. CN119938376A provides a system fault analysis and disposal method, device, equipment and computer program product, which automatically analyzes and disposes system faults through the combination of vector knowledge base and large language model, solves the problem of low fault disposal efficiency in the prior art, and realizes intelligent and automatic fault self-healing in the whole process. However, it has the following problems:

[0008] (1) It uses a single search-disposal process, which is difficult to cope with complex fault scenarios such as multi-component coupling and high concurrency in the transaction system, and lacks dynamic planning and multi-path adaptation capability for complex fault scenarios.

[0009] (2) Only the results of the disposal were summarized, without designing a closed-loop optimization process and lacking a reflection and iteration mechanism for uncertain conclusions.

[0010] (3) The lack of manual review process poses a risk of misoperation and lacks intervention and security guarantees.

[0011] In summary, there is currently a lack of a method for troubleshooting operational and maintenance faults in transaction systems based on intelligent agents, in order to solve or partially solve the aforementioned problems. Summary of the Invention

[0012] The purpose of this invention is to overcome the shortcomings of the existing technology by providing a method, system, and electronic device for troubleshooting the operation and maintenance of a trading system based on intelligent agents, so as to solve or partially solve the problem of low efficiency of existing troubleshooting methods when facing trading systems with high complexity and event density.

[0013] The objective of this invention can be achieved through the following technical solutions:

[0014] One aspect of the present invention provides a method for troubleshooting operational and maintenance faults in a transaction system based on intelligent agents, comprising the following steps:

[0015] In response to inputting operation and maintenance event alarms and / or operation and maintenance knowledge questions, the coordinating agent determines whether the input is related to operation and maintenance. If so, it calls the planning agent; otherwise, it calls the local large language model to generate response information for the input.

[0016] In response to the call to the planning agent, based on the pre-set system operation and maintenance knowledge base and preset prompt words, the operation and maintenance fault inspection plan process information is generated, and the execution management agent is called;

[0017] In response to the call to the execution management agent, based on the operation and maintenance fault inspection plan process information, the subordinate execution agent is called to execute the operation and maintenance fault inspection plan process, and the execution result is output after the execution is completed, thereby realizing the troubleshooting of operation and maintenance faults in the transaction system.

[0018] As a preferred technical solution, the process of generating operation and maintenance fault inspection plan process information by the planning intelligent agent includes the following steps:

[0019] The intelligent agent plans to access the system operation and maintenance knowledge base to obtain preliminary information;

[0020] Obtain capability boundary information for each executing agent;

[0021] Based on the aforementioned prerequisite information and capability boundary information, the intelligent agent executes the operation and maintenance fault inspection plan process based on preset prompt words.

[0022] As a preferred technical solution, the prompt word is configured to guide the planning agent to execute:

[0023] Determine if it is an operation and maintenance event. If yes, obtain the system-level component name information in the event. If no, skip the current step.

[0024] In each step of the plan, select the execution agent to be invoked and define the responsibility and output of the current step;

[0025] Combine consecutive steps assigned to the same executing agent into one step;

[0026] Generate operation and maintenance fault inspection plan process information.

[0027] As a preferred technical solution, the process of calling the subordinate execution intelligent agent to execute the operation and maintenance fault inspection plan and outputting the execution result after execution includes the following steps:

[0028] The preceding results obtained from the executed steps are passed as parameters to the executing agent to achieve dynamic context awareness;

[0029] Based on the implementation results returned by the executing agent, a branch path is triggered when preset conditions are met;

[0030] The completion rate is used to determine the coverage of verification steps and calculate the degree of goal achievement. Uncertain conclusions are identified, and the planning agent is called again to update the operation and maintenance fault inspection plan process information.

[0031] As a preferred technical solution, the executing intelligent agent includes:

[0032] The trading system check agent is invoked when the alarm includes the trading system and related components. It checks the status of the trading system and returns a list of tools that need to be called. The tools include trading market consistency check, seat quantity check, trading system status check, trading core process check, and pre-trading check.

[0033] The server inspection agent is used to check the server hardware status and return a list of tools that need to be invoked, including CPU usage check, memory usage check, disk space check, and SSH connectivity check.

[0034] The network inspection agent is used to check the server network status and return a list of tools that need to be invoked, including network health checks and database connection checks.

[0035] The information extraction agent is used to extract information from alarm events and return a list of tools to be invoked. These tools include system-level query, component name extraction, latest system log query, IP address extraction, and event time extraction.

[0036] As a preferred technical solution, after generating the operation and maintenance fault inspection plan process information, the method further includes:

[0037] Manual checks and adjustments are made to the maintenance and troubleshooting plan process information.

[0038] As a preferred technical solution, the system operation and maintenance knowledge base includes system definition, information on components within the system, relationships between components, and historical event processing procedures.

[0039] As a preferred technical solution, the operation and maintenance event alarms include trading system failures, trading blockages, and changes in the number of logged-in seats.

[0040] Another aspect of the present invention provides an agent-based transaction system operation and maintenance fault diagnosis system for implementing the aforementioned transaction system operation and maintenance fault diagnosis method, the system comprising:

[0041] The coordinating agent is used to respond to input operation and maintenance event alarms and / or operation and maintenance knowledge questions, determine whether the input is related to operation and maintenance, if so, call the planning agent, if not, call the local large language model to generate response information for the input;

[0042] The planning agent is used to respond to calls, generate maintenance fault inspection plan process information based on the pre-set system operation and maintenance knowledge base and preset prompt words, and call the execution management agent;

[0043] An execution management intelligent agent is used to respond to a call to the execution management intelligent agent, and based on the operation and maintenance fault inspection plan process information, to call the subordinate execution intelligent agent to execute the operation and maintenance fault inspection plan process, and output the execution result after the execution is completed, so as to realize the troubleshooting of operation and maintenance faults in the transaction system.

[0044] In another aspect, the present invention provides an electronic device including one or more processors, a memory, and one or more programs stored in the memory, said one or more programs including instructions for executing the aforementioned agent-based transaction system operation and maintenance fault diagnosis system.

[0045] Compared with the prior art, the present invention has at least one of the following beneficial effects:

[0046] (1) Fault scenario adaptation: The present invention calls the planning agent, based on the pre-set system operation and maintenance knowledge base and preset prompt words, to generate operation and maintenance fault inspection plan process information, dynamically generate execution process plan instead of manual pre-definition, fully consider the characteristics of the application scenario of transaction system fault investigation, and carry out targeted knowledge injection and prompt engineering for the agent, overcome the limitations of traditional static rule engine, realize fault scenario adaptation, improve path generation efficiency and reduce false alarm rate.

[0047] (2) Achieve cross-scenario fault diagnosis: In this invention, the management agent passes the previous results obtained from the executed steps as parameters to the execution agent. The execution agent provides feedback on the list of tools to be called, thereby achieving dynamic context awareness. Based on the implementation results returned by the execution agent, a branch path is triggered when the preset conditions are met. Through dynamic context awareness, condition triggering mechanism and precise tool matching, cross-scenario fault diagnosis is achieved, shortening the response time, improving resource utilization, and reducing the burden of manual operation and maintenance.

[0048] (3) Improve the accuracy of fault diagnosis: The management agent of the present invention covers and calculates the target achievement by judging the completion of the verification steps, identifies uncertain conclusions, and calls the planning agent again to update the operation and maintenance fault inspection plan process information, thereby reducing the probability of false positives in operation and maintenance fault identification and reducing the cost of manual operation and maintenance intervention.

[0049] (4) The generated operation and maintenance fault inspection plan process is detailed: In the process of generating the execution plan, the present invention introduces the automatic query of the system operation and maintenance knowledge base, which makes up for the shortcomings of the lack of business-related knowledge when using the general large model, so that a more targeted inspection plan can be formulated, the rationality is increased, the execution plan details are more detailed, and the component information in the knowledge base can be used to propose related inspection processes.

[0050] (5) Strong security: After generating the execution plan, the present invention can manually verify and adjust the capability, and can optimize and modify the inspection items and agent selection logic on the basis of the original plan, ensuring that the subsequent inspection process meets business expectations. This improves the accuracy of the subsequent agent execution and meets the security management process specifications. Attached Figure Description

[0051] Figure 1 This is a flowchart of the agent-based transaction system operation and maintenance fault diagnosis method in the embodiment;

[0052] Figure 2 This is a schematic diagram illustrating troubleshooting for operational and maintenance faults in the transaction system as described in the embodiment.

[0053] Figure 3 This is a schematic diagram of the agent-based transaction system operation and maintenance fault diagnosis system in the embodiment.

[0054] Figure 4 This is a schematic diagram of the electronic device in the embodiment. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0056] Example 1

[0057] To address the problems existing in the aforementioned prior art, this embodiment provides a method for troubleshooting operational and maintenance faults in a transaction system based on intelligent agents. The aim is to significantly improve the adaptability and robustness of the algorithm by automatically reflecting and adaptively adjusting during the troubleshooting process, thereby reducing manual operation and maintenance costs. When an incident occurs, it can assist operation and maintenance personnel in reducing troubleshooting time and quickly locating problems. (See also...) Figure 1 and Figure 2 The method includes the following steps:

[0058] Step S1: In response to inputting an operation and maintenance event alarm and / or operation and maintenance knowledge question, the coordinating agent determines whether the input is related to operation and maintenance. If so, the planning agent is invoked; otherwise, the local large language model is invoked to generate response information for the input.

[0059] In this step, the triggering conditions include the occurrence of operation and maintenance events as well as non-operation and maintenance events (such as asking operation and maintenance knowledge questions).

[0060] When an operational event occurs, it will be displayed on the alarm page, and the alarm content will include business-related alarms such as trading system failure, trading blockage, and changes in the number of logged-in seats.

[0061] In addition to the occurrence of operational events, operations and maintenance personnel hope to be able to simulate certain operational events and verify the inspection logic or create training cases.

[0062] The purpose of the coordinating agent is to determine the type of question and coordinate the response process. For example, when the question is unrelated to operations and maintenance knowledge, a casual conversation mode will be entered, and the regular large model will be invoked. When the question is related to operations and maintenance, it will fall into the subsequent process in this solution.

[0063] The coordinating agent can construct system prompts in the following format to perform tasks within a specified scope: "You are an operations and maintenance AI assistant. You excel at greetings and small talk, while delegating complex tasks to specialized planning agents."

[0064] General-purpose large language models are typically DeepSeek or Qwen-type large models. Due to the special nature of operation and maintenance scenarios where internet access is not possible, these large models are all locally deployed versions that conform to the OpenAI access format.

[0065] Step S2: In response to the call to the planning agent, based on the preset system operation and maintenance knowledge base and preset prompt words, generate operation and maintenance fault inspection plan process information, and call the execution management agent.

[0066] In this step, the main purpose of the planning agent is to break down the collected operational event alarms into execution flow plans. These execution flow plans are the core of this method, primarily relying on the system's operational knowledge base and being fed into the agent via system prompts.

[0067] Specifically, step S2 may include steps S201-S204:

[0068] S201, the planning agent will first call the system operation and maintenance knowledge base MCP tool. This knowledge base stores the association information of components and systems, as well as historical event content. This kind of preliminary information can help the planning agent make reasonable plans.

[0069] S202, the planned intelligent agent will collect the capability boundaries of other intelligent agents based on the system prompt words, and use them as part of the prompt words.

[0070] For example, the prompts for capability boundaries include:

[0071] The trading system check agent is used to check the status of the trading system. When the alarm includes the trading system and related components, this agent needs to be invoked. This agent will only return a list of tools that need to be invoked.

[0072] The server inspection agent checks the server hardware status, including hardware metrics such as disk, memory, and CPU. This agent only returns a list of tools that need to be invoked.

[0073] A network inspection agent is used to check the server's network status, including network latency, etc. This agent will only return a list of tools that need to be invoked.

[0074] An information extraction agent is used to extract information from alarm events. The tools this agent can access include, but are not limited to: system-level queries, component name extraction, latest system log queries, IP address extraction, and event time extraction.

[0075] S203, based on knowledge base information and the capabilities of other intelligent agents, automatically plans the execution plan for the overall inspection.

[0076] For example, system prompts include:

[0077] First, enter thinking mode and rephrase the user's needs in your own words.

[0078] - Create a step-by-step plan.

[0079] - If it is an operational event, the first step of the plan is to obtain necessary information such as the names of system-level components in the event. If it is not an operational event but an inspection requirement or question, this step can be skipped.

[0080] Next, select the inspection agent to be invoked. The available inspection agents are the transaction system inspection agent, the server inspection agent, and the network inspection agent.

[0081] - The final step is to summarize the output of the above steps and output the inspection conclusion. If the user needs to write an inspection report, the report writing agent will be called; otherwise, it is not necessary.

[0082] - Formatting requirements: The description of the step object needs to define the agent's responsibilities and outputs.

[0083] - Combine consecutive steps assigned to the same agent into one step.

[0084] - Generate plans using the same language as the user.

[0085] S204 outputs the generated execution plan and provides it to the next agent.

[0086] In this embodiment, the system operation and maintenance knowledge base is a large model knowledge base used to store system operation and maintenance information. It is used to provide auxiliary information for the planning and execution process of the intelligent agent. The knowledge base stores business-related information including system definitions, information on components within the system, relationships between components, and historical event handling processes. The operation and maintenance knowledge base can be updated and entered periodically according to business needs. The key processes of document cleaning, text vector generation, and vector database storage during the entry process will not be elaborated here.

[0087] Preferably, after generating the operation and maintenance fault inspection plan process information, step S2 also includes a manual adjustment step, which serves as an expert knowledge judgment and reliability review step. If manual confirmation is needed, the manual adjustment step is initiated. In this step, the called intelligent agents can be adjusted, and specific inspection items can be added or deleted. Subsequently, the execution management intelligent agent will call sub-intelligent agents according to the modified process.

[0088] Step S3: In response to the call to the execution management intelligent agent, based on the operation and maintenance fault inspection plan process information, the subordinate execution intelligent agent is called to execute the operation and maintenance fault inspection plan process, and the execution result is output after the execution is completed, so as to realize the troubleshooting of operation and maintenance faults in the transaction system.

[0089] The execution management agent primarily calls upon the execution agents it manages in an orderly manner according to the process orchestrated by the planning agent. The names of these execution agents must match the agent names mentioned in the planning agent's prompts. When executing tasks assigned by the planning agent, the management agent continuously checks whether the plan has been completed. If not, it continues to call subordinate execution agents based on the plan and context information until the plan is completed. When calling subordinate execution agents, the execution management agent determines which specific tools to use to complete the task. Different subordinate execution agents access different types of tools, therefore, it is necessary to select the appropriate agents to use.

[0090] After the management agent determines whether the plan has been completed and receives the completion result, it outputs a report and allows users to ask follow-up questions based on the results.

[0091] In this embodiment, the executing agent includes:

[0092] (1) Information extraction agent.

[0093] Primarily used for information extraction in alarm events, this agent integrates with tools including but not limited to: system-level queries, component name extraction, latest system log queries, IP address extraction, and event time extraction. The extracted information provides parameters and necessary details for calls to other agents.

[0094] (2) The transaction system checks the agent.

[0095] The main function is to conduct relevant checks on the trading system based on the components and system information extracted from alarm events. The tools accessed by this intelligent agent include, but are not limited to: trading data consistency checks, seat quantity checks, trading system status checks, core trading process checks, and pre-trading checks.

[0096] (3) The hardware host checks the intelligent agent.

[0097] The main function is to check the corresponding hardware host based on the IP address extracted from the alarm event. The tools accessed by this intelligent agent include, but are not limited to: CPU utilization check, memory utilization check, disk space check, SSH connectivity check, etc.

[0098] (4) Network inspection of intelligent agents.

[0099] The main function is to check the network of the corresponding host based on the IP address extracted from the alarm event. The tools accessed by this agent include, but are not limited to: network health status check, database connection check, etc.

[0100] (5) Report generated intelligent agent.

[0101] Generally, it is the agent called in the last step. The purpose of this agent is to summarize and analyze the call results of the above other subordinate agents and the planning and execution content of the planning agent, and generate a detection report.

[0102] In summary, this method has the following characteristics:

[0103] (1) Dynamic process planning mechanism based on the planning agent

[0104] Based on the operation and maintenance knowledge base and system prompt words, the planning agent dynamically generates an execution process plan, and fully considers the characteristics of the application scenario of transaction system fault troubleshooting, and injects targeted knowledge and performs prompt engineering on the agent. For example, according to the content of the alarm event, when the content includes a network fault, it intelligently judges whether to call the network inspection agent. This design breaks through the limitations of traditional static rule engines, realizes fault scenario adaptation, especially performs outstandingly in the case of multi-system linkage faults, improves the path generation efficiency, and reduces the false alarm rate.

[0105] (2) Multi-agent collaborative architecture based on the execution management agent

[0106] Dynamic context awareness: The execution management agent inherits the previous result as a parameter and passes it to the subordinate agent.

[0107] Condition trigger mechanism: Based on the real-time result, interrupt the preset process and trigger the branch path. For example, when the implementation result returns disk anomalies, trigger the call to the in-depth analysis tool.

[0108] Accurate tool matching: Select the exclusive tool according to the component type.

[0109] Through the above mechanisms, cross-scenario fault troubleshooting is jointly achieved, covering the transaction system, hardware, and network at the same time, shortening the response time to the second level, realizing the improvement of resource utilization rate, and reducing the burden of manual operation and maintenance.

[0110] (3) Reflection and iteration mechanism based on the execution management agent

[0111] Two-level verification: The execution management agent has a built-in completion judgment. First, verify the coverage of the previous steps, and then analyze the degree of goal achievement. Exemplarily, first analyze whether the previous steps are completed, and then judge the logical consistency of the evidence chain.

[0112] Iteration trigger: When an uncertain conclusion such as "possible memory leak" is identified, adjust the parameters and re-call the agent or require the planning agent to re-plan.

[0113] By setting the reflection and iteration mechanism, the probability of false positive judgment in operation and maintenance fault identification is reduced, and the cost of manual operation and maintenance intervention is reduced.

[0114] (5) System operation and maintenance knowledge base integration

[0115] During the execution plan generation process, the system operation and maintenance knowledge base is automatically consulted, which fills the gap in business-related knowledge when using a general large model, thereby enabling the formulation of more targeted inspection plans. After integrating the knowledge base, the process rationality increases compared to the original plan, the execution plan details are more comprehensive, and the component information within the knowledge base can be used to propose related inspection processes.

[0116] (6) Expert knowledge verification and adjustment

[0117] After the execution plan is generated, this method provides the ability for human experts to verify and adjust it, which can optimize and modify the inspection items and agent selection logic based on the original plan, ensuring that the subsequent inspection process meets business expectations.

[0118] While the execution plans generated by the planning agent can meet the expectations of the operation and maintenance event inspection in most cases, in order to avoid misoperation during the actual operation and maintenance process, it is necessary to add the ability of expert knowledge verification and adjustment before the formal execution of the inspection agent. This not only improves the accuracy of the subsequent execution of the agent, but also better meets the safety management process specifications.

[0119] Example 2

[0120] Based on Example 1, this example provides an agent-based transaction system operation and maintenance fault diagnosis system to implement the aforementioned transaction system operation and maintenance fault diagnosis method. (See also...) Figure 3 The system includes:

[0121] (1) Coordinating agent, used to respond to input operation and maintenance event alarms and / or operation and maintenance knowledge questions, determine whether the input is related to operation and maintenance, if so, call the planning agent, if not, call the local large language model to generate response information for the input;

[0122] (2) Planning agent, which is used to respond to the call, generate operation and maintenance fault inspection plan process information based on the pre-set system operation and maintenance knowledge base and preset prompt words, and call the execution management agent;

[0123] (3) Execution management intelligent agent, which is used to respond to the call to the execution management intelligent agent, and based on the operation and maintenance fault inspection plan process information, call the subordinate execution intelligent agent to execute the operation and maintenance fault inspection plan process, and output the execution result after the execution is completed, so as to realize the operation and maintenance fault inspection of the transaction system.

[0124] Example 3

[0125] Based on Embodiment 1, this embodiment provides an electronic device, including one or more processors, a memory, and one or more programs stored in the memory, wherein the one or more programs include instructions for executing the aforementioned agent-based transaction system operation and maintenance fault diagnosis system.

[0126] like Figure 4 At the hardware level, the electronic device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for the business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to achieve the above-mentioned functions. Figure 1 The method described herein. Of course, in addition to software implementation, this invention does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0127] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0128] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0129] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for troubleshooting operational and maintenance faults in a transaction system based on intelligent agents, characterized in that, Includes the following steps: In response to inputting operation and maintenance event alarms and / or operation and maintenance knowledge questions, the coordinating agent determines whether the input is related to operation and maintenance. If so, it calls the planning agent; otherwise, it calls the local large language model to generate response information for the input. In response to the call to the planning agent, based on the pre-set system operation and maintenance knowledge base and preset prompt words, the operation and maintenance fault inspection plan process information is generated, and the execution management agent is called; In response to the call to the execution management agent, based on the operation and maintenance fault inspection plan process information, the subordinate execution agent is called to execute the operation and maintenance fault inspection plan process, and the execution result is output after the execution is completed, thereby realizing the troubleshooting of operation and maintenance faults in the transaction system.

2. The method for troubleshooting operational and maintenance faults in a transaction system based on intelligent agents according to claim 1, characterized in that, The process of generating maintenance fault inspection plan flow information by the planning intelligent agent includes the following steps: The intelligent agent plans to access the system operation and maintenance knowledge base to obtain preliminary information; Obtain capability boundary information for each executing agent; Based on the aforementioned prerequisite information and capability boundary information, the intelligent agent executes the operation and maintenance fault inspection plan process based on preset prompt words.

3. The method for troubleshooting operational and maintenance faults in a transaction system based on intelligent agents according to claim 2, characterized in that, The prompt word is configured to guide the plan agent to execute: Determine if it is an operation and maintenance event. If yes, obtain the system-level component name information in the event. If no, skip the current step. In each step of the plan, select the execution agent to be invoked and define the responsibility and output of the current step; Combine consecutive steps assigned to the same executing agent into one step; Generate operation and maintenance fault inspection plan process information.

4. The method for troubleshooting operational and maintenance faults in a transaction system based on intelligent agents according to claim 1, characterized in that, The process of invoking the subordinate execution agent to execute the operation and maintenance fault inspection plan and outputting the execution result after execution includes the following steps: The preceding results obtained from the executed steps are passed as parameters to the executing agent to achieve dynamic context awareness; Based on the implementation results returned by the executing agent, a branch path is triggered when preset conditions are met; The completion rate is used to determine the coverage of verification steps and calculate the degree of goal achievement. Uncertain conclusions are identified, and the planning agent is called again to update the operation and maintenance fault inspection plan process information.

5. The method for troubleshooting operational and maintenance faults in a transaction system based on intelligent agents according to claim 1, characterized in that, The executing intelligent agent includes: The trading system check agent is invoked when the alarm includes the trading system and related components. It checks the status of the trading system and returns a list of tools that need to be called. The tools include trading market consistency check, seat quantity check, trading system status check, trading core process check, and pre-trading check. The server inspection agent is used to check the server hardware status and return a list of tools that need to be invoked, including CPU usage check, memory usage check, disk space check, and SSH connectivity check. The network inspection agent is used to check the server network status and return a list of tools that need to be invoked, including network health checks and database connection checks. The information extraction agent is used to extract information from alarm events and return a list of tools to be invoked. These tools include system-level query, component name extraction, latest system log query, IP address extraction, and event time extraction.

6. The method for troubleshooting operational and maintenance faults in a transaction system based on intelligent agents according to claim 1, characterized in that, After generating the operation and maintenance fault inspection plan process information, it also includes: Manual checks and adjustments are made to the maintenance and troubleshooting plan process information.

7. The method for troubleshooting operational and maintenance faults in a transaction system based on intelligent agents according to claim 1, characterized in that, The system operation and maintenance knowledge base includes system definition, information on components within the system, relationships between components, and historical event handling procedures.

8. The method for troubleshooting operational and maintenance faults in a transaction system based on intelligent agents according to claim 1, characterized in that, The aforementioned operational event alarms include trading system failures, trading blockages, and changes in the number of logged-in seats.

9. A transaction system operation and maintenance fault diagnosis system based on intelligent agents, characterized in that, For implementing the transaction system operation and maintenance fault diagnosis method as described in any one of claims 1-8, the system includes: The coordinating agent is used to respond to input operation and maintenance event alarms and / or operation and maintenance knowledge questions, determine whether the input is related to operation and maintenance, if so, call the planning agent, if not, call the local large language model to generate response information for the input; The planning agent is used to respond to calls, generate maintenance fault inspection plan process information based on the pre-set system operation and maintenance knowledge base and preset prompt words, and call the execution management agent; An execution management intelligent agent is used to respond to a call to the execution management intelligent agent, and based on the operation and maintenance fault inspection plan process information, to call the subordinate execution intelligent agent to execute the operation and maintenance fault inspection plan process, and output the execution result after the execution is completed, so as to realize the troubleshooting of operation and maintenance faults in the transaction system.

10. An electronic device, characterized in that, It includes one or more processors, a memory, and one or more programs stored in the memory, said one or more programs including instructions for executing the agent-based transaction system operation and maintenance fault diagnosis system as described in any one of claims 1-8.

Images