An ai model calling method and system of a 5g core network

Abstract

The application discloses an AI model calling method and system of a 5G core network, and the method comprises the following steps: establishing a communication connection between a client of the 5G core network and a MCP server; acquiring tools, resources and prompt words provided by the MCP server by the client of the 5G core network and injecting the tools, resources and prompt words into an AI model by using the MCP; constructing a model context according to the injected tools, resources and prompt words by the AI model, and returning the constructed feedback to the client of the 5G core network; injecting KPI data into the model context obtained by the AI model by the client of the 5G core network; returning a prediction result and the name and parameters of a tool required to be called to the client of the 5G core network by the AI model according to the model context and the KPI data; and calling the corresponding tool according to the name and parameters of the tool required to be called by the client of the 5G core network by using the MCP and executing the prediction result after the parameters are input into the tool.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and particularly relates to an AI model calling method and system of a 5G core network. BACKGROUND

[0002] In the process of evolution of 5G network to intelligence, the collaborative fusion of traditional 5G core network elements and artificial intelligence (AI) becomes the direction of improving network automation and optimizing resource scheduling. However, due to the differences between the traditional network element architecture and AI technology, the adaptability of the network element architecture and AI algorithm is insufficient in the collaborative process. SUMMARY

[0003] The technical problem to be solved by the present application is to provide an AI model calling method and system of a 5G core network to solve the problem of insufficient adaptability in the collaborative process of the network element architecture and AI algorithm of the existing 5G core network.

[0004] In a first aspect, the present application provides an AI model calling method of a 5G core network, comprising the following steps: establishing a communication connection between a client of a 5G core network and a MCP server; using MCP to obtain tools, resources and prompt words provided by the MCP server and inject them into an AI model through the client of the 5G core network; constructing a model context according to the injected tools, resources and prompt words through the AI model, and returning the constructed feedback to the client of the 5G core network; injecting KPI data into the model context obtained by the AI model through the client of the 5G core network; returning the prediction result and the name and parameters of the required called tool to the client of the 5G core network through the AI model according to the model context and KPI data; and calling the corresponding tool according to the name and parameters of the required called tool through the client of the 5G core network using MCP to the MCP server, and executing the prediction result after the parameter is entered.

[0005] In a second aspect, the present application also provides an AI model calling system of a 5G core network, comprising a communication connection module, a network analysis module, a model context construction module, a KPI data injection module, an AI prediction module and a tool calling module, the communication connection module is used to establish a communication connection between a client of the 5G core network and a MCP server; the network analysis module is used to obtain tools, resources and prompt words provided by the MCP server by the MCP through the client of the 5G core network and inject them into the AI model; the model context construction module is used to construct a model context according to the injected tools, resources and prompt words by the AI model, and return the construction feedback to the client of the 5G core network; the KPI data injection module is used to inject KPI data into the model context obtained by the client of the 5G core network; the AI prediction module is used to return a prediction result and the name and parameters of the required called tool to the client of the 5G core network according to the model context and the KPI data by the AI model; and the tool calling module is used to call the corresponding tool according to the name and parameters of the required called tool by the MCP server through the MCP by the client of the 5G core network, and execute the prediction result after the parameters are input.

[0006] The beneficial technical effects of the present application are that the AI model calling method of the 5G core network of the present application introduces MCP in the 5G core network, establishes a communication connection between the client of the 5G core network and the MCP server, the MCP realizes cross-session context persistence by encapsulating context data, ensures the logical coherence of the AI model in multiple rounds of interaction, realizes the collaborative work of the network element of the 5G core network and the AI model through the MCP, improves the adaptability between the network element of the 5G core network and the AI model, and provides a unified interface specification through the MCP, simplifies the integration process of the AI model and external data sources and / or tools. The AI model calling system of the 5G core network of the present application also has the above functions. BRIEF DESCRIPTION OF DRAWINGS

[0007] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0008] Figure 1 The flowchart of the AI model calling method of the 5G core network provided by the embodiments of the present application;

[0009] Figure 2 The sub-flowchart of the AI model calling method of the 5G core network provided by the embodiments of the present application;

[0010] Figure 3A framework schematic diagram of an AI model calling system of a 5G core network provided by an embodiment of the present application. DETAILED DESCRIPTION

[0011] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0012] Please refer to Figure 1 , Figure 1 A flow schematic diagram of an AI model calling method of a 5G core network provided by an embodiment of the present application, the AI model calling method of the 5G core network comprising the following steps:

[0013] S10: establishing a communication connection between a client of the 5G core network and a MCP server;

[0014] S20: acquiring tools, resources and prompt words provided by the MCP server by the client of the 5G core network using the MCP and injecting them into the AI model;

[0015] Among them, the client of the 5G core network can acquire resources and prompt words to preset specific scene injection AI model conversation; the client of the 5G core network can acquire tool description and inject AI model conversation to make the AI model know which tools can be called;

[0016] S30: constructing model context according to the injected tools, resources and prompt words by the AI model, and returning the construction feedback to the client of the 5G core network;

[0017] S40: injecting KPI (Key Performance Indicator) data into the model context obtained by the AI model construction by the client of the 5G core network; for example, when the model context is a card communication quality related index, the KPI data can be card communication state data, then the client of the 5G core network can inject the card communication state data into the model context about the card communication quality related index obtained by the AI model construction;

[0018] S50: returning the prediction result and the name and parameters of the required called tool to the client of the 5G core network by the AI model according to the model context and the KPI data;

[0019] Among them, the AI model responds to the prediction result according to the model context and the KPI data, judges whether the tool needs to be executed, and returns the prediction result and the related information of the required called execution tool to the client of the 5G core network;

[0020] S60: The 5G core network client uses the MCP to call the corresponding tool according to the name and parameters of the tool required to be called, and executes the prediction result after the parameters are input. The 5G core network client uses the MCP to call the corresponding tool according to the name and parameters of the tool required to be called, and inputs the parameters corresponding to the tool into the tool to execute the prediction result, and executes the prediction result through the tool after the parameters are input.

[0021] Wherein, MCP (Model Context Protocol) is an open protocol used to define the interaction between the model and the external environment. Tools are functions provided by the server that can be called, allowing the AI model to actively call corresponding operations, including searching, calculating, and querying databases, etc. The results of the corresponding operations performed through the tools can become new model contexts. Resources are static or semi-static data sets stored by the server, including code files, configuration files, documents, and templates. By reading the contents of the resources, project background information can be obtained. Prompts are standardized interaction templates used to define model roles and task processes to ensure consistent behavior. By dynamically injecting tools, resources, and prompts into the context through MCP, the model context becomes an active, extensible, and real-world-related information space. The 5G core network client can also be a client of MCP.

[0022] The AI model calling method of the 5G core network introduces MCP in the 5G core network, establishes a communication connection between the 5G core network client and the MCP server, and realizes cross-session context persistence through encapsulating context data, ensuring the logical coherence of the AI model in multiple rounds of interaction. The MCP enables the collaborative work of the network elements of the 5G core network and the AI model, improves the adaptability between the network elements of the 5G core network and the AI model, and provides a unified interface specification through the MCP, simplifying the integration process of the AI model and external data sources and / or tools, which is beneficial to solving the complex problems of deployment and operation, and improving the overall performance and service capability of the 5G core network. Context data includes historical dialogues, user preferences, and state information, external data sources include databases and APIs, and tools include calculators and code executors.

[0023] Specifically, in the embodiment, the 5G core network client in the step S10, that is, the 5G core network client in the communication connection between the 5G core network client and the MCP server, is a network data analyzer (Network Data Analytics Function, NWDAF) of the 5G core network, so as to dynamically discover the tools, resources and prompts related to network analysis provided by the MCP server through the established communication connection between the 5G core network client and the MCP server. The network data analyzer of the 5G core network has the function of 5G network data analysis, simplifies the generation and use mode of core network data, and generates insights and takes actions to enhance user experience. The MCP server can include a PCF (Policy Control Function), a SMF (Session Management Function) and an AMF (Access and Mobility Management Function) of the 5G core network, and the MCP is introduced into the 5G core network to take the network element of the 5G core network as the MCP server, so as to realize the cooperative work of the network element of the 5G core network and the AI model by using the MCP, and improve the adaptability.

[0024] Preferably, in the embodiment, the tools in the step S20, that is, the tools provided by the MCP server and injected into the AI model by the 5G core network client using the MCP, include a tool list, and the tool list includes tool information and dependency metadata between tools. The AI model can know the tools that can be called according to the tool list. The 5G core network client can construct a calling sequence of a directed acyclic graph (DAG) based on the dependency metadata between the tools, and execute the calling of the tools in the order of the calling sequence.

[0025] In combination with Figure 2 Preferably, in an embodiment, the step S10, that is, the establishment of the communication connection between the 5G core network client and the MCP server, includes:

[0026] S11: sending a token application request for accessing the MCP server to an authorization server by the 5G core network client; wherein the token application request includes a target resource declaration and a permission scope declaration;

[0027] S12: returning a token for authorized access to the MCP server to the 5G core network client by the authorization server;

[0028] S13: sending a token-carrying authentication request to the MCP server through the client of the 5G core network to establish a communication connection with the MCP server. The token is located in the request header of the authentication request.

[0029] Preferably, in the step S12, the token returned by the authorization server to the client of the 5G core network for authorized access to the MCP server includes a resource indicator and a dynamic permission declaration. The token is generated according to the target resource declaration and the permission scope declaration of the token application request. The resource indicator of the token corresponds to the target resource declaration of the token application request, and the dynamic permission declaration of the token corresponds to the permission scope declaration of the token application request. The resource indicator of the token is used to specify the requested MCP service, and the dynamic permission declaration of the token is used to declare the permission of the token, such as allowing only reading network data. The token can specify the MCP service requested by the client of the 5G core network and the corresponding permission, declare the target resource and the permission scope, and improve the security. The token adopts the JWT format.

[0030] Specifically, in the step S12, the token returned by the authorization server to the client of the 5G core network for authorized access to the MCP server further includes a one-time random number (Number once, Nonce). The one-time random number can be an AMF (Access and Mobility Management Function) ID. By adding the one-time random number in the token, the token can be prevented from being replayed after being hijacked.

[0031] Preferably, the type of the token can be a client credential or an authorization code.

[0032] Specifically, after the step S11, that is, after the client of the 5G core network sends a token application request for accessing the MCP server to the authorization server, the AI model calling method of the 5G core network further includes:

[0033] verifying the identity of the client of the 5G core network through the authorization server;

[0034] checking whether the target resource declaration and the permission scope declaration of the token application request are within the permission scope of the client of the 5G core network through the MCP server policy.

[0035] The authority scope of the client of the 5G core network can be formulated by the 3rd Generation Partnership Project (3GPP), which is a core formulating organization of global mobile communication technology standards. The MCP server includes a policy control function (PCF) of the 5G core network. When the target resource declaration and the authority scope declaration of the token application request are checked by the MCP server policy and are within the authority scope of the client of the 5G core network, step S12 is performed. When the target resource declaration and the authority scope declaration of the token application request are checked by the MCP server policy and are not within the authority scope of the client of the 5G core network, an error is returned.

[0036] Preferably, in the step S11, the authorization server in the token application request sent by the client of the 5G core network to the authorization server for accessing the MCP server is a network exposure function (NEF) server of the 5G core network. The network exposure function server of the 5G core network has a network exposure function and exposes network capabilities to third-party applications, realizes the docking of network capabilities and business needs, improves business experience, and optimizes network resource configuration. The network exposure function server of the 5G core network can use the OAuth2.0 open authorization framework to allow users to authorize third-party applications to access resources stored on service providers without directly providing a username and password, to realize authorization by generating a short-term valid token, and to ensure the security and controllability of user data.

[0037] Specifically, in some embodiments, before the step S12, that is, before the authorization server returns the token for accessing the MCP server to the client of the 5G core network, the AI model calling method of the 5G core network further includes:

[0038] Obtaining a real-time network load state through the MCP server, and sending an overload feedback to the authorization server when the real-time network load state exceeds a preset load threshold;

[0039] Returning an authority degradation suggestion to the client of the 5G core network according to the received overload feedback through the authorization server.

[0040] The MCP server includes a session manager (SMF) of the 5G core network, and can be used for being responsible for tunnel maintenance, IP address allocation and management, UP function selection, policy implementation, control in QoS (Quality of Service), charging data collection and roaming, etc. Real-time network load state is acquired through the MCP server, so as to adjust and suggest the authority of the token application request of the client of the 5G core network, so as to improve the reliability. The network load can be CPU utilization, and the preset load threshold can be that the CPU utilization is 70%. When the real-time CPU utilization exceeds 70%, an overload feedback is sent to the authorization server. When the authority range declaration of the token application request is data read-write authority, the authority downgrade suggestion can be that the authority range declaration is suggested to be downgraded to data read-only authority.

[0041] Specifically, after the authorization server returns the authority downgrade suggestion to the client of the 5G core network according to the received overload feedback, the AI model calling method of the 5G core network further includes:

[0042] The client of the 5G core network updates the token application request according to the received authority downgrade suggestion, and re-sends the updated token application request for accessing the MCP server to the authorization server.

[0043] Preferably, in other embodiments, the step S10, that is, the establishment of the communication connection between the client of the 5G core network and the MCP server, includes:

[0044] The client of the 5G core network sends a token-free authentication request to the MCP server;

[0045] The MCP server sends a token application request for accessing the MCP server to the authorization server according to the received token-free authentication request; wherein the token application request includes a target resource declaration and an authority range declaration;

[0046] The authorization server returns a token for authorized access to the MCP server to the client of the 5G core network;

[0047] The client of the 5G core network sends a token-bearing authentication request to the MCP server to establish a communication connection with the MCP server.

[0048] Preferably, in some embodiments, after the step S13, that is, after the client of the 5G core network sends a token-bearing authentication request to the MCP server to establish a communication connection with the MCP server, the AI model calling method of the 5G core network further includes:

[0049] sending, by the MCP server, a token validity verification request to the authorization server; wherein the MCP server can send the token validity verification request to a token introspection endpoint of the authorization server, and the token introspection endpoint allows a client to query the authorization server to verify an access token and retrieve metadata of the access token.

[0050] returning, by the authorization server, a token validity verification result to the MCP server;

[0051] when the token validity verification result is true, checking, by the MCP server, whether a uniform resource locator (URL) of the received authentication request with the token matches the resource indicator and the dynamic permission claim of the token.

[0052] wherein by verifying the validity of the token and checking the request and the dynamic permission claim of the token, it can be ensured that the token is only used for the target service, the token abuse is prevented, and the security is improved.

[0053] Preferably, after the authorization server returns the token validity verification result to the MCP server, the AI model calling method of the 5G core network further comprises:

[0054] when the token validity verification result is false, returning, by the MCP server, an error, recording the authentication request with the token as an illegal request, and synchronizing the illegal request to a service communication proxy (SCP) of the 5G core network for security analysis.

[0055] Specifically, after the MCP server checks whether the uniform resource locator of the received authentication request with the token matches the resource indicator and the dynamic permission claim of the token, the AI model calling method of the 5G core network further comprises:

[0056] when the uniform resource locator of the received authentication request with the token does not match the resource indicator and the dynamic permission claim of the token, returning an error, recording the authentication request with the token as an illegal request, and synchronizing the illegal request to a service communication proxy of the 5G core network for security analysis;

[0057] when the uniform resource locator of the received authentication request with the token matches the resource indicator and the dynamic permission claim of the token, returning an authentication pass feedback to a client of the 5G core network to complete authorization and release the request, so as to call the MCP server subsequently.

[0058] Referring to Figure 3 , Figure 3A framework schematic diagram of a 5G core network AI model calling system of the present application, the 5G core network AI model calling system can execute the 5G core network AI model calling method described above, the 5G core network AI model calling system includes a communication connection module 101, a network analysis module 102, a model context construction module 103, a KPI data injection module 104, an AI prediction module 105 and a tool calling module 106, the communication connection module 101 is used to establish the communication connection between the client of the 5G core network and the MCP server; the network analysis module 102 is used to obtain the tools, resources and prompt words provided by the MCP server through the MCP of the 5G core network client and inject them into the AI model; the model context construction module 103 is used to construct the model context according to the injected tools, resources and prompt words through the AI model, and return the construction feedback to the 5G core network client; the KPI data injection module 104 is used to inject the KPI data obtained by the 5G core network client into the model context of the AI model; the AI prediction module 105 is used to return the prediction result and the name and parameters of the required called tool to the 5G core network client according to the model context and KPI data through the AI model; the tool calling module 106 is used to call the corresponding tool according to the name and parameters of the required called tool through the MCP of the 5G core network client and execute the prediction result after the parameter is entered.

[0059] Among them, the 5G core network AI model calling system introduces MCP in the 5G core network, establishes the communication connection between the client of the 5G core network and the MCP server, MCP realizes the context persistence across sessions by encapsulating context data, ensures the logical coherence of AI model in multiple rounds of interaction, realizes the cooperative work of network elements of 5G core network and AI model through MCP, improves the adaptability between network elements of 5G core network and AI model, and provides unified interface specification through MCP, simplifies the integration process of AI model and external data source and / or tool, which is beneficial to solve the problem of deployment and operation, and improve the overall performance and service ability of 5G core network.

[0060] Preferably, in some embodiments, the communication connection module 101 is specifically used for:

[0061] sending a token application request for accessing the MCP server to the authorization server through the 5G core network client; wherein the token application request includes target resource declaration and permission scope declaration;

[0062] returning a token for authorized access to the MCP server to the 5G core network client through the authorization server;

[0063] sending a token-bearing authentication request to the MCP server by a client of the 5G core network to establish a communication connection with the MCP server.

[0064] Preferably, in some embodiments, the communication connection module 101 is further configured to:

[0065] verifying the identity of the client of the 5G core network by the authorization server;

[0066] checking whether the target resource declaration and the scope of authority declaration of the token application request are within the scope of authority of the client of the 5G core network by the MCP server policy.

[0067] Preferably, in the present embodiment, the communication connection module 101 is further configured to:

[0068] obtaining a real-time network load state by the MCP server, and sending an overload feedback to the authorization server when the real-time network load state exceeds a preset load threshold;

[0069] returning an authority downgrade suggestion to the client of the 5G core network by the authorization server according to the received overload feedback.

[0070] Preferably, in the present embodiment, the communication connection module 101 is further configured to:

[0071] updating the token application request by the client of the 5G core network according to the received authority downgrade suggestion, and re-sending the updated token application request to the authorization server to access the MCP server.

[0072] Preferably, in other embodiments, the communication connection module 101 is further configured to:

[0073] sending a token validity verification request to the authorization server by the MCP server;

[0074] returning a token validity verification result to the MCP server by the authorization server;

[0075] checking whether the uniform resource locator of the received token-bearing authentication request matches the resource indicator and dynamic authority declaration of the token by the MCP server when the token validity verification result is true.

[0076] Preferably, in other embodiments, the communication connection module 101 is further configured to:

[0077] returning an error when the uniform resource locator of the received token-bearing authentication request does not match the resource indicator and dynamic authority declaration of the token by the MCP server, recording the token-bearing authentication request as an illegal request, and synchronizing the illegal request to the service communication broker of the 5G core network for security analysis;

[0078] When the uniform resource locator of the received authentication request with the token is checked by the MCP server, the resource indicator of the token and the dynamic permission declaration match, and the authentication passes feedback is returned to the client of the 5G core network.

[0079] To sum up, the AI model calling method of the 5G core network of the present application introduces MCP in the 5G core network, establishes the communication connection between the client of the 5G core network and the MCP server, and realizes the cross-session context persistence through the encapsulation of context data by the MCP, ensures the logical coherence of the AI model in the multi-round interaction, realizes the collaborative work of the network element of the 5G core network and the AI model through the MCP, improves the adaptability between the network element of the 5G core network and the AI model, and provides a unified interface specification through the MCP, simplifies the integration process of the AI model and the external data source and / or tool. The AI model calling system of the 5G core network of the present application also has the above functions.

[0080] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of various equivalent modifications or replacements within the technical range disclosed by the present application, and these modifications or replacements should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1.An AI model calling method of a 5G core network, the method comprising: The method comprises the following steps: establishing a communication connection between a client of a 5G core network and a MCP server; using the MCP through the client of the 5G core network to obtain tools, resources and prompt words provided by the MCP server and inject them into an AI model; constructing a model context according to the injected tools, resources and prompt words through the AI model, and returning the constructed feedback to the client of the 5G core network; injecting KPI data into the model context obtained by the AI model construction through the client of the 5G core network; returning a prediction result and the name and parameters of a tool required to be called to the client of the 5G core network through the AI model according to the model context and the KPI data; calling the corresponding tool of the MCP server according to the name and parameters of the tool required to be called through the model context protocol by the client of the 5G core network, and executing the prediction result after the parameters are entered. 2.The AI model calling method of a 5G core network according to claim 1, wherein, The client of the 5G core network in the step of establishing a communication connection between the client of the 5G core network and the MCP server is a network data analyzer of the 5G core network. 3.The AI model calling method of a 5G core network according to claim 1, wherein, The step of establishing a communication connection between the client of the 5G core network and the MCP server comprises: sending a token application request for accessing the MCP server to an authorization server through the client of the 5G core network; wherein the token application request comprises a target resource declaration and a permission scope declaration; returning a token for authorized access to the MCP server to the client of the 5G core network through the authorization server; sending an authentication request with a token to the MCP server through the client of the 5G core network to establish a communication connection with the MCP server. 4.The AI model calling method of a 5G core network according to claim 3, characterized in that, After the step of sending a token application request for accessing the MCP server to the authorization server through the client of the 5G core network, the AI model calling method of the 5G core network further comprises: verifying the identity of the client of the 5G core network through the authorization server; checking whether the target resource declaration and the permission scope declaration of the token application request are within the permission scope of the client of the 5G core network through the MCP server policy. 5.The AI model invocation method of a 5G core network according to claim 3, characterized in that, Before the step of returning a token for authorized access to the MCP server to the client of the 5G core network through the authorization server, the AI model calling method of the 5G core network further comprises: obtaining a real-time network load state through the MCP server, and sending an overload feedback to the authorization server when the real-time network load state exceeds a preset load threshold; returning a permission downgrade suggestion to the client of the 5G core network through the authorization server according to the received overload feedback. 6.The AI model invocation method of a 5G core network according to claim 5, characterized in that, After the step of returning a permission downgrade suggestion to the client of the 5G core network through the authorization server according to the received overload feedback, the AI model calling method of the 5G core network further comprises: updating the token application request according to the received permission downgrade suggestion through the client of the 5G core network, and sending an updated token application request for accessing the MCP server to the authorization server again. 7.The AI model invocation method of a 5G core network according to claim 3, characterized in that, The token in the step of returning a token for authorized access to the MCP server to the client of the 5G core network through the authorization server comprises a one-time random number. 8.The AI model calling method of a 5G core network according to claim 3, characterized in that, After the 5G core network client sends a token-bearing authentication request to the MCP server to establish a communication connection with the MCP server, the AI model calling method of the 5G core network further comprises: sending a token validity verification request to the authorization server through the MCP server; returning the token validity verification result to the MCP server through the authorization server; when the token validity verification result is true, checking whether the uniform resource locator of the received token-bearing authentication request matches the resource indicator and dynamic permission declaration of the token through the MCP server. 9.The AI model invocation method of a 5G core network according to claim 8, characterized in that, After the 5G core network client sends a token-bearing authentication request to the MCP server to establish a communication connection with the MCP server, the AI model calling method of the 5G core network further comprises: when the uniform resource locator of the received token-bearing authentication request does not match the resource indicator and dynamic permission declaration of the token through the MCP server, return an error, record the token-bearing authentication request as an illegal request, and synchronize the illegal request to the service communication agent of the 5G core network for security analysis; when the uniform resource locator of the received token-bearing authentication request matches the resource indicator and dynamic permission declaration of the token through the MCP server, return an authentication pass feedback to the 5G core network client. 10.An AI model calling system of a 5G core network, characterized in that, comprises: a communication connection module for establishing a communication connection between the 5G core network client and the MCP server; a network analysis module for the 5G core network client to obtain tools, resources and prompt words provided by the MCP server through the MCP and inject them into the AI model; a model context construction module for the AI model to construct a model context according to the injected tools, resources and prompt words, and return a construction feedback to the 5G core network client; a KPI data injection module for the 5G core network client to inject KPI data into the model context obtained by the AI model; an AI prediction module for the AI model to return a prediction result and the name and parameters of the required called tool to the 5G core network client according to the model context and KPI data; a tool calling module for the 5G core network client to call the corresponding tool of the MCP server according to the name and parameters of the required called tool through the model context protocol, and execute the prediction result after the parameters are entered.

Images