Core network operation and maintenance system construction method based on large model and core network operation and maintenance system

Abstract

The invention provides a core network operation and maintenance system construction method based on a large model and a core network operation and maintenance system. The method comprises the following steps: creating sub-graph structures of a plurality of task execution agents and a graph structure of a main agent; establishing a calling relationship between the graph structure and / or the sub-graph structure and a large language model LLM; compiling the graph structure and the sub-graph structure to obtain a graph instance; constructing a core network operation and maintenance system based on the graph instance; wherein the graph instance is used for defining an operation process of the core network operation and maintenance system, and the operation process at least comprises the following steps: responding to a user instruction by a main agent, calling an LLM to analyze the user instruction so as to identify a task intention, and scheduling a corresponding task execution agent based on the task intention; and the task execution agent responds to the scheduling and calls the LLM to generate a task execution process, and executes the operation and maintenance task based on the task execution process. Therefore, a core network operation and maintenance system can be constructed, intelligentization and automation of core network operation and maintenance are realized, and operation and maintenance efficiency and accuracy are improved.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a method for constructing a core network operation and maintenance system based on a large model and a core network operation and maintenance system. Background Art

[0002] Mobile communication systems consist of multiple components, including the access network, core network, user equipment, and transport network. Through their collaboration, mobile communication systems enable efficient, reliable, and secure mobile communications. The core network is a key component of mobile communication systems, responsible for data exchange, routing, and network management.

[0003] With the rapid development of technologies such as 5G, the Internet of Things, and cloud computing, network architecture has gradually become cloud-based and virtualized. The types and number of network elements have increased significantly, and the complexity of interfaces has also increased, increasing the difficulty of core network operation and maintenance.

[0004] Traditional core network O&M methods rely primarily on manual monitoring and troubleshooting. These methods result in heavy network maintenance workloads, low automation, and poor accuracy, making it difficult to meet the stringent requirements of Service Level Agreements (SLAs) and failing to meet current demands for efficient and accurate O&M. Consequently, traditional core network O&M methods make it impossible to build automated core network O&M tools and achieve automated O&M. Summary of the Invention

[0005] The embodiments of the present application provide a method for constructing a core network operation and maintenance system based on a large model and a core network operation and maintenance system to solve the technical problem that traditional core network operation and maintenance methods cannot achieve automated operation and maintenance.

[0006] In the first aspect, an embodiment of the present application provides a method for constructing a core network operation and maintenance system based on a large model, the method comprising: based on the operation and maintenance tasks of the core network, creating a subgraph structure of multiple task execution agents using the LangGraph framework, and creating a graph structure of a supervisor agent using the LangGraph framework; wherein the supervisor agent is used to communicate with each task execution agent to schedule the task execution agent, and different task execution agents correspond to different operation and maintenance tasks; the task execution agent includes at least one of the following agents: a network monitoring agent, a fault detection and positioning agent, an alarm processing agent, a complaint processing agent, a configuration management agent, a network optimization agent, a data analysis agent and a knowledge base agent, the network monitoring agent is used to monitor the operating status of the core network, the fault detection and positioning agent is used to locate the cause and location of the fault in the core network, the alarm processing agent is used to process the alarm information in the core network, the complaint processing agent is used to process user complaints, and the configuration management agent is used to configure the network optimization agent, the data analysis agent and the ... The management agent is used to configure and manage the core network, the network optimization agent is used to optimize the core network, the data analysis agent is used to analyze the network data of the core network, and the knowledge base agent is used to manage and update the core network operation and maintenance knowledge base; establish a calling relationship between the graph structure and / or sub-graph structure and the large language model LLM; compile the graph structure and sub-graph structure to obtain a graph instance; build a core network operation and maintenance system based on the graph instance; wherein, the graph instance is used to define the operation process of the core network operation and maintenance system, and when the core network operation and maintenance system is running, the graph instance runs to execute the operation process; the operation process at least includes: when the user terminal issues a user instruction to perform an operation and maintenance task to the core network operation and maintenance system, the supervisor agent responds to the user instruction, calls LLM to parse the user instruction to identify the task intent, and schedules one or more corresponding task execution agents based on the task intent; and, the operation process also includes: the task execution agent responds to the scheduling and calls LLM to generate a task execution process, and executes the operation and maintenance task based on the task execution process.

[0007] In one possible implementation, the steps of creating a subgraph structure of multiple task execution agents using the LangGraph framework include: creating a subgraph structure based on the LangGraph framework for each task execution agent; creating one or more sub-agent nodes in the subgraph structure, and defining a graph state State object for the sub-agent node, the sub-agent node being used to update and view the graph state State object, and the number of sub-agent nodes being determined based on the task execution process corresponding to the task execution agent; and / or creating task executors and / or tool nodes in the subgraph structure, the number and type of task executors and tool nodes being determined based on the operation and maintenance tasks corresponding to the task execution agent; and creating an end node in the subgraph structure.

[0008] In one possible implementation, the step of creating a subgraph structure of multiple task execution agents using the LangGraph framework also includes: for each subgraph structure, setting its entry point, the entry point is the sub-agent node, and the entry point represents the starting point for the task execution agent to perform the operation and maintenance task; using edge components to connect the sub-agent nodes, task executors, tool nodes and / or end nodes, the edge components include conditional edges and unconditional edges, the conditional edges have routing conditions, and the routing conditions at least include that after the sub-agent node calls the LLM, if the return message of the LLM includes a tool call field, it is routed to the task executor and / or tool node, if the return message does not contain the tool call field, it is routed to the end node.

[0009] In one possible implementation, after the step of creating a task executor and / or tool node in the subgraph structure, the method further includes: creating a manual review node in the subgraph structure; and connecting the manual review node between the sub-agent node and the task executor and / or tool node using conditional edges.

[0010] In one possible implementation, the steps of creating a graph structure of a supervisory agent include: creating a graph structure based on the LangGraph framework; creating a master agent node in the graph structure, and defining a graph state State object for the master agent node, the master agent node is used to update and view the graph state State object; using edge components to connect the graph structure of the supervisory agent with each sub-graph structure; and the method also includes: creating an LLM and an external tool set, the external tool set includes multiple callable tools, and the tools are determined based on operation and maintenance tasks; binding each tool in the external tool set to the LLM through the LLM interface.

[0011] In a possible implementation, the steps of creating a subgraph structure of a network monitoring agent include: creating a network real-time monitoring agent node, a data acquisition tool node, an anomaly detection tool node, an alarm trigger tool node and an end node in the subgraph structure of the network monitoring agent; leading an unconditional edge pointing to the data acquisition tool node from the network real-time monitoring agent node, and leading an unconditional edge pointing to the anomaly detection tool node from the data acquisition tool node, and leading a conditional edge pointing to the alarm trigger tool node and the end node respectively from the anomaly detection tool node, and leading an unconditional edge pointing to the end node from the alarm trigger tool node; the steps of creating a subgraph structure of a fault detection and positioning agent include: in the fault In the subgraph structure of the fault detection and positioning agent, a fault detection and positioning agent node, a fault delimitation and positioning agent node, a fault report generation agent node, a fault detection task executor node and an end node are created; conditional edges pointing to the fault detection agent node, the fault delimitation and positioning agent node, the fault report generation agent node and the end node are derived from the fault detection and positioning agent node; and unconditional edges pointing to the fault detection and positioning agent node are derived from the fault detection agent node, the fault delimitation and positioning agent node and the fault report generation agent node respectively; and unconditional edges pointing to the fault detection and positioning agent node are derived from the fault detection agent node; and The fault detection task executor node leads to an unconditional edge pointing to the fault detection agent node; the step of creating a subgraph structure of the alarm processing agent includes: creating an alarm processing agent node, an alarm processing tool node, an alarm information feedback agent node, an alarm feedback tool node and an end node in the subgraph structure of the alarm processing agent; the alarm processing agent node leads to conditional edges pointing to the alarm processing tool node, the alarm information feedback agent node and the end node respectively; and, the alarm processing agent node and the alarm information feedback agent node lead to unconditional edges pointing to the alarm processing agent node respectively; and, the alarm information feedback agent node leads to a conditional edge pointing to the alarm feedback tool node, and , an unconditional edge pointing to the alarm information feedback agent node is drawn from the alarm feedback tool node; the step of creating a subgraph structure of the data analysis agent includes: creating a data analysis agent node, a device status query agent node, a device status query tool node, a performance indicator analysis agent node, a performance indicator analysis tool node, an alarm data query agent node, an alarm data query tool node, a log analysis agent node, a log analysis tool node and an end node in the subgraph structure of the data analysis agent; conditional edges pointing to the status query agent node, the performance indicator analysis agent node, the alarm data query agent node, the log analysis agent node and the end node are drawn from the data analysis agent node;And, the device status query agent node, the performance index analysis agent node, the alarm data query agent node and the log analysis agent node respectively lead to unconditional edges pointing to the data analysis agent node; and, the device status query agent node leads to conditional edges pointing to the device status query tool node; and, the device status query tool node leads to unconditional edges pointing to the device status query agent node; and, the performance index analysis agent node leads to conditional edges pointing to the performance index analysis tool node; and, the performance index analysis tool node leads to the performance index analysis agent node. An unconditional edge is derived from the alarm data query agent node and points to the alarm data query tool node; an unconditional edge is derived from the alarm data query tool node and points to the alarm data query agent node; a conditional edge is derived from the log analysis agent node and points to the log analysis tool node; and an unconditional edge is derived from the log analysis tool node and points to the log analysis agent node; the steps of creating a knowledge base agent include: creating a knowledge question answering agent node and an end node in the subgraph structure of the knowledge base agent; and an unconditional edge is derived from the knowledge question answering agent node and points to the end node.

[0012] In the second aspect, the embodiment of the present application provides a core network operation and maintenance system based on a large model. The system is constructed based on the core network operation and maintenance system construction method based on the large model in the aforementioned first aspect and its various implementation methods. The system includes at least a supervisor agent and a task execution agent. The supervisor agent is used to communicate with each task execution agent to schedule the task execution agent. Different task execution agents correspond to different operation and maintenance tasks; the supervisor agent is configured to: respond to the user instruction issued by the user end to execute the operation and maintenance task, call LLM to parse the user instruction to identify the task intent, and schedule one or more corresponding task execution agents based on the task intent; the task execution agent includes at least one of the following agents: a network monitoring agent, a fault detection and positioning agent , alarm processing agent, complaint processing agent, configuration management agent, network optimization agent, data analysis agent and knowledge base agent. The network monitoring agent is used to monitor the operating status of the core network. The fault detection and positioning agent is used to locate the cause and location of the fault in the core network. The alarm processing agent is used to process the alarm information in the core network. The complaint processing agent is used to handle user complaints. The configuration management agent is used to configure and manage the core network. The network optimization agent is used to optimize the core network. The data analysis agent is used to analyze the network data of the core network. The knowledge base agent is used to manage and update the core network operation and maintenance knowledge base. The task execution agent is configured to: respond to scheduling and call LLM to generate a task execution process, and execute operation and maintenance tasks based on the task execution process.

[0013] In one possible implementation, the supervisor agent is further configured to: generate operation and maintenance task information based on task intent and preset operation and maintenance knowledge, and update the operation and maintenance task information in the graph state State object to pass the operation and maintenance task information to the task execution agent through the graph state State object; the task execution agent is further configured to: view the graph state State object to obtain operation and maintenance task information, and based on the operation and maintenance task information and the task execution process, call its internal nodes to execute the operation and maintenance task, the nodes include sub-agent nodes, task executors, tool nodes, end nodes and / or manual review nodes; after the operation and maintenance task is completed, the task execution result is updated in the graph state State and routed to the supervisor agent or the user end; the task execution agent is further configured to: when the task execution process includes routing to the end node, route to the end node to end the task execution process; and / or, when the task execution process includes routing to the manual review node, respond to user input received by the manual review node, and route to its internal node, another task execution agent or the supervisor agent based on the user input.

[0014] In a possible implementation, the network monitoring agent includes: a network real-time monitoring agent node, a data acquisition tool node, an anomaly detection tool node, an alarm trigger tool node and an end node; the network real-time monitoring agent node is configured to: receive real-time monitoring task instructions, plan monitoring tasks, and sequentially call the data acquisition tool node, the anomaly detection tool node and the alarm trigger tool node to perform monitoring tasks, and the monitoring tasks include data acquisition monitoring tasks, anomaly detection monitoring tasks and alarm trigger monitoring tasks; the data acquisition tool node is configured to: integrate external data acquisition tools and perform data acquisition monitoring tasks; the anomaly detection tool node is configured to: integrate external anomaly detection tools and perform data acquisition monitoring tasks; Detection tool, and perform anomaly detection monitoring tasks. If the external anomaly detection tool detects an alarm event, it is routed to the alarm trigger tool node. If the external anomaly detection tool does not detect an alarm event, it is routed to the end node. The alarm trigger tool node is configured to: integrate the external alarm notification tool and perform alarm trigger monitoring tasks; the fault detection and positioning agent includes: fault detection and positioning agent node, fault detection agent node, fault delimitation and positioning agent node, fault report generation agent node, fault detection task executor node and end node; the fault detection and positioning agent node is configured to: receive alarm information triggered by the network monitoring process, and dispatch tasks to the fault detection agent Node, fault delimitation and positioning agent node and fault report generation agent node; the fault detection agent node is configured to: plan the detection task according to the business rules, and output the detection task to the fault detection task executor node for execution; the detection task includes at least one of the following tasks: equipment status inspection task, performance indicator inspection task, alarm inspection task, chr log inspection task; the fault detection task executor node is configured to: execute the detection task in a loop; the fault delimitation and positioning agent node is configured to: output the delimitation and positioning conclusion based on the fault delimitation and positioning rules according to the fault troubleshooting results, and the delimitation and positioning conclusion includes the fault cause and processing suggestions; the fault report generation ... fault detection task includes at least one of the following tasks: equipment status inspection task, performance indicator inspection task, alarm inspection task, chr log inspection task; the fault detection task executor node is configured to: execute the detection task in a loop; the fault delimitation and positioning agent node is configured to: output the delimitation and positioning conclusion based on the fault delimitation and positioning rules according to the fault delimitation and positioning results, and the delimitation and positioning conclusion includes the fault cause and processing suggestions; the fault report generation agent node is configured to: plan the detection task according to the business rules, and output the detection task to the fault detection task executor node for execution; the fault detection task includes at least one of The agent node is configured to generate a fault analysis report based on the fault troubleshooting results and demarcation and positioning conclusions. The fault analysis report includes at least the operating status of the equipment, fault statistics, and performance trends. The alarm processing agent includes an alarm processing agent node, an alarm processing tool node, an alarm information feedback agent node, an alarm feedback tool, and an end node. The alarm processing agent node is configured to receive alarm information and processing suggestions, and dispatch alarm tasks to the alarm processing tool node. The alarm processing tool node is configured to execute alarm tasks based on alarm classification and alarm level. The alarm information feedback agent node is configured to feed back processing results and related information to operation and maintenance personnel or related systems.The data analysis agent includes: a data analysis agent node, a device status query agent, a device status query tool, a performance indicator analysis agent, a performance indicator analysis tool, an alarm data query agent, an alarm data query tool, a log analysis agent, a log analysis tool, and an end node. The data analysis agent node is configured to detect user questions and dispatch tasks to the device status query agent node, the performance indicator analysis agent node, the alarm data query agent node, and / or the log analysis agent node. The device status query agent node is configured to integrate external tools for device status query and perform device status data queries. The performance indicator analysis agent node is configured to integrate external tools for performance indicator analysis and perform performance indicator data query analysis and indicator prediction. The alarm data query agent node is configured to integrate external tools for alarm data query and perform alarm data query and statistics. The log analysis agent node is configured to integrate external tools for log data query and perform log data analysis. The knowledge base agent includes: a knowledge question and answer agent node and an end node.

[0015] In the third aspect, an embodiment of the present application provides a core network operation and maintenance system construction device based on a large model, the device including: a first construction module, which is used to create a subgraph structure of multiple task execution agents based on the LangGraph framework, and a graph structure of a supervisor agent based on the LangGraph framework based on the operation and maintenance tasks of the core network; wherein the supervisor agent is used to communicate with each task execution agent to schedule the task execution agent, and different task execution agents correspond to different operation and maintenance tasks; the task execution agent includes at least one of the following agents: a network monitoring agent, a fault detection and positioning agent, an alarm processing agent, a complaint processing agent, a configuration management agent, a network optimization agent, a data analysis agent and a knowledge base agent, the network monitoring agent is used to monitor the operating status of the core network, the fault detection and positioning agent is used to locate the cause and location of the fault in the core network, the alarm processing agent is used to process the alarm information in the core network, the complaint processing agent is used to process user complaints, and the configuration management agent is used to Configuration management core network, network optimization agent is used to optimize the core network, data analysis agent is used to analyze the network data of the core network, knowledge base agent is used to manage and update the core network operation and maintenance knowledge base; relationship establishment module, used to establish the calling relationship between graph structure and / or sub-graph structure and large language model LLM; compilation module, used to compile graph structure and sub-graph structure to obtain graph instance; second construction module, used to build core network operation and maintenance system based on graph instance; wherein, graph instance is used to define the operation process of core network operation and maintenance system, when the core network operation and maintenance system is running, graph instance runs to execute the operation process; the operation process at least includes: when the user terminal issues a user instruction to perform an operation and maintenance task to the core network operation and maintenance system, the supervisor agent responds to the user instruction, calls LLM to parse the user instruction to identify the task intent, and schedules one or more corresponding task execution agents based on the task intent; and, the operation process also includes: the task execution agent responds to the scheduling and calls LLM to generate a task execution process, and executes the operation and maintenance task based on the task execution process.

[0016] From the above content, it can be seen that the embodiment of the present application provides a method for constructing a core network operation and maintenance system based on a large model and a core network operation and maintenance system, the method including: based on the operation and maintenance tasks of the core network, creating a subgraph structure of multiple task execution agents using the LangGraph framework, and creating a graph structure of a supervisor agent using the LangGraph framework; wherein the supervisor agent is used to communicate with each task execution agent to schedule the task execution agent, and different task execution agents correspond to different operation and maintenance tasks; the task execution agent includes at least one of the following agents: a network monitoring agent, a fault detection and positioning agent, an alarm processing agent, a complaint processing agent, a configuration management agent, a network optimization agent, a data analysis agent and a knowledge base agent, the network monitoring agent is used to monitor the operating status of the core network, the fault detection and positioning agent is used to locate the cause and location of the fault in the core network, the alarm processing agent is used to process the alarm information in the core network, and the complaint processing agent is used to process the user The core network operation and maintenance system comprises a plurality of intelligent agents, a configuration management agent, a network optimization agent, a data analysis agent, and a knowledge base agent. The plurality of intelligent agents are configured to configure and manage the core network, a network optimization agent, a data analysis agent, and a knowledge base agent. The plurality of intelligent agents are configured to configure and manage the core network, a network optimization agent, a data analysis agent, and a knowledge base agent. The plurality of intelligent agents are configured to analyze the network data of the core network, and a knowledge base agent is configured to manage and update the core network operation and maintenance knowledge base. The plurality of intelligent agents are configured to establish a calling relationship between the graph structure and / or subgraph structure and the large language model (LLM). The graph structure and subgraph structure are compiled to obtain a graph instance. The core network operation and maintenance system is constructed based on the graph instance. The graph instance is used to define the operation process of the core network operation and maintenance system. When the core network operation and maintenance system is running, the graph instance runs to execute the operation process. The operation process at least includes: when the user terminal issues a user instruction to execute an operation and maintenance task to the core network operation and maintenance system, the supervisor agent responds to the user instruction, calls the LLM to parse the user instruction to identify the task intent, and schedules one or more corresponding task execution agents based on the task intent. The operation process also includes: the task execution agent responds to the scheduling and calls the LLM to generate a task execution process, and executes the operation and maintenance task based on the task execution process. In this way, a core network operation and maintenance system can be constructed, and based on the core network operation and maintenance system, the intelligent and automated core network operation and maintenance can be realized, thereby improving the efficiency and accuracy of operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of the framework of the core network operation and maintenance system provided in an embodiment of the present application;

[0018] Figure 2 A flowchart of a method for constructing a core network operation and maintenance system based on a large model provided in an embodiment of the present application;

[0019] Figure 3 A schematic diagram of the graph structure of the supervisor agent provided in an embodiment of the present application;

[0020] Figure 4A schematic diagram of the network monitoring agent subgraph structure provided in an embodiment of the present application;

[0021] Figure 5 A schematic diagram of the subgraph structure of the fault detection and positioning agent provided in an embodiment of the present application;

[0022] Figure 6 A schematic diagram of the subgraph structure of the alarm processing agent provided in an embodiment of the present application;

[0023] Figure 7 A schematic diagram of the subgraph structure of the data analysis agent provided in an embodiment of the present application;

[0024] Figure 8 A schematic diagram of the subgraph structure of the knowledge base agent provided in an embodiment of the present application;

[0025] Figure 9 A schematic diagram of a subgraph structure including a tool review node provided in an embodiment of the present application;

[0026] Figure 10 A schematic diagram of a multi-round dialogue architecture provided in an embodiment of the present application;

[0027] Figure 11 A schematic diagram of the structure of a core network operation and maintenance system based on a large model provided in an embodiment of the present application;

[0028] Figure 12 A schematic diagram of the structure of a core network operation and maintenance system construction device based on a large model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0030] Before introducing the technical solutions of the embodiments of the present application, an exemplary introduction to the terms involved in the embodiments of the present application is first given.

[0031] 1. Core network: An important component of the mobile communication system, connected to external network parts such as the access network, transmission network, and wireless access network, responsible for carrying and managing the control, signaling, data processing of the communication system, and connecting exchanges between different networks.

[0032] 2. Agent: A system or entity that can autonomously perceive the environment, make decisions, and perform actions to achieve specific goals.

[0033] In order to improve the accuracy and timeliness of core network operation and maintenance, an embodiment of the present application provides a core network operation and maintenance system construction method and a core network operation and maintenance system based on a large model.

[0034] 3. Language Graph Framework (LangGraph): A model or tool for structured representation and processing of language information, used to build stateful, multi-actor applications based on large language models (LLMs) by modeling steps as edges and nodes in a graph.

[0035] Specifically, the construction method provided in the embodiments of this application can combine large model technology and LangGraph technology to build a core network intelligent operation and maintenance framework. This framework uses the language understanding and generation capabilities of the large model to accurately analyze user operation and maintenance requirements. Through the intelligent proxy function of the LangGraph framework, it can realize the automated execution and real-time monitoring of operation and maintenance tasks. This improves operation and maintenance efficiency and accuracy, reduces operation and maintenance costs, and enhances user experience.

[0036] Figure 1 A schematic diagram of the framework of the core network operation and maintenance system provided in an embodiment of the present application.

[0037] like Figure 1 As shown, the core network operation and maintenance system constructed based on the construction method provided in the embodiment of the present application can include multiple architectural layers. Specifically, the core network operation and maintenance system provided in the embodiment of the present application uses infrastructure (Infrastructure) or computer (Computer) as the underlying framework, and the infrastructure and computer can be used to provide computing and storage resources. Furthermore, the core network operation and maintenance system provided in the embodiment of the present application can construct a multi-agent orchestration and scheduling framework based on the LangGraph framework. The multi-agent orchestration and scheduling framework can be combined with the API layer to communicate with a large language model (Large Language Model, LLM) or a collection of multiple large language models (Large Language Models, LLMs) to integrate the intelligent question-answering, text analysis and other functions of the large language model LLM. Furthermore, based on the multi-agent orchestration and scheduling framework, a supervisor agent and professional agents can be formed. The supervisor agent is used to receive instructions to schedule professional agents. Professional agents are also called task execution agents, which can include network monitoring agents, detection and positioning agents, alarm processing agents, complaint handling agents, fault management agents, network optimization agents, knowledge base agents, etc. Each professional agent is specifically related to the operation and maintenance tasks performed by the core network operation and maintenance system.

[0038] Furthermore, the core network operation and maintenance system framework integrates multiple layers, including external tool sets, data management, and large-scale model training. The external tool set includes operation and maintenance automation tools, such as status monitoring tools, performance monitoring tools, alarm monitoring tools, and log analysis tools, which are used to obtain core network status information and handle operation and maintenance tasks. The data management layer is responsible for data storage, retrieval, and management, providing data support for intelligent agents. The large-scale model training layer is used to train and optimize large language models.

[0039] As can be seen, the core network operation and maintenance system provided by the embodiment of the present application, supported by its framework, can realize the collaborative work of intelligent agents and tool sets at different levels, thereby achieving efficient operation and management of the core network. For example, it can be applied to scenarios such as knowledge question answering, monitoring and troubleshooting, and performance optimization.

[0040] It is worth noting that the core network operation and maintenance system provided in the embodiment of the present application may include the following parts:

[0041] ① Large Language Model (LLM): including large language model, embedding model and multimodal model. The embodiment of the present application can build a unified large model access API layer, which can realize unified access to multiple models and flexible switching.

[0042] ② Data management: Stores and stores various intermediate and persistent data in the Agent management and operation process, including structured and unstructured knowledge documents, vector databases, analytical data, message history, log data, etc., and provides necessary data maintenance and management tools, such as cleaning, vectorization, import and export of private knowledge data.

[0043] ③ Large model operation and maintenance management: including the operation and maintenance management of LLM and various intelligent entities built on LLM, specifically including management work at different stages of the application life cycle.

[0044] ④ Development and orchestration framework: The LangGraph framework is a low-level library designed specifically for building applications with complex state management and multi-role interaction capabilities. Based on the LangGraph framework, the embodiments of this application can simplify the complexity and workload of upstream intelligent agent construction and reduce risks.

[0045] ⑤ Core network operation and maintenance agent (including supervisor agent and task execution agent): An agent built on the development and orchestration framework. The core network operation and maintenance agent collaborates with external tools through APIs or code interpreters to complete tasks. Task execution agents specifically include:

[0046] Network monitoring agent: monitors the status and performance indicators of the core network and equipment in real time, detects abnormal conditions and errors in the network, and triggers alarms.

[0047] Fault detection and location agent: locates the cause and location of the fault, outputs solutions or handling suggestions, and guides operation and maintenance personnel to carry out repair work.

[0048] Alarm processing agent: Receives alarm information and triggers the corresponding processing flow to quickly restore the normal operation of the equipment.

[0049] Complaint Handling Agent: This agent is primarily used to handle user complaints. It automatically completes complaint classification, diagnosis, signaling analysis, and work order completion in a single step. This automated process reduces manual intervention and improves the speed and accuracy of complaint handling.

[0050] Configuration Management Agent: The Configuration Management Agent is responsible for core network configuration management, including device configuration and network topology configuration. Automated configuration management reduces human error and improves configuration efficiency and accuracy.

[0051] Network Optimization Agent: This agent is responsible for optimizing the core network, including performance and resource optimization. By intelligently analyzing network data, it identifies potential performance bottlenecks and resource waste, proposes optimization suggestions, and implements them to improve network performance and resource utilization.

[0052] Data analysis agent: Analyzes and mines various types of collected network data, discovers network operation trends and potential problems, and helps operation and maintenance personnel better understand the network operation status.

[0053] Knowledge base agent: stores and manages knowledge and cases related to core network operation and maintenance, and provides a knowledge question and answer interface for other agents.

[0054] Task scheduling agent: Identifies and perceives the needs of operation and maintenance personnel, and dispatches operation and maintenance needs to relevant agents based on the capabilities of the agents.

[0055] Operation and Maintenance Supervisor Agent (abbreviated as Supervisor Agent): The Operation and Maintenance Supervisor Agent acts as a bridge between humans and machines, allowing operation and maintenance personnel to easily communicate with the intelligent assistant through the natural language processing interface, and delegate the operation and maintenance tasks proposed by users to various professional agents for collaborative processing, and return the results to the user after the task processing is completed.

[0056] ⑥ External tool set: including equipment status monitoring and query tools, alarm monitoring and query tools, log analysis tools, performance monitoring and query tools, etc., used to obtain core network status information, handle operation and maintenance tasks, etc.

[0057] Figure 2 A flowchart of a method for constructing a core network operation and maintenance system based on a large model provided in an embodiment of the present application.

[0058] See also Figure 1and Figure 2 , an embodiment of the present application provides a method for constructing a core network operation and maintenance system based on a large model, which may include steps S100-S400.

[0059] S100: Based on the operation and maintenance tasks of the core network, a subgraph structure of multiple task execution agents is created using the LangGraph framework, and a graph structure of a supervisor agent is created using the LangGraph framework; wherein the supervisor agent is used to communicate with each task execution agent to schedule the task execution agents, and different task execution agents correspond to different operation and maintenance tasks.

[0060] The steps of creating a subgraph structure for task execution agents and a graph structure for supervisor agents refer to building a graph structure for agents. This specifically involves defining components such as nodes, edges, and states, and determining the connections and interactions between agents. Nodes represent specific steps or operations in an operation and maintenance task, edges represent the logical relationships or data flows between steps, and states represent globally shared information during process execution.

[0061] Task execution agents are used to perform specific operations and maintenance tasks, such as network monitoring, alarm handling, complaint processing, configuration management, and network optimization, thereby improving the professionalism and accuracy of task execution. This ensures that each task is completed by the relevant agent and avoids performance degradation caused by overloading a single agent.

[0062] The operation and maintenance supervisor agent is used to act as a bridge between people (operation and maintenance personnel) and machines, allowing operation and maintenance personnel to communicate with the operation and maintenance system through the natural language processing interface, and delegate the operation and maintenance tasks proposed by users to various professional agents for collaborative processing, and return the results to the user after the task processing is completed.

[0063] It is understandable that the specific composition of the task execution agent subgraph structure and the supervisor agent graph structure can be determined based on actual needs. In the embodiment of the present application, the task execution agent includes at least one of the following agents: a network monitoring agent, a fault detection and positioning agent, an alarm processing agent, a complaint processing agent, a configuration management agent, a network optimization agent, a data analysis agent, and a knowledge base agent. The network monitoring agent is used to monitor the operating status of the core network, the fault detection and positioning agent is used to locate the cause and location of the fault in the core network, the alarm processing agent is used to process the alarm information in the core network, the complaint processing agent is used to handle user complaints, the configuration management agent is used to configure and manage the core network, the network optimization agent is used to optimize the core network, the data analysis agent is used to analyze the network data of the core network, and the knowledge base agent is used to manage and update the core network operation and maintenance knowledge base. The specific structure of each task execution agent will be described in detail below and will not be repeated here.

[0064] S200: Establishing a calling relationship between a graph structure and / or a subgraph structure and a large language model LLM.

[0065] It is understandable that the large language model LLM can understand natural language instructions and can be used to understand the requirements of operation and maintenance tasks and generate corresponding task execution processes. The embodiment of the present application can connect the LLM with the graph structure and / or subgraph structure through the API interface. In this way, during the application stage of the core network operation and maintenance system, the supervisor agent and / or task execution agent can interact with the large language model LLM to achieve the purpose of guiding the agent to complete complex operation and maintenance tasks.

[0066] S300: Compile the graph structure and subgraph structure to obtain a graph instance.

[0067] It can be understood that by compiling the graph structure and subgraph structure, the graph structure and subgraph structure can be converted into an executable graph instance, which is used to represent the operation process of the core network operation and maintenance system. In this way, when the core network operation and maintenance system receives the user's operation and maintenance task request, it can start executing the task based on the compiled graph instance. For example, by inputting the initial state (such as user instructions, the current state of the core network, etc.), the graph instance can automatically schedule each task execution agent to perform the operation and maintenance task according to the defined relationship between nodes and edges.

[0068] S400: Build a core network operation and maintenance system based on graph instances.

[0069] Among them, when the core network operation and maintenance system is running, the graph instance can run to execute the operation process.

[0070] Furthermore, the step of obtaining a core network operation and maintenance system based on the graph instance includes at least building a user end for the core network operation and maintenance system. Specifically, this may include building a user interface, which is the front-end entry point for users to interact with the core network operation and maintenance system and through which users can input user commands. Furthermore, the interface between the user interface and the graph instance may be defined, such as the format, protocol, and method for data transmission to ensure accurate data transmission and stable interaction, thereby integrating the user interface with the graph instance.

[0071] Furthermore, the graph instance encapsulates the logic of the operation and maintenance system, including task scheduling, state management, and agent collaboration. After receiving user instructions through the user interface, the graph instance executes and obtains the operation results. Afterwards, the operation results can be fed back to the user through the user interface.

[0072] In some implementations, step S400 may also include deploying the core network operation and maintenance system on a suitable server, and performing configuration and optimization steps to ensure that the system can run stably. This embodiment of the present application does not specifically limit this.

[0073] It is worth noting that, based on the graph example, the operation process of the core network operation and maintenance system at least includes: when the user terminal issues a user instruction to the core network operation and maintenance system to perform an operation and maintenance task, the supervisor agent calls the LLM to parse the user instruction to identify the task intent, and schedules one or more corresponding task execution agents based on the task intent; and the operation process also includes: the task execution agent responds to the scheduling and calls the LLM to generate a task execution process, and executes the operation and maintenance task based on the task execution process.

[0074] In other words, the compiled graph instance can be used to implement the following process:

[0075] ① Respond to user command input: When the user enters a command, the supervisor agent can parse the command through the LLM node and identify the task intent.

[0076] ②Task scheduling: The supervisor agent can schedule the corresponding task execution agent to perform operation and maintenance tasks based on the task intention analyzed by LLM.

[0077] ③Task execution analysis: Each task execution agent can call LLM to further analyze the task execution process and obtain the execution steps or execution process.

[0078] ④Task execution: The task execution agent can execute tasks based on the process of LLM feedback, and finally return the results or complete the task.

[0079] From the above content, it can be seen that the embodiment of the present application provides a method for constructing a core network operation and maintenance system based on a large model, which includes: based on the operation and maintenance tasks of the core network, creating a subgraph structure of multiple task execution agents using the LangGraph framework, and creating a graph structure of a supervisor agent using the LangGraph framework; wherein the supervisor agent is used to communicate with each task execution agent to schedule the task execution agent, and different task execution agents correspond to different operation and maintenance tasks; the task execution agent includes at least one of the following agents: a network monitoring agent, a fault detection and positioning agent, an alarm processing agent, a complaint processing agent, a configuration management agent, a network optimization agent, a data analysis agent and a knowledge base agent, the network monitoring agent is used to monitor the operating status of the core network, the fault detection and positioning agent is used to locate the cause and location of the fault in the core network, the alarm processing agent is used to process the alarm information in the core network, and the complaint processing agent is used to process user complaints. The configuration management agent is used to configure and manage the core network, the network optimization agent is used to optimize the core network, the data analysis agent is used to analyze the network data of the core network, and the knowledge base agent is used to manage and update the core network operation and maintenance knowledge base; a calling relationship between a graph structure and / or a subgraph structure and a large language model (LLM) is established; the graph structure and subgraph structure are compiled to obtain a graph instance; and a core network operation and maintenance system is constructed based on the graph instance. The graph instance is used to define the operation process of the core network operation and maintenance system. When the core network operation and maintenance system is running, the graph instance is run to execute the operation process. The operation process at least includes: when a user terminal issues a user instruction to execute an operation and maintenance task to the core network operation and maintenance system, the supervisor agent responds to the user instruction, calls the LLM to parse the user instruction to identify the task intent, and schedules one or more corresponding task execution agents based on the task intent; and the operation process also includes: the task execution agent responds to the scheduling and calls the LLM to generate a task execution process, and executes the operation and maintenance task based on the task execution process. In this way, a core network operation and maintenance system can be constructed, and based on the core network operation and maintenance system, intelligent and automated core network operation and maintenance can be achieved, thereby improving operation and maintenance efficiency and accuracy.

[0080] Furthermore, in the embodiment of the present application, the task execution agent may specifically include the following types:

[0081] ① Network Monitoring Agent: This agent monitors the core network's operational status in real time, including key indicators such as device status, network performance, network traffic, and the number of user connections. Once an anomaly is detected, it immediately triggers an alert processing process to ensure that the problem is resolved promptly.

[0082] ② Fault detection and positioning agent: It is used to detect network status, performance, alarms, and logs, locate the cause and location of the fault, output solutions or processing suggestions, and guide operation and maintenance personnel to carry out repair work.

[0083] ③ Alarm Processing Agent: This agent processes alarms in the core network and connects to the work order system to provide one-click alarm Q&A, case recommendations, and intelligent diagnosis. Through intelligent analysis of alarm data, it quickly locates the root cause of the problem, reduces manual intervention, shortens alarm processing time, and improves operation and maintenance efficiency.

[0084] ④ Complaint Handling Agent: This agent handles user complaints and performs complaint classification, complaint diagnosis, signaling analysis, and work order completion. This automated process reduces manual intervention and improves the speed and accuracy of complaint handling.

[0085] ⑤ Configuration Management Agent: This agent manages the core network's configuration, including device configuration and network topology configuration. Automated configuration management reduces human error and improves configuration efficiency and accuracy.

[0086] ⑥ Network Optimization Agent: This agent optimizes the core network, including performance and resource optimization. It intelligently analyzes network data to identify potential performance bottlenecks and resource waste, proposes optimization suggestions, and implements them to improve network performance and resource utilization.

[0087] ⑦ Knowledge base agent: used to manage and update the internal core network operation and maintenance knowledge base to ensure that all agents can access the latest and most accurate operation and maintenance knowledge.

[0088] It is worth noting that if the graph structure or subgraph structure of the embodiment of the present application is visualized, then the graph structure or subgraph structure can be a set of interconnected nodes and edges, the nodes represent the supervisory intelligent agent, task execution intelligent agent and LLM, etc., and the edges represent the communication and control relationship between the nodes.

[0089] Figure 3 A schematic diagram of the graph structure of the supervisor agent provided in an embodiment of the present application.

[0090] like Figure 3 As shown, the graph structure of the supervisor agent provided by the embodiment of the present application includes a supervisor agent node and multiple task execution agent nodes, and the supervisor agent node can lead to conditional edges pointing to each task execution agent node ( Figure 3 In the figure, conditional edges are represented by dashed lines with arrows), and each task execution agent can lead to unconditional edges pointing to the supervisor agent ( Figure 3 (Unconditional edges are represented by arrows in the diagram.) Conditional and unconditional edges are edge components, and conditional edges have routing conditions. Based on conditional edges, the supervisor agent node can route to various task execution agent nodes to perform operation and maintenance tasks. The diagram only shows some task execution agent nodes by way of example; the number of task execution agents in the graph structure can be increased or decreased based on actual needs.

[0091] The supervisor agent can be called a Supervisor node. In the embodiment of the present application, prompts and routes can be designed for the supervisor agent so that the supervisor agent can manage each task execution agent.

[0092] Among them, the code for designing prompt can be:

[0093] members = ["Real-time Monitoring Agent", "Fault Detection and Location Agent", "Alarm Processing Agent", "Data Analysis Agent", "Knowledge Question Answering Agent"]

[0094] system_prompt=

[0095] "You are an Agent Supervisor managing the conversation between the following Agents: {members}. Given the following user request, the Agents respond together to take the next action. Each Agent will perform a task and respond with its results and status. Upon completion, respond with "FINISH."."

[0096] The routing design is used to define the Supervisor node routing results so that the Supervisor node can dispatch the user's question to the next Agent node by calling llm.with_structured_output.

[0097] For example, the routing design code may be:

[0098]

[0099]

[0100] Based on this code, an LLM-based task routing controller can be implemented to build a supervisor agent node (Supervisor) in a multi-agent collaborative system, so that the supervisor agent can implement the following functions: dynamically decide which sub-agent (task execution agent) to call next or end the task based on the current dialogue status.

[0101] Figure 4 The figure shows an example of a visual display of the subgraph structure of a network monitoring agent. The network monitoring agent can specifically include a network real-time monitoring Agent node, a data acquisition tool (Tool) node, an anomaly detection Tool node, an alarm trigger Tool node and an end (end) node, etc., and can also include a start (start) node. The composition of the specific subgraph structure can be designed based on actual needs, and the embodiments of the present application do not make specific limitations on this.

[0102] In an embodiment of the present application, the intelligent agent construction process may include four key steps: LangGraph graph structure design, large model integration, external tool interface development and state management implementation.

[0103] ① LangGraph graph structure design: Constructing the graph structure of the agent involves defining components such as nodes, edges, and states, and determining the connections and interactions between them. Nodes represent specific steps or operations in an operation and maintenance task, edges represent the logical relationships or data flows between steps, and states represent globally shared information during process execution.

[0104] ② Large Model Integration: Large language models (LLMs) are integrated into the agent as key nodes within the graph structure. LLMs are responsible for processing natural language instructions, understanding the requirements of the operation and maintenance task, and generating corresponding execution plans. By interacting with other nodes in the graph structure, LLMs can guide the agent to complete complex operation and maintenance tasks.

[0105] ③ External Tool Interface Development: To automate O&M tasks, agents need to interact with external tools. Therefore, corresponding interfaces must be developed to connect the agents and external tools. These interfaces can be customized based on the type and functionality of the external tools, ensuring that the agents can correctly call external tools and perform the corresponding O&M operations.

[0106] ④ State Management Implementation: The state management functionality of the LangGraph framework ensures that agents maintain consistent contextual understanding and decision-making capabilities when performing operations and maintenance tasks. By defining state diagrams and state transition rules, agents can perform different operations in different states and dynamically adjust workflows as needed.

[0107] Furthermore, in order to build a core network operation and maintenance system based on a big model, the method provided in the embodiment of the present application can first initialize the big model and tools, wherein initializing the big model can refer to instantiating an LLM object, and initializing the tool can refer to each tool in the external tool set and binding it to the LLM through the LLM interface.

[0108] Furthermore, before step S200, the method provided in the embodiment of the present application may also include the following steps S501-S502.

[0109] S501: Create LLM and external tool set, the external tool set includes multiple callable tools, and the tools are determined based on operation and maintenance tasks.

[0110] In this embodiment of the present application, the core network operation and maintenance assistant can integrate multiple external tools to complete specific operation and maintenance tasks. By integrating external tools into the graph structure and writing corresponding calling code, the operation and maintenance assistant can use these tools to complete specific operation and maintenance tasks. For example, a database can be used to store and query core network status information, log analysis tools can be used to analyze device log information, and performance monitoring tools can be used to monitor core network performance indicators. These tools can be integrated and called through the LangGraph API.

[0111] In an embodiment of the present application, the step of instantiating an LLM object to create an LLM may include the step of loading a pre-trained model. The pre-trained model may adopt an open source model, and the embodiment of the present application does not make any specific limitations on this.

[0112] S502: Bind each tool in the external tool set to the LLM through the LLM interface.

[0113] Among them, the interface of LLM is, for example, the bind_tools() interface, which is not specifically limited in the embodiment of the present application.

[0114] Furthermore, in the embodiment of the present application, step S100 may specifically include steps S101-S102.

[0115] S101: For each task execution agent, a subgraph structure is created based on the LangGraph framework.

[0116] It can be understood that based on the LangGraph framework, a subgraph structure is a Graph.

[0117] S102: Create one or more sub-agent nodes in the sub-graph structure, and define a graph state State object for the sub-agent node. The sub-agent node is used to update and view the graph state State object. The number of sub-agent nodes is determined based on the task execution process corresponding to the task execution agent.

[0118] For example, Figure 3 As shown, the subgraph structure of the network monitoring agent may include only one sub-agent node, namely the network real-time monitoring Agent.

[0119] Figure 5 A schematic diagram of the subgraph structure of the fault detection and positioning agent provided in an embodiment of the present application.

[0120] like Figure 5As shown, the subgraph structure of the alarm processing agent may include four sub-agent nodes, namely, the fault detection and positioning Agent node, the fault detection Agent node, the fault delimitation and positioning Agent node, and the fault report generation Agent node, and may include a start node and an end node. The embodiment of the present application does not make specific limitations on this.

[0121] It's also important to note that defining a State object for a sub-agent node involves initializing the sub-graph structure using LangGraph's built-in MessagesState object. State objects are used to maintain status information related to the execution of maintenance tasks, such as the task's current progress, data input and output, and processing results.

[0122] For example, the State design of the network monitoring agent can be:

[0123] By inheriting the MessagesState class of the LangGraph framework, three new attributes are added: collect_data, abnormal_info, and alarm_info. collect_data is used to collect data, abnormal_info is used to record abnormal information, and alarm_info is used to process alarm information.

[0124] collect_data:Annotated[List[Tuple],operator.add] / / Collect data;

[0125] abnormal_info:Annotated[List[Tuple],operator.add] / / abnormal information;

[0126] alarm_info:Annotated[List[Tuple],operator.add] / / alarm information;

[0127] And / or, step S100 may further include the following steps S103-S104.

[0128] S103: Create task executors and / or tool nodes in the subgraph structure. The number and type of task executors and tool nodes are determined based on the operation and maintenance tasks corresponding to the task execution agent.

[0129] The task executor can refer to a collection of tools, such as Figure 5As shown, the fault detection and location agent can include a fault detection executor. This fault detection executor can integrate external tools such as device status checks, performance indicator checks, alarm checks, and chr log checks to execute detection tasks in a loop. It is understood that tools can be integrated and called through the API provided by the LangGraph architecture.

[0130] S104: Create an end node.

[0131] Furthermore, after step S104, the following steps S105-S106 may be included.

[0132] S105: For each subgraph structure, set its entry point, which is the sub-agent node.

[0133] Among them, the entry point represents the starting point where the task execution agent performs the operation and maintenance task, that is, where the task begins.

[0134] S106: Use edge components to connect sub-agent nodes, task executors, tool nodes and / or end nodes. Edge components include conditional edges and unconditional edges. Conditional edges have routing conditions. The routing conditions at least include that after the sub-agent node calls LLM, if the return message of LLM includes a tool call field, it is routed to the task executor and / or tool node. If the return message does not include a tool call field, it is routed to the end node.

[0135] It can be understood that the tool call field is used to indicate the need to further call a tool or task executor, etc.

[0136] It is understandable that the routing conditions can be designed based on the actual task objectives of the operation and maintenance tasks, and the embodiments of the present application do not make specific limitations on this.

[0137] The following is an exemplary introduction to the process of creating the subgraph structure of each task execution agent.

[0138] Continue to see Figure 4 The network real-time monitoring intelligent agent plays the role of real-time network monitoring and status feedback in the core network operation and maintenance assistant. It monitors the operating status and performance parameters of core network equipment in real time, and promptly discovers and reports equipment failures or abnormal conditions. Core network equipment includes routers, switches, gateways, servers, controllers, and security devices.

[0139] The steps of creating a subgraph structure of a network monitoring agent may include:

[0140] S100-a: Create a network real-time monitoring agent node, a data acquisition tool node, an anomaly detection tool node, an alarm triggering tool node, and an end node in the subgraph structure of the network monitoring agent;

[0141] S100-b: An unconditional edge pointing to the data acquisition tool node is derived from the network real-time monitoring intelligent agent node, and an unconditional edge pointing to the anomaly detection tool node is derived from the data acquisition tool node, and conditional edges pointing to the alarm trigger tool node and the end node are derived from the anomaly detection tool node, and an unconditional edge pointing to the end node is derived from the alarm trigger tool node.

[0142] In this way, the anomaly detection tool can be routed to the alarm triggering tool to trigger an alarm when an alarm event occurs, and can be routed to the end node to end the task process when no alarm event occurs.

[0143] Continue to see Figure 5 ,The fault detection and location agent is used to detect the network status, ,performance, alarms, and logs, locate the cause and location of the fault, ,output solutions or processing suggestions to guide the operation and ,maintenance personnel to perform repair work.

[0144] The steps of creating a subgraph structure of a fault detection and localization agent may include:

[0145] S100-c: creating a fault detection and location agent node, a fault detection agent node, a fault delimitation and location agent node, a fault report generation agent node, a fault detection task executor node, and an end node in the subgraph structure of the fault detection and location agent;

[0146] S100-d: Conditional edges are derived from the fault detection and locating agent node, pointing to the fault detection agent node, the fault delimitation and locating agent node, the fault report generation agent node and the end node respectively; and unconditional edges are derived from the fault detection agent node, the fault delimitation and locating agent node and the fault report generation agent node, pointing to the fault detection and locating agent node respectively; and unconditional edges are derived from the fault detection agent node, pointing to the fault detection task executor node; and unconditional edges are derived from the fault detection task executor node, pointing to the fault detection agent node.

[0147] Figure 6 A schematic diagram of the subgraph structure of the alarm processing agent provided in an embodiment of the present application.

[0148] like Figure 6As shown, the alarm processing agent can include an alarm processing agent node, an alarm processing tool node, an alarm information feedback agent node, an alarm feedback tool, an end node, and a start node. The alarm processing agent is used to receive alarm information sent by the network monitoring agent and processing suggestions sent by the troubleshooting and location agent, trigger the corresponding processing flow, such as restarting the device and adjusting the configuration, and feedback the processing results to the operation and maintenance personnel or related systems.

[0149] The steps of creating a subgraph structure of an alarm processing agent may include:

[0150] S100-e: creating an alarm processing agent node, an alarm processing tool node, an alarm information feedback agent node, an alarm feedback tool node, and an end node in the subgraph structure of the alarm processing agent;

[0151] S100-f: The alarm processing agent node leads to conditional edges pointing to the alarm processing tool node, the alarm information feedback agent node, and the end node respectively; and, the alarm processing agent node and the alarm information feedback agent node lead to unconditional edges pointing to the alarm processing agent node respectively; and, the alarm information feedback agent node leads to a conditional edge pointing to the alarm feedback tool node, and, the alarm feedback tool node leads to an unconditional edge pointing to the alarm information feedback agent node.

[0152] Figure 7 A schematic diagram of the subgraph structure of the data analysis agent provided in an embodiment of the present application.

[0153] like Figure 7 As shown, the data analysis agent may include a data analysis agent node, a device status query agent, a device status query tool, a performance indicator analysis agent, a performance indicator analysis tool, an alarm data query agent, an alarm data query tool, a log analysis agent, a log analysis tool, and an end node, and may also include a start node. The data analysis agent is used to analyze and mine collected network data, identify potential network failures or performance bottlenecks, provide optimization suggestions, or predict future network trends.

[0154] The steps of creating a subgraph structure of a data analysis agent may include:

[0155] S100-g: Create a data analysis agent node, a device status query agent node, a device status query tool node, a performance indicator analysis agent node, a performance indicator analysis tool node, an alarm data query agent node, an alarm data query tool node, a log analysis agent node, a log analysis tool node, and an end node in the subgraph structure of the data analysis agent;

[0156] S100-h: The data analysis agent node leads to conditional edges pointing to the status query agent node, the performance indicator analysis agent node, the alarm data query agent node, the log analysis agent node, and the end node respectively; and the device status query agent node, the performance indicator analysis agent node, the alarm data query agent node, and the log analysis agent node lead to unconditional edges pointing to the data analysis agent node respectively; and the device status query agent node leads to a conditional edge pointing to the device status query tool node; and the device status query tool node leads to an unconditional edge pointing to the device status query agent node. The unconditional edge of the node; and the conditional edge directed to the performance indicator analysis tool node induced by the performance indicator analysis agent node; and the unconditional edge directed to the performance indicator analysis agent node induced by the performance indicator analysis tool node; and the conditional edge directed to the alarm data query tool node induced by the alarm data query agent node; and the unconditional edge directed to the alarm data query agent node induced by the alarm data query tool node; and the conditional edge directed to the log analysis tool node induced by the log analysis agent node; and the unconditional edge directed to the log analysis agent node induced by the log analysis tool node.

[0157] Figure 8 A schematic diagram of the subgraph structure of the knowledge base agent provided in an embodiment of the present application.

[0158] like Figure 8 As shown, the knowledge base agent can include a knowledge question and answer agent node and an end node, and can also include a start node. The knowledge base agent is used to store and manage knowledge and experience related to core network operation and maintenance, and provide an operation and maintenance knowledge and operation and maintenance case query interface for other agents.

[0159] The steps of creating a subgraph structure of a knowledge base agent may include:

[0160] S100-i: Create a knowledge question-answering agent node and an end node in the subgraph structure of the knowledge base agent;

[0161] S100-j: An unconditional edge pointing to the end node is drawn from the knowledge question answering agent node.

[0162] It should be noted that Figure 4-Figure 8 In the figure, solid lines with arrows represent unconditional edges, and dashed lines with arrows represent conditional edges.

[0163] Furthermore, in an embodiment of the present application, a human-machine collaborative working mode can be set up for the core network operation and maintenance system. The "Human-in-the-Loop" human-machine collaborative working mode based on the LangGraph framework integrates human (operation and maintenance personnel) input into the automated process through an embedded memory mechanism.

[0164] Specifically, step S102 may further include the following steps S107-S108.

[0165] S107: Create a human review node (human_review) in the subgraph structure.

[0166] It's worth noting that the manual review node interacts with operators through the LLM, enabling natural language interaction between operators and the manual review node. This node can be specifically applied to tool invocation review scenarios, where operators review, edit, or approve tool invocations requested by the LLM before execution. It can also be applied to multi-turn conversations, which involve multiple rounds of back-and-forth between the agent and the user, allowing the agent to gather information and make informed decisions until the agent hands the conversation off to another agent or other system components.

[0167] Figure 9 A schematic diagram of a subgraph structure including a tool review node provided in an embodiment of the present application.

[0168] like Figure 9 As shown, the tool review node can be used to review, edit, or approve tool calls requested by the LLM before they are executed. Specifically, a human_review node can be added between the Agent and tools nodes. If the Agent node returns "END" as the route, the session ends. Otherwise, the session enters the human_review node, where the session pauses and waits for user input. The user can perform several different actions: approve the tool call and continue; manually modify the tool call and continue, such as correcting or supplementing the tool call parameters; provide natural language feedback, which is then passed to the agent to reject the tool call.

[0169] S108: Use conditional edges to connect the manual review node between the sub-agent node and the task executor and / or tool node.

[0170] Alternatively, the manual review node may be connected only to the sub-agent node, and this embodiment of the present application does not impose any specific limitation on this.

[0171] In this way, human-machine collaboration can be achieved.

[0172] Furthermore, the step of creating a graph structure of the supervisor agent may specifically include steps S109-S111.

[0173] S109: Create a graph structure based on the LangGraph framework;

[0174] S110: Create a master agent node in the graph structure, and define a graph state State object for the master agent node. The master agent node is used to update and view the graph state State object.

[0175] It can be understood that the graph structure of the supervisory agent is a Graph.

[0176] S111: Use edge components to connect the graph structure of the supervisor agent with each subgraph structure.

[0177] In this way, a graph representing the complete operation and maintenance process can be formed.

[0178] As can be seen, the large-model-based core network operation and maintenance system construction method provided by the embodiments of this application can decompose the core network operation and maintenance system into multiple smaller independent agents, combine these agents into a multi-agent system, and manage each task execution agent with a supervisor agent to achieve centralized scheduling. Furthermore, based on the method provided by this application, the agents can use LLM to determine control flow, achieving high operation and maintenance efficiency and accuracy.

[0179] It should also be noted that the embodiment of the present application can also build a multi-round dialogue architecture for the core network operation and maintenance system.

[0180] Figure 10 A schematic diagram of the multi-round dialogue architecture provided in an embodiment of the present application.

[0181] like Figure 10 As shown, a multi-turn dialogue architecture can involve multiple rounds of interaction between the agent and the user. This allows the agent to call the LLM to gather more information from the user in a conversational manner, enabling it to make informed decisions until it decides to hand the conversation off to another agent or another part of the system. One or more agent nodes can engage in multi-turn dialogues with the user, with the user providing input or feedback at different stages of the conversation. Based on the results of these actions, the agent node determines its next action: waiting for user input to continue the conversation; routing the request to another agent (or returning to itself, for example, in a loop).

[0182] In some implementations, the fault report generation agent can adopt the above-mentioned multi-round dialogue mode. The user provides feedback on the generated fault report and interacts with the large model multiple times to ultimately generate a high-quality fault analysis report.

[0183] Figure 11 A schematic diagram of the structure of a core network operation and maintenance system based on a large model provided in an embodiment of the present application.

[0184] like Figure 11As shown, the embodiment of the present application provides a core network operation and maintenance system based on a large model, which can also be called a core network operation and maintenance assistant based on a large model. The operation and maintenance system can be constructed based on the aforementioned core network operation and maintenance system construction method based on a large model.

[0185] In the embodiment of the present application, the core network operation and maintenance system based on the large model includes at least a supervisory agent, namely Figure 11 The operation and maintenance supervisor agent shown in FIG can also include multiple task execution agents. Task execution agents can include a network monitoring agent, a fault detection and location agent, an alarm processing agent, a complaint handling agent, a configuration management agent, a network optimization agent, and a knowledge base agent. The supervisor agent is used to communicate with each task execution agent to schedule the task execution agents. Different task execution agents correspond to different operation and maintenance tasks.

[0186] Furthermore, the supervisor agent is specifically configured to execute the following steps S600.

[0187] S600: In response to a user instruction issued by a user terminal to execute an operation and maintenance task, the LLM is called to parse the user instruction to identify the task intent, and one or more corresponding task execution agents are scheduled based on the task intent. The task execution agent includes at least one of the following agents: a network monitoring agent, a fault detection and location agent, an alarm processing agent, a complaint handling agent, a configuration management agent, a network optimization agent, a data analysis agent, and a knowledge base agent. The network monitoring agent is used to monitor the operating status of the core network, the fault detection and location agent is used to locate the cause and location of faults in the core network, the alarm processing agent is used to process alarm information in the core network, the complaint handling agent is used to handle user complaints, the configuration management agent is used to configure and manage the core network, the network optimization agent is used to optimize the core network, the data analysis agent is used to analyze network data of the core network, and the knowledge base agent is used to manage and update the core network operation and maintenance knowledge base.

[0188] The supervisor agent serves as the interface between humans (operation and maintenance personnel) and the intelligent assistant. It can receive operation and maintenance instructions expressed by users in natural language. Afterwards, the supervisor agent can use the large-scale model language perception capability and combine its own operation and maintenance expertise to accurately analyze the operation and maintenance needs of the operation and maintenance personnel, identify the operation and maintenance intentions, and distribute the operation and maintenance instructions to relevant professional agents.

[0189] The task execution agent is specifically configured to execute the following steps S700.

[0190] S700: In response to the scheduling, the LLM is called to generate a task execution process, and the operation and maintenance task is executed based on the task execution process.

[0191] In this way, through the collaboration of multiple intelligent agents, operation and maintenance task perception, operation and maintenance task scheduling, operation and maintenance task execution and execution result return can be achieved.

[0192] Furthermore, the supervisor agent is also configured to perform the following step S800.

[0193] S800: Generate operation and maintenance task information based on task intention and preset operation and maintenance knowledge, and update the operation and maintenance task information in the graph state State object to pass the operation and maintenance task information to the task execution intelligent agent through the graph state State object.

[0194] Illustratively, the operation and maintenance task information may be a name of a network element device in the core network, or the operation and maintenance task information may be used to represent instruction information, such as: performing fault diagnosis on a certain network element.

[0195] Furthermore, the task execution agent is also configured to perform the following steps S901-S902.

[0196] S901: Check the graph state State object to obtain operation and maintenance task information. Based on the operation and maintenance task information and task execution process, call its internal nodes to execute the operation and maintenance tasks. The nodes include sub-agent nodes, task executors, tool nodes, end nodes and / or manual review nodes.

[0197] It is understandable that the graph state State object is continuously updated and is used to track and manage the flow of messages or states between the agent.

[0198] S902: After the operation and maintenance task is completed, the task execution result is updated in the graph state State and routed to the responsible intelligent agent or user end.

[0199] It can be understood that routing to the supervisor agent means returning the control flow to the supervisor agent, while routing to the user end means returning the control flow to the operation and maintenance personnel.

[0200] It should also be noted that in the embodiment of the present application, after the task execution agent obtains the operation and maintenance task information, if cooperation with other agents is required, the control flow return value will be sent to the responsible agent, which will be scheduled again.

[0201] Furthermore, the task execution agent is also configured to perform the following step S903.

[0202] S903: When the task execution process includes routing to an end node, routing to the end node to end the task execution process.

[0203] And / or, the task execution agent is further configured to perform the following step S904.

[0204] S904: When the task execution process includes routing to a manual review node, respond to user input received by the manual review node and route to a node within it, another task execution agent, or a supervisor agent based on the user input.

[0205] Here, routing to a node inside it, another task execution agent or a supervisory agent means transferring the control flow to a node inside it or to another task execution agent or a supervisory agent.

[0206] Exemplarily, in a review tool call scenario, user input may be approving the tool call and continuing, manually modifying the tool call and continuing, where manual modification refers to correcting or supplementing tool call parameters, and user input may also be providing natural language feedback to refuse to execute the call tool.

[0207] In some implementations, such as Figure 11 As shown, the Assistant node represents a manual review node. Conventional tools may not be reviewed, but important tools must be reviewed. This embodiment of the present application does not make specific limitations on this.

[0208] Alternatively, in a multi-round dialogue scenario, user input can be user feedback. For example, the user provides feedback on a fault report generated by the task execution agent, and through multiple interactions with the LLM, a high-quality fault analysis report is finally generated.

[0209] It should be noted that Figure 11 The interaction process shown in includes at least the following three aspects.

[0210] ① Interaction between operation and maintenance personnel and operation and maintenance supervisor agent: The operation and maintenance personnel send a message (Message) to the operation and maintenance supervisor agent, and the operation and maintenance supervisor agent gives a response (Response) after processing.

[0211] ② Task allocation of the operation and maintenance supervisor agent: The operation and maintenance supervisor agent assigns specific tasks to different task execution agents, such as network monitoring agents, fault detection and location agents, etc., through delegation.

[0212] ③ Direct interaction: There is direct interaction (Directly Interact) between the task execution agent and the operation and maintenance personnel, which facilitates collaborative work.

[0213] In an embodiment of the present application, a network monitoring agent includes: a network real-time monitoring agent node, a data collection tool node, an anomaly detection tool node, an alarm trigger tool node and an end node; the network real-time monitoring agent node is configured to: receive real-time monitoring task instructions, plan monitoring tasks, and sequentially call the data collection tool node, the anomaly detection tool node and the alarm trigger tool node to perform monitoring tasks, and the monitoring tasks include data collection monitoring tasks, anomaly detection monitoring tasks and alarm trigger monitoring tasks; the data collection tool node is configured to: integrate external data collection tools and perform data collection monitoring tasks; the anomaly detection tool node is configured to: integrate external anomaly detection tools and perform anomaly detection monitoring tasks, if the external anomaly detection tool detects an alarm event, it is routed to the alarm trigger tool node, if the external anomaly detection tool does not detect an alarm event, it is routed to the end node; the alarm trigger tool node is configured to: integrate external alarm notification tools and perform alarm trigger monitoring tasks;

[0214] The fault detection and positioning agent includes: fault detection and positioning agent node, fault detection agent node, fault delimitation and positioning agent node, fault report generation agent node, fault detection task executor node and end node; the fault detection and positioning agent node is configured to: receive alarm information triggered by the network monitoring process, dispatch tasks to the fault detection agent node, fault delimitation and positioning agent node and fault report generation agent node; the fault detection agent node is configured to: plan the detection task according to business rules, and output the detection task to the fault detection task executor node for execution; the detection task package The system includes at least one of the following tasks: a device status check task, a performance indicator check task, an alarm check task, and a chr log check task; the fault detection task executor node is configured to execute the detection task in a loop; the fault delimitation and positioning agent node is configured to output a delimitation and positioning conclusion based on the fault troubleshooting results and the fault delimitation and positioning rules, and the delimitation and positioning conclusion includes the fault cause and treatment suggestions; the fault report generation agent node is configured to generate a fault analysis report based on the fault troubleshooting results and the delimitation and positioning conclusion, and the fault analysis report includes at least the operating status of the equipment, fault statistics, and performance trends;

[0215] The alarm processing agent includes: alarm processing agent node, alarm processing tool node, alarm information feedback agent node, alarm feedback tool and end node; the alarm processing agent node is configured to: receive alarm information and processing suggestions, and dispatch alarm tasks to the alarm processing tool node; the alarm processing tool node is configured to: execute alarm tasks according to alarm classification and alarm level; the alarm information feedback agent node is configured to: feed back processing results and related information to operation and maintenance personnel or related systems; the data analysis agent includes: data analysis agent node, equipment status query agent, equipment status query tool, performance indicator analysis agent, performance indicator analysis tool, alarm data query agent, alarm data query tool, log analysis agent, Log analysis tool and end node; the data analysis agent node is configured to: perceive user questions and dispatch tasks to the device status query agent node, performance indicator analysis agent node, alarm data query agent node and / or log analysis agent node; the device status query agent node is configured to: integrate external tools of device status query type to execute device status data query; the performance indicator analysis agent node is configured to: integrate external tools of performance indicator analysis type to execute performance indicator data query analysis and indicator prediction; the alarm data query agent node is configured to: integrate external tools of alarm data query type to execute alarm data query and statistics; the log analysis agent node is configured to: integrate external tools of log data query type to execute log data analysis;

[0216] The knowledge base agent includes: knowledge question and answer agent node and end node.

[0217] It can be seen that the core network operation and maintenance system (also referred to as the core network operation and maintenance assistant) provided in the embodiment of the present application can realize multi-agent communication and collaboration design. The core network operation and maintenance assistant is divided according to roles, including multiple agents such as network real-time monitoring, fault detection and positioning, alarm processing, data analysis, knowledge base, etc. These agents complete the core network operation and maintenance tasks together by perceiving the environment, making decisions, and communicating and collaborating with other agents, thereby realizing efficient and accurate management of core network operation and maintenance.

[0218] In summary, the large-model-based core network operation and maintenance system construction method and core network operation and maintenance system provided in the embodiments of this application, by integrating large-model and LangGraph technology, construct an intelligent operation and maintenance framework, improving operation and maintenance efficiency and accuracy. Specifically, it can achieve: ① using large-model and LangGraph to implement operation and maintenance demand analysis and automated task execution; ② based on multimodal data (operation and maintenance knowledge) to perform intelligent analysis to support decision-making; ③ through real-time monitoring and dynamic adjustment, optimize the execution of operation and maintenance tasks; ④ in a multi-agent system, agent collaboration and task scheduling improve work efficiency, realizing intelligent and automated operation and maintenance.

[0219] In general, the embodiments of the present application provide a method for constructing a core network operation and maintenance system based on a large model and a core network operation and maintenance system based on the large model. The method realizes the intelligence and automation of core network operation and maintenance work by constructing a LangGraph graph structure with a large language model and external tool integration. It has the characteristics of modularity, specialization and controllability. Modularity means: dividing into multiple intelligent agents, which can make it easier to develop, test and maintain intelligent agents. Specialization means: creating intelligent agents that focus on specific fields, which helps to improve the overall system performance. Controllability means: the communication method of the intelligent agent can be explicitly controlled.

[0220] This application integrates big models with LangGraph technology to construct an intelligent core network O&M framework. This framework leverages the powerful language understanding and generation capabilities of big models to accurately analyze user O&M requirements. Through the LangGraph framework's intelligent proxy functionality, it enables automated execution and real-time monitoring of O&M tasks. This not only improves O&M efficiency but also significantly enhances O&M accuracy and reliability. Furthermore, based on intelligent analysis of multidimensional data, this application leverages big models to integrate and analyze various types of data, including network performance indicators, logs, and alarm events, to provide a scientific basis for O&M decision-making. Furthermore, this application leverages the planning capabilities of the LangGraph framework to optimize O&M strategies and improve O&M effectiveness. Furthermore, this application enables adaptive dynamic adjustment of O&M tasks. During the O&M process, this application monitors the execution of O&M tasks and system status in real time, leveraging the dynamic adjustment capabilities of the LangGraph framework to adaptively adjust O&M tasks. This adaptive adjustment mechanism ensures efficient execution of O&M tasks while reducing O&M costs and improving O&M efficiency. Furthermore, the present application provides a multi-agent communication and collaboration mechanism, in which the core network operation and maintenance assistant agents exchange information and share resources by sharing graph states, using standardized communication protocols and data formats to ensure accurate transmission and parsing of information. The task scheduling agent (which can be integrated with the supervisor agent into the same agent) assigns tasks to each agent based on the requirements of the operation and maintenance tasks and the capabilities of the agent. Each agent works together according to the assigned tasks to jointly complete the task objectives. In summary, the method provided by this application can realize the intelligence and automation of core network operation and maintenance, and improve the efficiency and accuracy of operation and maintenance.

[0221] Figure 12 A schematic diagram of the structure of a core network operation and maintenance system construction device based on a large model provided in an embodiment of the present application.

[0222] like Figure 12 As shown, an embodiment of the present application provides a device for constructing a core network operation and maintenance system based on a large model, which may include:

[0223] The first building block 100 is configured to create a subgraph structure of multiple task execution agents based on a core network operation and maintenance task using a LangGraph framework, and to create a graph structure of a supervisor agent using a LangGraph framework. The supervisor agent is configured to communicate with each task execution agent to schedule the task execution agents, and different task execution agents correspond to different operation and maintenance tasks. The task execution agents include at least one of the following agents: a network monitoring agent, a fault detection and location agent, an alarm processing agent, a complaint handling agent, a configuration management agent, a network optimization agent, a data analysis agent, and a knowledge base agent. The network monitoring agent is configured to monitor the operating status of the core network, the fault detection and location agent is configured to locate the cause and location of faults in the core network, the alarm processing agent is configured to process alarm information in the core network, the complaint handling agent is configured to process user complaints, the configuration management agent is configured to configure and manage the core network, the network optimization agent is configured to optimize the core network, the data analysis agent is configured to analyze network data of the core network, and the knowledge base agent is configured to manage and update the core network operation and maintenance knowledge base.

[0224] A relationship establishment module 200 is used to establish a calling relationship between a graph structure and / or a subgraph structure and a large language model LLM;

[0225] Compilation module 300, used to compile the graph structure and subgraph structure to obtain a graph instance;

[0226] The second construction module 400 is used to construct a core network operation and maintenance system based on a graph instance; wherein the graph instance is used to define the operation process of the core network operation and maintenance system. When the core network operation and maintenance system is running, the graph instance runs to execute the operation process; the operation process at least includes: when the user terminal issues a user instruction to the core network operation and maintenance system to execute an operation and maintenance task, the supervisor agent responds to the user instruction, calls the LLM to parse the user instruction to identify the task intent, and schedules one or more corresponding task execution agents based on the task intent; and the operation process also includes: the task execution agent responds to the scheduling and calls the LLM to generate a task execution process, and executes the operation and maintenance task based on the task execution process.

[0227] In some implementations, the first construction module 100 is specifically used to: create a subgraph structure based on the LangGraph framework for each task execution agent; create one or more sub-agent nodes in the subgraph structure, and define a graph state State object for the sub-agent node, the sub-agent node is used to update and view the graph state State object, and the number of sub-agent nodes is determined based on the task execution process corresponding to the task execution agent; and / or create task executors and / or tool nodes in the subgraph structure, the number and type of task executors and tool nodes are determined based on the operation and maintenance tasks corresponding to the task execution agent; create an end node in the subgraph structure.

[0228] In some implementations, the first construction module 100 is specifically used to: set an entry point for each subgraph structure, where the entry point is a sub-agent node, and the entry point represents the starting point for the task execution agent to perform the operation and maintenance task; use edge components to connect the sub-agent nodes, task executors, tool nodes and / or end nodes, and the edge components include conditional edges and unconditional edges. The conditional edges have routing conditions, and the routing conditions at least include that after the sub-agent node calls the LLM, if the return message of the LLM includes a tool call field, it is routed to the task executor and / or tool node; if the return message does not contain a tool call field, it is routed to the end node.

[0229] In some implementations, the first construction module 100 is further used to: create a manual review node in the subgraph structure; and connect the manual review node between the sub-agent node and the task executor and / or tool node using conditional edges.

[0230] In some implementations, the first construction module 100 is also used to: create a graph structure based on the LangGraph framework; create a main agent node in the graph structure, and define a graph state State object for the main agent node, and the main agent node is used to update and view the graph state State object; use edge components to connect the graph structure of the supervisor agent with each sub-graph structure; and the first construction module 100 is also used to: create an LLM and an external tool set, the external tool set includes multiple callable tools, and the tools are determined based on operation and maintenance tasks; bind each tool in the external tool set to the LLM through the LLM interface.

[0231] In some implementations, the first construction module 100 is also used to: create a network real-time monitoring agent node, a data collection tool node, an anomaly detection tool node, an alarm trigger tool node and an end node in the subgraph structure of the network monitoring agent; lead an unconditional edge pointing to the data collection tool node from the network real-time monitoring agent node, and lead an unconditional edge pointing to the anomaly detection tool node from the data collection tool node, and lead conditional edges pointing to the alarm trigger tool node and the end node respectively from the anomaly detection tool node, and lead an unconditional edge pointing to the end node from the alarm trigger tool node.

[0232] In some implementations, the first building module 100 is also used to: create a fault detection and location agent node, a fault detection agent node, a fault delimitation and location agent node, a fault report generation agent node, a fault detection task executor node and an end node in the subgraph structure of the fault detection and location agent; derive conditional edges from the fault detection and location agent node pointing to the fault detection agent node, the fault delimitation and location agent node, the fault report generation agent node and the end node respectively; and, derive unconditional edges from the fault detection agent node, the fault delimitation and location agent node and the fault report generation agent node pointing to the fault detection and location agent node respectively; and, derive an unconditional edge from the fault detection agent node pointing to the fault detection task executor node; and, derive an unconditional edge from the fault detection task executor node pointing to the fault detection agent node.

[0233] In some implementations, the first construction module 100 is also used to: create an alarm processing agent node, an alarm processing tool node, an alarm information feedback agent node, an alarm feedback tool node and an end node in the subgraph structure of the alarm processing agent; derive conditional edges from the alarm processing agent node pointing to the alarm processing tool node, the alarm information feedback agent node and the end node respectively; and, derive unconditional edges from the alarm processing agent node and the alarm information feedback agent node pointing to the alarm processing agent node respectively; and, derive a conditional edge from the alarm information feedback agent node pointing to the alarm feedback tool node, and, derive an unconditional edge from the alarm feedback tool node pointing to the alarm information feedback agent node.

[0234] In some implementations, the first construction module 100 is also used to: create a data analysis agent node, a device status query agent node, a device status query tool node, a performance indicator analysis agent node, a performance indicator analysis tool node, an alarm data query agent node, an alarm data query tool node, a log analysis agent node, a log analysis tool node and an end node in the subgraph structure of the data analysis agent; lead conditional edges from the data analysis agent node pointing to the status query agent node, the performance indicator analysis agent node, the alarm data query agent node, the log analysis agent node and the end node respectively; and lead conditional edges from the device status query agent node, the performance indicator analysis agent node, the alarm data query agent node and the log analysis agent node pointing to the data analysis agent node respectively. Unconditional edge; and, conditional edge directed to device status query tool node induced by device status query agent node; and, unconditional edge directed to device status query agent node induced by device status query tool node; and, conditional edge directed to performance indicator analysis tool node induced by performance indicator analysis agent node; and, unconditional edge directed to performance indicator analysis agent node induced by performance indicator analysis tool node; and, conditional edge directed to alarm data query tool node induced by alarm data query agent node; and, unconditional edge directed to alarm data query agent node induced by alarm data query tool node; and, conditional edge directed to log analysis tool node induced by log analysis agent node; and, unconditional edge directed to log analysis agent node induced by log analysis tool node.

[0235] In some implementations, the first construction module 100 is further used to: create a knowledge question and answer agent node and an end node in the subgraph structure of the knowledge base agent; and draw an unconditional edge from the knowledge question and answer agent node to the end node.

[0236] The embodiment of the present application further provides a core network operation and maintenance method based on a large model, including the following steps S1001-S1002:

[0237] S1001: The supervisor agent responds to a user instruction issued by a user terminal to execute an operation and maintenance task, calls the LLM to parse the user instruction to identify the task intent, and schedules one or more corresponding task execution agents based on the task intent;

[0238] S1002: The task execution agent responds to the scheduling and calls the LLM to generate a task execution process, and executes the operation and maintenance task based on the task execution process.

[0239] Among them, the supervisor agent is used to communicate with each task execution agent, and different task execution agents correspond to different operation and maintenance tasks.

[0240] In some implementations, step S1002 may specifically include the following steps.

[0241] S1002-1: The supervisor agent generates operation and maintenance task information based on the task intention and preset operation and maintenance knowledge, and updates the operation and maintenance task information in the graph state State object, so as to pass the operation and maintenance task information to the task execution agent through the graph state State object;

[0242] S1002-2: The task execution agent checks the graph state State object to obtain the operation and maintenance task information. Based on the operation and maintenance task information and the task execution process, it calls its internal nodes to execute the operation and maintenance task. The nodes include sub-agent nodes, task executors, tool nodes, end nodes, and / or manual review nodes.

[0243] S1002-3: After the operation and maintenance task is completed, the task execution agent updates the task execution result in the graph state State and routes it to the supervisor agent or the user end.

[0244] In some implementations, step S1002 may further include the following steps.

[0245] S1002-4: When the task execution process includes routing to an end node, the task execution agent routes to the end node to end the task execution process;

[0246] And / or, S1002-5: When the task execution agent includes routing to a manual review node in the task execution process, it responds to user input received by the manual review node and routes to a node within it, another task execution agent or a supervisor agent based on the user input.

[0247] In some implementations, the network monitoring agent includes: a network real-time monitoring agent node, a data collection tool node, an anomaly detection tool node, an alarm triggering tool node and an end node; the network real-time monitoring agent node is configured to: receive real-time monitoring task instructions, plan monitoring tasks, and sequentially call the data collection tool node, the anomaly detection tool node and the alarm triggering tool node to perform monitoring tasks, the monitoring tasks including data collection monitoring tasks, anomaly detection monitoring tasks and alarm triggering monitoring tasks; the data collection tool node is configured to: integrate external data collection tools and perform data collection monitoring tasks; the anomaly detection tool node is configured to: integrate external anomaly detection tools and perform anomaly detection monitoring tasks, if the external anomaly detection tool detects an alarm event, it is routed to the alarm triggering tool node, if the external anomaly detection tool does not detect an alarm event, it is routed to the end node; the alarm triggering tool node is configured to: integrate external alarm notification tools and perform alarm triggering monitoring tasks;

[0248] The fault detection and positioning agent includes: fault detection and positioning agent node, fault detection agent node, fault delimitation and positioning agent node, fault report generation agent node, fault detection task executor node and end node; the fault detection and positioning agent node is configured to: receive alarm information triggered by the network monitoring process, dispatch tasks to the fault detection agent node, fault delimitation and positioning agent node and fault report generation agent node; the fault detection agent node is configured to: plan the detection task according to business rules, and output the detection task to the fault detection task executor node for execution; the detection task package The system includes at least one of the following tasks: a device status check task, a performance indicator check task, an alarm check task, and a chr log check task; the fault detection task executor node is configured to execute the detection task in a loop; the fault delimitation and positioning agent node is configured to output a delimitation and positioning conclusion based on the fault troubleshooting results and the fault delimitation and positioning rules, and the delimitation and positioning conclusion includes the fault cause and treatment suggestions; the fault report generation agent node is configured to generate a fault analysis report based on the fault troubleshooting results and the delimitation and positioning conclusion, and the fault analysis report includes at least the operating status of the equipment, fault statistics, and performance trends;

[0249] The alarm processing agent includes: alarm processing agent node, alarm processing tool node, alarm information feedback agent node, alarm feedback tool and end node; the alarm processing agent node is configured to: receive alarm information and processing suggestions, and dispatch alarm tasks to the alarm processing tool node; the alarm processing tool node is configured to: execute alarm tasks according to alarm classification and alarm level; the alarm information feedback agent node is configured to: feed back processing results and related information to operation and maintenance personnel or related systems; the data analysis agent includes: data analysis agent node, equipment status query agent, equipment status query tool, performance indicator analysis agent, performance indicator analysis tool, alarm data query agent, alarm data query tool, log analysis agent, Log analysis tool and end node; the data analysis agent node is configured to: perceive user questions and dispatch tasks to the device status query agent node, performance indicator analysis agent node, alarm data query agent node and / or log analysis agent node; the device status query agent node is configured to: integrate external tools of device status query type to execute device status data query; the performance indicator analysis agent node is configured to: integrate external tools of performance indicator analysis type to execute performance indicator data query analysis and indicator prediction; the alarm data query agent node is configured to: integrate external tools of alarm data query type to execute alarm data query and statistics; the log analysis agent node is configured to: integrate external tools of log data query type to execute log data analysis;

[0250] The knowledge base agent includes: knowledge question and answer agent node and end node.

[0251] In a specific implementation, the present invention further provides a computer storage medium, wherein the computer storage medium may store a program that, when executed, may include some or all of the steps of each embodiment of the method for constructing a core network operation and maintenance system based on a large model and / or the core network operation and maintenance method based on a large model provided by the present invention. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0252] It is easy to understand that those skilled in the art can combine, split, reorganize, etc. the embodiments of the present application based on the several embodiments provided in the present application to obtain other embodiments, and these embodiments do not exceed the scope of protection of the present application.

[0253] The above specific implementation methods further explain in detail the purpose, technical solutions and beneficial effects of the embodiments of the present application. It should be understood that the above are only specific implementation methods of the embodiments of the present application and are not intended to limit the scope of protection of the embodiments of the present application. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solutions of the embodiments of the present application should be included in the scope of protection of the embodiments of the present application.

Claims

1. A method for constructing a core network operation and maintenance system based on a large model, characterized in that: The method comprises: Based on the operation and maintenance tasks of the core network, a subgraph structure of multiple task execution agents is created using the LangGraph framework, and a graph structure of a supervisor agent is created using the LangGraph framework; wherein the supervisor agent is used to communicate with each of the task execution agents to schedule the task execution agents, and different task execution agents correspond to different operation and maintenance tasks; the task execution agents include at least one of the following agents: a network monitoring agent, a fault detection and location agent, an alarm processing agent, a complaint handling agent, a configuration management agent, a network optimization agent, a data analysis agent, and a knowledge base agent, wherein the network monitoring agent is used to monitor the operating status of the core network, the fault detection and location agent is used to locate the cause and location of the fault in the core network, the alarm processing agent is used to process the alarm information in the core network, the complaint handling agent is used to handle user complaints, the configuration management agent is used to configure and manage the core network, the network optimization agent is used to optimize the core network, the data analysis agent is used to analyze the network data of the core network, and the knowledge base agent is used to manage and update the core network operation and maintenance knowledge base; Establishing a calling relationship between the graph structure and / or the subgraph structure and a large language model LLM; Compiling the graph structure and the subgraph structure to obtain a graph instance; The core network operation and maintenance system is constructed based on the graph instance; wherein, the graph instance is used to define the operation process of the core network operation and maintenance system, and when the core network operation and maintenance system is running, the graph instance runs to execute the operation process; the operation process at least includes: when the user terminal issues a user instruction to the core network operation and maintenance system to execute the operation task, the supervisor agent responds to the user instruction, calls the LLM to parse the user instruction to identify the task intent, and schedules the corresponding one or more task execution agents based on the task intent; and the operation process also includes: the task execution agent responds to the scheduling and calls the LLM to generate a task execution process, and executes the operation and maintenance task based on the task execution process.

2. The method for constructing a core network operation and maintenance system based on a large model according to claim 1, characterized in that: The steps to create a subgraph structure of multiple task execution agents using the LangGraph framework include: For each of the task execution agents, create the subgraph structure based on the LangGraph framework; Creating one or more sub-agent nodes in the sub-graph structure, and defining a graph state State object for the sub-agent nodes, wherein the sub-agent nodes are used to update and view the graph state State object, and the number of the sub-agent nodes is determined based on the task execution process corresponding to the task execution agent; and / or, creating task executors and / or tool nodes in the subgraph structure, wherein the number and type of the task executors and the tool nodes are determined based on the operation and maintenance tasks corresponding to the task execution agent; An end node is created in the subgraph structure.

3. The method for constructing a core network operation and maintenance system based on a large model according to claim 2, characterized in that: The steps of creating a subgraph structure of multiple task execution agents using the LangGraph framework also include: For each of the subgraph structures, an entry point is set, where the entry point is the sub-agent node, and the entry point represents the starting point for the task execution agent to execute the operation and maintenance task; The sub-agent node, the task executor, the tool node and / or the end node are connected by edge components, and the edge components include conditional edges and unconditional edges. The conditional edges have routing conditions, and the routing conditions at least include that after the sub-agent node calls the LLM, if the return message of the LLM includes a tool call field, it is routed to the task executor and / or the tool node; if the return message does not include a tool call field, it is routed to the end node.

4. The method for constructing a core network operation and maintenance system based on a large model according to claim 3, characterized in that: After the step of creating a task executor and / or tool node in the subgraph structure, the method further comprises: creating a manual review node in the subgraph structure; The conditional edge is used to connect the manual review node between the sub-agent node and the task executor and / or the tool node.

5. The method for constructing a core network operation and maintenance system based on a large model according to claim 3, characterized in that: The steps to create the graph structure of the supervisor agent include: Creating the graph structure based on the LangGraph framework; Creating a master agent node in the graph structure, and defining the graph state State object for the master agent node, wherein the master agent node is used to update and view the graph state State object; connecting the graph structure of the supervisor agent with each of the subgraph structures using the edge components; And, the method further comprises: Creating the LLM and an external tool set, wherein the external tool set includes a plurality of callable tools, and the tools are determined based on the operation and maintenance task; Each of the tools in the external tool set is bound to the LLM through the interface of the LLM.

6. The method for constructing a core network operation and maintenance system based on a large model according to claim 2, characterized in that: The steps of creating the subgraph structure of the network monitoring agent include: Creating a network real-time monitoring agent node, a data acquisition tool node, an anomaly detection tool node, an alarm triggering tool node, and an end node in a subgraph structure of the network monitoring agent; An unconditional edge pointing to the data acquisition tool node is derived from the network real-time monitoring agent node, an unconditional edge pointing to the anomaly detection tool node is derived from the data acquisition tool node, and conditional edges pointing to the alarm triggering tool node and the end node are derived from the anomaly detection tool node, and an unconditional edge pointing to the end node is derived from the alarm triggering tool node; The steps of creating the subgraph structure of the fault detection and location agent include: Creating a fault detection and positioning agent node, a fault detection agent node, a fault delimitation and positioning agent node, a fault report generation agent node, a fault detection task executor node, and an end node in a subgraph structure of the fault detection and positioning agent; Conditional edges pointing to the fault detection agent node, the fault delimitation and positioning agent node, the fault report generation agent node and the end node are derived from the fault detection and positioning agent node; and unconditional edges pointing to the fault detection and positioning agent node are derived from the fault detection agent node, the fault delimitation and positioning agent node and the fault report generation agent node respectively; and unconditional edges pointing to the fault detection task executor node are derived from the fault detection agent node; and unconditional edges pointing to the fault detection agent node are derived from the fault detection task executor node; The steps of creating the subgraph structure of the alarm processing agent include: Creating an alarm processing agent node, an alarm processing tool node, an alarm information feedback agent node, an alarm feedback tool node, and an end node in a subgraph structure of the alarm processing agent; Conditional edges pointing to the alarm processing tool node, the alarm information feedback agent node, and the end node are derived from the alarm processing agent node; and unconditional edges pointing to the alarm processing agent node are derived from the alarm processing agent node and the alarm information feedback agent node respectively; and a conditional edge pointing to the alarm feedback tool node is derived from the alarm information feedback agent node, and an unconditional edge pointing to the alarm information feedback agent node is derived from the alarm feedback tool node; The steps of creating the subgraph structure of the data analysis agent include: Creating a data analysis agent node, a device status query agent node, a device status query tool node, a performance indicator analysis agent node, a performance indicator analysis tool node, an alarm data query agent node, an alarm data query tool node, a log analysis agent node, a log analysis tool node, and an end node in the subgraph structure of the data analysis agent; Conditional edges are derived from the data analysis agent node, pointing to the status query agent node, the performance indicator analysis agent node, the alarm data query agent node, the log analysis agent node, and the end node respectively; and unconditional edges are derived from the device status query agent node, the performance indicator analysis agent node, the alarm data query agent node, and the log analysis agent node, pointing to the data analysis agent node respectively; and conditional edges are derived from the device status query agent node, pointing to the device status query tool node; and an edge is derived from the device status query tool node, pointing to the device status query agent node. unconditional edge; and, a conditional edge pointing to the performance indicator analysis tool node is derived from the performance indicator analysis agent node; and, an unconditional edge pointing to the performance indicator analysis agent node is derived from the performance indicator analysis tool node; and, a conditional edge pointing to the alarm data query tool node is derived from the alarm data query agent node; and, an unconditional edge pointing to the alarm data query agent node is derived from the alarm data query tool node; and, a conditional edge pointing to the log analysis tool node is derived from the log analysis agent node; and, an unconditional edge pointing to the log analysis agent node is derived from the log analysis tool node; The steps of creating the knowledge base agent include: Creating a knowledge question-answering agent node and an end node in the subgraph structure of the knowledge base agent; An unconditional edge pointing to the end node is drawn from the knowledge question and answer agent node.

7. A core network operation and maintenance system based on a large model, characterized in that: The system is constructed based on the method for constructing a core network operation and maintenance system based on a large model according to any one of claims 1 to 6, and the system includes at least a supervisor agent and a task execution agent, wherein the supervisor agent is used to communicate with each of the task execution agents to schedule the task execution agents, and different task execution agents correspond to different operation and maintenance tasks; The supervisor agent is configured to: respond to a user instruction issued by a user terminal to execute the operation and maintenance task, call the LLM to parse the user instruction to identify the task intent, and schedule one or more corresponding task execution agents based on the task intent; The task execution agent includes at least one of the following agents: a network monitoring agent, a fault detection and positioning agent, an alarm processing agent, a complaint handling agent, a configuration management agent, a network optimization agent, a data analysis agent, and a knowledge base agent. The network monitoring agent is used to monitor the operating status of the core network, the fault detection and positioning agent is used to locate the cause and location of the fault in the core network, the alarm processing agent is used to process the alarm information in the core network, the complaint handling agent is used to handle user complaints, the configuration management agent is used to configure and manage the core network, the network optimization agent is used to optimize the core network, the data analysis agent is used to analyze the network data of the core network, and the knowledge base agent is used to manage and update the core network operation and maintenance knowledge base; The task execution agent is configured to: generate a task execution process in response to scheduling and calling the LLM, and execute the operation and maintenance task based on the task execution process.

8. The core network operation and maintenance system based on a large model according to claim 7, characterized in that: The supervisor agent is further configured to: Generate operation and maintenance task information based on the task intention and preset operation and maintenance knowledge, and update the operation and maintenance task information in the graph state State object, so as to transmit the operation and maintenance task information to the task execution agent through the graph state State object; The task execution agent is further configured to: view the graph state State object to obtain the operation and maintenance task information, and based on the operation and maintenance task information and the task execution process, call its internal nodes to execute the operation and maintenance task, wherein the nodes include sub-agent nodes, task executors, tool nodes, end nodes and / or manual review nodes; After the operation and maintenance task is completed, the task execution result is updated in the graph state State and routed to the supervisor agent or the user end; The task execution agent is further configured to: When routing to the end node is included in the task execution process, routing to the end node to end the task execution process; And / or, when the task execution process includes routing to the manual review node, in response to user input received by the manual review node, routing to the node within it, another task execution agent or the supervisor agent based on the user input.

9. The core network operation and maintenance system based on a large model according to claim 8, characterized in that: The network monitoring agent includes: a network real-time monitoring agent node, a data acquisition tool node, an anomaly detection tool node, an alarm trigger tool node and an end node; the network real-time monitoring agent node is configured to: receive real-time monitoring task instructions, plan monitoring tasks, and sequentially call the data acquisition tool node, the anomaly detection tool node and the alarm trigger tool node to perform the monitoring tasks, and the monitoring tasks include data acquisition monitoring tasks, anomaly detection monitoring tasks and alarm trigger monitoring tasks; the data acquisition tool node is configured to: integrate external data acquisition tools and perform the data acquisition monitoring tasks; the anomaly detection tool node is configured to: integrate external anomaly detection tools and perform the anomaly detection monitoring tasks, if the external anomaly detection tool detects an alarm event, it is routed to the alarm trigger tool node, if the external anomaly detection tool does not detect the alarm event, it is routed to the end node; the alarm trigger tool node is configured to: integrate external alarm notification tools and perform the alarm trigger monitoring tasks; The fault detection and positioning agent includes: a fault detection and positioning agent node, a fault detection agent node, a fault delimitation and positioning agent node, a fault report generation agent node, a fault detection task executor node and an end node; the fault detection and positioning agent node is configured to: receive alarm information triggered by a network monitoring process, and dispatch tasks to the fault detection agent node, the fault delimitation and positioning agent node and the fault report generation agent node; the fault detection agent node is configured to: plan the detection task according to business rules, and output the detection task to the fault detection task executor node for execution; the detection task The system comprises at least one of the following tasks: a device status check task, a performance indicator check task, an alarm check task, and a chr log check task; the fault detection task executor node is configured to execute the detection task in a loop; the fault delimitation and positioning agent node is configured to output a delimitation and positioning conclusion based on the fault troubleshooting results and the fault delimitation and positioning rules, the delimitation and positioning conclusion including the fault cause and treatment suggestions; the fault report generation agent node is configured to generate a fault analysis report based on the fault troubleshooting results and the delimitation and positioning conclusion, the fault analysis report including at least the operating status of the device, fault statistics, and performance trends; The alarm processing agent includes: an alarm processing agent node, an alarm processing tool node, an alarm information feedback agent node, an alarm feedback tool and an end node; the alarm processing agent node is configured to: receive alarm information and processing suggestions, and dispatch alarm tasks to the alarm processing tool node; the alarm processing tool node is configured to: execute the alarm task according to the alarm classification and alarm level; the alarm information feedback agent node is configured to: feed back the processing results and related information to the operation and maintenance personnel or related systems; the data analysis agent includes: a data analysis agent node, an equipment status query agent, an equipment status query tool, a performance indicator analysis agent, a performance indicator analysis tool, an alarm data query agent, an alarm data query tool, a log analysis agent ... Analysis tools and end nodes; the data analysis agent node is configured to: perceive user questions and dispatch tasks to the device status query agent node, the performance indicator analysis agent node, the alarm data query agent node and / or the log analysis agent node; the device status query agent node is configured to: integrate external tools of the device status query type to execute device status data query; the performance indicator analysis agent node is configured to: integrate external tools of the performance indicator analysis type to execute performance indicator data query analysis and indicator prediction; the alarm data query agent node is configured to: integrate external tools of the alarm data query type to execute alarm data query and statistics; the log analysis agent node is configured to: integrate external tools of the log data query type to execute log data analysis; The knowledge base agent includes: a knowledge question and answer agent node and an end node.

10. A core network operation and maintenance system construction device based on a large model, characterized in that: The device comprises: The first building module is configured to create a subgraph structure of multiple task execution agents based on the core network operation and maintenance tasks using the LangGraph framework, and to create a graph structure of a supervisor agent using the LangGraph framework; wherein the supervisor agent is configured to communicate with each of the task execution agents to schedule the task execution agents, and different task execution agents correspond to different operation and maintenance tasks; the task execution agents include at least one of the following agents: a network monitoring agent, a fault detection and location agent, an alarm processing agent, a complaint processing agent, a configuration management agent, a network optimization agent, a data analysis agent, and a knowledge base agent; the network monitoring agent is configured to monitor the operation status of the core network, the fault detection and location agent is configured to locate the cause and location of faults in the core network, the alarm processing agent is configured to process alarm information in the core network, the complaint processing agent is configured to process user complaints, the configuration management agent is configured to configure and manage the core network, the network optimization agent is configured to optimize the core network, the data analysis agent is configured to analyze network data of the core network, and the knowledge base agent is configured to manage and update the core network operation and maintenance knowledge base; A relationship establishment module, configured to establish a calling relationship between the graph structure and / or the subgraph structure and the large language model LLM; A compiling module, configured to compile the graph structure and the subgraph structure to obtain a graph instance; The second construction module is used to construct the core network operation and maintenance system based on the graph instance; wherein, the graph instance is used to define the operation process of the core network operation and maintenance system, and when the core network operation and maintenance system is running, the graph instance runs to execute the operation process; the operation process at least includes: when the user terminal issues a user instruction to the core network operation and maintenance system to execute the operation task, the supervisor agent responds to the user instruction, calls the LLM to parse the user instruction to identify the task intent, and schedules the corresponding one or more task execution agents based on the task intent; and the operation process also includes: the task execution agent responds to the scheduling and calls the LLM to generate a task execution process, and executes the operation and maintenance task based on the task execution process.

Images