Dynamic agent management method and system based on intelligent terminal AI expansion, terminal equipment and medium

Abstract

The invention discloses a dynamic agent management method and system based on intelligent terminal AI expansion, terminal equipment and a medium, and relates to the technical field of intelligent terminal AI agent management. The method comprises the following steps: receiving an AI expansion capability request sent by an intelligent terminal through a preset communication protocol, calling a preset AI large model, and reasoning based on related information of the intelligent terminal to obtain a reasoning result of an application scene; based on a reasoning result of the application scene, obtaining a demand specification of an AI agent through reasoning, and constructing a corresponding AI agent based on the demand specification; and receiving and responding to a service request initiated by a user and / or the intelligent terminal, and providing application capability of AI expansion for the intelligent terminal through the AI intelligent agent. According to the method, the AI expansion requirement of the intelligent terminal is dynamically adapted without presetting the intelligent agent, the adaptability of the intelligent agent and the terminal scene is improved, the service requirement is efficiently met, and the management capability of the intelligent control system on the AI expansion of the terminal is enhanced.

Description

Technical Field

[0001] This invention relates to the field of intelligent terminal AI agent management technology, and in particular to a dynamic agent management method, system, terminal device and medium based on intelligent terminal AI extension. Background Technology

[0002] With the widespread application of smart home devices, smart office terminals, and other smart terminals in daily life and work, users' demand for AI-enhanced services is increasing. However, most smart terminals only have basic control and data collection functions and lack the ability to provide complex scenario services driven by large AI models, making it difficult to meet personalized and intelligent needs.

[0003] In existing AI extension solutions for smart terminals, intelligent agents are mostly pre-defined and fixed types, unable to be dynamically adjusted according to the terminal's business capabilities and interaction characteristics, resulting in poor adaptability. Furthermore, when facing multi-terminal collaborative scenarios, it is difficult to integrate the comprehensive capabilities of each terminal to achieve integrated services, leading to low efficiency and incomplete scenario coverage of AI extension services, failing to realize the value of multi-terminal collaboration. In addition, existing solutions lack flexibility in responding to terminal AI extension needs, making it difficult to quickly match dynamically changing service requirements of terminals, thus hindering the effective implementation of AI extensions for smart terminals.

[0004] Therefore, there is an urgent need for an AI extension management method that can dynamically construct AI agents based on the characteristics of smart terminals and adapt to single or multiple smart terminal scenarios to fill the gaps in existing technologies. Summary of the Invention

[0005] The technical problem this invention aims to solve is that, in the field of AI agent management technology for smart terminals, existing technologies cannot provide users with the application capabilities of a wide variety of smart terminals under a unified large model capability. Therefore, an effective solution is urgently needed to address the aforementioned technical problem.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a dynamic intelligent agent management method based on AI extension of a smart terminal, applied to an intelligent control system, wherein the intelligent control system has a pre-set large AI model, and the method includes: Receive an AI extension capability request sent by a smart terminal through a preset communication protocol, wherein the AI ​​extension capability request carries relevant information of the smart terminal; The system invokes a pre-defined AI model and, based on the relevant information of the smart terminal, infers the reasoning results of the application scenarios. The reasoning results include the reasoning results of a single smart terminal application scenario and the reasoning results of multiple smart terminal application scenarios. Based on the reasoning results of the application scenario, the requirement specifications of the AI ​​agent are obtained, and the corresponding AI agent is constructed based on the requirement specifications. Receive and respond to service requests initiated by users and / or the smart terminal, and provide the smart terminal with AI-extended application capabilities through the AI ​​agent.

[0007] In one implementation, the relevant information of the smart terminal includes the smart terminal's attribute information, business capability information, and interaction capability information; The attribute information includes the device name and the device unique identifier; the service capability information includes the control instruction set, application dataset, and built-in service scenario information; and the interaction capability information is used to express whether the smart terminal has user interaction functions.

[0008] In one implementation, the invocation of a preset AI model, based on relevant information from the smart terminal, yields inference results for application scenarios. These inference results include inference results for a single smart terminal application scenario and inference results for multiple smart terminal application scenarios, including: By invoking a preset AI big model, and based on the application dataset, built-in business scenario information, and interaction capability information in the relevant information, the AI ​​application scenarios and corresponding service capabilities that a single smart terminal can provide are inferred and determined. Integrate the relevant information of all connected smart terminals to obtain comprehensive information of multiple terminals, and establish an association mapping of the capabilities of multiple terminals based on the comprehensive information of multiple terminals. The association mapping includes the complementary relationship of the business capabilities of different smart terminals and the data interaction logic. By invoking a preset AI big model and performing fusion reasoning on the comprehensive information of the multiple terminals based on the association mapping, the AI ​​application scenarios and corresponding collaborative service capabilities that can be provided by multiple smart terminals in collaboration are obtained.

[0009] In one implementation, the step of reasoning the requirement specifications of the AI ​​agent based on the reasoning results of the application scenario, and constructing the corresponding AI agent based on the requirement specifications, includes: The system invokes a pre-defined AI model and, based on the reasoning results of the application scenario, infers the required specifications of the AI ​​agent. The required specifications of the AI ​​agent include the functions of the AI ​​agent and the interaction methods with the user and / or the smart terminal. Based on the requirements and specifications of the AI ​​agent, AI agents adapted to a single smart terminal and AI agents adapted to multi-terminal collaboration are constructed respectively.

[0010] In one implementation, receiving and responding to service requests initiated by the user and / or the smart terminal, and providing the smart terminal with AI-extended application capabilities through the AI ​​agent, includes: If the smart terminal has interactive capabilities, then the service request initiated by the smart terminal is received, and the corresponding AI agent is invoked to respond to the service request. If the smart terminal does not have interactive capabilities, then based on the business needs of the smart terminal, the corresponding AI agent is invoked to complete the business request response and feedback to the smart terminal.

[0011] In one implementation, the method further includes: Based on the pre-set custom intelligent agent creation interface, it receives custom configuration information input by the user based on the business capabilities of the intelligent terminal; Based on the custom configuration information and the relevant information of the smart terminal, a user-defined AI agent is constructed, and the user-defined AI agent responds to the business requests initiated by the smart terminal.

[0012] In one implementation, the method further includes: Receive instructions sent by the user and invoke a large AI model to infer the intent of the instructions; If the instruction is a smart terminal control instruction, then the control instruction is issued to the corresponding smart terminal; If the instruction is an AI extension instruction, then the corresponding AI agent will be matched and a response will be made; If the business has the capability to respond to the AI ​​extension instructions, but there is no matching AI agent, the instruction data processed by the preset AI big model will be fed back to the user in a dialogue manner, and the user will be asked to confirm whether to create a new AI agent based on the instructions.

[0013] Secondly, embodiments of the present invention also provide a dynamic intelligent agent management system based on AI extensions of smart terminals, the system comprising: The AI ​​extension capability request acquisition module is used to receive AI extension capability requests sent by a smart terminal through a preset communication protocol, wherein the AI ​​extension capability request carries relevant information of the smart terminal. The application scenario reasoning module is used to call a preset AI large model and, based on the relevant information of the smart terminal, reason to obtain the reasoning result of the application scenario. The reasoning result includes the reasoning result of a single smart terminal application scenario and the reasoning result of multiple smart terminal application scenarios. The AI ​​agent construction module is used to infer the requirement specifications of the AI ​​agent based on the inference results of the application scenario, and to construct the corresponding AI agent based on the requirement specifications. The business request response module is used to receive and respond to business requests initiated by users and / or the smart terminal, and to provide the smart terminal with AI-extended application capabilities through the AI ​​agent.

[0014] Thirdly, embodiments of the present invention also provide a terminal device, the terminal device including a memory, a processor, and a dynamic intelligent agent management program based on intelligent terminal AI extension stored in the memory and executable on the processor. When the processor executes the dynamic intelligent agent management program based on intelligent terminal AI extension, it implements the steps of the dynamic intelligent agent management method based on intelligent terminal AI extension as described in any of the above schemes.

[0015] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a dynamic intelligent agent management program based on AI extensions of a smart terminal. When the dynamic intelligent agent management program based on AI extensions of a smart terminal is executed by a processor, it implements the steps of the dynamic intelligent agent management method based on AI extensions of a smart terminal as described in any of the above schemes.

[0016] Beneficial Effects: This invention discloses a dynamic intelligent agent management method, system, terminal device, and medium based on AI extension of intelligent terminals, relating to the field of intelligent terminal AI intelligent agent management technology. The method first receives an AI extension capability request sent by the intelligent terminal through a preset communication protocol, wherein the AI ​​extension capability request carries relevant information about the intelligent terminal. Then, a preset AI large model is invoked, and based on the relevant information of the intelligent terminal, inference results for application scenarios are obtained. These inference results include inference results for a single intelligent terminal application scenario and inference results for multiple intelligent terminal application scenarios. Next, based on the inference results of the application scenarios, the requirement specifications for AI intelligent agents are inferred, and corresponding AI intelligent agents are constructed based on these requirements. Finally, service requests initiated by users and / or the intelligent terminal are received and responded to, providing AI extension application capabilities to the intelligent terminal through the AI ​​intelligent agents. This invention does not require the preset of fixed AI intelligent agents; it can dynamically construct corresponding AI intelligent agents based on relevant information sent by the intelligent terminal through a preset communication protocol, and infer suitable application scenario results through an AI large model. This design adapts to scenarios involving single smart terminals or multi-terminal collaboration, improving the compatibility between the AI ​​agent and terminal requirements. Simultaneously, by efficiently responding to user and terminal business requests, it provides AI-enhanced application capabilities, addressing the issues of poor adaptability and low efficiency in existing solutions, and enhancing the management efficiency of the intelligent control system for terminal AI expansion. Attached Figure Description

[0017] Figure 1A flowchart illustrating a specific implementation of the dynamic intelligent agent management method based on AI extensions of a smart terminal provided in this invention.

[0018] Figure 2 This is a flowchart illustrating a specific implementation of the dynamic intelligent agent management method based on AI extension of a smart terminal provided in an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram of a dynamic intelligent agent management device based on AI extension of a smart terminal, provided in an embodiment of the present invention.

[0020] Figure 4 This is a block diagram illustrating the internal structure of the terminal device provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0022] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content, operations, or steps, nor does it require execution in the described order. For example, some operations or steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0023] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. For example, "first control information" and "second control information" are only used to distinguish different control information and do not limit their order.

[0025] Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or the order of execution, and that the words "first" and "second" do not necessarily imply that they are different.

[0026] It should also be understood that the term “and / or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0027] With the popular application of large AI models, the demand of intelligent terminals for the capabilities of large AI models is also growing stronger. It has become an irresistible trend for intelligent terminal devices to possess AI intelligent capabilities. There are numerous intelligent terminals in smart homes, and their intelligent capabilities vary widely. The application capabilities of each large model also differ significantly, and some do not even have the capabilities of large models. Moreover, the interaction methods between such terminals and users are also diverse. How to enable these various intelligent terminals to provide users with the application capabilities of large AI models under a unified large model capability is an urgent problem to be solved currently.

[0028] Therefore, the present invention proposes a comprehensive intelligent agent management method based on the AI expansion of intelligent terminal devices, and takes a commercial intelligent control system named AI Smart Home System as an example to demonstrate this method. As Figure 2 shown, this method specifically uses the Smart Screen AI Smart Home System as the carrier of the large AI model capabilities of intelligent terminals. Through the request for the AI expansion capabilities of intelligent terminals, it completes the inference of the AI intelligent agent requirements for the business of intelligent terminals, completes the construction of the AI intelligent agent system based on the business capabilities and application scenarios of intelligent terminals, and realizes the platformization of the application capabilities of large AI models for intelligent terminals and the consistency construction of AI capabilities.

[0029] Among them, the AI Smart Home System is a commercial intelligent control system that pre-sets large AI models, can connect various intelligent home appliances, and realizes device linkage through AI algorithms, such as automatically turning on the lights when opening the door and scene customization. At the same time, it can receive user instructions and overall coordinate the operation of home appliances. In addition, the AI Smart Home System also has the functions of receiving requests from intelligent terminals, inferring application scenarios, constructing AI intelligent agents, and providing AI expansion services. The user interaction entry name of the AI Smart Home System is the Smart Screen, which is a large-screen device with intelligent interaction capabilities. It can receive user-defined intelligent agent configuration information, display the results of instruction processing, and support user interaction with the system. At the same time, the Smart Screen can also support conventional functions such as integrating voice assistants, networking capabilities, and multi-device linkage functions. It can be used as a home control center and a multimedia playback terminal, and supports users to perform operations such as home appliance control and content on-demand through voice or touch. An intelligent terminal is a general term for intelligent devices with data collection, processing, or interaction capabilities, covering smart home devices such as mobile phones, smart watches, body fat scales, and smart bulbs. It can interact with other devices or platforms through the network. At the same time, an intelligent terminal can send an AI expansion request with its own information and rely on the system to obtain AI capabilities.

[0030] Furthermore, AI-enhanced capabilities refer to the functions added to devices beyond their original basic functions by integrating AI technology. For example, ordinary cameras can gain facial recognition and abnormal behavior detection capabilities through AI algorithms, or ordinary home appliances can gain voice control and habit prediction functions through AI modules. Smart terminals themselves do not provide complex AI services; specific AI services, such as health analysis and multi-terminal collaborative control, need to be provided by AI agents. AI agents are AI entities with autonomous perception, decision-making, and execution capabilities. They can autonomously plan and complete tasks based on environmental data or user needs, such as intelligent customer service robots and decision-making modules in autonomous driving systems. AI agents are built based on scenario-based reasoning results and are divided into two categories: adaptable to single terminals and multi-terminal collaboration, used to respond to business requests. Pre-set large-scale AI models refer to AI models with large parameter scales and massive training data, possessing cross-task processing capabilities. They can be applied to scenarios such as text generation, image understanding, voice interaction, and logical reasoning, as well as scenario reasoning, agent specification reasoning, and user command intent judgment; they are the core reasoning tool of the system.

[0031] This embodiment provides a dynamic intelligent agent management method based on AI extensions of smart terminals, which is applied to intelligent control systems, such as... Figure 1 As shown, the specific steps include the following: Step S100: Receive an AI extension capability request sent by a smart terminal through a preset communication protocol, wherein the AI ​​extension capability request carries relevant information of the smart terminal.

[0032] In this embodiment, receiving AI extension capability requests sent by smart terminals through a preset communication protocol is the initial step of this management method, used to collect complete and accurate basic information about the smart terminals. This embodiment uses a commercial smart control system called AI Smart Home System as an example. First, the communication protocol used by the smart terminal and the smart control system is defined. In this embodiment, the AIoT (Artificial Intelligence Internet of Things) protocol is specifically adopted. This protocol is not a general communication protocol, but rather an information interaction architecture specifically defined for AI extension scenarios of smart terminals. This protocol standardizes the communication format between the smart terminal and the smart control system, and also determines the information categories and data standards that the AI ​​extension capability request must include, avoiding information transmission chaos or data loss due to protocol inconsistencies. In practical applications, before sending a request, the smart terminal first establishes an encrypted communication connection with the smart control system through this protocol. The smart control system will authenticate the smart terminal's device identity, such as verifying whether the terminal's unique device identifier has been registered in the system. Only authenticated terminals can initiate AI extension capability requests, preventing unauthorized device access.

[0033] After a connection is established between the smart terminal and the smart control system, the smart control system receives the AI ​​extended capability request sent by the smart terminal. This request carries relevant information about the smart terminal. In practical applications, the relevant information carried by different types of smart terminals, such as body fat scales, smart refrigerators, and smart sockets, varies, but all include information on attributes, business capabilities, and interaction capabilities. For example, the specific data for the attributes of a body fat scale includes "Body Fat Scale - Device Name" and "SN202405001 - Unique Device Identifier"; the specific data for business capabilities includes "Weight and Body Fat Percentage Acquisition Command Set - Control Command Set", "User's Weight and Body Fat Data for the Past 30 Days - Application Dataset", and "Basic Body Fat Management - Built-in Business Scenario Information"; and the specific data for interaction capabilities is "Has Screen Display and Voice Broadcast Functions - Has User Interaction Functions". After being organized and packaged according to the format defined by the AIoT protocol, the smart terminal sends this information to the smart control system.

[0034] Furthermore, the smart terminal automatically filters out duplicate information already stored in the system. For example, the control command set of the body fat scale has already been uploaded during the initial connection, and subsequent requests will not carry it again, only supplementing updated dynamic information, such as the latest user body fat data. Through the filtering process, the amount of data transmitted is reduced, communication efficiency is improved, and redundant storage caused by the system receiving duplicate data is avoided.

[0035] After receiving a request for AI expansion capabilities, the intelligent control system first performs a preliminary completeness check on the relevant information in the request. In this embodiment, it checks whether the body fat scale is missing body fat percentage data from the "application dataset" or whether the smart light is not labeled with "interactive capability information." If missing information is found, the intelligent control system sends a request to the terminal to supplement the relevant information. If the information is complete, it temporarily stores the relevant information in the intelligent terminal information database of the intelligent control system. Through standardized protocol communication and information verification, the system ensures that the terminal information obtained by the intelligent control system is complete, avoiding issues caused by missing or incorrect information.

[0036] In one implementation, the relevant information of the smart terminal includes the smart terminal's attribute information, business capability information, and interaction capability information; wherein, the attribute information includes the device name and the device's unique identifier; the business capability information includes the control instruction set, application dataset, and built-in business scenario information; and the interaction capability information is used to express whether the smart terminal has user interaction functions.

[0037] In this embodiment, attribute information is used for the identification and management of smart terminals to prevent confusion when multiple terminals access the system. Attribute information may include device name, unique device identifier, etc. The device name is used by the intelligent control system to quickly identify the terminal type, such as a smart living room light or a body fat scale in the bedroom. This allows the system to directly determine the scenario and basic functions of the terminal when receiving a request through semantic recognition. The unique device identifier can use identifiers such as the device MAC address to locate the target terminal in multi-terminal collaborative scenarios. For example, if the system needs to send a data collection command to a specific body fat scale, the unique device identifier is used to locate the device, preventing the command from being sent incorrectly. In practical applications, the system binds and stores attribute information with the terminal's network address, forming a mapping relationship between terminal identity and network location. In subsequent steps of receiving requests or issuing commands, this mapping can be used to quickly locate the terminal, improving system processing efficiency.

[0038] Business capability information serves as the data source for subsequent scenario inference steps, determining the scalable AI service scope of the smart terminal. This information can include control instruction sets, application datasets, and built-in business scenario information. The control instruction set contains all operational commands the terminal can respond to, such as temperature adjustment commands or food sorting and storage commands for a smart refrigerator, or data acquisition commands or unit switching commands for a body fat scale. When building intelligent agents, the system designs the agent's control logic based on the control instruction set. For example, if the agent needs to automatically collect body fat data, it calls the data acquisition commands from the body fat scale's control instruction set. It's important to note that the terminal only uploads this information if the system hasn't stored its control instruction set. For instance, a newly connected smart light needs to upload instruction sets for on / off control and brightness adjustment. If a previously connected smart light requests the same instruction set again, the system automatically skips receiving that instruction set and only receives the updated color temperature adjustment command. This design reduces non-redundant data transmission and lowers system storage pressure. Application datasets are business data generated by the terminal during actual use, such as historical data on user weight and body fat percentage from body fat scales, and food storage lists and expiration dates from smart refrigerators. These data are used by AI models to perform scenario-based reasoning. Specifically, for example, the system can infer AI application scenarios such as health warnings and weight loss suggestions based on data from a body fat scale indicating a 5% increase in user body fat percentage over the past week, or it can infer food expiration reminder scenarios based on data from a smart refrigerator indicating that milk has only 2 days left to shelf life. Built-in business scenario information refers to the basic service scenarios preset at the time of terminal manufacturing, such as basic body fat analysis from body fat scales and baking mode selection from smart ovens. The system can then expand more complex AI services based on these basic scenarios. For example, in the basic body fat analysis scenario, it can combine user health goals to generate personalized weight loss plans, thus upgrading from basic functions to AI-enhanced functions.

[0039] Interactive capability information is used to support business response methods, enabling terminals with different interactive capabilities to adapt to user needs. For example, a smart refrigerator with screen and voice functions, upon receiving responses such as ingredients and recipes from an intelligent agent, can display recipe steps on the screen and provide cooking precautions via voice, eliminating the need for users to operate other devices. Conversely, for smart sockets marked as lacking user interaction capabilities, the response results must be fed back to the user through the smart control system's interactive interface on the smart screen, preventing users from being unable to access service status due to the terminal's lack of interactive capabilities. The system associates interactive capability information with subsequent business response modules, automatically matching the appropriate response method when receiving business requests. Specifically, for example, if a user initiates a power-on request to a smart socket, the system will invoke the corresponding intelligent agent to execute the command and display "Socket is powered on" feedback to the user in the smart screen's service card.

[0040] By clarifying the specific content and function of various types of information, the relevant information of intelligent terminals enables intelligent control systems to obtain information such as the identity, capabilities, and interaction methods of intelligent terminals, providing a data foundation for subsequent reasoning application scenarios and the construction of adapted intelligent agents.

[0041] Step S200: Call the preset AI big model, and based on the relevant information of the smart terminal, infer the reasoning results of the application scenario. The reasoning results include the reasoning results of a single smart terminal application scenario and the reasoning results of multiple smart terminal application scenarios.

[0042] In this embodiment, the pre-defined AI large model is a collaborative architecture between the edge and the cloud. The edge-side large model is deployed on the local hardware of the intelligent control system, responsible for handling inference tasks with high real-time requirements and small data volumes. This ensures fast response times and avoids latency caused by data uploads to the cloud when handling simple scenarios such as inference from a single smart terminal. The cloud-side large model is deployed on a remote server, has a larger parameter scale than the edge-side model, and is trained using a larger dataset. It is responsible for handling complex multi-terminal fusion inference tasks, such as collaborative scenario inference integrating the capabilities of multiple terminals, resulting in higher inference accuracy and the ability to handle complex logic. In the scenario of interpreting health data from a single body fat scale, the edge-side model completes inference within seconds, meeting the user's real-time needs. For personalized exercise plan scenarios combining body fat scales, treadmills, and smart bracelets, the cloud-side model combines data from all three sources and completes fusion inference within several seconds, ensuring the comprehensiveness of the inference results.

[0043] Specifically, the edge-side large model has billions of parameters and is deployed on terminal devices such as smart TVs and gateways, supporting local real-time inference. The cloud-based model has hundreds of billions of parameters and is used for inference in complex scenarios. For the cloud-based model, training data includes device perception data, user behavior data, and multimodal labeled data. Device perception data includes operation logs and environmental sensor data from tens of millions of smart terminals. This training dataset corresponds to the terminal attribute information and application datasets defined by the AIoT protocol. User behavior data is constructed based on user home interaction records from the intelligent control system, including voice commands, gesture operations, and scene mode triggers, used to train the model's ability to map natural language to device actions. Multimodal labeled data includes aligned labeled data that integrates visual, voice, and device status information, adapting to the multi-dimensional information transmission of the AIoT protocol. The training process employs a federated learning and supervised fine-tuning strategy. Data is preprocessed locally on the smart terminal side, and only feature vectors are uploaded to prevent privacy leaks. The cloud updates model parameters by aggregating gradients. Meanwhile, a mapping dataset of instructions and actions is built for high-frequency home commands such as device control and scene switching, and the attention layer of the model is fine-tuned to improve the inference accuracy of AIoT protocol data.

[0044] In this embodiment, application scenario inference is divided into two categories: single smart terminal application scenario inference and multi-smart terminal application scenario inference. Both types of inference rely on relevant information from the smart terminal, but their processing logic differs. For single smart terminal inference, the system extracts three types of data from the terminal's relevant information: application dataset, built-in business scenario information, and interaction capability information. These are then input into the AI ​​large model for information reconstruction and logical inference. Taking a body fat scale as an example, the system combines the user's application dataset (weight fluctuation within 2kg and body fat percentage stable at 22% over the past 30 days) with the built-in business scenario information of basic body fat management. Based on the information regarding screen and voice interaction capabilities, the system infers two AI application scenarios—real-time body fat analysis and voice broadcast, and weekly health report screen display—through the edge-side large model. Simultaneously, it determines the service capabilities corresponding to each scenario. For example, the service capabilities of the real-time body fat analysis scenario include automatically calculating body fat percentage levels after data collection and providing voice alerts for abnormal indicators. The purpose of this inference process is to explore the capabilities of a single terminal, avoiding the terminal only providing basic functions such as data collection, and further expanding intelligent services through AI, thereby enhancing the usability of the smart terminal.

[0045] For application scenarios involving multiple smart terminals, it's necessary to integrate the capabilities of these terminals and construct collaborative logic. The system first integrates relevant information from all connected terminals and builds a comprehensive multi-terminal information database. For example, it integrates health data from body fat scales, exercise modes and historical exercise data from treadmills, and heart rate and sleep data from smart bracelets, categorizing the data according to the complementarity of their business capabilities. Subsequently, the system establishes a multi-terminal capability association mapping based on this comprehensive information. This mapping clarifies the complementary relationships of business capabilities and data interaction logic between different terminals. For instance, data showing a high body fat percentage on a body fat scale is complementary to the fat-burning exercise mode on a treadmill; the data interaction logic involves the body fat scale transmitting its body fat percentage data to the treadmill in real time, and the treadmill adjusting the exercise intensity based on the data. Similarly, data showing a heart rate exceeding 120 beats per minute on a smart bracelet is complementary to the safety protection mode on a treadmill; the specific data interaction logic involves the bracelet transmitting its heart rate data to the treadmill, and the treadmill automatically reducing its speed and alerting the user when the heart rate exceeds the limit.

[0046] After establishing the association mapping, the system calls upon a cloud-based AI model to perform fusion reasoning on comprehensive information from multiple terminals. For example, by combining body fat data from a body fat scale, exercise data from a treadmill, and heart rate data from a fitness tracker, it infers a collaborative health and weight loss scenario. The service capabilities of this scenario include recommending treadmill exercise modes based on body fat data, real-time heart rate monitoring and intensity adjustment during exercise, and generating an exercise effect analysis report based on changes in body fat after exercise. The purpose of multi-terminal collaborative reasoning is to overcome the limitations of individual terminals. For example, a standalone treadmill can only provide exercise functions, and a standalone body fat scale can only provide data collection functions. Through collaborative reasoning, the combination of the two can provide AI-enhanced services such as data-guided exercise and exercise feedback data, improving the user experience.

[0047] By leveraging the reasoning capabilities of large AI models, static information from smart terminals is transformed into dynamic scenes, expanding the AI ​​service directions for individual terminals and enabling multi-terminal capability collaboration. This provides functional positioning for the subsequent construction of two types of intelligent agents: those adapted to single terminals and those adapted to multiple terminals.

[0048] In one implementation, the step of calling a preset AI large model and, based on relevant information of the smart terminal, infers the reasoning result of the application scenario. The reasoning result includes the reasoning result of a single smart terminal application scenario and the reasoning result of multiple smart terminal application scenarios. Specifically, this includes the following steps: Step S210: Call the preset AI big model, and based on the application dataset, built-in business scenario information and interaction capability information in the relevant information, infer and determine the AI ​​application scenarios and corresponding service capabilities that a single smart terminal can provide. Step S220: Integrate the relevant information of all connected smart terminals to obtain comprehensive information of multiple terminals, and establish a correlation mapping of the capabilities of multiple terminals based on the comprehensive information of multiple terminals. The correlation mapping includes the complementary relationship of the business capabilities of different smart terminals and the data interaction logic. Step S230: Call the preset AI big model, and perform fusion reasoning on the comprehensive information of the multiple terminals based on the association mapping to obtain the AI ​​application scenarios and corresponding collaborative service capabilities that multiple smart terminals can provide collaboratively.

[0049] In this embodiment, during the inference step for a single smart terminal application scenario, the system first cleans and filters the relevant information of the terminal. Specifically, invalid data is removed from the application dataset, and core scenarios are extracted from the built-in business scenario information. Subsequently, the system inputs the filtered information into the edge AI model, which reconstructs the information. For example, it organizes the scattered weight data from the body fat scale into a daily weight change curve for the user, and breaks down the built-in body fat management scenario into three sub-steps: data collection, data interpretation, and result feedback. Finally, scenario inference is performed based on the reconstructed information.

[0050] Taking a smart refrigerator as an example, the application dataset includes a food list, food expiration dates, and built-in business scenario information for food storage management. Interactive capabilities include touchscreen and voice functionality. Based on this information from the smart terminal, the edge-side model infers two AI application scenarios: one is a food expiration warning scenario, where the service capability is to check food expiration dates and display information for foods with expiration dates below a threshold on the screen with a voice reminder; the other is a food pairing recommendation scenario, where, when a user searches for egg recipes, the model recommends milk-based egg stew recipes based on the milk available in the refrigerator, displaying the steps on the screen and providing voice prompts. During this inference process, the model designs the scenario feedback methods in conjunction with the terminal's interactive capabilities, such as utilizing the refrigerator's screen and voice functions to ensure users can intuitively and conveniently access scenario services.

[0051] In the reasoning steps of multiple smart terminal application scenarios, the intelligent control system first integrates comprehensive information from multiple terminals, establishing a classification index according to terminal type and business capabilities. For example, it categorizes terminals into health monitoring terminals, sports terminals, and home control terminals, extracting the main business capabilities of each type for association mapping. Subsequently, it establishes a multi-terminal capability association mapping, constructing mapping relationships based on business objectives and capability matching logic. Taking health and weight loss as the business objective, the system first determines the capabilities required to achieve this objective, including data acquisition capabilities, exercise control capabilities, and physiological monitoring capabilities. Then, it matches terminals with corresponding capabilities, including body fat scales that can collect weight and body fat data, treadmills for exercise control, and smart bracelets that monitor physiological information such as heart rate and calorie consumption, determining the complementary relationship of their business capabilities and data interaction logic. The body fat scale's data guides the treadmill's exercise mode, and the smart bracelet's heart rate data ensures exercise safety. For example, the body fat scale transmits body fat data to the treadmill once per hour, and the smart bracelet transmits heart rate data to the treadmill once per second. The association mapping can change dynamically. If a new terminal is connected, the system will automatically update the terminal information and data interaction logic in the mapping to ensure the real-time nature of the mapping.

[0052] After determining the association mapping, multi-terminal fusion inference is performed. The intelligent control system inputs the association mapping and comprehensive information from multiple terminals into the cloud-based AI model, which then logically associates the information from multiple terminals. Specifically, for example, it associates the body fat percentage data from the body fat scale with the fat-burning exercise mode of the treadmill, and the heart rate threshold of the smart bracelet with the safety protection mechanism of the treadmill, inferring collaborative scenarios based on the association logic. Taking the health and fat-burning collaborative scenario as an example, the service capabilities inferred by the cloud model specifically include: when the user starts the treadmill, the body fat scale automatically transmits the current body fat data to the treadmill, and the treadmill recommends an initial exercise mode based on the data. During exercise, the smart bracelet transmits heart rate data to the treadmill in real time; if the heart rate exceeds the threshold, the treadmill automatically reduces its speed and provides a voice reminder that the heart rate is too high. After the exercise, the treadmill combines the exercise data with the post-exercise body fat data from the body fat scale to generate an exercise effect analysis report, which is displayed to the user through a smart screen.

[0053] By refining the steps of data processing, information reconstruction, and logical association in the reasoning process, we ensure that the scenario reasoning results not only conform to the actual capabilities of the terminal but also meet the actual needs of the user.

[0054] Step S300: Based on the reasoning results of the application scenario, the requirement specifications of the AI ​​agent are obtained, and the corresponding AI agent is constructed based on the requirement specifications.

[0055] In this embodiment, the inference AI agent requirement specifications are also implemented based on a preset large AI model inference, and its inference logic must correspond to the application scenarios obtained from the above inference. Optionally, the preset large AI model used for the inference AI agent requirement specifications can be the same large model as the large model for inferring the application scenarios. Alternatively, a separate large model architecture can be deployed specifically for the inference AI agent requirement specifications, which will not be elaborated further. The intelligent control system inputs the scenario inference results, including single-terminal scenarios and multi-terminal collaborative scenarios, into the preset large AI model to determine the service objectives for each scenario. For example, in the above-mentioned single-terminal scenario of food expiration warning, the service objective is to promptly remind users of expired food and reduce waste. In the above-mentioned multi-terminal collaborative scenario of healthy weight loss collaboration, the service objective is to provide safe and effective weight loss services based on multi-terminal data. Subsequently, the model infers the necessary intelligent terminal functional modules for the agent based on the service objectives and determines the interaction method in combination with the interactive capabilities of the intelligent terminal. The necessary functional modules and interaction methods together constitute the requirement specifications for the AI ​​agent.

[0056] Taking the aforementioned food expiration warning scenario as an example, the necessary functional modules of the required specifications derived by the model include: a data acquisition module for daily reading of food shelf-life data from the smart refrigerator; a warning judgment module for calculating and determining whether the food has entered the warning period threshold; and a result feedback module for sending the warning information to the refrigerator. The interaction method of the required specifications applies to the refrigerator's screen and voice capabilities, i.e., the screen displays the list of warning food items and the voice announces the warning content.

[0057] Taking the aforementioned smart agent in the health and weight loss collaborative scenario as an example, the necessary functional modules in the requirement specification derived by the model include: a data integration module for receiving real-time data from the body fat scale, treadmill, and wristband; a mode recommendation module for recommending treadmill modes based on body fat data; a safety control module for adjusting treadmill parameters based on heart rate data; and a report generation module for integrating pre- and post-exercise data to generate analysis reports. The interaction methods are adapted to multiple terminals, namely, voice broadcast of exercise suggestions from the treadmill, report display on the smart screen, and vibration alerts from the wristband when the heart rate exceeds the limit. During the requirement specification reasoning process, the model avoids functional redundancy; for example, it does not add a recipe recommendation function to the smart agent in the food expiration warning scenario, allowing the smart agent to focus on the core functions of the application scenario.

[0058] Subsequently, in the step of building the AI ​​agent based on the requirements specifications, the system calls a pre-defined agent component library. This library contains various reusable functional and interactive components, such as data interface components adapting to different terminal communication protocols, business logic components, and interactive adaptation components like screen display or voice broadcasting. Based on the required functional modules and interaction methods in the requirements specifications, suitable components are selected from the component library and combined.

[0059] For a single-terminal AI agent, such as the refrigerator food safety warning agent mentioned above, the system selects a data interface component to establish communication with the refrigerator, and performs food shelf-life collection and warning judgment through a business logic component that reads and judges data. Subsequently, it interacts with the user through a screen display component and a voice broadcast component. These components are assembled in the order of data collection, judgment, and feedback to form a complete agent. During assembly, the system configures the parameters of the components, such as setting the trigger time for the voice broadcast component and setting the display effects for the screen display component, to ensure that the agent's operating logic is consistent with the scenario requirements.

[0060] For AI agents that support multi-terminal collaboration, such as the aforementioned health and weight loss collaborative agent, the system selects a multi-protocol data interface component to communicate simultaneously with the body fat scale, treadmill, and wristband, and selects a multi-terminal data integration component to unify the data format across different terminals. Next, it selects a combination of business logic components, including a mode recommendation component, a security control component, and a report generation component. Then, it sets the triggering logic between components; for example, when heart rate data exceeds a threshold, the safety control component automatically adjusts the treadmill speed. Finally, it selects multi-terminal interaction components, including a treadmill voice component, a smart screen display component, and a wristband vibration component, to ensure that each terminal can provide service results in an adapted manner.

[0061] After the AI ​​agent is built, the system performs functional verification on it, checking whether the components communicate normally, whether the functions meet the requirements specifications, and whether the interaction methods are compatible with the terminal. If the verification passes, the AI ​​agent will be incorporated into the system's AI agent management center for unified management and can be called to respond to business requests subsequently. If the verification fails, the system returns to the requirements specification reasoning stage, re-checks the adaptability of the interface components, optimizes the requirements specifications, and rebuilds the agent until verification passes.

[0062] Through the steps of demand specification reasoning, modular construction, and functional verification, each AI agent can be accurately matched with application scenarios and terminal capabilities.

[0063] In one implementation, the step of reasoning the requirements of the AI ​​agent based on the inference results of the application scenario, and constructing the corresponding AI agent based on the requirements, specifically includes the following steps: Step S310: Call the preset AI big model, and based on the reasoning results of the application scenario, reason the requirements specifications of the AI ​​agent, wherein the requirements specifications of the AI ​​agent include the functions of the AI ​​agent and the interaction methods with the user and / or the smart terminal. Step S320: Based on the requirements and specifications of the AI ​​agent, construct an AI agent adapted to a single smart terminal and an AI agent adapted to multi-terminal collaboration.

[0064] In this embodiment, the requirement specification reasoning of the AI ​​agent needs to determine performance indicators and data security requirements to ensure the reliable operation of the agent. Performance indicators mainly include three categories: response time, data processing accuracy, and operational stability. Specifically, for a single terminal agent of the refrigerator food warning agent, the response time is relatively low, requiring only responses on an hourly or daily basis. However, for safety warning agents such as those for gas safety, the response time needs to be on a second or minute basis to ensure timely issuance of warnings. Similarly, the data processing accuracy requirement for a single terminal agent of the refrigerator food warning agent is not high; date calculation on a daily basis is sufficient. Operational stability requires continuous, fault-free operation for the refrigerator food warning agent to ensure service continuity. For multi-terminal collaborative agents, the response time requirement is that the time from receiving heart rate data to adjusting the treadmill speed should not exceed one second to avoid safety risks caused by untimely adjustments after the heart rate exceeds the limit. The data processing accuracy requirement is that the integration error between body fat data and exercise data should not exceed a preset threshold to ensure the accuracy of the exercise effect analysis report. Operational stability requires no data loss when multiple terminals communicate simultaneously to avoid interruption of collaborative services.

[0065] Data security requirements for sensitive user data processed by the intelligent agent necessitate encryption of data transmission, access control for data storage, and data anonymization. These security requirements are automatically generated by the AI ​​model during the inference of requirement specifications, incorporating industry data security standards to ensure user data security while the intelligent agent provides services.

[0066] In the process of building an AI agent based on requirements specifications, the components of the multi-terminal collaborative agent need to be adapted. The system establishes an adaptation list of components and terminal capabilities. For example, the AIoT data interface component adapts to body fat scales and smart refrigerators, the Bluetooth data interface component adapts to treadmills and smart bracelets, and the WiFi data interface component adapts to smart screens and smart lights. During construction, the corresponding component is selected from the list according to the terminal type to avoid interface incompatibility issues. Taking the health and weight loss collaborative agent as an example, the system selects the AIoT data interface component for the body fat scale, the Bluetooth interface component for the treadmill, and the WiFi interface component for the bracelet. At the same time, a protocol conversion component is added so that terminal data with different protocols can be uniformly converted into a format that the system can recognize. In this embodiment, JSON format is used to realize data interaction.

[0067] For business logic components, the system customizes component parameters according to the functional details of the requirements specifications. For example, the mode recommendation component recommends treadmill modes based on body fat data. The system configures body fat percentage and mode mapping parameters for this component. These parameters are generated by an AI model based on professional data in the field of sports and health, and the parameter settings refer to the safety standards of sports medicine to ensure user exercise safety.

[0068] The intelligent control system uses a combination of simulation and real-world testing for functional verification. In simulation testing, the system builds a virtual terminal environment to test whether the intelligent agent can correctly receive data, execute logic, and provide feedback. In real-world testing, the system can deploy the intelligent agent in a real terminal environment, invite users to participate in the testing, and collect user feedback. Verification metrics include functional correctness, response timeliness compliance, and data security compliance. Only when all metrics are met will the intelligent agent be incorporated into the business response system.

[0069] By clearly defining performance indicators and data security requirements, the reliability and security of the intelligent agent are improved. Component adaptation and parameter customization ensure that the agent's functionality matches the scenario requirements. Simulation and real-world testing guarantee that the intelligent agent can operate stably and efficiently in real-world environments.

[0070] Step S400: Receive and respond to service requests initiated by the user and / or the smart terminal, and provide the smart terminal with AI-extended application capabilities through the AI ​​agent.

[0071] In this embodiment, two types of request sources are received: those initiated by users and those initiated by terminals. User-initiated requests can be made through voice input, touch input, or remote input via a mobile app on a smart screen. Terminal-initiated requests are automatically triggered based on the terminal's own business needs; for example, a smart refrigerator might automatically issue a food expiration warning request at regular intervals, or a body fat scale might automatically issue a body fat data interpretation request after the user completes their measurement. The intelligent control system is equipped with a request receiving module that uniformly receives all types of requests and performs preliminary classification, including labeling the request type, associating it with a smart terminal, and specifying the request content, for subsequent intelligent agent matching.

[0072] Based on this, the system matches the corresponding intelligent agent from the intelligent agent management center according to the associated terminal and request content. For example, if a user initiates a request to analyze body fat data through a smart screen, with the associated terminal being a body fat scale and the request content being data interpretation, the system will match the body fat scale as a single-terminal intelligent agent. If a user initiates a request to activate a health and weight loss service, with the associated terminals being a body fat scale, treadmill, and fitness tracker, and the request content being multi-terminal collaborative service, the system will match a health and weight loss collaborative intelligent agent. If a smart refrigerator initiates a food expiration warning request, with the associated terminal being the refrigerator and the request content being a warning reminder, the system will match the refrigerator food expiration warning intelligent agent. If multiple intelligent agents are available for matching, the system will select one based on the request complexity and the online status of the terminals.

[0073] Subsequently, the system employs different service feedback methods based on the varying interactive capabilities of each terminal, ensuring that every request receives an appropriate response. For terminals with interactive capabilities, the agent's response results are directly fed back to the user through the terminal's own interactive functions. For example, after analyzing data, the body fat scale agent displays the results on the scale's screen and simultaneously provides a voice announcement; similarly, after issuing an expiration warning for food, the refrigerator agent displays a warning list on the refrigerator screen and provides a voice reminder. This direct feedback from the terminal eliminates the need for users to rely on other devices, enhancing service convenience and catering to the real-time needs of users near the terminal.

[0074] For terminals lacking interactive capabilities, such as smart sockets, smart light strips, and smart curtains, the response results to their service requests cannot be fed back by the terminals themselves. The system then relays the results to the smart screen, which informs the user through its interactive functions. For example, if a user initiates a power-on request for a smart socket through the smart screen, the system invokes the socket's intelligent agent to execute the command, and the smart screen's service card displays that the smart socket is powered on. If the user enables voice prompts, the smart screen will simultaneously announce that the socket is powered on. Similarly, if a smart light strip automatically turns on according to scene settings, the light strip's intelligent agent feeds back the result to the smart screen, allowing the user to view the status. If adjustments are needed, the user can directly issue commands through the smart screen, and the system will then invoke the light strip's intelligent agent to execute the adjustments. This feedback through the smart screen solves the problem of informing users of results for terminals without interactive capabilities, enabling users to promptly grasp the service execution status of the terminals and providing a convenient secondary control entry point.

[0075] In addition, the system incorporates request timeout handling and execution failure retry mechanisms during the response process. If the intelligent agent does not receive execution feedback from the terminal within a preset time after making a call, the system will automatically retry the call once. If there is still no feedback, the user will be notified of a request timeout. If the terminal fails to execute the instruction, the system will display "Instruction Execution Failure" on the smart screen.

[0076] By covering multiple types of request sources, matching intelligent agents, and adapting to terminal interaction capabilities, a comprehensive and accurate response to user and terminal business requests is achieved.

[0077] In one implementation, receiving and responding to service requests initiated by the user and / or the smart terminal, and providing the smart terminal with AI-extended application capabilities through the AI ​​agent, specifically includes the following steps: Step S410: If the smart terminal has interactive capabilities, then receive the service request initiated by the smart terminal and call the corresponding AI agent to respond to the service request; Step S420: If the smart terminal does not have interactive capabilities, then based on the business needs of the smart terminal, the corresponding AI agent is invoked to complete the business request response and feedback to the smart terminal.

[0078] In this embodiment, for smart terminals with interactive capabilities, the response results of services are fed back through their interactive capabilities. If the smart terminal has multiple interactive capabilities, a combination of multiple interactions is used for feedback. Taking a smart refrigerator as an example, when a user initiates a request for a recommended dinner recipe through the refrigerator's touchscreen, the system calls the refrigerator's intelligent agent. The intelligent agent filters suitable recipes based on the food data inside the refrigerator and the cloud-based recipe library, and then provides feedback through a multimodal approach. Specifically, the recipes are displayed on the refrigerator screen in order of recommendation level, and corresponding voice broadcasts are provided. In addition, the system can further guide the user to perform further operations. If the user clicks on a recipe, the screen displays the cooking process step by step, and voice reminders are given at key cooking steps. Through multimodal feedback, the system adapts to different user states. For example, when a user is busy preparing ingredients and cannot touch the screen, they can obtain the recipe steps by listening to the voice. If the user needs to confirm details, they can check the screen at any time. If the user has questions about the recipe, they can also ask follow-up questions through the refrigerator's voice function, and the intelligent agent will answer in real time, providing a service of request, response, and interaction. In addition, the intelligent agent will optimize the feedback method based on the user's usage habits. For example, if the user selects voice broadcast as the priority multiple times, the system will reduce the amount of text displayed on the screen by default, increase the detail of the voice broadcast, and improve the personalization of the user experience.

[0079] For smart terminals lacking interactive capabilities, the smart screen serves as a platform for feedback and an entry point for initiating control commands. Taking smart light strips as an example, users set specific lighting parameters on the smart screen's home scene page. At the designated trigger time, the system invokes the light strip's intelligent agent to execute the command and provides feedback on the execution result through the smart screen. If users want to further adjust the light strip after seeing changes in its lighting status, they can do so directly through the light strip service card on the smart screen.

[0080] In one implementation, the dynamic intelligent agent management method based on AI extension of smart terminals further includes the following steps: Step S510: Based on the pre-set custom intelligent agent creation interface, receive user-inputted custom configuration information based on the business capabilities of the intelligent terminal; Step S520: Based on the custom configuration information and the relevant information of the smart terminal, construct a user-defined AI agent, and respond to the service requests initiated by the smart terminal through the user-defined AI agent.

[0081] In this embodiment, the user-defined AI agent is implemented through a visual low-code custom interface, which is integrated into the operation interface of the intelligent control system. The interface's intelligent terminal selection area lists the connected intelligent terminals and core business capabilities. The user selects the required associated intelligent terminals and sets specific trigger conditions, such as time, events, and data. Subsequently, the user sets the terminal action sequence and parameters, and configures the result feedback method. After receiving the user's configuration information, the system converts it into a structured parameter table, associating it with the control command set and interaction capability information of the selected terminal. During construction, the system calls the intelligent agent component library, selects basic components suitable for the terminal, and connects them according to user logic, customizing parameters. The completed custom intelligent agent is included in the intelligent agent center, where the user can edit, enable, and disable it. When responding to business requests, the agent is automatically invoked if the trigger conditions are met, or manually invoked by the user. The intelligent agent controls the terminal according to the set logic, and the result is fed back through a preset method, adapting to the user's personalized scenario needs.

[0082] In one implementation, the dynamic intelligent agent management method based on AI extension of smart terminals further includes the following steps: Step S610: Receive the instruction sent by the user and call the AI ​​big model to infer the intent of the instruction; Step S620: If the instruction is a smart terminal control instruction, then issue a control instruction to the corresponding smart terminal; Step S630: If the instruction is an AI extension instruction, then match the corresponding AI agent and respond. Step S640: If the business capability to respond to the AI ​​extension command is available, but there is no matching AI agent, the command data processed by the preset AI big model will be fed back to the user in a dialogue manner, and the user will be confirmed whether to create a new AI agent based on the command.

[0083] In this embodiment, the large model identifies the specific intent of user commands by obtaining the user's desired purpose through keyword recognition and scene matching. For example, if the command contains keywords such as "open" or "adjust," it is determined to be a smart terminal control command. If the command contains AI-enhanced instructions such as "analyze" or "recommend," the intent is verified by combining the connected terminal and the smart agent scene. Specifically, for example, the command to turn on the bedroom air conditioner contains the keyword "open," and the system has a corresponding smart terminal, so it is determined to be a control command. If the command is to analyze body fat data, it contains the keyword "analysis," and there is a corresponding smart agent, so it is determined to be an AI-enhanced command.

[0084] During categorized responses, control commands are directly matched with terminal control commands, and results are fed back through input channels. AI-extended commands are matched with corresponding intelligent agents, and analysis results are fed back after invocation. If the business capability exists but there is no matching intelligent agent, the system first generates a temporary result using the AI ​​large model, feeds it back in a dialogue manner, and asks whether to create a new intelligent agent. If the user confirms, an intelligent agent is constructed based on the command and terminal information and incorporated into the intelligent agent center; otherwise, only a temporary result is fed back.

[0085] In summary, this invention discloses a dynamic intelligent agent management method based on AI extensions from smart terminals. It eliminates the need for pre-setting fixed AI agents and dynamically constructs corresponding AI agents based on relevant information sent by the smart terminal through a pre-defined communication protocol, using a large AI model to infer and adapt to the application scenario. This design adapts to both single and multi-terminal collaborative scenarios, improving the adaptability of AI agents to terminal requirements. Simultaneously, by efficiently responding to user and terminal business requests and providing AI extension application capabilities, it addresses the problems of poor adaptability and low efficiency in existing solutions, enhancing the management efficiency of intelligent control systems for terminal AI extensions.

[0086] like Figure 3 As shown in the figure, this embodiment of the invention provides a dynamic intelligent agent management system based on AI extension of a smart terminal. The system includes: an AI extension capability request acquisition module 10, an application scenario reasoning module 20, an AI intelligent agent construction module 30, and a business request response module 40.

[0087] Specifically, the AI ​​expansion capability request acquisition module 10 is used to receive an AI expansion capability request sent by a smart terminal through a preset communication protocol, wherein the AI ​​expansion capability request carries relevant information of the smart terminal; the application scenario reasoning module 20 is used to call a preset AI large model, and based on the relevant information of the smart terminal, to reason about the application scenario, wherein the reasoning result includes the reasoning result of a single smart terminal application scenario and the reasoning result of multiple smart terminal application scenarios; the AI ​​intelligent agent construction module 30 is used to reason about the requirement specification of the AI ​​intelligent agent based on the reasoning result of the application scenario, and to construct the corresponding AI intelligent agent based on the requirement specification; the business request response module 40 is used to receive and respond to business requests initiated by users and / or the smart terminal, and to provide the smart terminal with AI expansion application capabilities through the AI ​​intelligent agent.

[0088] Based on the above embodiments, the present invention also provides a terminal device, the principle block diagram of which can be as follows: Figure 4As shown, the terminal device includes a processor, memory, network interface, display screen, and temperature sensor connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a dynamic intelligent agent management method based on AI extensions of the intelligent terminal. The display screen can be an LCD screen or an e-ink screen. The temperature sensor is pre-installed inside the terminal device to detect the operating temperature of the internal components.

[0089] Those skilled in the art will understand that Figure 4 The schematic diagram shown is only a partial structural diagram related to the present invention and does not constitute a limitation on the terminal device to which the present invention is applied. The specific terminal device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0090] In one embodiment, a terminal device is provided, including a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors. The one or more programs include instructions for performing the following operations: Receive an AI extension capability request sent by a smart terminal through a preset communication protocol, wherein the AI ​​extension capability request carries relevant information of the smart terminal; The system invokes a pre-defined AI model and, based on the relevant information of the smart terminal, infers the reasoning results of the application scenarios. The reasoning results include the reasoning results of a single smart terminal application scenario and the reasoning results of multiple smart terminal application scenarios. Based on the reasoning results of the application scenario, the requirement specifications of the AI ​​agent are obtained, and the corresponding AI agent is constructed based on the requirement specifications. Receive and respond to service requests initiated by users and / or the smart terminal, and provide the smart terminal with AI-extended application capabilities through the AI ​​agent.

[0091] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0092] In summary, this invention discloses a dynamic intelligent agent management method, system, terminal device, and medium based on AI extensions of smart terminals, relating to the field of AI intelligent agent management technology for smart terminals. The method first receives an AI extension capability request sent by a smart terminal through a preset communication protocol, wherein the AI ​​extension capability request carries relevant information about the smart terminal. Then, a preset AI large model is invoked, and based on the relevant information of the smart terminal, inference results for application scenarios are obtained. These inference results include inference results for a single smart terminal application scenario and inference results for multiple smart terminal application scenarios. Next, based on the inference results of the application scenarios, the requirement specifications for AI intelligent agents are inferred, and corresponding AI intelligent agents are constructed based on these requirements. Finally, a service request initiated by a user and / or the smart terminal is received and responded to, providing the smart terminal with AI-extended application capabilities through the AI ​​intelligent agents. This invention does not require the preset of fixed AI intelligent agents; it can dynamically construct corresponding AI intelligent agents based on relevant information sent by the smart terminal through a preset communication protocol, inferring suitable application scenario results through an AI large model. This design adapts to scenarios involving single smart terminals or multi-terminal collaboration, improving the compatibility between the AI ​​agent and terminal requirements. Simultaneously, by efficiently responding to user and terminal business requests, it provides AI-enhanced application capabilities, addressing the issues of poor adaptability and low efficiency in existing solutions, and enhancing the management efficiency of the intelligent control system for terminal AI expansion.

[0093] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0094] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A dynamic agent management method based on AI expansion of a smart terminal, applied to a smart control system, characterized in that, The intelligent control system is provided with an AI large model, and the method comprises: receiving an AI extension capability request sent by the intelligent terminal through a preset communication protocol, wherein the AI extension capability request carries relevant information of the intelligent terminal; calling a preset AI large model, and reasoning an application scenario reasoning result based on the relevant information of the intelligent terminal, wherein the reasoning result comprises a single intelligent terminal application scenario reasoning result and a plurality of intelligent terminal application scenario reasoning results; based on the application scenario reasoning result, reasoning an AI agent demand specification, and constructing a corresponding AI agent based on the demand specification; receiving and responding to a service request initiated by a user and / or the intelligent terminal, and providing the intelligent terminal with AI-extended application capabilities through the AI agent. 2.The dynamic agent management method based on intelligent terminal AI expansion according to claim 1, characterized in that, The relevant information of the intelligent terminal comprises attribute information, service capability information and interaction capability information of the intelligent terminal; wherein the attribute information comprises a device name and a device unique identifier; the service capability information comprises a control instruction set, an application data set and self-provided business scenario information; and the interaction capability information is used to express whether the intelligent terminal has user interaction function. 3.The dynamic agent management method based on intelligent terminal AI expansion of claim 2, wherein, The calling of the preset AI large model, based on the relevant information of the intelligent terminal, to reason an application scenario reasoning result, wherein the reasoning result comprises a single intelligent terminal application scenario reasoning result and a plurality of intelligent terminal application scenario reasoning results, comprises: calling a preset AI large model, and reasoning and determining the AI application scenario and the corresponding service capability that can be provided by a single intelligent terminal based on the application data set, the self-provided business scenario information and the interaction capability information in the relevant information; integrating the relevant information of all the intelligent terminals that have been accessed to obtain comprehensive information of multiple terminals, and establishing an association mapping of multiple terminal capabilities based on the comprehensive information of the multiple terminals, wherein the association mapping comprises a business capability complementary relationship and a data interaction logic of different intelligent terminals; calling a preset AI large model, and performing fusion reasoning on the comprehensive information of the multiple terminals based on the association mapping to obtain AI application scenarios and corresponding collaborative service capabilities that can be provided by the multiple intelligent terminals in cooperation. 4.The dynamic agent management method based on intelligent terminal AI expansion of claim 3, wherein, The reasoning of an AI agent demand specification based on the application scenario reasoning result, and the construction of a corresponding AI agent based on the demand specification, comprises: calling a preset AI large model, and reasoning an AI agent demand specification based on the application scenario reasoning result, wherein the AI agent demand specification comprises the function of the AI agent and the interaction mode with the user and / or the intelligent terminal; based on the AI agent demand specification, constructing an AI agent adapted to a single intelligent terminal and an AI agent adapted to multiple terminals in cooperation, respectively. 5.The dynamic agent management method based on intelligent terminal AI expansion of claim 4, wherein, The receiving and responding to a service request initiated by a user and / or the intelligent terminal, and providing the intelligent terminal with AI-extended application capabilities through the AI agent, comprises: if the intelligent terminal has interaction capability, receiving the service request initiated by the intelligent terminal, and calling the corresponding AI agent to respond to the service request; If the intelligent terminal does not have interaction capability, based on the service demand of the intelligent terminal, a corresponding AI agent is called to complete the service request response and feedback to the intelligent terminal. 6.The dynamic agent management method based on intelligent terminal AI expansion of claim 1, wherein, The method further comprises: Based on the pre-set custom agent creation interface, the custom configuration information based on the service capability of the intelligent terminal is received; Based on the custom configuration information and the related information of the intelligent terminal, the AI agent customized by the user is constructed, and the service request initiated by the intelligent terminal is responded through the AI agent customized by the user. 7.The method of claim 1, wherein, The method further comprises: Receiving the instruction sent by the user, calling the AI large model to infer the intention of the instruction; If the instruction is an intelligent terminal control instruction, the control instruction is issued to the corresponding intelligent terminal; If the instruction is an AI expansion instruction, the corresponding AI agent is matched and responded; If it has the service capability to respond to the AI expansion instruction, but there is no matching AI agent, the instruction data processed by the pre-set AI large model is fed back to the user in the form of a dialogue, and it is confirmed whether the user creates a new AI agent based on the instruction.

8. A dynamic intelligent agent management system based on intelligent terminal AI expansion, characterized in that, The system comprises: An AI expansion capability request acquisition module for receiving an AI expansion capability request sent by an intelligent terminal through a pre-set communication protocol, wherein the AI expansion capability request carries the related information of the intelligent terminal; An application scenario inference module for calling a pre-set AI large model to infer an application scenario inference result based on the related information of the intelligent terminal, wherein the inference result includes a single intelligent terminal application scenario inference result and a plurality of intelligent terminal application scenario inference results; An AI agent construction module for inferring the demand specification of an AI agent based on the application scenario inference result, and constructing a corresponding AI agent based on the demand specification; A service request response module for receiving and responding to a service request initiated by a user and / or the intelligent terminal, and providing AI expansion application capability for the intelligent terminal through the AI agent.

9. A terminal device, comprising: The terminal device comprises a memory, a processor, and a smart terminal AI expansion based dynamic agent management program stored in the memory and executable on the processor. When the processor executes the smart terminal AI expansion based dynamic agent management program, the steps of the smart terminal AI expansion based dynamic agent management method according to any one of claims 1-7 are implemented.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a smart terminal AI expansion based dynamic agent management program, which is executed by the processor to implement the steps of the smart terminal AI expansion based dynamic agent management method according to any one of claims 1-7.

Images