Cloud mobile phone control system based on large language model
By using a cloud phone control system based on a large language model, intelligent control of cloud phones is achieved through natural language commands and the MCP protocol, solving the problem of low management efficiency in existing technologies and realizing efficient operation without programming.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUNAN XIAOSUAN TECH INFORMATION CO LTD
- Filing Date
- 2025-07-21
- Publication Date
- 2026-07-21
AI Technical Summary
Existing cloud phone control methods lack intelligence, resulting in low management efficiency. Users need to operate manually or write scripts, which is also unfriendly.
A cloud phone control system based on a large language model is adopted. By receiving natural language commands, the system analyzes and constructs tool call commands using a large language model intelligent agent, and interacts with the cloud phone control platform in conjunction with the MCP protocol to achieve intelligent control of the cloud phone.
Users can efficiently control cloud phones using natural language without any programming skills, simplifying the operation process and improving management efficiency.
Smart Images
Figure CN120658720B_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of interdisciplinary technology of artificial intelligence and cloud computing, and in particular relates to a cloud mobile phone control system based on a large language model. Background Technology
[0002] Currently, cloud phone control is mainly achieved through the following methods: 1. Application Programming Interface (API) control: Basic operations of the cloud phone, such as restart and reset, are controlled through API interfaces. However, these API operations usually rely on explicit instructions or commands provided by the user and lack the ability to interact with natural language; 2. Script automation: Automated operations are performed using scripting languages (such as Python or Shell). These methods usually require users to write and maintain code and are not user-friendly for new users; 3. Device management platform: Some companies provide cloud phone management platforms with the ability to control devices in batches, but most of these platforms are based on graphical interfaces or predefined commands and do not support intelligent interaction based on natural language.
[0003] As can be seen from the above, the control of cloud phones currently usually relies on manual operation or preset script tools, lacking intelligent control, resulting in low management efficiency of cloud phones. Summary of the Invention
[0004] This application provides a cloud phone control system based on a large language model, which can solve the problem of low management efficiency caused by the lack of intelligent control in cloud phones.
[0005] This application provides a cloud mobile phone control system based on a large language model, including:
[0006] An interactive intelligent agent is used to receive cloud phone control commands input by the user in natural language and send the cloud phone control commands to a large language model intelligent agent;
[0007] The large language model intelligent agent is used to analyze the cloud phone control commands after receiving them, construct tool invocation commands, and send the tool invocation commands to the MCP client; the tool invocation commands include the tool name and tool parameters;
[0008] The MCP client is used to initiate a remote procedure call request to the MCP server via the MCP protocol after receiving a tool invocation instruction; the remote procedure call request includes the tool name and tool parameters;
[0009] The MCP server is used to send task instructions containing the tool name and tool parameters to the cloud phone control platform after receiving a remote procedure call request.
[0010] The cloud phone control platform is used to determine the device information of the target cloud phone to be controlled by the task instruction based on the task instruction, and send the task instruction and device information to the edge server to which the target cloud phone belongs.
[0011] Edge servers are used to transmit task instructions and device information to the cloud phone management service node to which the target cloud phone belongs, so that the cloud phone management service node can perform the operation corresponding to the tool name on the target cloud phone.
[0012] Optionally, the MCP client is also used to perform the following steps:
[0013] Establish a connection with the MCP server through the MCP protocol, and obtain the list of metadata from the MCP server through the MCP protocol;
[0014] The metadata list is converted into a format recognizable by the large language model, and the converted metadata list is registered with the large language model agent.
[0015] Optionally, the metadata list includes multiple tool names, a function description for each tool name, and the tool parameters required to perform the operation corresponding to each tool name.
[0016] Optionally, the large language model agent is specifically used to perform the following steps:
[0017] By analyzing the cloud phone control commands and the list of registered metadata using a large language model, the tool names and tool parameters corresponding to the cloud phone control commands can be obtained.
[0018] The tool invocation instructions are constructed using a large language model. These instructions include the tool name and parameters corresponding to the cloud phone control instructions.
[0019] Optionally, the cloud phone control platform is specifically used to perform the following steps:
[0020] The device information corresponding to the target cloud phone is determined based on the tool parameters in the task instructions; the device information includes the edge server, cloud phone cluster, cloud phone management service node, instance IP address, and device running status;
[0021] If the device corresponding to the target cloud phone is in normal operating status, the device information and task instructions will be encapsulated and sent to the edge server to which the target cloud phone belongs.
[0022] If the device status corresponding to the target cloud phone is abnormal, the abnormal information will be returned to the user in sequence through the MCP server, MCP client, large language model agent and interactive agent; the abnormal information is used to indicate that the operation corresponding to the cloud phone control command failed to be executed.
[0023] Optionally, the edge server is specifically used to perform the following steps:
[0024] Based on the cloud phone cluster and cloud phone management service node corresponding to the target cloud phone, the packaged device information and task instructions are sent to the cloud phone management service node corresponding to the target cloud phone.
[0025] Optionally, the cloud phone management service node is specifically used to perform the following steps:
[0026] Based on the instance IP address of the target cloud phone, execute the operation corresponding to the tool name in the task instruction on the target cloud phone.
[0027] Optionally, the cloud phone management service node is also used to perform the following steps:
[0028] After executing the operation corresponding to the tool name in the task instruction, the execution result information is returned to the large language model agent in sequence through the edge server, cloud phone control platform, MCP server and MCP client. The large language model agent then converts the execution result information into natural language and returns it to the user through the interactive agent. The execution result information is used to indicate the execution status of the operation corresponding to the cloud phone control instruction.
[0029] Optionally, the edge server is also used to collect the operating status of each cloud phone management service node and each cloud phone in the cloud phone control system, and to report the collected operating status to the cloud phone control platform in real time.
[0030] Optionally, cloud phone control commands are sent by the human-computer interaction software to the interactive intelligent agent after receiving cloud phone control commands input by the user in natural language.
[0031] The above-mentioned solution in this application has the following beneficial effects:
[0032] In the embodiments of this application, cloud phone control commands input by the user in natural language are received. Then, a large language model agent analyzes these commands to identify the user's intent and key parameters. A tool call command conforming to computer language specifications is then constructed. Next, the MCP client initiates a remote procedure call request to the MCP server based on this tool call command via the MCP protocol. The MCP server then sends a task command to the cloud phone control platform based on the remote procedure call request. Finally, the cloud phone control platform transmits the task command to the cloud phone management service node via an edge server, enabling the cloud phone management service node to perform corresponding operations on the target cloud phone, thus completing the cloud phone control. Since users do not need complex programming skills and can achieve efficient control of the cloud phone using simple natural language commands, this cloud phone control system greatly simplifies the cloud phone control process and significantly improves cloud phone management efficiency.
[0033] Other beneficial effects of this application will be described in detail in the following detailed description section. Attached Figure Description
[0034] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0035] Figure 1 This is a schematic diagram of the structure of a cloud mobile phone control system based on a large language model, provided as an embodiment of this application. Detailed Implementation
[0036] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.
[0037] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.
[0038] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0039] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."
[0040] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0041] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.
[0042] To address the current problem of low management efficiency caused by the lack of intelligent control in cloud phones, this application provides a cloud phone control system based on a large language model. This control system receives cloud phone control commands input by the user in natural language, analyzes these commands using a large language model agent to identify the user's intent and key parameters, and constructs tool call commands conforming to computer language specifications. Then, the MCP client initiates a remote procedure call request to the MCP server based on these tool call commands via the MCP protocol. The MCP server then sends task commands to the cloud phone control platform based on the remote procedure call request. Finally, the cloud phone control platform transmits the task commands to the cloud phone management service node via an edge server, enabling the cloud phone management service node to perform corresponding operations on the target cloud phone, thus completing the cloud phone control. Since users do not need complex programming skills and can achieve efficient control of the cloud phone using simple natural language commands, this cloud phone control system greatly simplifies the control process and significantly improves cloud phone management efficiency.
[0043] The cloud mobile phone control system based on a large language model provided in this application will be illustrated below with reference to specific embodiments.
[0044] like Figure 1 As shown in the embodiments of this application, the cloud mobile phone control system based on a large language model includes an interactive intelligent agent, a large language model intelligent agent, an MCP client, an MCP server, a cloud mobile phone control platform, an edge server, and a cloud mobile phone management service node.
[0045] The aforementioned interactive intelligent agent is used to receive cloud phone control commands input by the user in natural language and send the cloud phone control commands to the large language model intelligent agent.
[0046] In some embodiments of this application, the aforementioned cloud phone control instructions are sent by the human-computer interaction software to the interactive intelligent agent after receiving cloud phone control instructions input by the user using natural language. These cloud phone control instructions are used to indicate the need for control operations on the target cloud phone, such as restarting or shutting down.
[0047] Specifically, users can use natural language to input cloud phone control commands into human-computer interaction software (such as the DingTalk application on the integration platform) (such as restarting cloud phone AC1231, where AC1231 is the cloud phone number, and cloud phone AC1231 is the aforementioned target cloud phone). After receiving the cloud phone control command, the human-computer interaction software triggers the preset web hook interface, and then the web hook calls the external interface (API) exposed by the interactive intelligent agent in the cloud phone control system to pass the cloud phone control command to the interactive intelligent agent.
[0048] The aforementioned large language model agent analyzes the cloud phone control commands upon receiving them, constructs tool invocation commands, and sends these commands to the MCP client. The tool invocation commands include the tool name and tool parameters.
[0049] In some embodiments of this application, upon receiving a cloud phone control command, the Large Language Model (LLM) agent analyzes and interprets the command to identify the user intent (e.g., restart, which corresponds to the tool name mentioned above; that is, the tool corresponding to the tool name is used to implement the user intent) and key parameters (i.e., the tool parameters mentioned above, such as the cloud phone ID being AC1231). In other words, the LLM can recognize that the user wants to control the target cloud phone.
[0050] In some embodiments of this application, after recognizing the user's intent and key parameters, the large language model agent determines whether the conditions for executing the user's intent are met (e.g., the "restart" tool exists and the parameters (i.e., the aforementioned key parameters) meet the requirements). If met, a tool invocation instruction conforming to computer language specifications (e.g., function call specifications) is constructed to instruct control operations on the target cloud phone. For example, the tool invocation instruction could be: {"name":"restart","arguments":{"device ID":"AC1231"}}. Here, name represents the tool name, restart represents restarting, arguments represent the tool parameters, device ID represents the device number (i.e., the cloud phone's ID), and AC1231 is the cloud phone's ID. It is understood that if the conditions are not met, an operation failure message will be returned to the human-computer interaction software through the interactive agent to inform the user that the current operation has failed.
[0051] The aforementioned Model Context Protocol (MCP) client is used to initiate a remote procedure call request to the MCP server via the MCP protocol after receiving a tool call instruction.
[0052] The aforementioned remote procedure call request includes a tool name and tool parameters to instruct control operations to be performed on the target cloud phone (the cloud phone corresponding to the tool parameters) (the operation performed by the tool corresponding to the tool name, such as restarting).
[0053] In some embodiments of this application, the MCP client will initiate a remote procedure call request to the MCP server by specifying the tool name in accordance with the specifications of the MCP protocol, so as to instruct the target cloud phone (the cloud phone corresponding to the tool parameter) to perform control operations (the operation performed by the tool corresponding to the tool name, such as restart).
[0054] The aforementioned MCP server is used to send task instructions containing the tool name and tool parameters to the cloud phone control platform after receiving a remote procedure call request.
[0055] The above task instructions are used to instruct control operations on the target cloud phone (the cloud phone corresponding to the tool parameters) (the operations performed by the tool corresponding to the tool name, such as restarting).
[0056] In some embodiments of this application, after receiving a remote procedure call request, the MCP server will call a specified interface of the cloud phone platform to pass the constructed task instructions to the cloud phone control platform.
[0057] It should be noted that, to facilitate the large language model agent in determining whether the conditions for executing user intent are met, the aforementioned MCP client is also used to perform the following steps in order to register a list of metadata for the large language model agent:
[0058] First, a connection is established with the MCP server via the MCP protocol, and a metadata list is obtained from the MCP server via the MCP protocol. Then, the metadata list is converted into a format recognizable by the large language model, and the converted metadata list is registered with the large language model agent. The metadata list includes multiple tool names, a functional description for each tool name (i.e., a description of the function of the tool corresponding to that tool name), and tool parameters required to execute the operation corresponding to each tool name (e.g., the tool parameter for the restart tool is the device ID).
[0059] After registering the metadata list, the large language model agent can specifically use the large language model to analyze the cloud phone control commands and the registered metadata list to obtain the tool name and tool parameters corresponding to the cloud phone control commands; then, it can use the large language model to construct tool invocation commands, which contain the tool name and tool parameters corresponding to the cloud phone control commands.
[0060] The aforementioned cloud phone control platform is used to determine the device information of the target cloud phone to be controlled by the task instruction based on the task instruction, and to send the task instruction and device information to the edge server to which the target cloud phone belongs.
[0061] In some embodiments of this application, the cloud phone control platform is specifically used to perform the following steps:
[0062] The device information corresponding to the target cloud phone is determined based on the tool parameters in the task instructions; the device information includes the edge server, cloud phone cluster, cloud phone management service node, instance IP address, and device running status;
[0063] If the device corresponding to the target cloud phone is in a normal operating state (including but not limited to online, idle, etc.), the device information and task instructions will be packaged and sent to the edge server to which the target cloud phone belongs (i.e. the edge server corresponding to the target cloud phone; it can be understood that the edge server has the ability to depackage the packaged device information and task instructions according to the pre-agreed packaging method).
[0064] If the target cloud phone's device is in an abnormal operating state (including but not limited to offline, faulty, etc.), no further instructions will be issued. Instead, the system will sequentially return exception information to the user through the MCP server, MCP client, large language model agent, and interactive agent. This exception information indicates that the operation corresponding to the cloud phone control instruction has failed (or the device is unavailable). In other words, the cloud phone control platform returns exception information to the interactive agent through a layer-by-layer feedback mechanism. The interactive agent then returns the exception information to the human-computer interaction software to inform the user that the operation has failed. This mechanism ensures state pre-checking before task distribution, improving system stability and task execution success rate.
[0065] The aforementioned edge server is used to transmit task instructions and device information to the cloud phone management service node to which the target cloud phone belongs, so that the cloud phone management service node can perform the operation corresponding to the tool name on the target cloud phone.
[0066] In some embodiments of this application, the edge server is specifically used to perform the following steps: Based on the cloud phone cluster and cloud phone management service node corresponding to the target cloud phone, the encapsulated device information and task instructions are sent to the cloud phone management service node corresponding to the target cloud phone. It is understood that a cloud phone management service node typically manages multiple cloud phones; therefore, it is necessary to first locate the cloud phone management service node to which the target cloud phone belongs based on the device information, and then send the device information and task instructions to that cloud phone management service node. It should be noted that the cloud phone management service node has the ability to decapsulate the encapsulated device information and task instructions according to a pre-agreed encapsulation method.
[0067] The aforementioned cloud phone management service nodes are deployed on ARM servers (ARM servers refer to servers using processors based on the ARM architecture, which is an architecture based on Reduced Instruction Set Computing (RISC)). When the edge server sends a request to the cloud phone management service node to which the target cloud phone belongs, it first locates the cloud phone cluster corresponding to the target cloud phone. Then, it sends task instructions and device information to the cloud phone management service node in the cloud phone cluster, so that the cloud phone management service node can accurately determine the target cloud phone based on the instance IP address corresponding to the target cloud phone. Then, it executes the operation corresponding to the tool name on the target cloud phone, that is, it executes the tool corresponding to the tool name on the target cloud phone (such as restarting).
[0068] The aforementioned edge servers are long-connection edge servers, maintaining a continuous bidirectional connection with the cloud phone control platform and the cloud phone management service nodes in the cloud phone cluster. They bear the dual responsibility of task relay and status synchronization. In addition to quickly and accurately routing tasks to the corresponding cloud phone management service nodes based on the target cloud phone cluster and instance location information provided by the cloud phone control platform during task distribution, and relaying task instructions to these nodes, the edge servers are also used to collect the operational status (including but not limited to online, offline, and abnormal operation status) of each cloud phone management service node and each cloud phone in the cloud phone control system. This collected operational status is reported to the cloud phone control platform in real time, allowing the platform to clearly understand the operational status of each cloud phone management service node and each cloud phone. This mechanism not only ensures reliable relay of control tasks but also provides fundamental support for the upper-layer platform's functions such as device scheduling, task fault tolerance, and status monitoring.
[0069] In some embodiments of this application, the aforementioned cloud phone management service node is specifically used to perform the following steps: based on the instance IP address corresponding to the target cloud phone, execute the operation corresponding to the tool name in the task instruction on the target cloud phone, that is, execute the tool corresponding to the tool name on the target cloud phone. It is understood that after executing the operation corresponding to the tool name in the task instruction, the cloud phone management service node will sequentially return execution result information to the large language model agent through the edge server, cloud phone control platform, MCP server, and MCP client, so that the large language model agent can convert the execution result information into natural language form and return it to the user through the interactive agent; this execution result information is used to indicate the execution status of the operation corresponding to the cloud phone control instruction.
[0070] After receiving the execution result information returned by the MCP client (which includes structured status codes / messages, or lower-level log fragments), the aforementioned large language model agent analyzes the received execution result information using the large language model to determine whether the operation on the cloud phone was successfully executed (if unsuccessful, it also understands the reason for failure), and generates an execution summary in natural language that is easy for users to understand (e.g., "Cloud phone AC1231 has been successfully restarted" or "Restarting cloud phone AC1231 failed, reason for failure: device is currently unavailable"). It is understood that the execution result information received by the large language model agent is in computer language format, while the execution result information sent to the interactive agent (i.e., the aforementioned execution summary) is in natural language format.
[0071] Understandably, after receiving the execution result information, the interactive intelligent agent will return the execution result information in natural language format to the human-computer interaction software to inform the user of the execution status of the current operation, and realize feedback to the user on the execution status of the current operation through the original path (or the notification channel set by the system) (for example, replying to the user who sent the instruction via DingTalk message).
[0072] It should be noted that, in order to ensure that the large language model accurately converts information in computer language format and information in natural language format, it can be trained with a large number of training samples (including information in computer language format and the corresponding information in natural language format) before use.
[0073] The cloud mobile phone control system based on a large language model of this application will be illustrated below with specific examples.
[0074] In this example, the test environment is as follows:
[0075] The large language model uses qwen-plus, the cloud phone control platform uses the cloud phone PAAS platform, the communication protocol uses the MCP protocol, and the toolset includes cloud phone restart, reset, etc.
[0076] Typical implementation cases are as follows:
[0077] User command (i.e. cloud phone control command): Send "Restart cloud phone AC1231" via DingTalk.
[0078] Instruction parsing:
[0079] The MCP client obtains tool metadata (including tool parameter rules: device ID cannot be empty) from the MCP server.
[0080] After LLM parses the instructions, it generates a structured call: {"name":"restart","arguments":{"device ID":"AC1231"}. Instruction execution: The MCP protocol encapsulates the parameters and calls the API of the cloud phone PaaS platform.
[0081] The execution result of the cloud phone PaaS platform is: {"code":"200","msg":"Operation successful"}; "code":"200" represents the code indicating successful execution, and msg represents the Chinese description. Closed-loop feedback: LLM converts the result into a natural language response: "Cloud phone AC1231 restart operation executed successfully".
[0082] Furthermore, this application tested 200 commands containing ambiguous expressions to control the cloud phone, finding a command parsing accuracy rate of 98.5%, an end-to-end latency of an average of 3–10 seconds (from command input to user feedback), and excellent anomaly handling capabilities: returning "Device ID cannot be empty" (e.g., AC1231) when the device is empty, and a 100% illegal command interception rate (e.g., "Delete system file" returning "No operation permission"). These test data demonstrate that the MCP protocol effectively bridges the LLM intelligent layer and the cloud phone execution layer (100% correct command routing), the LLM dual parsing mechanism (command understanding + result translation) significantly lowers the barrier to entry, and dynamic toolset management supports seamless business expansion (new tools take effect immediately).
[0083] In summary, the cloud mobile phone control system based on a large language model provided in this application has the following advantages:
[0084] I. Natural Language Control: Compared to existing API-based control or script-based automation methods, this application utilizes the natural language understanding capabilities of Large Language Models (LLM) to parse user-input natural language commands into executable cloud phone control operations. These operations interact with the cloud phone control backend via a standardized MCP protocol, achieving an intelligent closed-loop control process. Maintenance personnel do not need complex programming skills; they can efficiently operate the cloud phone using simple natural language commands, greatly simplifying the operation process.
[0085] II. Intelligent Parsing and High-Efficiency Automation: Existing technologies largely rely on predefined commands or codes, lacking flexible intelligent parsing capabilities. This application utilizes the Function Calling capability of a large language model to accurately parse the user's natural language input into structured control instructions. The system can intelligently understand based on context and language, avoiding errors that may occur when manually writing scripts, and improving the accuracy and automation of operations.
[0086] III. Operation Feedback and Intelligent Adjustment Mechanism: Compared to existing technologies that simply record or ignore operation feedback, this application achieves real-time transmission and intelligent adjustment of cloud phone operation results through a feedback mechanism. After an operation, the system will determine whether the operation was successful based on the feedback, and automatically execute subsequent operations or provide prompts when necessary. This automated feedback process greatly improves the system's reliability and self-correction capabilities.
[0087] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.
Claims
1. A cloud mobile phone control system based on a large language model, characterized in that, include: An interactive intelligent agent is used to receive cloud phone control commands input by the user in natural language and send the cloud phone control commands to a large language model intelligent agent; The large language model agent is used to analyze the cloud phone control command after receiving it, construct a tool invocation command, and send the tool invocation command to the MCP client; the tool invocation command includes the tool name and tool parameters; The MCP client is used to initiate a remote procedure call request to the MCP server via the MCP protocol after receiving the tool call instruction. The remote procedure call request includes the tool name and the tool parameters; The MCP server is used to send a task instruction containing the tool name and the tool parameters to the cloud phone control platform after receiving the remote procedure call request. The cloud phone control platform is used to determine the device information of the target cloud phone to be controlled by the cloud phone control instruction according to the task instruction, and send the task instruction and the device information to the edge server to which the target cloud phone belongs; The edge server is used to transmit the task instruction and the device information to the cloud phone management service node to which the target cloud phone belongs, so that the cloud phone management service node can perform the operation corresponding to the tool name on the target cloud phone.
2. The cloud phone control system according to claim 1, characterized in that, The MCP client is also used to perform the following steps: A connection is established with the MCP server via the MCP protocol, and a list of metadata is obtained from the MCP server via the MCP protocol. The metadata list is converted into a format recognizable by the large language model, and the converted metadata list is registered with the large language model agent.
3. The cloud phone control system according to claim 2, characterized in that, The metadata list includes multiple tool names, a function description for each tool name, and tool parameters required to perform the operation corresponding to each tool name.
4. The cloud phone control system according to claim 3, characterized in that, The large language model agent is specifically used to perform the following steps: The cloud phone control commands and the registered metadata list are analyzed using a large language model to obtain the tool names and tool parameters corresponding to the cloud phone control commands. The tool invocation instruction is constructed using the large language model, and the tool invocation instruction includes the tool name and tool parameters corresponding to the cloud phone control instruction.
5. The cloud phone control system according to claim 1, characterized in that, The cloud-based mobile phone control platform is specifically used to perform the following steps: The device information corresponding to the target cloud phone is determined based on the tool parameters in the task instruction; the device information includes the edge server, cloud phone cluster, cloud phone management service node, instance IP address, and device operating status; If the device corresponding to the target cloud phone is in a normal operating state, the device information and the task instruction are encapsulated and sent to the edge server to which the target cloud phone belongs. If the device operating status corresponding to the target cloud phone is abnormal, then abnormal information is returned to the user in sequence through the MCP server, the MCP client, the large language model agent, and the interactive agent; the abnormal information is used to indicate that the operation corresponding to the cloud phone control command failed to execute.
6. The cloud phone control system according to claim 5, characterized in that, The edge server is specifically used to perform the following steps: Based on the cloud phone cluster and cloud phone management service node corresponding to the target cloud phone, the encapsulated device information and task instructions are sent to the cloud phone management service node corresponding to the target cloud phone.
7. The cloud phone control system according to claim 6, characterized in that, The cloud phone management service node is specifically used to perform the following steps: Based on the instance IP address corresponding to the target cloud phone, perform the operation corresponding to the tool name in the task instruction on the target cloud phone.
8. The cloud phone control system according to claim 7, characterized in that, The cloud phone management service node is also used to perform the following steps: After executing the operation corresponding to the tool name in the task instruction, the execution result information is returned to the large language model agent in sequence through the edge server, the cloud phone control platform, the MCP server, and the MCP client, so that the large language model agent can convert the execution result information into natural language form and return it to the user through the interactive agent; the execution result information is used to indicate the execution status of the operation corresponding to the cloud phone control instruction.
9. The cloud phone control system according to claim 1, characterized in that, The edge server is also used to collect the operating status of each cloud phone management service node and each cloud phone in the cloud phone control system, and to report the collected operating status to the cloud phone control platform in real time.
10. The cloud phone control system according to claim 1, characterized in that, The cloud phone control commands are sent by the human-computer interaction software to the interactive intelligent agent after receiving cloud phone control commands input by the user in natural language.