Methods and related products for implementing AI Agents to call RPA services based on AI and LLM.

Abstract

This disclosure provides a method and related products for implementing AI Agent invoking RPA services based on AI and LLM. The method includes: receiving a semantic instruction; determining at least one automation bearer unit corresponding to the semantic instruction, each automation bearer unit carrying at least one automation service; obtaining a list of automation tools corresponding to each automation bearer unit, each automation tool list including descriptive information of each automation service carried by the automation bearer unit; determining at least one target automation service from each automation tool list whose descriptive information matches the semantic instruction; and invoking the corresponding target automation service from each automation bearer unit. This implementation allows users to directly invoke the corresponding target automation service through semantic instructions.

Description

Technical Field

[0001] This application relates to the technical field of service automation, and more particularly to a method and related products for implementing AIAgent to call RPA services based on AI and LLM. Background Technology

[0002] Robotic Process Automation (RPA) uses specific "robot software" to simulate human operations on a computer and automatically execute process tasks according to rules.

[0003] Artificial intelligence (AI) is a technical science that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence.

[0004] Large Language Models (LLMs) are models trained on massive amounts of text that can recognize human language, perform language-related tasks, and have a large number of parameters.

[0005] Artificial Intelligence Agents (AI Agents) are capable of perceiving their environment, making decisions, and executing actions. Unlike traditional artificial intelligence, they possess the ability to think and act independently, and can utilize tools to achieve given goals. AI Agents are based on Large Language Models (LLMs) as their core computing engine, enabling them to engage in dialogue, perform tasks, reason, and exhibit a degree of autonomy. They possess the ability to autonomously understand, perceive, plan, remember, and use tools, and can automate complex tasks. Specifically, LLM-driven AI Agents, composed of various AI capabilities, can interact with employees using natural language, understand employee instructions and needs, and provide feedback and responses; they can acquire domain-specific knowledge relevant to the business to complete complex professional tasks; they can break down complex tasks into several executable tasks and use data and tools to complete them; they can also collaborate with employees, and AI Agents can collaborate with each other to complete complex tasks, enabling digital employees to leap from automation to intelligence, helping employees complete their work more efficiently, and fully realizing human-machine collaboration.

[0006] In relevant RPA service scenarios, such as when an enterprise's business system uses RPA services to automate the control of various desktop applications, it is usually necessary to trigger the RPA service based on fixed conditions or manually trigger the RPA service and intelligently execute pre-designed fixed processes. However, it lacks the ability to flexibly understand user intent, that is, it cannot identify user intent and actively trigger one or more RPA services to realize that intent based on the identified information.

[0007] In some cases, enterprises have accumulated a large number of different RPA services. However, since each RPA service needs to be triggered by fixed conditions or manually, even with an AI Agent, there is no unified and routable interaction path between the AI ​​Agent and the various RPA services. Therefore, it is impossible to realize the intelligent driving of one or more RPA implementations to automate UI (User Interface) services according to the user's intentions.

[0008] Specifically, in the relevant methods of calling RPA through AI Agent, each RPA can usually only be modified one by one to give each RPA a corresponding Application Programming Interface (API), so that the AI ​​Agent can call RPA through API, but cannot call RPA directly.

[0009] Furthermore, such calls can only invoke a single RPA service one-to-one. In other words, each RPA service still runs independently, and the AI ​​Agent is isolated from each RPA service. The AI ​​Agent cannot arbitrarily invoke the corresponding RPA service according to the user's intention. Moreover, once the process changes, the API layer also needs to be adjusted. RPA assets are not uniformly abstracted into "tool capabilities", resulting in low reuse efficiency.

[0010] Furthermore, the agent cannot obtain a complete list of RPA services for each desktop. Therefore, by only triggering a few fixed APIs, it is difficult to dynamically select multiple RPA services and combine them into an end-to-end solution based on the current user intent.

[0011] On the other hand, in environments with multiple desktops, the RPA service usually solves the scheduling problem internally, and the agent often cannot precisely control the desktops executing the RPA service. Summary of the Invention

[0012] This application provides a method and related products for implementing AI Agent calling RPA services based on AI and LLM, in order to solve the problems existing in related technologies. The technical solution is as follows: In a first aspect, embodiments of this application provide a method for implementing AI Agent invoking RPA services based on AI and LLM, including: Receive semantic instructions; Identify at least one automated bearer unit corresponding to the semantic instruction, and each automated bearer unit carries at least one automated service; Obtain the list of automation tools corresponding to each automation bearer unit. Each automation tool list includes at least one description of a Model Control Protocol (MCP). Each description is obtained by encapsulating the service information of the corresponding automation service in the form of an MCP tool. Identify at least one target automation service from the list of automation tools whose description information matches the semantic instructions; The corresponding target automation service is invoked from each automation carrier unit.

[0013] In some alternative implementations, the semantic instruction includes at least one unit identifier; Accordingly, semantic instructions are received, including: Utilize an AI Agent to receive semantic instructions; Determining at least one automated bearer unit corresponding to a semantic instruction includes: The large language model (LLM) in the AI ​​Agent is used to identify at least one unit identifier in the semantic information, and the corresponding automated carrier unit for each unit identifier.

[0014] In some optional implementations, before obtaining the list of automation tools mapped to each automation bearer unit, the method further includes: Register the description information of each automated bearer unit and the corresponding preset unit identifier to the MCP route.

[0015] In some optional implementations, a list of automation tools mapped to each automation bearer unit is obtained, including: The AI ​​Agent is used to send the unit identifiers in the semantic instructions to the MCP route; The MCP routing is used to obtain the description information of the automated bearer unit corresponding to each unit identifier in the semantic instructions, and then feeds it back to the AI ​​Agent.

[0016] In some alternative implementations, semantic instructions include task information; Accordingly, at least one target automation service whose description information conforms to semantic instructions is identified from the list of automation tools, including: Using an AI Agent, identify at least one target description that matches the task information from the description information of the automated carrier unit corresponding to each unit identifier in the semantic instructions; Determine the target automation service corresponding to the description information of each target.

[0017] In some optional implementations, the corresponding target automation service is invoked from each automation bearer unit, including: The AI ​​Agent sends a call request to the MCP route, and the call request points to the automated services for each target. Use MCP routing to invoke automated services for each target.

[0018] The call request includes the unit identifier corresponding to each target automation service; Accordingly, the MCP routing is used to invoke various target automation services, including: In response to receiving a call request from the MCP route, the corresponding target automation service is invoked from the automation bearer unit corresponding to each unit identifier in the call request.

[0019] In some optional implementations, after invoking the corresponding target automation service from each automation bearer unit, the method further includes: Execute the pre-set automated processes in the automated services for each objective; The execution results of each automated process are fed back to the AI ​​Agent.

[0020] Secondly, embodiments of this application provide an apparatus for implementing AIAgent to call RPA services based on AI and LLM, comprising:

[0021] The receiving module is configured to receive semantic instructions; The unit determination module is configured to determine at least one automated bearer unit corresponding to the semantic instruction, and each automated bearer unit carries at least one automated service; The list acquisition module is configured to acquire the list of automation tools mapped to each automation carrier unit. Each automation tool list includes at least one description of a Model Control Protocol (MCP). Each description is obtained by encapsulating the service information of the corresponding automation service in the form of an MCP tool. The service module is configured to identify at least one target automation service from a list of automation tools whose description information matches the semantic instructions. The calling module is configured to call the corresponding target automation service from each automation bearer unit.

[0022] In some alternative implementations, the semantic instruction includes at least one unit identifier;

[0023] Accordingly, the receiving module is also specifically configured as follows:

[0024] Utilize an AI Agent to receive semantic instructions; Accordingly, the unit determination module is also specifically configured as follows: The large language model (LLM) in the AI ​​Agent is used to identify at least one unit identifier in the semantic information, and the corresponding automated carrier unit for each unit identifier.

[0025] In some optional implementations, the list retrieval module is further configured as follows:

[0026] The AI ​​Agent is used to send the unit identifiers in the semantic instructions to the MCP route; The MCP routing is used to obtain the description information of the automated bearer unit corresponding to each unit identifier in the semantic instructions, and then feeds it back to the AI ​​Agent.

[0027] In some alternative implementations, the apparatus for implementing AI Agent invocation of RPA services based on AI and LLM also includes a registration module configured to: Before obtaining the list of automation tools mapped to each automation carrier unit, the service information of each automation service in each automation carrier unit is encapsulated into the description information of the corresponding Model Control Protocol (MCP) tool. Register the description information of each automated bearer unit and the corresponding preset unit identifier to the MCP route.

[0028] In some alternative implementations, semantic instructions include task information; Accordingly, the service module is also specifically configured as follows: The AI ​​Agent sends a call request to the MCP route, and the call request points to the automated services for each target. Use MCP routing to invoke automated services for each target.

[0029] This includes using MCP routing to invoke automated services for various objectives, including: In response to receiving a call request from the MCP route, the corresponding target automation service is invoked from the automation bearer unit corresponding to each unit identifier in the call request.

[0030] In some alternative implementations, the apparatus for implementing AI Agent invocation of RPA services based on AI and LLM further includes an execution module configured to: Execute the pre-set automated processes in the automated services for each objective; The execution results of each automated process are fed back to the AI ​​Agent.

[0031] In some optional implementations, the registration module is further configured as follows: Establish remote control protocols between the AI ​​Agent and each automated carrier unit; Register each remote control protocol to the MCP route; In some alternative implementations, the execution module is further configured as follows: AI Agent remotely controls the execution of target automated services through remote control protocols.

[0032] Thirdly, embodiments of this application provide a device for implementing AI Agent invoking RPA services based on AI and LLM. The device includes a memory and a processor. The memory and the processor communicate with each other via an internal connection. The memory stores instructions, and the processor executes the instructions stored in the memory. When the processor executes the instructions stored in the memory, it causes the processor to perform the method described in any of the above embodiments.

[0033] Fourthly, embodiments of this application provide a computer-readable storage medium that stores a computer program, wherein when the computer program is run on a computer, the methods in any of the above-described embodiments are executed.

[0034] Fifthly, embodiments of this disclosure provide a computer program product, including a computer program / instructions that, when executed by a processor, implement the method described in any of the implementations of the first aspect.

[0035] To address the issue of AI Agents being unable to directly and effectively invoke automated services according to task intent, embodiments of this disclosure provide an AI Agent-based method and related products for invoking RPA services. By receiving semantic instructions and accurately locating the corresponding automated service provider, and combining this with a list of automated tools to filter target automated services and complete the invocation, an end-to-end connection from semantic instructions to automated service execution is achieved. This eliminates the need for fixed triggering conditions or manual operations, allowing AI Agents to drive required automated services solely through semantic instructions in natural language form, significantly improving the flexibility and convenience of automated service invocation. Furthermore, semantic matching is achieved through service description information included in the automated tool list, ensuring accurate selection of target automated services. Intelligent invocation can be achieved without additional modifications to existing automated services, effectively reusing existing automated service resources and reducing integration and usage costs.

[0036] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of this application will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0037] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in this application and should not be construed as limiting the scope of this application.

[0038] Figure 1 This is an exemplary system architecture diagram to which one embodiment of this disclosure may be applied; Figure 2A A flowchart illustrating an embodiment of an AI Agent invoking RPA service method based on AI and LLM according to this disclosure; Figure 2B This is a flowchart of an embodiment of the decomposition process 2010 according to an embodiment of step 203 of this disclosure; Figure 2C This is a flowchart illustrating an embodiment of the decomposition process 2050 according to step 205 of this disclosure; Figure 3 A flowchart illustrating one embodiment of performing RPA services according to this disclosure; Figure 4 The execution flow diagram of execution flow 400 is a specific example of implementing an AI Agent to call an RPA service method based on AI and LLM according to this disclosure; Figure 5 This is a schematic diagram of a structure of an embodiment of an AI Agent calling RPA service device based on AI and LLM according to the present disclosure; Figure 6 This is a schematic diagram of a device for implementing AI Agent invoking RPA service based on AI and LLM, suitable for carrying out embodiments of this disclosure. Detailed Implementation

[0039] The present disclosure will now be described in further detail 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 not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0040] In the description of this disclosure, the term "multiple" means two or more.

[0041] In the description of this disclosure, the term "semantic instruction" refers to an instruction presented in a human-understandable natural language form that accurately conveys task intent, constraints, and / or expected output requirements, and can be parsed and transformed into specific execution logic by an intelligent system.

[0042] In the description of this disclosure, the term "Model Control Protocol (MCP)" refers to a standard interaction protocol for AI Agents and external functional units, such as RPA engines, desktop automation tools, cloud service interfaces, etc. It is used to establish a unified calling standard between AI Agents and various functional units. By defining a unified tool description standard, tool registration and discovery mechanism, and routing and secure transmission rules for cross-module calls, the protocol shields the underlying implementation differences of different functional units, so that AI Agents do not need to adapt to the exclusive APIs of specific functional modules, but can complete tool queries, calls and result reception through a unified protocol format.

[0043] In the description of this disclosure, the term "Application Programming Interface (API)" refers to a predefined set of standardized rules for interaction between two or more software components to achieve specific functionalities. By specifying the input parameter format, output data structure, calling method, and error handling mechanism, API establishes communication specifications between components.

[0044] The implementation of the Digital Employee Platform (WEP) has gone through three stages. The first stage is automation: targeting RPA (Robotic Process Automation) with low business complexity, it utilizes software automation technology to automate rule-based, predefined procedural tasks. The second stage is intelligence: leveraging AI to extend the boundaries of RPA, such as processing unstructured documents and making data-driven decisions. The third stage is human-machine collaboration: utilizing the understanding, planning, and execution capabilities of large-scale models to automate complex tasks end-to-end.

[0045] In the human-machine collaboration phase, the digital employee platform serves as a bridge connecting workers and systems, workers and data, and systems and data. It is capable of: operating complex systems, processing various types of data, and interacting and collaborating with employees. The digital employee platform helps industries build large-scale, model-enabled digital employees (i.e., intelligent agents), achieving automation, intelligence, and human-machine collaboration in business processes.

[0046] The digital employee platform can seamlessly integrate multiple capabilities such as Agentic Process Automation (APA), Agentic Document Processing (ADP), and Agentic Business Insights (ABI). It has five major functions: "business understanding", "process creation", "run anywhere", "centralized management and control" and "human-machine collaboration". It enables enterprises to achieve end-to-end intelligent automation of business processes, replace manual operations, further improve business efficiency, and accelerate digital transformation.

[0047] It should be noted that, unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other. This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0048] Figure 1 An exemplary system architecture 100 is shown, which can be implemented using the AI ​​and LLM-based methods and related products disclosed herein to enable AI Agents to invoke RPA services.

[0049] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0050] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various artificial intelligence (AI) based models and services can be installed on terminal devices 101, 102, and 103, such as AI Agents, LLM, and RPA services. These AI-based models and services collectively form an intelligent agent (AI Agent) capable of performing predetermined tasks.

[0051] Terminal devices 101, 102, and 103 can be hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with service operation functions to carry and run various RPA services. Each terminal device can have a display function and display one or more desktops during operation. Terminal devices 101, 102, and 103 include, but are not limited to, smartphones, tablets, e-book readers, laptops, and desktop computers. When terminal devices 101, 102, and 103 are hardware, each terminal device can independently carry a WEP, on which an AI Agent composed of the aforementioned AI-based models and services is deployed. Alternatively, each terminal device can carry one or more of the aforementioned AI-based models and / or services, and the AI ​​Agent can be composed of the AI-based models and services carried by each of terminal devices 101, 102, and 103.

[0052] When terminal devices 101, 102, and 103 are software, they can be installed on the terminal devices listed above and used to run various RPA services in processes, displaying the RPA services running on each desktop through the screen. They can be implemented as multiple software programs or software modules (e.g., to carry the aforementioned RPA services), or as a single software program or software module. That is, each terminal device as software can be a module composed of one or more RPA services. No specific limitations are made here.

[0053] In some cases, the method for implementing AI Agent invoking RPA services based on AI and LLM provided in this disclosure can be executed individually by terminal devices 101, 102, and 103, or jointly by terminal devices 101, 102, and 103. Correspondingly, the apparatus for implementing AI Agent invoking RPA services based on AI and LLM can be set in terminal devices 101, 102, and 103. In this case, system architecture 100 may not include server 105.

[0054] In some cases, the method for implementing AI Agent invoking RPA services based on AI and LLM provided in this disclosure can be jointly executed by terminal devices 101, 102, and 103 and server 105. For example, server 105 carries a WEP, on which an AI Agent composed of the aforementioned models and services is deployed. Terminal devices 101, 102, and 103 invoke and run various RPA services intended for use by server 105. For instance, the step of "invoking the corresponding target automation service from each automation bearer unit" can be executed by terminal devices 101, 102, or 103, and the step of "determining at least one automation bearer unit corresponding to the semantic instruction" can be executed by server 105. This disclosure does not limit this. Correspondingly, the apparatus for implementing AI Agent invoking RPA services based on AI and LLM can also be respectively located in terminal devices 101, 102, and 103 and server 105.

[0055] In some cases, the method provided in this disclosure for implementing AI Agent to call RPA services based on AI and LLM can be executed by server 105 alone, and the execution results of server 105 can be displayed through terminal devices 101, 102, and 103. In this case, server 105 carries WEP, on which AI Agent composed of the above-mentioned models and services is deployed, and carries various RPA services. Server 105 can call the corresponding RPA service according to the specific intent. Correspondingly, terminal devices 101, 102, and 103 can also be set in server 105. In this case, system architecture 100 may not include terminal devices 101, 102, and 103.

[0056] It should be noted that server 105 can be either hardware or software. When server 105 is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, each of which can call the corresponding RPA service, or it can be implemented as a single server. When server 105 is software, it can be implemented as multiple software programs or software modules (for example, to provide the aforementioned RPA services, or to provide an AI Agent for calling the various RPA services), or it can be implemented as a single software program or software module. No specific limitations are made here.

[0057] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0058] In some alternative implementations, refer to Figure 2AThis document illustrates a flow 200 of an embodiment of the method for implementing AIAgent to call RPA services based on AI and LLM. Referring to the example of the system architecture 100 described above, in this flow 200, server 105 hosts the AI ​​Agent, and terminal devices 101, 102, and 103 respectively host various RPA services. These RPA services can be, for example, an RPA engine, an RPA editor, and / or an RPA executor. The method for implementing AIAgent to call RPA services based on AI and LLM includes the following steps 201 to 205: Step 201: Receive semantic instructions.

[0059] Semantic instructions can be natural language-compliant commands issued by the user to the AI ​​Agent in the form of speech or text, used to convey the user's task intent.

[0060] In this step, the AI ​​Agent can receive semantic commands issued by the user.

[0061] Step 202: Determine at least one automated bearer unit that corresponds to the semantic instructions in the MCP tool list.

[0062] Each automation bearer unit is used to carry one or more RPA services. For example, each automation bearer unit can be a terminal device or a desktop in a terminal device.

[0063] When each automation host unit is a desktop, the desktop can display various applications and / or functional units in the graphical user interface (GUI). Each RPA service on the desktop can automatically control the operation of one or more applications and / or functional units in the GUI.

[0064] Semantic instructions may include at least task information and at least one unit identifier, wherein each unit identifier represents a corresponding automated carrier unit. For example, when the automated carrier unit is a desktop, the unit identifier corresponding to the automated carrier unit may be, for example, a desktop identification document (ID).

[0065] Furthermore, the AI ​​Agent is configured with an LLM.

[0066] Accordingly, based on the semantic instructions received by the AI ​​Agent in step 201 above, the AI ​​Agent can call the LLM to perform semantic recognition on the semantic information of the received semantic instructions, thereby determining the task information and unit identifiers contained in the semantic information of the semantic instructions.

[0067] Furthermore, since each unit identifier represents a corresponding automated carrier unit, the AI ​​Agent can determine the automated carrier unit corresponding to that unit identifier based on each unit identifier.

[0068] As an example, when the semantic command is a voice command and the automated carrier unit is a desktop, the AI ​​Agent can identify various desktop IDs from the voice command through LLM. After determining each desktop ID, the AI ​​Agent can determine the desktop corresponding to each desktop ID.

[0069] Step 203: Obtain the list of automation tools corresponding to each automation support unit.

[0070] Each automated service unit is pre-configured with a corresponding list of MCP tools. Each MCP tool list is used to represent the description information of each automated service carried in the corresponding automated service unit. Thus, the tasks that each RPA service carried by the automated service unit can perform can be determined through each MCP tool list.

[0071] In this step, the list of MCP tools for each automated carrier unit can be pre-set.

[0072] Accordingly, based on the automated carrier units determined in step 202 above, the AI ​​Agent can determine the MCP tool list of the automated carrier unit pointed to by the semantic instruction from the list of MCP tools according to the unit identifiers determined in the semantic instructions.

[0073] Thus, the AI ​​Agent can determine the various RPA services provided by each automated carrier unit to which the semantic instructions are directed.

[0074] As an example, when the automation host unit is a desktop, each desktop is set up with an MCP tool list. The MCP tool list contains multiple descriptive information items. Each descriptive information item specifically describes the relevant information of an RPA service hosted by the desktop, such as the parameters of the RPA service, the functions it can perform, or the tasks it can perform.

[0075] Step 204: Identify at least one target automation service from each list of automation tools whose description information conforms to the semantic instructions of the MCP tool list.

[0076] In this step, based on the list of MCP tools corresponding to the semantic instructions determined in step 203 above, the RPA services that each automated bearer unit pointed to by the semantic instructions can execute can be determined.

[0077] Accordingly, the AI ​​Agent can use LLM to perform semantic recognition on the description information of each MCP tool list corresponding to the semantic instruction, thereby determining the semantic information of each description.

[0078] Furthermore, based on the task information identified from the semantic instructions in step 201 above, the AI ​​Agent can determine one or more descriptive information whose semantic information matches the task information from the various descriptive information, and use them as the target descriptive information.

[0079] Furthermore, the AI ​​Agent can directly identify the RPA service specifically described by the identified target description information based on the determined target description information, and use that RPA service as the target RPA service.

[0080] Step 205: Invoke the corresponding target automation service from each automation carrier unit.

[0081] Based on the target RPA services determined in step 204 above, the AI ​​Agent can accurately call the target RPA services that match the semantic instructions and can meet the task information, thereby realizing the effective automatic selection of applicable RPA services through the AI ​​Agent.

[0082] Based on this, the system locates the automation carrier unit by receiving semantic commands, then filters and invokes the target automation service by retrieving the automation tool list, achieving end-to-end intelligent connection from semantic commands to automation service execution. This process does not rely on fixed triggering conditions or manual operation; users can accurately invoke the required automation service simply through semantic commands in natural language form, significantly improving the flexibility and convenience of automation service invocation. Simultaneously, the accurate matching of description information in the automation tool list with the semantic commands ensures the accurate selection of the target automation service. Furthermore, the process directly completes service invocation based on the automation carrier unit and its corresponding tool list, without requiring additional modifications to existing automation services. This effectively reuses existing automation service resources, reduces integration costs and operational barriers, and significantly improves the efficiency and adaptability of automation services.

[0083] In some alternative implementations, a registration service and MCP route can be pre-configured before the AI ​​Agent receives semantic instructions. The MCP route can be used to establish associations between the AI ​​Agent and various automated bearer units.

[0084] Continue to refer to Figure 2B The diagram illustrates a breakdown flowchart of an embodiment of the breakdown process 2030 for step 203 of this disclosure. The breakdown process 2030 includes steps 2031 to 2034: Step 2031: Encapsulate the service information of each automation service in each automation bearer unit into the description information of the corresponding Model Control Protocol (MCP) tool.

[0085] In this step, for each RPA service carried in each automatic hosting unit, various relevant information about each RPA service can be obtained in advance, such as the name, parameters, function description, and task examples of the RPA service.

[0086] Furthermore, all relevant information can be combined to form the service information for this RPA service.

[0087] Furthermore, the service information for each RPA service can be encapsulated.

[0088] Specifically, service information can be encapsulated in the form of a pre-defined MCP tool to obtain information described in the form of an MCP tool, thereby representing the RPA service as an MCP tool. Step 2032: Register the description information of each automated bearer unit and the corresponding preset unit identifier to the MCP route.

[0089] Based on the description information of each MCP tool form encapsulated in step 301 above, and based on the preset registration service, each description information can be registered to the MCP route using the registration service.

[0090] In order to associate each description information with the corresponding automated bearer unit, during the registration of each description information, a corresponding unit identifier can be assigned to each description information, and each description information and the corresponding unit identifier can be registered together to the MCP route.

[0091] Based on this, the MCP route can store the unit identifier of each automated bearer unit and combine the description information of each RPA service corresponding to each automated bearer unit into the MCP tool list in the aforementioned steps.

[0092] Furthermore, during the registration process of the registration service, a mapping relationship between each unit identifier and the corresponding MCP tool list can be established, and the registration service can be used to maintain each unit identifier and its mapping relationship.

[0093] Step 2033: Use the MCP tool list AI Agent to send the unit identifiers of each item in the MCP tool list semantic instructions to the MCP tool list MCP route.

[0094] In this step, during the process of obtaining the MCP tool list, based on the unit identifiers registered in the MCP route in step 2032 above, and the MCP tool list that has a mapping relationship with the unit identifiers, the AI ​​Agent can determine the unit identifier of each automated bearer unit based on the automated bearer units identified from the semantic instructions, and send the unit identifier to the MCP route.

[0095] Step 2034: Use the MCP routing in the MCP tool list to obtain the description information of each unit identifier in the semantic instructions of the MCP tool list corresponding to the automated carrier unit, and feed it back to the MCP tool list AI Agent.

[0096] In this step, the MCP route can receive the unit identifier from the AI ​​Agent. Based on the mapping relationship corresponding to the unit identifier stored in step 2032 above, the MCP route can determine the corresponding MCP tool list from each of the stored unit identifiers through the mapping relationship.

[0097] Furthermore, the MCP routing can feed back the determined list of MCP tools to the AI ​​Agent, enabling the AI ​​Agent to determine the RPA service that conforms to the semantic instructions based on the description information of each RPA service in the MCP tool list.

[0098] Based on this, a standardized implementation process for obtaining the list of automated tools is provided by encapsulating descriptive information, registering it with the MCP route, and then using the sending unit identifier to correspond to the descriptive information. By first encapsulating the service information of automated services into unified MCP tool descriptive information, and then registering it with the MCP route in conjunction with the unit identifier, the accurate association between the descriptive information and the automated carrying unit is ensured, laying the foundation for subsequent semantic matching. At the same time, through the interaction logic of AIAgent sending the unit identifier to the MCP route and the MCP route responding with the corresponding descriptive information, the efficient acquisition of the list of automated tools is achieved. The AI ​​Agent can accurately obtain the service capability description of the target automated carrying unit without manual intervention, which greatly improves the accuracy and efficiency of subsequent target automated service screening. Moreover, the entire process is based on standardized information encapsulation and interaction processes, ensuring the standardization and scalability of the entire process.

[0099] In some alternative implementations, refer to Figure 2C The diagram illustrates a breakdown flowchart of an embodiment of the breakdown process 2050 for step 205 of this disclosure. The breakdown process 2050 includes steps 2051 to 2052: Step 2051: Use the MCP tool list AI Agent to send a call request to the MCP tool list MCP route.

[0100] Based on the target RPA service determined in step 204 above, during the process of calling the target RPA service, since each RPA service cannot be directly called by the AI ​​Agent, and setting up the corresponding API for each RPA service is very cumbersome, the AI ​​Agent can send a call request to the target RPA service to the MCP route through the aforementioned MCP route.

[0101] The call request can not only point to the target RPA service, but also include the specific execution content required from the target RPA service.

[0102] Step 2052: Use the MCP tool list to route and call the automated services for each target.

[0103] In this step, the MCP route can establish an interaction relationship with each automation bearer unit in advance. Thus, when the MCP route receives a call request, it can directly call the target RPA service to the automation bearer unit corresponding to the unit identifier according to the target RPA service pointed to by the call request and the unit identifier corresponding to the target RPA service.

[0104] Based on this, the interaction process of first sending a call request through the AI ​​Agent and then calling the target automation service through MCP routing achieves precise and efficient invocation of the target automation service. This process involves the AI ​​Agent sending a call request to the target automation service via MCP routing, eliminating the need for direct interaction with the automation support unit and simplifying the call logic. Simultaneously, MCP routing can directly associate with the automation support unit and execute the call operation, avoiding additional modifications to the target automation service and ensuring the convenience and smoothness of the call process. Furthermore, the call request explicitly targets the automation service, accurately triggering the corresponding automation process, reducing call deviations, improving the execution efficiency and reliability of the automation service, and providing crucial support for the closed loop from semantic instructions to service execution.

[0105] In some alternative implementations, further reference is made. Figure 3 This illustrates a flow 300 of an embodiment of the execution of RPA services according to this disclosure. The flow 300 includes the following steps 301 to 302: Step 301: Execute the preset automated processes in the automated services for each objective.

[0106] Based on the aforementioned step 2052, the target RPA service is invoked to the automated bearer unit, and the corresponding automated bearer unit can execute the target RPA service.

[0107] During the execution of various target RPA services, the execution process of each target RPA service can be recorded and the recorded content can be stored in a preset storage location.

[0108] Step 302: Feed back the execution results of each automated process to the MCP tool list AI Agent.

[0109] During the execution of the target RPA service by the automated host unit, the automated host unit can report the execution result of the target RPA service to AIAgent.

[0110] The execution result can be, for example, the recorded content of the execution process of the target RPA service.

[0111] In some cases, the automated bearer unit can first feed back the execution result of the target RPA service to the MCP route, and then the MCP route can feed back the execution result of the target RPA service to the AI ​​Agent.

[0112] Based on this, the execution of semantic command-driven automated services is completed by executing the automated process of the target automated service and then feeding back the execution results to the AI ​​Agent. This process ensures the precise execution of the preset automated process of the target automated service, guaranteeing that the task requirements corresponding to the user's semantic commands are effectively met. Simultaneously, the execution results are directly fed back to the AI ​​Agent, enabling real-time feedback of service execution status and results. This allows users to promptly understand the task completion status through the AI ​​Agent, significantly improving the completeness and user experience of semantic command-driven automated service invocation, and ensuring the traceability and reliability of the entire invocation process.

[0113] In some optional implementations, after registering the various descriptive information and corresponding unit identifiers with the MCP route, a remote control protocol can also be established between the AI ​​Agent and each automated bearer unit, so that the AI ​​Agent can directly control the various PRAs in each automated bearer unit.

[0114] The remote control protocol can be, for example, Secure Shell (SSH).

[0115] Based on this, the remote control protocol between the AI ​​Agent and each automated carrier unit can be registered to the MCP route. Accordingly, for each automated carrier unit in the MCP route, a corresponding unit identifier, corresponding descriptive information and corresponding remote control protocol are registered.

[0116] Furthermore, after calling the corresponding target automation service from each automation bearer unit, during the process of running the corresponding target RPA service in each automation bearer unit, the AI ​​Agent can control the running target RPA service based on the remote control protocol established with the automation bearer unit.

[0117] During this process, the AI ​​Agent can display the operation of the target RPA service in the automated hosting unit through its display device.

[0118] For example, for Figure 1 In the example shown, the AI ​​Agent can be deployed on server 105. During the process of running the target RPA service on the desktop of terminal device 101, the AI ​​Agent can retrieve the remote control protocol between server 105 and terminal device 101 from the MCP route, and control the target RPA service running on the desktop of terminal device 101 through the remote control protocol.

[0119] In some alternative implementations, refer to Figure 4 This illustrates the execution flow 400 of a specific example of implementing an AIAgent to invoke an RPA service method based on AI and LLM, as disclosed herein. This decomposed flow 400 includes the following steps 401 to 40X: As an example, the automated host unit can be, for example, a desktop, and the corresponding unit identifier can be, for example, a desktop ID.

[0120] Step 401: Engage in language interaction with the user.

[0121] In this step, the AI ​​Agent can engage in natural language conversations with the user and extract semantic instructions containing desktop IDs and task information from these conversations.

[0122] Step 402: Register for various RPA services.

[0123] In this step, you can obtain information such as the name, parameters, function descriptions, and task examples of the various RPA services hosted on each desktop.

[0124] Furthermore, the relevant information of each PRA service on the desktop can be encapsulated into corresponding description information in the form of an MCP tool, and the description information of each PRA service on the same desktop can be pre-assembled into an MCP tool list.

[0125] Furthermore, the MCP tool list for each desktop and its corresponding desktop ID can be registered to the preset MCP route through the registration service.

[0126] Step 403: Obtain the list of MCP tools from the MCP route.

[0127] In this step, after obtaining the semantic instructions, the AI ​​Agent can identify the desktop ID and send the desktop ID to the MCP router.

[0128] Based on this, the MCP router can obtain information such as the name, parameters, function description, and task examples of the various RPA services carried by the desktop corresponding to that desktop ID, which is the list of MCP tools for that desktop that has been pre-registered in the MCP router.

[0129] Accordingly, the MCP router can send the list of MCP tools corresponding to each desktop ID to the AI ​​Agent.

[0130] Step 404: The AI ​​Agent determines the target RPA service based on the description information and invokes it.

[0131] In this step, after the AI ​​Agent obtains the list of MCP tools corresponding to each desktop ID from the MCP route, it can determine the target RPA service suitable for semantic instructions based on the description information of each item in the MCP tool list.

[0132] Accordingly, the AI ​​Agent can send a call request to the MCP route pointing to the target RPA service, so that the MCP route can determine the target RPA services that the AI ​​Agent needs to call.

[0133] Step 405: Invoke the target RPA service.

[0134] In this step, MCP routing is based on the identified target RAP services and the desktop IDs corresponding to each target RAP service. MCP routing can directly call the target RAP services from the corresponding desktops based on each desktop ID.

[0135] Based on this, semantic commands are obtained through language interaction and RPA services are registered, thereby directly obtaining a list of MCP tools containing RPA service description information. The target RPA service is then determined accordingly, realizing a complete process from natural language interaction to precise RPA service invocation. This process allows users to convey their needs, including desktop ID and task information, simply through language interaction with the AI ​​Agent, without relying on fixed triggering conditions or manual operation, significantly lowering the barrier to entry for RPA services. Simultaneously, by encapsulating RPA services into MCP tool description information and registering them with the MCP route, combined with the desktop ID, the accurate acquisition of the MCP tool list and target service filtering are achieved, ensuring the accuracy and relevance of RPA service invocation. Furthermore, no additional modifications to existing RPA services are required, effectively reusing existing service resources and significantly improving the efficiency and adaptability of RPA service invocation.

[0136] Further reference Figure 5 As an implementation of the methods shown in the above figures, this disclosure provides an embodiment of an AI Agent calling RPA service device based on AI and LLM. This device embodiment is similar to... Figure 2A Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0137] like Figure 5 As shown, the AI ​​Agent calling RPA service device 500 based on AI and LLM in this embodiment includes: a receiving module 501, a unit determination module 502, a list acquisition module 503, a service module 504, and a calling module 505; The receiving module 501 is configured to receive semantic instructions; The unit determination module 502 is configured to determine at least one automated bearer unit corresponding to the semantic instruction, and each automated bearer unit carries at least one automated service; The list acquisition module 503 is configured to acquire the list of automation tools mapped to each automation carrier unit. Each automation tool list includes at least one description of a Model Control Protocol (MCP). Each description is obtained by encapsulating the service information of the corresponding automation service in the form of an MCP tool. Service module 504 is configured to identify at least one target automation service from a list of automation tools whose description information matches the semantic instructions. Module 505 is configured to invoke the corresponding target automation service from each automation bearer unit.

[0138] In this embodiment, the specific processing of the receiving module 501, unit determination module 502, list acquisition module 503, service module 504, and calling module 505 of the AI ​​Agent calling the RPA service device 500 based on AI and LLM, and the resulting technical effects, can be found in the following references. Figure 2A The relevant descriptions of steps 201, 202, 203, 204 and 205 in the corresponding embodiments will not be repeated here.

[0139] In some alternative implementations, the semantic instruction includes at least one unit identifier;

[0140] Accordingly, the receiving module 501 is also specifically configured as follows:

[0141] Utilize an AI Agent to receive semantic instructions; Accordingly, the unit determination module 502 is further configured as follows: The large language model (LLM) in the AI ​​Agent is used to identify at least one unit identifier in the semantic information, and the corresponding automated carrier unit for each unit identifier.

[0142] In some alternative implementations, the list retrieval module 503 is further configured as follows:

[0143] The AI ​​Agent is used to send the unit identifiers in the semantic instructions to the MCP route; The MCP routing is used to obtain the description information of the automated bearer unit corresponding to each unit identifier in the semantic instructions, and then feeds it back to the AI ​​Agent.

[0144] In some alternative implementations, the apparatus for implementing AI Agent invocation of RPA services based on AI and LLM further includes a registration module 506, configured to: Before obtaining the list of automation tools mapped to each automation carrier unit, the service information of each automation service in each automation carrier unit is encapsulated into the description information of the corresponding Model Control Protocol (MCP) tool. Register the description information of each automated bearer unit and the corresponding preset unit identifier to the MCP route.

[0145] In some alternative implementations, semantic instructions include task information; Accordingly, service module 504 is also specifically configured as follows: The AI ​​Agent sends a call request to the MCP route, and the call request points to the automated services for each target. Use MCP routing to invoke automated services for each target.

[0146] This includes using MCP routing to invoke automated services for various objectives, including: In response to receiving a call request from the MCP route, the corresponding target automation service is invoked from the automation bearer unit corresponding to each unit identifier in the call request.

[0147] In some alternative implementations, the apparatus for implementing AI Agent invocation of RPA services based on AI and LLM further includes an execution module 507 configured to: Execute the pre-set automated processes in the automated services for each objective; The execution results of each automated process are fed back to the AI ​​Agent.

[0148] In some alternative implementations, the registration module 506 is further configured as follows: Establish remote control protocols between the AI ​​Agent and each automated carrier unit; Register each remote control protocol to the MCP route; In some alternative implementations, execution module 507 is further configured to: AI Agent remotely controls the execution of target automated services through remote control protocols.

[0149] It should be noted that the implementation details and technical effects of each module in the device for implementing AI Agent to call RPA services based on AI and LLM provided in the embodiments of this disclosure can be referred to the descriptions of other embodiments in this disclosure, and will not be repeated here.

[0150] The following is for reference. Figure 6 It shows a schematic diagram of the structure of a computer system 600 suitable for implementing the electronic device of the present disclosure. Figure 6 The computer system 600 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0151] like Figure 6 As shown, the computer system 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. The RAM 603 also stores various programs and data required for the operation of the computer system 600. The processing device 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0152] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 608 including, for example, magnetic tapes, hard disks, etc.; and communication devices 609. Communication device 609 allows computer system 600 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 A computer system 600 with various electronic devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0153] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 609, or installed from a storage device 608, or installed from a ROM 602. When the computer program is executed by a processing device 601, it performs the functions defined in the methods of embodiments of this disclosure.

[0154] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0155] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0156] The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the following functions: Figure 2A The illustrated embodiments and their alternative implementations demonstrate a method for implementing AI Agent invoking RPA services based on AI and LLM.

[0157] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0158] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0159] The units or modules described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units or modules do not necessarily limit the unit itself; for example, a receiving module can also be described as a "module for receiving semantic instructions".

[0160] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

Claims

1. A method for implementing AI Agent calling RPA services based on AI and LLM, characterized in that, The AI ​​Agent is used to receive semantic instructions, which include at least one unit identifier. The semantic instruction corresponds to at least one automated bearer unit, and each automated bearer unit carries at least one automated service; The step of determining the at least one automated bearer unit corresponding to the semantic instruction includes: The AI ​​Agent is used to call the Large Language Model (LLM) to perform semantic recognition on the semantic information of the received semantic instructions, determine the task information and unit identifiers contained in the semantic information of the semantic instructions, and determine the automated carrier unit corresponding to the unit identifier based on the unit identifier. Obtain multiple relevant information items for each RPA service carried in each automatic bearer unit, and combine these relevant information items into service information for that RPA service; The service information is encapsulated in the form of a pre-defined MCP tool to obtain information described in the form of an MCP tool. Register the description information of each automated bearer unit and the corresponding preset unit identifier to the MCP route; Obtain the list of automation tools corresponding to each automation bearer unit. Each automation tool list includes at least one description of a Model Control Protocol (MCP). Each description is obtained by encapsulating the service information of the corresponding automation service in the form of an MCP tool. The step of obtaining the list of automation tools corresponding to each automation bearer unit includes: The AI ​​Agent is used to send the unit identifiers of each element in the semantic instruction to the MCP route; The MCP routing is used to determine the corresponding MCP tool list from each of the stored unit identifiers based on the mapping relationship of the unit identifier. The determined MCP tool list is then fed back to the AI ​​Agent. Identify at least one target automation service from the list of automation tools whose description information matches the semantic instructions; The semantic instructions include task information; The step of determining at least one target automation service from the list of automation tools whose description information matches the semantic instructions includes: The AI ​​Agent uses a Large Language Model (LLM) to perform semantic recognition on the description information of each MCP tool list corresponding to the semantic instruction, determines the semantic information of each description, and determines one or more descriptions that match the task information from each description and uses them as the target description information. Determine the target automation service corresponding to the description information of each target; The corresponding target automation service is invoked from each automation carrier unit.

2. The method according to claim 1, characterized in that, The step of calling the corresponding target automation service from each automation bearer unit includes: The AI ​​Agent sends a call request to the MCP route, and the call request points to the automated service for each target. The MCP routing is used to invoke the automated services for each target.

3. The method according to claim 2, characterized in that, The invocation request includes the unit identifier corresponding to each target automated service; The process of using the MCP routing to invoke various target automation services includes: In response to the MCP route receiving the call request, the corresponding target automation service is invoked from the automation bearer unit corresponding to each unit identifier of the call request.

4. The method according to claim 1, characterized in that, After invoking the corresponding target automation service from each automation bearer unit, the method further includes: Execute the pre-set automated processes in the automated services for each objective; The execution results of each automated process are fed back to the AI ​​Agent.

5. The method according to claim 1, characterized in that, After registering the description information of each automated bearer unit and the corresponding preset unit identifier to the MCP route, the method further includes: Establish remote control protocols between the AI ​​Agent and each automated carrier unit; Register each remote control protocol to the MCP route; After invoking the corresponding target automation service from each automation bearer unit, the method further includes: The AI ​​Agent remotely controls the execution of the target automated service through the remote control protocol.

6. A device for implementing AI Agent invoking RPA services based on AI and LLM, characterized in that, include: The receiving module is configured to receive semantic instructions using an AI Agent, the semantic instructions including at least one unit identifier; The unit determination module is configured to determine at least one automated bearer unit corresponding to the semantic instruction, and each automated bearer unit carries at least one automated service; The step of determining the at least one automated bearer unit corresponding to the semantic instruction includes: The AI ​​Agent is used to call the Large Language Model (LLM) to perform semantic recognition on the semantic information of the received semantic instructions, determine the task information and unit identifiers contained in the semantic information of the semantic instructions, and determine the automated carrier unit corresponding to the unit identifier based on the unit identifier. The list retrieval module is configured to retrieve multiple related information items for each RPA service carried in each automatic bearer unit, and combine these related information items into service information for that RPA service. The service information is encapsulated in the form of a pre-defined MCP tool to obtain information described in the form of an MCP tool. Register the description information of each automated bearer unit and its corresponding preset unit identifier to the MCP route. Obtain the list of automation tools corresponding to each automation bearer unit. Each automation tool list includes at least one description of a Model Control Protocol (MCP). Each description is obtained by encapsulating the service information of the corresponding automation service in the form of an MCP tool. The step of obtaining the list of automation tools corresponding to each automation bearer unit includes: The AI ​​Agent is used to send the unit identifiers of each element in the semantic instruction to the MCP route; The MCP routing is used to determine the corresponding MCP tool list from each of the stored unit identifiers based on the mapping relationship of the unit identifier. The determined MCP tool list is then fed back to the AI ​​Agent. The service module is configured to determine at least one target automation service from a list of automation tools whose description information matches the semantic instructions; The semantic instructions include task information; The step of determining at least one target automation service from the list of automation tools whose description information matches the semantic instructions includes: The AI ​​Agent uses a Large Language Model (LLM) to perform semantic recognition on the description information of each MCP tool list corresponding to the semantic instruction, determines the semantic information of each description, and determines one or more descriptions that match the task information from each description and uses them as the target description information. Determine the target automation service corresponding to the description information of each target; The calling module is configured to call the corresponding target automation service from each automation bearer unit.

7. A device for implementing AI Agent invoking RPA services based on AI and LLM, characterized in that, include: A processor and a memory, wherein instructions are stored in the memory and are loaded and executed by the processor to implement the method as claimed in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method as described in any one of claims 1-5.

9. A computer program product comprising computer program instructions, characterized in that, When the computer program instructions are executed on the computer, the computer causes the computer to perform the method as described in any one of claims 1-5.

Images