Task processing method and device, computer equipment, storage medium and computer program product

Abstract

The invention relates to a task processing method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a sent query request; wherein the query request comprises query content; performing task planning on the query content to obtain at least one task step; for the task step, under the condition that execution of the task step depends on a target local tool of a corresponding associated terminal, sending a calling request of the target local tool to the associated terminal; determining a processing result of the task step based on a calling result corresponding to the calling request returned by the associated terminal; and based on the processing result corresponding to the at least one task step, determining reply content for replying the query content. The task processing capability of the large model can be improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a task processing method, apparatus, computer equipment, storage medium, and computer program product. Background Technology

[0002] Large Language Models (LLMs) are AI models capable of large-scale natural language understanding and generation. However, due to limitations in computing resources, network bandwidth, and functional boundaries, large models have limited task processing capabilities and struggle to solve complex tasks. Summary of the Invention

[0003] Therefore, it is necessary to provide a task processing method, apparatus, computer device, computer-readable storage medium, and computer program product that can improve the task processing capabilities of large models in response to the above-mentioned technical problems.

[0004] Firstly, this application provides a task processing method. Applied to a server, the method includes:

[0005] Obtain the sent query request; wherein the query request includes query content;

[0006] The query content is used for task planning to obtain at least one task step;

[0007] For the task step, if the execution of the task step depends on a target local tool on an associated terminal, a request to invoke the target local tool is sent to the associated terminal; wherein, the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal;

[0008] Based on the call result corresponding to the call request returned by the associated terminal, the processing result of the task step is determined;

[0009] Based on the processing results corresponding to the at least one task step, the response content for replying to the query content is determined.

[0010] Secondly, this application provides a task processing method applied to an associated terminal, the method comprising:

[0011] Obtain the query content, and determine the query request based on the query content;

[0012] The query request is sent to the server to instruct the server to perform task planning on the query content and obtain at least one task step;

[0013] The server displays the step nodes corresponding to each of the at least one task step; the server is also configured to send a call request for the target local tool to the associated terminal when the execution of the task step depends on the target local tool of the associated terminal; and determine the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal; wherein the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal.

[0014] The server displays the result nodes corresponding to the processing results of the task steps; the server is also used to determine the response content to the query content based on the processing results corresponding to the at least one task step.

[0015] The response content is displayed.

[0016] Thirdly, this application also provides a task processing apparatus. The apparatus includes:

[0017] The first acquisition module is used to acquire the sent query request; wherein the query request includes query content;

[0018] The planning module is used to plan tasks for the query content and obtain at least one task step;

[0019] The invocation module is configured to, for the task step, send an invocation request to the associated terminal for the target local tool when the execution of the task step depends on the target local tool of the associated terminal; wherein the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal;

[0020] The first determining module is used to determine the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal.

[0021] The second determining module is used to determine the response content to the query content based on the processing results corresponding to the at least one task step.

[0022] In one embodiment, the planning module is further configured to:

[0023] Parse the query request to determine the target account information corresponding to the query request;

[0024] Based on the association between account information, terminal information and local tool information, at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information are determined.

[0025] Task planning is performed on the query content based at least on the at least one terminal information and the at least one local tool information.

[0026] In one embodiment, the planning module is further configured to:

[0027] For each task step, a target execution agent whose functional scope matches the task step is determined from the agent resource pool, thus obtaining the target execution agent corresponding to the task step;

[0028] The calling module is also used for:

[0029] Based on the target execution agent corresponding to the task step, the target terminal information and target local tool information corresponding to the task step are determined from at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information; wherein, the target account information is obtained by parsing the query request;

[0030] If the target terminal information corresponds to an associated terminal and the target local tool information corresponds to a target local tool, a request to invoke the target local tool is sent to the associated terminal.

[0031] In one embodiment, the calling module is further configured to:

[0032] The target prompt is determined based on at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information.

[0033] The target prompt is input into the target execution agent corresponding to the task step, and the target terminal information and target local tool information corresponding to the task step are output.

[0034] In one embodiment, the calling module is further configured to:

[0035] Identify candidate local tools whose functional scope satisfies the task steps, and query the online status information of the candidate local tools;

[0036] Based on the online status information, the target local tool and the associated terminal where the target local tool is located are determined from the candidate local tools;

[0037] Determine the target local tool information corresponding to the target local tool and the target terminal information corresponding to the associated terminal.

[0038] In one embodiment, the calling module is further configured to:

[0039] Identify candidate local tools whose functional scope satisfies the task steps, and query the performance information and historical execution result accuracy of the terminal where the candidate local tools are located;

[0040] Based on the performance information and the accuracy of the execution results, a target local tool and the associated terminal where the target local tool is located are determined from the candidate local tools.

[0041] Determine the target local tool information corresponding to the target local tool and the target terminal information corresponding to the associated terminal.

[0042] In one embodiment, the planning module is further configured to:

[0043] The query content is decomposed into tasks to obtain at least one task objective;

[0044] For the task objective, at least based on the tool information corresponding to the query request, determine the target local tool whose functional scope meets the task objective;

[0045] Based on the task objective, the tool identifier of the target local tool, and the terminal identifier of the associated terminal where the target local tool is located, at least one task step is determined.

[0046] In one embodiment, the planning module is further configured to:

[0047] If the target local tool is a standard protocol tool, obtain the identification information of the agent executing the standard protocol tool;

[0048] Based on the task objective, the tool identifier of the target local tool corresponding to the task objective, the terminal identifier of the target local tool, and the identification information of the agent executing the standard protocol tool, at least one task step is determined.

[0049] In one embodiment, the calling module is further configured to:

[0050] If the target local tool is a standard protocol tool, the input parameters of the target local tool are queried from the target local tool information corresponding to the target local tool;

[0051] Based on the task steps corresponding to the target local tool, determine the parameter content of the input parameters;

[0052] Based on the input parameters of the target local tool and the corresponding parameter content, the invocation request of the target local tool is determined.

[0053] In one embodiment, the apparatus further includes a task decomposition module, the task decomposition module being used to:

[0054] For the task steps, if the task steps meet preset conditions, the task steps that meet the preset conditions are decomposed to obtain at least one subtask; wherein, the subtask tool information corresponding to the matching of the at least one subtask is at least one type of tool information.

[0055] In one embodiment, the task decomposition module is used to:

[0056] For the subtask, if the execution of the subtask depends on the target local tool of the corresponding associated terminal, a subtask invocation request of the target local tool is sent to the associated terminal;

[0057] Based on the call result corresponding to the call request of the subtask returned by the associated terminal, the processing result of the subtask is determined;

[0058] Based on the processing results of the at least one sub-task, the processing results of the task steps that meet the preset conditions are determined.

[0059] In one embodiment, the calling module is further configured to:

[0060] When the execution of the task step depends on the target local tool of the associated terminal by the intelligent agent executing the standard protocol tool, the terminal identifier of the associated terminal corresponding to the task step and the tool identifier corresponding to the target local tool are encapsulated in the data format of the standard protocol to obtain the invocation request corresponding to the task step.

[0061] The call request is sent to the associated terminal.

[0062] In one embodiment, the calling module is further configured to:

[0063] Obtain the terminal identifier corresponding to the associated terminal and the account information corresponding to the query request;

[0064] Determine the long-connection communication interface corresponding to the terminal identifier and the account information;

[0065] Based on the long-connection communication interface, a request to invoke the target local tool is sent.

[0066] In one embodiment, the device further includes a communication module, the communication module being used for:

[0067] Obtain the account information corresponding to the query request;

[0068] Determine the terminal information corresponding to at least one terminal associated with the account information;

[0069] Establish long-connection communication interfaces for the terminals corresponding to the account information and the at least one terminal information, respectively.

[0070] In one embodiment, the apparatus further includes a registration module, the registration module being used for:

[0071] Obtain the registration information of local tools reported by each terminal; wherein, the registration information includes account information, terminal information, and local tool information;

[0072] Based on the registration information, the relationships between account information, terminal information, and local tool information are determined.

[0073] In one embodiment, the first determining module is further configured to:

[0074] If the call result corresponding to the call request returned by the associated terminal is a call failure, the target local tool is updated to a backup local tool;

[0075] Based on the processing results of the backup local tool, the processing result of the task step is determined.

[0076] In one embodiment, the second determining module is further configured to:

[0077] If the backup local tool fails to be invoked, the query content is re-planned to obtain at least one updated task step;

[0078] The step of determining the response content based on the processing results corresponding to the at least one task step includes:

[0079] Based on the processing results corresponding to the at least one updated task step, the response content for replying to the query content is determined.

[0080] In one embodiment, the planning module is further configured to:

[0081] Retrieve historical query content that matches the query request from the experience knowledge base;

[0082] If a historical query that matches the query request is found, the historical task steps corresponding to the historical query are obtained; based on the historical task steps, at least one task step to reply to the query is determined.

[0083] If no historical query content matching the query request is found, task planning is performed on the query content to obtain at least one task step.

[0084] In one embodiment, the associated terminal includes at least one of a mobile terminal, a computer terminal, and a browser terminal, and the calling module is further configured to:

[0085] If the execution of the task step depends on the target local tool of the corresponding mobile terminal, a request to invoke the target local tool is sent to the mobile terminal.

[0086] If the execution of the task step depends on the target local tool of the corresponding computer terminal, a request to invoke the target local tool is sent to the computer terminal.

[0087] If the execution of the task step depends on the target local tool of the corresponding browser terminal, a request to invoke the target local tool is sent to the browser terminal.

[0088] In one embodiment, the second determining module is further configured to:

[0089] Based on the invocation result of the first local tool returned by the first terminal, the processing result of the first task step is determined; wherein, the first task step includes using the first local tool to search for information from the Internet;

[0090] Based on the invocation result of the second local tool returned by the second terminal, the processing result of the second task step is determined; wherein, the second task step includes using the second local tool to convert the searched information into a target format file;

[0091] Based on the invocation result of the third local tool returned by the third terminal, the processing result of the third task step is determined; wherein, the third task step includes using the third local tool to send the target format file to the communication account via an instant messaging tool.

[0092] Fourthly, this application also provides a task processing apparatus. The apparatus includes:

[0093] The second acquisition module is used to determine the query request based on the query content;

[0094] The sending module is used to send the query request to the server to instruct the server to perform task planning on the query content and obtain at least one task step;

[0095] The first display module is used to display the step nodes corresponding to the at least one task step respectively; the server is also used to send a call request for the target local tool to the associated terminal when the execution of the task step depends on the target local tool of the corresponding target terminal; and to determine the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal; wherein, the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal;

[0096] The second display module is used to display the result nodes corresponding to the processing results of the task steps; the server is also used to determine the response content to the query content based on the processing results corresponding to the at least one task step.

[0097] The third display module is used to display the response content.

[0098] In one embodiment, the apparatus further includes a configuration module, the configuration module being configured to:

[0099] Display the target page, on which configuration controls for the local tools are displayed;

[0100] In response to a triggered event on the configuration control, display a list of local tools;

[0101] In response to a selection event for a target local tool from the list of local tools, the target local tool is installed to at least one of the associated terminals.

[0102] In one embodiment, the device further includes a sharing module, the sharing module being used for:

[0103] In response to a share event for a target local tool from the list of local tools, the link information of the target local tool is sent to other accounts.

[0104] In one embodiment, the configuration module is further configured to:

[0105] Display the target page, on which a new control for local tools is displayed;

[0106] In response to the trigger event for the newly created control, display the local tool creation page;

[0107] The local tool import page receives an input event for the link information of a local tool, and downloads the local tool from a third-party platform based on the link information to obtain a newly created local tool.

[0108] In one embodiment, the sending module is further configured to send the query request to the server based on the associated terminal where the target local tool is located.

[0109] The device further includes a processing module, the processing module being used for:

[0110] Parse the call request to obtain the call instructions for the target local tool to execute the task steps;

[0111] Based on the invocation instruction and the target local tool, the task steps are processed to obtain the processing result, and the processing result is sent to the server.

[0112] Fifthly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method described in any of the embodiments of this disclosure.

[0113] Sixthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the methods described in any of the embodiments of this disclosure.

[0114] In a seventh aspect, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the methods described in any of the embodiments of this disclosure.

[0115] The aforementioned task processing method, apparatus, computer equipment, storage medium, and computer program product, in response to a sent query request, perform task planning on the query content through a server to obtain at least one task step, breaking down complex query content into task steps and solving each task step one by one. Optionally, if the task steps are relatively complex, they can be further broken down. Therefore, this application can handle relatively complex query content. Furthermore, when executing task steps, in addition to calling cloud tools, more importantly, it can call various associated terminals related to the query request and the target local tools of each associated terminal, making full use of the rich localization capabilities of each associated terminal, such as calling professional software, accessing hardware sensors, and processing local data. In this process, it fully utilizes the advantage that each target local tool can access the corresponding terminal environment, greatly expanding the capability boundary for solving large-scale model tasks, and can solve extremely complex tasks involving cross-terminal collaboration, possessing heterogeneous dedicated capabilities. Furthermore, by sending a call request for the target local tool to the associated terminal; based on the call result corresponding to the call request returned by the associated terminal, the processing result of the task step is determined. This approach eliminates the need to access end-user data, requiring only the retrieval of the call result, thus protecting end-user data security. Furthermore, it enables comprehensive utilization of data across various associated devices, achieving cross-terminal resource aggregation. Additionally, when a new terminal is added to an account, it can be aggregated into the existing terminal list corresponding to that account, demonstrating strong scalability and dynamically increasing the system's capacity ceiling. Attached Figure Description

[0116] Figure 1 This is an application environment diagram of a task processing method in one embodiment;

[0117] Figure 2 This is a flowchart illustrating a task processing method in one embodiment;

[0118] Figure 3 This is a user interface diagram of a task processing method in one embodiment;

[0119] Figure 4 This is a schematic diagram showing the task steps in one embodiment;

[0120] Figure 5 This is a flowchart illustrating the task processing method in another embodiment;

[0121] Figure 6 This is a user interface diagram of a task processing method in another embodiment;

[0122] Figure 7 A schematic diagram of the user interface settings for a local tool in one embodiment;

[0123] Figure 8 A schematic diagram of the user interface settings for a local tool in one embodiment;

[0124] Figure 9 A schematic diagram of the user interface settings for a local tool in one embodiment;

[0125] Figure 10 This is a flowchart illustrating a task processing method in one embodiment;

[0126] Figure 11 This is a system architecture diagram of a task processing method in one embodiment;

[0127] Figure 12 This is a structural block diagram of a task processing device in one embodiment;

[0128] Figure 13 This is a structural block diagram of the task processing device in another embodiment;

[0129] Figure 14 This is an internal structural diagram of a computer device in one embodiment;

[0130] Figure 15 This is a diagram of the internal structure of a computer device in another embodiment. Detailed Implementation

[0131] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0132] The task processing method provided in this application embodiment can be applied to, for example, Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store data that server 104 needs to process. The data storage system can be integrated onto server 104 or located in the cloud or on another server. Terminal 102 sends a query request to server 104, the query request including query content. Server 104 performs task planning on the query content, obtaining at least one task step. For the task step, if the execution of the task step depends on a target local tool of the corresponding target terminal 106, a call request for the target local tool is sent to the target terminal 106. Based on the call result corresponding to the call request returned by the target terminal 106, the processing result of the task step is determined. Based on the processing results corresponding to the at least one task step, the response content for replying to the query content is determined. Terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc. Alternatively, the terminal can be any hardware device, as long as it can perform specific functions, such as a browser terminal that enables browsing. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0133] In one embodiment, such as Figure 2 As shown, a task processing method is provided, which is applied to Figure 1 Taking server 104 as an example, the following steps are included:

[0134] Step S201: Obtain the sent query request.

[0135] The query request includes query content. Optionally, the query request may carry a terminal identifier and an account identifier, wherein the terminal identifier is used to uniquely identify the corresponding terminal, and the account identifier is used to uniquely identify the corresponding user account. The query content may include query content from any field where an answer is desired, such as resource interaction, office work, medical care, daily life, document editing, etc. For example, in the field of resource interaction, it could be "Which should I buy, phone A or phone B?" In the field of document editing, it could be "Extract screenshots from a PPT video and identify the title document." In the field of daily life, it could be "Please help me count the photos on my computer," or "Please help me retweet a Weibo post." In the field of office work, it could be "Please help me collect the latest viewpoints and data on 'Artificial Intelligence Trends in 2025,' create a simple PPT, and send it to our team's instant messaging group via instant messaging software A," etc. In one exemplary embodiment, the query content can be entered via text. In another exemplary embodiment, the query content can also be entered via voice. In a specific implementation, a query content input box can be displayed on the query content input page, and the query content can be entered in the query content input box as: "Please help me count the number of files on my computer desktop."

[0136] Step S203: Perform task planning on the query content to obtain at least one task step.

[0137] In the specific implementation process, the query content can be transmitted to an intelligent agent, which then parses the query content and performs task planning on the parsed content to obtain at least one task step to respond to the query content. The intelligent agent can be an agent built on a large language model, possessing the ability to autonomously understand, perceive, plan, remember, and use tools. In an exemplary embodiment, the intelligent agent can adopt a "thinking + action" reasoning mode when understanding the query content. Thinking refers to the intelligent agent analyzing the current situation and considering what to do next; action refers to the action performed by the intelligent agent, usually by calling a tool. During the intelligent agent's task planning phase, the intelligent agent performs a round of thinking beforehand, generating at least one task step to respond to the query content. In the specific implementation process, different query content can correspond to different prompt templates, and a prompt is constructed based on the prompt template. For example, when the query content is: "How tall is A?", the prompt may include: User: How tall is B? Intelligent agent thinking: I need to search for B's height; Intelligent agent action: Search online; Intelligent agent output: B's height. Now let's start processing the new task: the user's query content: what is Xiao A's height, etc., where the agent's actions can be the task steps for the above query content.

[0138] In an exemplary embodiment, the aforementioned intelligent agent can be an AI agent focused on task planning. The intelligent agent responsible for task planning is used to break down complex user needs, understand task objectives and contextual information, and dynamically formulate and adjust a distributed multi-agent collaborative execution plan. In this embodiment, the intelligent agent responsible for task planning can be obtained by training an initial intelligent agent with task steps. In an exemplary embodiment, to improve the accuracy of the initial task planning, prompts can be generated based on candidate task steps in the target domain. The prompts and query content are input to the intelligent agent responsible for task steps, and at least one corresponding task step is output. In a specific implementation, for example, in the domain of resource interaction, the query content is input to the intelligent agent responsible for task planning, and at least one corresponding task step is output, such as: analysis of the query object in the query content, recommendation of specific products of the query object, and comparison between various recommended products. The prompt can be input text provided by the intelligent agent system to the large language model in the intelligent agent, used to guide the large language model to produce the desired output or behavior.

[0139] Step S205: For the task step, if the execution of the task step depends on a target local tool on an associated terminal, send a request to the associated terminal to invoke the target local tool; the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal.

[0140] The terminal can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle systems, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Optionally, the terminal can also be any hardware device, as long as it can perform specific functions, such as a browser terminal that performs browsing functions.

[0141] The associated terminal refers to at least one of the terminals associated with the query request. In one exemplary embodiment, the query request can be sent by any of the associated terminals where the target local tool is located. For example, the terminals associated with the query request include terminal A, terminal B, and terminal C. In this particular query request, the associated terminals where the target local tool is located include terminal A and terminal B, and terminal A sends the query request. In another exemplary embodiment, the query request can also be sent by other associated terminals. For example, in the above example, terminal C can also send the query request.

[0142] The local tools can include tools installed on the terminal. These local tools can have various functions. For example, tools can be developed based on proprietary protocols to obtain custom local tools. Alternatively, tools can be developed based on public standard protocols. When using tools developed with public standard protocols, they can be downloaded and used from a platform tool store by searching for keywords such as "query train ticket information," "generate PPT," or "generate code." In a specific embodiment, local tools can be configured on the terminal, and the terminal downloads the corresponding tools based on the configuration. Optionally, a list of completed local tools can be displayed on the terminal, for example, referring to... Figure 3 As shown, when a local tool 301 is triggered, the local tool is automatically downloaded and installed on the terminal. This embodiment of the disclosure... Figure 3 The source of the local tools shown is not restricted. The target local tool can be one or more of the local tools.

[0143] Optionally, the query request may include an account identifier. By establishing the relationship between account information, terminal information, and local tool information, the terminal information associated with the account identifier and the local tool information corresponding to the local tools installed on the terminal can be determined. For example, account A has a mobile terminal, a computer terminal, and a vehicle terminal. The mobile terminal has instant messaging tools and video software tools, the computer terminal has PPT creation tools and workflow system tools, and the vehicle terminal has browser tools, navigation tools, audio playback tools, etc.

[0144] The local tool information may also include the parameters required to call each local tool. In specific implementations, different development methods may correspond to different tool information. For example, tool information may include the tool's identifier, input parameters, parameter types, and function descriptions. This disclosure does not limit the content of the tool information.

[0145] In one exemplary embodiment, the execution agent responsible for executing the task steps can determine the associated terminal and target local tool for executing the specific task steps based on local tool information corresponding to the query request. In another exemplary embodiment, the local tool information can be input to the agent responsible for the task steps, which can then filter target local tools that match the task steps based on the description of the local tool's function in the tool information.

[0146] It is understood that in at least one task step, some task steps need to be implemented by calling local tools, while others can be implemented by calling cloud tools. Therefore, in an exemplary embodiment, local tools or cloud tools that match the task steps can be selected based on the tool information of the cloud tools and the description of the functions of the cloud tools.

[0147] In an exemplary embodiment, for example, the execution of task step V depends on a local tool t1 on the mobile terminal z. The tool identifier and input parameters of local tool t1 can be queried, and the content corresponding to the task step can be used as the parameter content corresponding to the input parameters to obtain a request to invoke local tool t1. For example, the function description of local tool t1 is "open URL," and the input parameters include: url:. The task objective of the task step is: "Use local tool t1 to open the target website, the address is https: / / www.yyy.com." Based on the task objective and input parameters, a request to invoke can be obtained, for example: invoke local tool t1, the parameters are {"tool name: "local tool t1""; url: https: / / www.yyy.com}. Next, the above request to invoke is sent to the mobile terminal z. Optionally, the execution of task step B depends on a cloud tool t2 on the cloud. When executing task step B, cloud tool t2 is invoked to implement the processing of task step B. Optionally, the execution of task step N may also be implemented without relying on a tool. For example, task step N is a summary task, and the summary task processing is implemented by invoking a large language model. If task step M is relatively complex, it can be further decomposed to obtain at least one subtask corresponding to task step M.

[0148] Step S207: Based on the call result corresponding to the call request returned by the associated terminal, determine the processing result of the task step.

[0149] Specifically, the call result returned by the associated terminal where the target local tool is located can be used as the processing result of the task step. In one exemplary embodiment, for example, if the task step is to count the number of photos in folder F on the computer terminal, and the corresponding file counting tool counts one thousand photos, then one thousand photos are used as the processing result of the task step. In the specific implementation process, the call result may return an error, such as an error in the execution of the target local tool, or the target local tool not being found. Exemplarily, the call request can be resent. Another exemplary embodiment, a waiting delay can be set; for example, if the call result is not received from the associated terminal within 5 minutes, failure processing can be performed, and the call request can be resent. In another exemplary embodiment, if the number of times the task step fails to execute meets a preset requirement, the task step can be replanned.

[0150] Step S209: Based on the processing results corresponding to the at least one task step, determine the response content to reply to the query content.

[0151] In this embodiment of the disclosure, some of the at least one task steps may be implemented by calling cloud tools during execution, and the results of the cloud tool calls are obtained. Exemplarily, some of the at least one task steps may not require calling tools, for example, they can be obtained through model inference or by querying a database. Another example is that some task steps are implemented by calling local tools, and the results of the local tool calls are obtained.

[0152] In this embodiment of the disclosure, each task step in the at least one task step corresponds to a processing result. For example, an intelligent agent can summarize the at least one processing result, such as extracting a summary of the at least one processing result and outputting it according to a preset output format. For example, a report corresponding to the response content can also be generated and downloaded. In a specific embodiment, for example, the user's query content includes: "Check the weather, please book train tickets for next week on the ticketing system." The at least one task step includes: Task step a: Check the cloud weather tool; the corresponding query result is sunny and suitable for travel. Task step b: Use the local ticketing tool to book tickets on the ticketing system; the booking result is successful ticket booking. The processing results of each task step are summarized to obtain a response content such as: "Sunny and suitable for travel; I have booked your train ticket for tomorrow; ticket information is as follows," etc.

[0153] In the above task processing method, for the sent query request, the server performs task planning on the query content to obtain at least one task step, breaking down complex query content into task steps and solving each task step one by one. Optionally, if the task steps are more complex, they can be further broken down. Therefore, this application can handle relatively complex query content. Furthermore, when executing task steps, in addition to calling cloud tools, more importantly, it can call various associated terminals related to the query request and the target local tools of each associated terminal, making full use of the rich localization capabilities of each associated terminal, such as calling professional software, accessing hardware sensors, and processing local data. In this process, the advantage that each target local tool can access the corresponding terminal environment is fully utilized, greatly expanding the capability boundary for solving large-scale model tasks, enabling the solution of extremely complex tasks involving cross-terminal collaboration, and possessing heterogeneous dedicated capabilities. Furthermore, by sending a call request for the target local tool to the associated terminal; based on the call result corresponding to the call request returned by the associated terminal, the processing result of the task step is determined. This approach eliminates the need to access end-user data, requiring only the retrieval of the call result, thus protecting end-user data security. Furthermore, it enables comprehensive utilization of data across various associated devices, achieving cross-terminal resource aggregation. Additionally, when a new terminal is added to an account, it can be aggregated into the existing terminal list corresponding to that account, demonstrating strong scalability and dynamically increasing the system's capacity ceiling.

[0154] In one embodiment, the task planning for the query content includes:

[0155] The query request is parsed to determine the target account information corresponding to the query request.

[0156] Based on the association between account information, terminal information and local tool information, at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information are determined.

[0157] Task planning is performed on the query content based at least on the at least one terminal information and the at least one local tool information.

[0158] The query request includes target account information, which may include an account identifier. The query request may also include the identifier of the terminal where the account is logged in. Optionally, the query request may also include information about local tools installed on the terminal receiving the query request or updated local tool information.

[0159] In the specific implementation process, local tool information can be received and reported, establishing associations between account information, terminal information, and local tool information. For example, account A has terminals a, b, and c. Terminal a has tools t1, t2, and t3 installed, terminal b has tools t2 and t4 installed, and terminal c has tools t5, t6, and t7 installed. Account B has terminals d, f, and e. Terminal d has tools t1, t3, and t4 installed, terminal f has tools t4 and t5 installed, and terminal e has tools t1, t3, and t7 installed. For example, the corresponding terminal information and the local tool information corresponding to each terminal are indexed through the account identifier in the target account information.

[0160] In one exemplary embodiment, the query content can be task-planned according to the method described in any of the above embodiments to obtain at least one task step. The local or cloud tools on which the execution of the task step depends are determined. Similarly, if the task step is complex, it can be further decomposed to obtain at least one sub-task.

[0161] In the above embodiments, by leveraging the association between account information, terminal information, and local tool information, at least one piece of local tool information reported by various terminals under the same account name is received, and dynamically maintained and updated. This enhances the heterogeneous and dedicated capabilities of multi-device AI agents centered on accounts, giving the system elastic scalability similar to a "personal computing cluster." Each device executes the task steps it is best at, and the results are aggregated, greatly improving the efficiency of completing complex queries.

[0162] In one embodiment, after performing task planning on the query content to obtain at least one task step, the method further includes:

[0163] For each task step, a target execution agent whose functional scope matches the task step is determined from the agent resource pool, thus obtaining the target execution agent corresponding to the task step.

[0164] In the case where the execution of the task step depends on a target local tool on the associated terminal, sending a request to invoke the target local tool to the associated terminal includes:

[0165] Based on the target execution agent corresponding to the task step, the target terminal information and target local tool information corresponding to the task step are determined from at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information; wherein, the target account information is obtained by parsing the query request.

[0166] If the target terminal information corresponds to an associated terminal and the target local tool information corresponds to a target local tool, a request to invoke the target local tool is sent to the associated terminal.

[0167] In the specific implementation process, for each task step, candidate agents whose functional scope matches the task step are determined from the agent resource pool. The function and purpose of each agent in the agent resource pool are clearly defined; for example, agent A is responsible for domestic weather queries, and agent B is responsible for global weather queries. For instance, if the task step is to query tomorrow's train ticket information, agents with retrieval capabilities should be selected from the agent resource pool to form candidate agents. Examples include candidate agents C1, C2, and C3. Optionally, based on their functional scope, agents can also be divided into agents that execute standard protocol tools and agents that cannot execute standard protocol tools. It is understood that agents executing standard protocol tools can be further divided into agents executing cloud-based standard protocol tools and agents executing terminal-based standard protocol tools.

[0168] In an exemplary embodiment, capability dimension information of each candidate execution agent is obtained. Optionally, the capability dimension information of the agent can be updated in real time or periodically to ensure the reliability of the information. Exemplarily, capability dimension information includes accuracy, load, domain, etc. It should be noted that the setting method of the capability dimension information is not limited to the above examples. For example, agent input / output support, response latency, context window, multimodal capabilities, etc., can also be used as the preset trigger event. Those skilled in the art may make other changes based on the essence of this application, but as long as the implemented function and effect are the same as or similar to this application, they should all be covered within the scope of protection of this application. In one possible implementation, one of the capability dimension information can be selected as a filtering condition according to task needs; for example, the target execution agent with higher accuracy is selected first. In another possible implementation, the various capability dimension information can be weighted and summed to obtain the comprehensive capability of the agent, and the target execution agent with higher comprehensive capability is selected.

[0169] In an exemplary embodiment, at least one terminal information corresponding to the target account information, and at least one local tool information corresponding to each terminal information, can be provided to the target execution agent to facilitate the target execution agent in determining the target local tool for executing the corresponding task step from multiple local tool information. Optionally, cloud tool information can also be provided to the target execution agent to facilitate the target execution agent in selecting from it. For example, the content of the task step is to query the latest event information from the Internet. The target execution agent discovers that search engine A is installed on the terminal a associated with the target account, and search engine B is installed on the associated terminal b. It further discovers that search engine A is online, thereby determining that the target terminal information is associated terminal a and the target local tool information is search engine A. Then, a request to call search engine A is sent to associated terminal a.

[0170] In the above embodiments, for each task step, a target execution agent whose functional scope matches the task step is determined from the agent resource pool, thus obtaining the target execution agent corresponding to the task step. By describing the functional scope and capability dimension information of the agent, the target execution agent that best matches the task step can be selected, improving the success rate and efficiency of task step execution. Next, based on the target execution agent corresponding to the task step, target terminal information and target local tool information corresponding to the task step are determined from at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information, limiting the tool selection range of the target execution agent and improving the task execution accuracy of the target execution agent.

[0171] In one embodiment, determining the target terminal information and target local tool information corresponding to the task step from at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information, based on the execution agent corresponding to the task step, includes:

[0172] The target prompt is determined based on at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information.

[0173] The target prompt is input into the target execution agent corresponding to the task step, and the target terminal information and target local tool information corresponding to the task step are output.

[0174] In the specific implementation process, determining the target terminal information and target local tool information corresponding to the task steps can be achieved through prompts. The prompt can be input text provided by the intelligent agent system to the large language model within the agent, guiding the large language model to produce the desired output or behavior. In an exemplary embodiment, the target prompt might be, for example: "Target account information includes terminal information ID1, terminal information ID2, and terminal information ID3, where terminal information ID1 corresponds to a browser tool and an instant messaging tool; terminal information ID2 corresponds to a weather query tool, a ticket information query tool, and a mini-program tool; and terminal information ID3 corresponds to an office software creation tool and a text-to-image tool." It can be understood that the target prompt and corresponding task steps are input to the target execution intelligent agent, which then calls the large language model to output the target terminal information and target local tool information corresponding to the task steps.

[0175] In the above embodiments, the target prompt is determined based on at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information. The target prompt is input to the target execution agent corresponding to the task step, and the target terminal information and target local tool information corresponding to the task step are output. This eliminates the need to adjust the parameters of the large model, improving the accuracy of determining the target terminal information and target local tool information.

[0176] In one embodiment, determining the target terminal information and target local tool information corresponding to the task step includes:

[0177] Identify candidate local tools whose functional scope satisfies the task steps, and query the online status information of the candidate local tools.

[0178] Based on the online status information, the target local tool and the associated terminal where the target local tool is located are determined from the candidate local tools.

[0179] Determine the target local tool information corresponding to the target local tool and the target terminal information corresponding to the associated terminal.

[0180] In one exemplary embodiment, the task steps include, for example, "searching for the latest football matches," where candidate local tools with search functionality include search engines a and b on a mobile terminal; search engine c on a tablet terminal; and search engine d on a specific device terminal. In one exemplary embodiment, instructions can be sent to the mobile terminal, tablet terminal, and specific device respectively. If no response instruction is received within a preset time, it can be determined that the local tool on the corresponding device is offline. In one exemplary embodiment, if a response instruction is received, it is determined that the corresponding terminal device is online.

[0181] In one exemplary embodiment, for example, if a search engine a on a mobile terminal is found to be online, while the search engines on a tablet terminal and a specific device are not online, then the search engine a on the mobile terminal is determined as the target local tool.

[0182] In the above embodiments, the online status information of the candidate local tools is queried. Based on the online status information, the target local tool is determined from the candidate local tools. This ensures that the selected target local tool is online and can be invoked.

[0183] In one embodiment, determining the target terminal information and target local tool information corresponding to the task step includes:

[0184] Identify candidate local tools whose functional scope satisfies the task steps, and query the performance information and historical execution result accuracy of the terminal where the candidate local tools are located.

[0185] Based on the performance information and the accuracy of the execution results, the target local tool and the associated terminal where the target local tool is located are determined from the candidate local tools.

[0186] Determine the target local tool information corresponding to the target local tool and the target terminal information corresponding to the associated terminal.

[0187] In one exemplary embodiment, the task steps include, for example, “searching for the latest football matches,” wherein candidate local tools with search functionality include search engines a and b on a mobile terminal; search engine c on a tablet terminal; and search engine d on a specific device terminal.

[0188] In one exemplary embodiment, performance information of mobile terminals, tablet terminals, and specific device terminals is queried, such as memory size, CPU utilization, and whether GPU is supported, and the associated terminal with the highest performance information is selected. In another exemplary embodiment, the historical execution result accuracy of each candidate local tool is queried, and the target local tool with the highest execution result accuracy is selected. Optionally, the number of times each candidate local tool is called is queried, and the candidate local tool with the highest number of calls is selected as the target local tool. For example, when multiple selection dimensions are included, a corresponding weight is determined for each selection dimension to comprehensively determine the target local tool. For example, terminal performance information is represented as X, historical execution result accuracy as Y, and the number of calls as Z, with corresponding weight coefficients a, b, and c, respectively. The comprehensive score of the candidate local tool can be represented as a×X+b×Y+c×Z. It is understood that the online status information of the candidate local tools can also be combined to select online candidate local tools.

[0189] In the above embodiments, the performance information and historical execution result accuracy of the terminal where the candidate local tools are located are queried. Based on the performance information and the execution result accuracy, the target local tool and the terminal where the target local tool is located are determined from the candidate local tools. This allows for the selection of target local tools with a high success rate. In one embodiment, the step of performing task planning on the query content to obtain at least one task step includes:

[0190] The query content is decomposed into tasks to obtain at least one task objective.

[0191] For the stated task objective, at least based on the tool information corresponding to the query request, a target local tool whose functional scope satisfies the stated task objective is determined.

[0192] Based on the task objective, the tool identifier of the target local tool, and the terminal identifier of the associated terminal where the target local tool is located, at least one task step is determined.

[0193] In this context, the task objective describes the task content of the task step. In this embodiment, the query content can be broken down or planned into at least one task objective according to the method described in the above embodiments. In an exemplary embodiment, the query content is, for example: "Help me collect the latest viewpoints and data on 'Artificial Intelligence Trends in 2025,' create a simple PPT, and send it to our team's instant messaging group via instant messaging software 'a'." Exemplarily, the agent of the task step plans the query content into the following three task objectives: Task objective a: "Match the web_search_and_summarize tool of a browser to obtain the latest authoritative information from the internet." Task objective b: "Match the ppt_generation tool of a PC to convert the collected structured information into PPT format." Task objective c: "Match the wechat_file_sending tool of a mobile phone to send the final result to the target group via instant messaging software 'a'."

[0194] The tool information corresponding to the query request may include at least one terminal information corresponding to the target account information, and at least one local tool information corresponding to each terminal information. The target account information can be parsed from the query request. In an exemplary embodiment, for the content of the task objective, a target tool whose functional scope matches the task objective can be determined from the tool's functional description. The target tool can be a cloud tool or a local tool. The tool information of a cloud tool can be stored on a server. Optionally, if multiple tools simultaneously satisfy the same task objective, the tool with the highest performance can be selected from the multiple tools, or an online target tool can be selected. Exemplarily, the task objective and the tool identifier for executing the task objective are written into the task steps to obtain the task steps. When the target tool is a target local tool, the terminal identifier can also be written into the corresponding task steps. It is understood that if the execution of the task steps does not depend on a tool, the tool identifier and terminal identifier may not be filled in.

[0195] In the above embodiments, the query content is decomposed into tasks to obtain at least one task objective. For each task objective, a target local tool whose functional scope meets the task objective is determined based on at least the tool information corresponding to the query request. Based on the task objective, the tool identifier of the target local tool, and the terminal identifier of the terminal where the target local tool is located, at least one task step is determined. The tool identifier and terminal identifier are written into the corresponding task step to facilitate the selection of a matching target tool during task step execution. This allows the planning end to focus on planning and the execution end to focus on execution, simplifying the execution process and improving processing efficiency.

[0196] In one embodiment, determining at least one task step based on the task objective, the tool identifier of the terminal local tool, the input parameters of the terminal local tool, and the corresponding parameter content includes:

[0197] If the target local tool is a standard protocol tool, obtain the identification information of the agent executing the standard protocol tool.

[0198] Based on the task objective, the tool identifier of the target local tool corresponding to the task objective, the terminal identifier of the target local tool, and the identification information of the agent executing the standard protocol tool, at least one task step is determined.

[0199] The tools involved in this application may include cloud tools or target local tools. The identification information of the intelligent agent may include intelligent agents executing standard protocol tools and intelligent agents executing non-standard protocol tools. In an exemplary embodiment, if a standard protocol tool is used, an intelligent agent executing the standard protocol tool is used. The identification information of the intelligent agent executing the standard protocol tool, such as "General ReACT," is written into the corresponding task steps. In an exemplary embodiment, if a non-standard protocol tool is used, the identification information of the intelligent agent executing the non-standard protocol tool, such as "Websearch," is written into the corresponding task steps. It should be noted that an intelligent agent executing a standard protocol tool can call either a cloud tool or a target local tool, and an intelligent agent executing a non-standard protocol tool can call either a cloud tool or a target local tool. This disclosure does not impose any limitations on this.

[0200] In the above embodiments, based on the task objective, the tool identifier of the target local tool corresponding to the task objective, the terminal identifier of the target local tool, and the identifier information of the intelligent agent executing the standard protocol tool, corresponding task steps are written. During the execution of the task steps, the scheduler can automatically route the execution to the corresponding intelligent agent. Different tools correspond to different intelligent agents, which improves the professionalism of the intelligent agent in completing the task, saves processes, and improves execution efficiency and accuracy.

[0201] In one embodiment, prior to sending the invocation request for the target local tool to the associated terminal, the method further includes:

[0202] If the target local tool is a standard protocol tool, the input parameters of the target local tool are queried from the target local tool information corresponding to the target local tool.

[0203] Based on the task steps corresponding to the target local tool, the parameter content of the input parameters is determined.

[0204] Based on the input parameters of the target local tool and the corresponding parameter content, the invocation request of the target local tool is determined.

[0205] Standard protocol tools can include tools developed using open, shared protocols, and these tools must conform to the aforementioned protocol standards. For example, MCP (Model Context Protocol) is an open communication protocol standard designed to provide a standardized and secure interaction framework between artificial intelligence models and external tools, data sources, and services. These services can include, for example, databases, applications, and calculators. Various tools can run on different user terminals. As long as developers adhere to the MCP protocol, the tool will operate according to a standard, and large language models can utilize these tools to execute tasks and complete instructions.

[0206] In this embodiment, the MCP server is a program or service running on the terminal or server, responsible for providing access interfaces to local tools and resources available on the terminal device, or access interfaces to cloud tools and resources. The MCP client can be a program or service running on the server, responsible for actively discovering connections and calling the tools and resources provided by the MCP server on the terminal device, or calling the tools and resources provided by the MCP server on the server.

[0207] In one exemplary embodiment, after completing the development of a standard protocol tool, the developer can agree on the functions implemented by the tool, the tool's calling interface, the tool's identification information, and the tool's input parameters. The tool's input parameters describe the parameters required to call the tool, such as parameter URL: [ ]; parameter type: [ ], etc. In one exemplary embodiment, the parameter content is determined based on the task steps. For example, if the task step is to search for the homepage of Company A, then the parameter content is exemplarily: "https: / / www.companyA.com".

[0208] In one exemplary embodiment, the invocation request of the target local tool is determined based on the input parameters of the target local tool and the parameter content corresponding to the input parameters.

[0209] In the above embodiments, standard protocol tools are used to expand the scope of tool access. This allows the use of publicly available tools from third-party standard protocol tools, lowers the barrier to entry for tool access, and lays the technical foundation for building a user-led, device-independent personal distributed computing ecosystem, which has enormous platform potential.

[0210] In one embodiment, after performing task planning on the query content to obtain at least one task step, the method further includes:

[0211] For the task steps, if the task steps meet preset conditions, the task steps that meet the preset conditions are decomposed into at least one subtask. The subtask tool information corresponding to the matching of the at least one subtask is at least one type of tool information.

[0212] In the actual implementation process, the task steps are quite complex. For example, if the answer cannot be obtained by calling tools or relying on knowledge base content, it is necessary to decompose the task steps into at least one subtask. Understandably, if the subtask is still quite complex, it can be further decomposed into at least one second-level subtask, until the decomposed subtask can be processed.

[0213] In this embodiment of the disclosure, when a subtask can be processed by calling a tool, the subtask carries corresponding subtask tool information. The subtask tool information can be tool information corresponding to a local tool, or tool information corresponding to a cloud tool, etc. Optionally, processing the subtask can be achieved by calling multiple local tools, or it can be achieved by calling one local tool; this embodiment of the disclosure does not impose any limitations on this.

[0214] In the above embodiments, when the task steps meet the preset conditions, the task steps that meet the preset conditions are decomposed into at least one sub-task, which realizes the layer-by-layer decomposition of complex query content, with each device focusing on executing the corresponding task steps or sub-tasks, and finally aggregating the results, which greatly improves the efficiency and accuracy of completing complex tasks.

[0215] In one embodiment, after decomposing the task steps that meet preset conditions to obtain at least one sub-task, the method further includes:

[0216] For the subtask, if the execution of the subtask depends on the target local tool of the corresponding associated terminal, a subtask invocation request of the target local tool is sent to the associated terminal;

[0217] Based on the call result corresponding to the call request of the subtask returned by the associated terminal, the processing result of the subtask is determined;

[0218] Based on the processing results of the at least one sub-task, the processing results of the task steps that meet the preset conditions are determined.

[0219] In one exemplary embodiment, for example, if the execution of a subtask depends on a target local tool a1 of the associated terminal a, the input parameters required by the target local tool a1 can be queried from the tool information, and the content of the subtask can be used as the parameter content corresponding to the input parameters. The input parameters and parameter content are encapsulated to obtain a call request. Based on the call result corresponding to the call request of the subtask returned by the associated terminal a, the processing result of the subtask is determined. In one exemplary embodiment, for example, if the execution of a subtask depends on a target local tool b1 of the associated terminal b, the input parameters required by the target local tool b1 can be queried from the tool information, and the content of the subtask can be used as the parameter content corresponding to the input parameters. The input parameters and parameter content are encapsulated to obtain a call request. Based on the call result corresponding to the call request of the subtask returned by the associated terminal b, the processing result of the subtask is determined. In one exemplary embodiment, for example, if the execution of a subtask corresponds to a cloud tool, the subtask is processed by calling the cloud tool to obtain the processing result of the subtask. Finally, based on the processing result of each subtask, the processing result corresponding to the task step that meets the preset conditions is determined.

[0220] In the above embodiments, the processing result of the subtask is determined by the call result corresponding to the call request of the subtask returned by the associated terminal. Based on the processing result of at least one subtask, the processing result of the task step that meets the preset conditions is determined. This achieves the processing of complex subtasks.

[0221] In one embodiment, when the execution of the task step depends on a target local tool on the associated terminal, sending a request to invoke the target local tool to the associated terminal includes:

[0222] When the execution of the task step depends on the target local tool of the associated terminal by the intelligent agent executing the standard protocol tool, the terminal identifier of the associated terminal corresponding to the task step and the tool identifier corresponding to the target local tool are encapsulated in the data format of the standard protocol to obtain the invocation request corresponding to the task step.

[0223] The call request is sent to the associated terminal.

[0224] In the specific implementation process, it can be determined whether the tool corresponding to the task tool information of the task step is a local tool or a cloud tool by querying. In this embodiment of the disclosure, the local tool adopts a standard protocol tool, and the agent that calls the local tool selects the agent that executes the standard protocol tool. By executing the agent that executes the standard protocol, the input parameters of the target local tool corresponding to the task step are written into the call request. For example, the tool identifier being called is puppeteer_navigate, and the input parameters of the target local tool are specified. Thus, the call request can be represented as: {"tool_name": "puppeteer_navigate", "url": "https: / / www.yyy.com"}.

[0225] In one exemplary embodiment, the input parameters of the target local tool in the task step can be input into the agent executing the task step as prompt words, and the agent executing the task step is required to output in a specified JSON format, such as: {tool_name:xxx, input:xxx}, and finally these fields are parsed by the program.

[0226] In the above embodiments, standard protocol local tools are employed. These standard protocol local tools can be user-defined, obtained from third-party platforms, or downloaded from the tool set applied in this application, thus providing more tool sources and enriching the variety of local tools. Furthermore, the intelligent agent executing the standard protocol tool can directly output a call request that satisfies the standard protocol, improving the convenience of tool invocation and the execution efficiency of task steps. Each terminal can report tool information to the server registry center, completely breaking down resource silos.

[0227] In one embodiment, sending the invocation request for the target local tool to the associated terminal includes:

[0228] Obtain the terminal identifier corresponding to the associated terminal and the account information corresponding to the query request.

[0229] Determine the long-connection communication interface corresponding to the terminal identifier and the account information.

[0230] Based on the long-connection communication interface, a request to invoke the target local tool is sent.

[0231] In the specific implementation, when a terminal sends a query request to the server, the server obtains the account information from the query request, including an account identifier, and establishes a corresponding long-connection communication interface. The server can also proactively send connection requests to each terminal under the aforementioned account information. Therefore, an association between the account identifier, terminal identifier, and long-connection communication interface can be established. When it is necessary to send a message to an associated terminal, the long-connection communication interface corresponding to the terminal identifier din and account identifier uin of the associated terminal is determined. Optionally, the aforementioned long-connection communication interface can be a WebSocket connection. Optionally, when the local tool is a standard protocol tool, the association between the account identifier, terminal identifier, and long-connection communication interface can be maintained through the MCP Transport Svr.

[0232] In the actual implementation, when a terminal (e.g., JavaScript code in a browser) executes `newWebSocket("ws: / / server.example.com")`, it sends a special HTTP request to the server. The key to this request is that it includes a header requesting an upgrade protocol. Upon receiving the request, the server performs a series of checks, such as whether WebSocket is supported. If everything is normal, the server returns an HTTP 101 SwitchingProtocols status code response, indicating a successful protocol switch. Once the terminal verifies that the server's `Sec-WebSocket-Accept` response is correct, the TCP connection is not closed, but rather formally upgraded from the HTTP protocol to the WebSocket protocol. At this point, the application layer protocol has changed. Both parties no longer use the HTTP request / response model, but instead use frames defined by the WebSocket protocol for full-duplex communication.

[0233] In the above embodiments, the terminal identifier corresponding to the associated terminal and the account information corresponding to the query request are obtained; a long-connection communication interface corresponding to the terminal identifier and the account information is determined; and a call request for the target local tool is sent based on the long-connection communication interface. This enables full-duplex communication, facilitates re-initiating the call or switching to a new task step when the call fails, reduces the time consumption of traditional request-response methods, and improves the execution efficiency of local tool calls.

[0234] In one embodiment, before sending the invocation request for the target local tool based on the long-connection communication interface, the method further includes:

[0235] Obtain the account information corresponding to the query request.

[0236] Determine the terminal information corresponding to at least one terminal that corresponds to the account information.

[0237] Establish long-connection communication interfaces for the terminals corresponding to the account information and the at least one terminal information, respectively.

[0238] The account information may include an account identifier and other relevant information, such as a communication number and associated login account. In an exemplary embodiment, at least one piece of terminal information corresponding to the account identifier can be queried from the association between account information, terminal information, and local tool information. The terminal information may include a terminal identifier.

[0239] In an exemplary embodiment, the server may proactively send long connection establishment requests to each terminal and maintain the association between account identifier (uin, user identify number), terminal identifier (din, device identify number) and long connection communication interface.

[0240] In the above embodiments, by establishing a long-connection communication interface with the terminal corresponding to the account information and the at least one terminal information, accurate communication between the server and various heterogeneous login devices under the account name can be achieved, laying the foundation for cross-device tool invocation.

[0241] In one embodiment, before performing task planning on the query content to obtain at least one task step, the method further includes:

[0242] Obtain the registration information of local tools reported by each terminal; wherein, the registration information includes account information, terminal information and local tool information.

[0243] Based on the registration information, the relationships between account information, terminal information, and local tool information are determined.

[0244] In the specific implementation process, each terminal can include any of the terminals in the above embodiments, such as computer terminals, mobile phone terminals, and tablet terminals. Taking a mobile phone terminal as an example, the required local tools can be imported from the tool list in the application interface of the query request, or custom local tools can be created based on tool creation software. If the local tool is a standard protocol tool, the input parameters corresponding to the standard protocol tool can also be configured from a third-party platform, and the standard protocol tool can be downloaded to the mobile phone terminal. In an exemplary embodiment, the terminal can report locally installed local tools, such as tool identifiers and input parameters, thereby forming registration information. For example, a browser terminal reports a web search and information summary tool, a computer terminal reports an office document generation and editing tool, and a mobile phone terminal reports an instant messaging file sending tool.

[0245] In the above embodiments, registration information of local tools reported by each terminal is obtained; wherein, the registration information includes account information, terminal information, and local tool information. Based on the registration information, the association between account information, terminal information, and local tool information is determined. In this way, users do not need to purchase more expensive single devices; by increasing the number of logged-in devices, new tools and computing power are linearly injected into their entire AI system. The unique capabilities of new devices are automatically registered to the server and immediately used by subsequent servers, giving the system elastic scalability similar to a "personal computing cluster."

[0246] In one embodiment, determining the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal includes:

[0247] If the call result corresponding to the call request returned by the associated terminal is a call failure, the target local tool is updated to a backup local tool;

[0248] Based on the processing results of the backup local tool, the processing result of the task step is determined.

[0249] In the specific implementation process, the call result of the associated terminal may be a call failure, such as an error in the execution of the target local tool or the target local tool not being found. In an exemplary embodiment, the task tool information of the task step may include the preferred target local tool 'a', and may also include selecting backup local tools 'b' and 'c' if target local tool 'a' fails. In the specific implementation process, the processing result of the task step can be determined based on the processing result of the backup local tools. In a specific application scenario, for example, the target local tool uses browser 'a' to search for the target event. If the search using browser 'a' fails, browser 'b' can be used to search for the target event.

[0250] If the call result corresponding to the call request returned by the associated terminal is a call failure, the target local tool is updated to a backup local tool; based on the processing result of the backup local tool, the processing result of the task step is determined. This can improve the success rate of task step execution.

[0251] In one embodiment, after updating the execution of the target local tool to an alternate local tool, the method further includes:

[0252] If the backup local tool fails to be invoked, the query content is re-planned to obtain at least one updated task step.

[0253] The step of determining the response content based on the processing results corresponding to the at least one task step includes:

[0254] Based on the processing results corresponding to the at least one updated task step, the response content for replying to the query content is determined.

[0255] In the specific implementation process, refer to Figure 4 As shown, during the execution of task step 401, multiple tools can be invoked, such as local tool 403 and local tool 405. If execution fails, the local tool is replaced with a backup local tool, and task step 401 is re-executed. If execution fails again, this embodiment of the disclosure can return to the previous step, and the task planning agent can re-formulate the task steps to obtain updated task steps.

[0256] In the specific implementation process, the task steps corresponding to the query content may include task step A, task step B, and task step C. In an exemplary embodiment, all task steps A, B, and C can be updated, or only some of the task steps can be updated, such as updating task step A, or updating task steps A and B.

[0257] In one exemplary embodiment, when only some task steps are updated, the response content for the query is determined based on the processing results of the partially updated task steps. In another exemplary embodiment, when all task steps are updated, the response content for the query is determined based on the processing results of all updated task steps.

[0258] In the above embodiments, if the backup local tool fails to be invoked, the query content is re-planned to obtain at least one updated task step. Based on the processing results corresponding to each of the at least one updated task step, the response content to the query content is determined. In this way, by first using local remediation—for example, selecting a backup local tool to execute task steps if the local tool fails—and then deploying a large-scale task replanning if the backup tool fails, the success rate of the response content can be comprehensively improved.

[0259] In one embodiment, the task planning for the query content to obtain at least one task step includes:

[0260] Retrieve historical query content that matches the query request from the experience knowledge base;

[0261] If a historical query that matches the query request is found, the historical task steps corresponding to the historical query are obtained; based on the historical task steps, at least one task step to reply to the query is determined.

[0262] If no historical query content matching the query request is found, task planning is performed on the query content to obtain at least one task step. The experience knowledge base may include a database storing historical query content and corresponding task steps. Optionally, the experience knowledge base may also include the execution results of the task steps. In an exemplary embodiment, the experience knowledge base may include a context manager for managing the short-term working memory of the current session, responsible for continuously updating the context window of the dialogue to achieve coherent multi-turn dialogue. In another exemplary embodiment, the experience knowledge base may also include a memory for persistently storing task steps related to the query content corresponding to an account identifier, thereby extracting information of interest corresponding to that account identifier, such as being more interested in lively answers. Optionally, the memory may also store tools of interest corresponding to the account identifier, updating and replacing local tools on a computer terminal with those on a mobile terminal, etc., providing technical support for subsequent task steps.

[0263] In the specific implementation process, historical query content matching the query request can be queried from the context manager first. For example, the historical query content is similar to the query content in the query request. If historical query content matching the query request is found, the historical task steps corresponding to the historical query content are obtained. Optionally, the execution success rate of the historical task steps can be selected. If the execution success rate reaches a preset requirement, the task steps for the current query content can be determined based on the historical task steps. In an exemplary embodiment, if no historical query content matching the query request is found, task planning is performed on the query content to obtain at least one task step.

[0264] In the above embodiments, an experience knowledge base is established, and historical query content matching the query request is queried from the experience knowledge base. If historical query content matching the query request is found, the historical task steps corresponding to the historical query content are obtained; based on the historical task steps, at least one task step to reply to the query content is determined. The experience knowledge base helps the large language model to autonomously analyze and read, enabling it to formulate task steps more accurately and quickly.

[0265] In one embodiment, the associated terminal includes at least one of a mobile terminal, a computer terminal, and a browser terminal. When the execution of the task step depends on a target local tool on the corresponding associated terminal, sending a request to the associated terminal to invoke the target local tool includes:

[0266] If the execution of the task step depends on the target local tool of the corresponding mobile terminal, a request to invoke the target local tool is sent to the mobile terminal.

[0267] If the execution of the task step depends on a target local tool on the corresponding computer terminal, a request to invoke the target local tool is sent to the computer terminal.

[0268] If the execution of the task step depends on the target local tool of the corresponding browser terminal, a request to invoke the target local tool is sent to the browser terminal.

[0269] Among them, the local tools on mobile terminals run on the mobile terminal, the local tools on computer terminals run on the computer, and the local tools on browser terminals run on any operating system on which the browser is located, without any hardware device restrictions.

[0270] In one exemplary embodiment, when the execution of the task step depends on a target local tool on a corresponding mobile terminal, a corresponding long-connection communication interface is determined based on the account identifier and the terminal identifier of the mobile terminal, for example, long-connection communication interface a, and a request to invoke the target local tool is sent to the mobile terminal based on long-connection communication interface a. When the execution of the task step depends on a target local tool on a corresponding computer terminal, a corresponding long-connection communication interface is determined based on the account identifier and the terminal identifier of the computer terminal, for example, long-connection communication interface b, and a request to invoke the target local tool is sent to the computer terminal based on long-connection communication interface b. When the execution of the task step depends on a target local tool on a corresponding browser terminal, a corresponding long-connection communication interface is determined based on the account identifier and the terminal identifier of the browser terminal, for example, long-connection communication interface c, and a request to invoke the target local tool is sent to the browser terminal based on long-connection communication interface c.

[0271] In the above embodiments, the associated terminals include mobile terminals, computer terminals, and browser terminals, etc. Based on the global "device-capability" view, the server accurately distributes the task steps to the corresponding dedicated capability devices for execution, thereby improving the accuracy of task execution.

[0272] In one embodiment, determining the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal includes:

[0273] Based on the invocation result of the first local tool returned by the first terminal, the processing result of the first task step is determined; wherein, the first task step includes using the first local tool to search for information from the Internet;

[0274] Based on the invocation result of the second local tool returned by the second terminal, the processing result of the second task step is determined; wherein, the second task step includes using the second local tool to convert the searched information into a target format file;

[0275] Based on the invocation result of the third local tool returned by the third terminal, the processing result of the third task step is determined; wherein, the third task step includes using the third local tool to send the target format file to the communication account via an instant messaging tool.

[0276] In the specific implementation process, the query content could be: "Help me collect the latest viewpoints and data on 'Artificial Intelligence Trends in 2025' and create a simple PowerPoint presentation." The agent responsible for the task steps understands the above query content and formulates the following task steps: First task step, for example, using the browser tool of the first terminal (browser terminal) to search for information on the Internet; Second task step, using the slide creation tool of the second terminal (computer terminal) to convert the searched information into a slide format; Third task step, using the mobile terminal to send the slide to the communication account via an instant messaging tool.

[0277] Optionally, the server call center schedules and executes each task step separately. Specifically, the browser tool on the browser terminal searches multiple authoritative technology media and research institution websites to retrieve key information and data, generating a structured content summary, thus obtaining the processing result of the first task step. The computer terminal's slide generator is then used, taking the structured summary returned by the browser terminal as input, to automatically generate a PPT document containing a title, table of contents, core viewpoints, and data charts. This yields the processing result of the second task step. Finally, the mobile terminal's instant messaging tool sends the PPT document generated on the computer terminal to a designated communication group, along with a message: "This is a newly compiled 2025 AI trend report; please refer to it." Optionally, the execution status and results of the entire task step can be notified via the mobile terminal or computer terminal.

[0278] In the above embodiments, by invoking multiple heterogeneous devices, such as mobile terminals, computer terminals, and browser terminals, the system shields all the technical complexities of cross-device invocation. Users do not need to manually switch between different devices and applications; they can mobilize all their devices to work together with just a single natural language command. This greatly lowers the user threshold and provides an unprecedented level of convenience and smoothness.

[0279] In one embodiment, such as Figure 5 As shown, a task processing method is provided. Taking the application of this method to an associated terminal as an example, the method includes the following steps:

[0280] Step S501: Obtain the query content and determine the query request based on the query content.

[0281] The query content can include queries from any area where an answer is expected, such as resource interaction, office work, medical care, daily life, document editing, etc. For example, the terminal identifier and account identifier are obtained, and a query request is determined based on the query content and the corresponding terminal identifier and account identifier. In an exemplary embodiment, reference is made to... Figure 6 As shown, you can create a new query result 601 in the terminal, or view historical query results 603.

[0282] Step S503: Send the query request to the server to instruct the server to perform task planning on the query content and obtain at least one task step.

[0283] In the specific implementation process, the query content can be transmitted to an intelligent agent, which then parses the query content and performs task planning on the parsed content to obtain at least one task step to respond to the query content. The intelligent agent can be an agent built on a large language model, possessing the ability to autonomously understand, perceive, plan, remember, and use tools. In an exemplary embodiment, the intelligent agent can adopt a "thinking + action" reasoning mode when understanding the query content. Thinking refers to the intelligent agent analyzing the current situation and considering what to do next; action refers to the action performed by the intelligent agent, usually by calling a tool. During the intelligent agent's task planning phase, the intelligent agent performs a round of thinking beforehand, generating at least one task step to respond to the query content. In the specific implementation process, different query content can correspond to different prompt templates, and a prompt is constructed based on the prompt template. For example, when the query content is: "How tall is A?", the prompt may include: User: How tall is B? Intelligent agent thinking: I need to search for B's height; Intelligent agent action: Search online; Intelligent agent output: B's height. Now let's start processing the new task: the user's query content: what is Xiao A's height, etc., where the agent's actions can be the task steps for the above query content.

[0284] In an exemplary embodiment, the aforementioned intelligent agent can be an AI agent focused on task planning. The intelligent agent responsible for task planning is used to break down complex user needs, understand task objectives and contextual information, and dynamically formulate and adjust a distributed multi-agent collaborative execution plan. In this embodiment, the intelligent agent responsible for task planning can be obtained by training an initial intelligent agent with task steps. In an exemplary embodiment, to improve the accuracy of the initial task planning, prompts can be generated based on candidate task steps in the target domain. The prompts and query content are input to the intelligent agent responsible for task steps, and at least one corresponding task step is output. In a specific implementation, for example, in the domain of resource interaction, the query content is input to the intelligent agent responsible for task planning, and at least one corresponding task step is output, such as: analysis of the query object in the query content, recommendation of specific products of the query object, and comparison between various recommended products. The prompt can be input text provided by the intelligent agent system to the large language model in the intelligent agent, used to guide the large language model to produce the desired output or behavior.

[0285] Step S505: Display the step nodes corresponding to the at least one task step; the server is further configured to send a call request for the target local tool to the associated terminal when the execution of the task step depends on the target local tool of the associated terminal; and determine the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal.

[0286] The associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal.

[0287] In one exemplary embodiment, the step node corresponding to at least one task step can be as follows: Figure 6 In area 605, the step nodes can be presented in a tree structure. The terminal can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Optionally, the terminal can also be any hardware device, as long as it can perform a specific function, such as a browser terminal that performs browsing functionality.

[0288] The associated terminal refers to at least one of the terminals associated with the query request. In one exemplary embodiment, the query request can be sent by any of the associated terminals where the target local tool is located. For example, the terminals associated with the query request include terminal A, terminal B, and terminal C. In this particular query request, the associated terminals where the target local tool is located include terminal A and terminal B, and terminal A sends the query request. In another exemplary embodiment, the query request can also be sent by other associated terminals. For example, in the above example, terminal C can also send the query request.

[0289] The local tools can include tools installed on the terminal. These local tools can have various functions. For example, tools can be developed based on proprietary protocols to obtain custom local tools. Alternatively, tools can be developed based on public standard protocols. When using tools developed with public standard protocols, they can be downloaded and used from a platform tool store by searching for keywords such as "query train ticket information," "generate PPT," or "generate code." In a specific embodiment, local tools can be configured on the terminal, and the terminal downloads the corresponding tools based on the configuration. Optionally, a list of completed local tools can be displayed on the terminal, for example, referring to... Figure 3 As shown, when a local tool 301 is triggered, the local tool is automatically downloaded and installed on the terminal. This embodiment of the disclosure... Figure 3 The source of the local tools shown is not restricted. The target local tool can be one or more of the local tools.

[0290] Optionally, the query request may include an account identifier. By establishing the relationship between account information, terminal information, and local tool information, the terminal information associated with the account identifier and the local tool information corresponding to the local tools installed on the terminal can be determined. For example, account A has a mobile terminal, a computer terminal, and a vehicle terminal. The mobile terminal has instant messaging tools and video software tools, the computer terminal has PPT creation tools and workflow system tools, and the vehicle terminal has browser tools, navigation tools, audio playback tools, etc.

[0291] The local tool information may also include the parameters required to call each local tool. In specific implementations, different development methods may correspond to different tool information. For example, tool information may include the tool's identifier, input parameters, parameter types, and function descriptions. This disclosure does not limit the content of the tool information.

[0292] In one exemplary embodiment, the execution agent responsible for executing the task steps can determine the associated terminal and target local tool for executing the specific task steps based on local tool information corresponding to the query request. In another exemplary embodiment, the local tool information can be input to the agent responsible for the task steps, which can then filter target local tools that match the task steps based on the description of the local tool's function in the tool information.

[0293] It is understood that in at least one task step, some task steps need to be implemented by calling local tools, while others can be implemented by calling cloud tools. Therefore, in an exemplary embodiment, local tools or cloud tools that match the task steps can be selected based on the tool information of the cloud tools and the description of the functions of the cloud tools.

[0294] Step S507: Display the result node corresponding to the processing result of the task step; the server is also used to determine the reply content to the query content based on the processing results corresponding to the at least one task step.

[0295] In one exemplary embodiment, the result node may be displayed on Figure 6 In area 605, for example, the leaf nodes of the tree structure correspond to the result nodes of the task steps or the sub-result nodes of the subtasks. Alternatively, area 609 can display detailed information about the processing results.

[0296] In this embodiment of the disclosure, some of the at least one task steps may be implemented by calling cloud tools during execution, and the results of the cloud tool calls are obtained. Exemplarily, some of the at least one task steps may not require calling tools, for example, they can be obtained through model inference or by querying a database. Another example is that some task steps are implemented by calling local tools, and the results of the local tool calls are obtained.

[0297] In this embodiment of the disclosure, each task step in the at least one task step corresponds to a processing result. For example, an intelligent agent can summarize the at least one processing result, such as extracting a summary of the at least one processing result and outputting it according to a preset output format. For example, a report corresponding to the response content can also be generated and downloaded. In a specific embodiment, for example, the user's query content includes: "Check the weather, please book train tickets for next week on the ticketing system." The at least one task step includes: Task step a: Check the cloud weather tool; the corresponding query result is sunny and suitable for travel. Task step b: Use the local ticketing tool to book tickets on the ticketing system; the booking result is successful ticket booking. The processing results of each task step are summarized to obtain a response content such as: "Sunny and suitable for travel; I have booked your train ticket for tomorrow; ticket information is as follows," etc.

[0298] Step S509: Display the response content.

[0299] In one exemplary embodiment, reference is made to Figure 6 As shown, area 605 can display task steps and subtasks within those steps. Area 609 can display detailed information about the task steps. Response content can also be displayed in area 605. Optionally, the response content can be displayed as a report file, which can be downloaded by triggering the report file.

[0300] In the above task processing method, for the sent query request, the server performs task planning on the query content to obtain at least one task step, breaking down complex query content into task steps and solving each task step one by one. Optionally, if the task steps are more complex, they can be further broken down. Therefore, this application can handle relatively complex query content. Furthermore, when executing task steps, in addition to calling cloud tools, more importantly, it can call various associated terminals related to the query request and the target local tools of each associated terminal, making full use of the rich localization capabilities of each associated terminal, such as calling professional software, accessing hardware sensors, and processing local data. In this process, the advantage that each target local tool can access the corresponding terminal environment is fully utilized, greatly expanding the capability boundary for solving large-scale model tasks, enabling the solution of extremely complex tasks involving cross-terminal collaboration, and possessing heterogeneous dedicated capabilities. Furthermore, by sending a call request for the target local tool to the associated terminal; based on the call result corresponding to the call request returned by the associated terminal, the processing result of the task step is determined. This approach eliminates the need to access end-user data, requiring only the retrieval of the call result, thus protecting end-user data security. Furthermore, it enables comprehensive utilization of data across various associated devices, achieving cross-terminal resource aggregation. Additionally, when a new terminal is added to an account, it can be aggregated into the existing terminal list corresponding to that account, demonstrating strong scalability and dynamically increasing the system's capacity ceiling.

[0301] In one embodiment, prior to retrieving the query content, the method further includes:

[0302] Display the target page, which shows the configuration controls for the local tools.

[0303] In response to a triggered event for the configuration control, display a list of local tools.

[0304] In response to a selection event for a target local tool from the list of local tools, the target local tool is installed to at least one of the associated terminals.

[0305] In the specific implementation process, refer to Figure 6 As shown, a configuration control 607 for local tools can be displayed on the target page. This configuration control 607 is triggered, for example, by clicking it. (See reference) Figure 7 As shown, it can be displayed as follows Figure 7 The page displays a list of local tools. In one exemplary embodiment, each local tool has an enable control 701. Triggering the enable control 701 selects the corresponding target local tool and downloads and installs it to the terminal.

[0306] In the above embodiments, by providing a list of local tools, which includes tools with various functions for users to choose from, the convenience of installing and downloading local tools is improved, and the usage rate of local tools is increased.

[0307] In one embodiment, following the display of the list of local tools, the following is also included:

[0308] In response to a share event for a target local tool from the list of local tools, the link information of the target local tool is sent to other accounts.

[0309] In the specific implementation process, refer to Figure 7 As shown, the local tools list can display dropdown control 703. Triggering this dropdown control is indicated in the reference. Figure 8 As shown, a sharing control 801 can be displayed. Triggering this sharing control 801 can share the corresponding local tool to other accounts. In an exemplary embodiment, refer to... Figure 7 As shown, the page also allows you to set an import configuration control 705. When the import configuration control 705 is triggered, you can paste the sharing password, and the system will automatically download the corresponding local tool based on the password.

[0310] In the above embodiments, by sending the link information of the target local tool to other terminal accounts through the sharing event of the target local tool in the local tool list, the use and dissemination of local tools are facilitated, the types of tools are enriched, and the calling capability of large models is improved.

[0311] In one embodiment, prior to retrieving the query content, the method further includes:

[0312] Display the target page, on which a new control for local tools is displayed;

[0313] In response to the trigger event for the newly created control, display the local tool creation page;

[0314] The local tool import page receives an input event for the link information of a local tool, and downloads the local tool from a third-party platform based on the link information to obtain a newly created local tool.

[0315] In the specific implementation process, the target page displays a new control for local tools, see reference. Figure 7 As shown, this page has a control for adding a new local tool (705). Triggering this control will display the following: Figure 9The local tool creation page 901 is shown. This page receives link information for local tools, indicating how to obtain the tool. This link information can be generated during the development of the local tool. For example, a third-party platform can be used to select the tool to be installed, and the tool's information can be copied to the link information field here. In an exemplary embodiment, the creation page 901 can also include a tool identifier, a description of the tool's corresponding functions, and configuration of the corresponding service type.

[0316] In the above embodiments, by setting up a new local tool page, the corresponding local tool can be automatically downloaded from a third-party platform by simply inputting link information, thus expanding the channels for obtaining local tools and enriching the local tools available on the terminal.

[0317] In one embodiment, sending the query request to the server includes:

[0318] The query request is sent to the server based on the associated terminal where the target local tool is located;

[0319] The document further includes, after displaying the step nodes corresponding to the at least one task step;

[0320] Parse the call request to obtain the call instructions for the target local tool to execute the task steps;

[0321] Based on the invocation instruction and the target local tool, the task steps are processed to obtain the processing result, and the processing result is sent to the server.

[0322] In the specific implementation process, the call request is parsed to obtain the calling instructions needed to invoke the local tool. For example, this might involve processing files in the target area of ​​the terminal, such as counting desktop photos and automatically printing them. After receiving the relevant calling instructions, the local tool processes the desktop photo files, determines the processing result (e.g., successful printing or some photos failing to print), and then sends the processing result to the server.

[0323] In the above embodiments, by executing the call command through local tools, the runtime environment can be accessed in a real way without uploading the terminal's files to the server, thus ensuring the data security of the terminal, while also enabling related processing of the terminal's files.

[0324] In one specific embodiment, the method of this application can be applied to scenarios involving responding to query requests. In the prior art, after receiving a user's query request, the server uses an intelligent agent to plan the task for the query request, obtains task steps, and executes the task steps by calling cloud tools. This approach is inadequate for complex query requests, such as cross-device collaborative requests; therefore, the task processing capacity of large models is limited. This application provides a task processing method that can improve the processing capacity of large models.

[0325] The task processing method described in this application may include a local tool configuration stage, a task planning stage, a task execution stage, and a response content generation stage.

[0326] Specifically, during the local tool configuration phase, refer to Figure 10 The user terminal registration method 1001 includes: displaying a target page, on which configuration controls for local tools are displayed; in response to a trigger event for the configuration controls, displaying a list of local tools; and in response to a selection event for a target local tool from the list of local tools, installing the target local tool to at least one of the associated terminals. (Reference) Figure 11 As shown, the terminal can include various types, such as computer terminals, browser terminals, and mobile terminals. After displaying the list of local tools, the process further includes: in response to a share event for a target local tool from the list of local tools, sending the link information of the target local tool to other accounts. Before obtaining the query content, the process further includes: displaying a target page, showing a new local tool control on the target page; in response to a trigger event for the new control, displaying a new local tool page; receiving an input event for the link information of a local tool on the local tool import page, and downloading the local tool from a third-party platform based on the link information to obtain the newly created local tool. (Reference) Figure 11 As shown, the terminal is equipped with user identifiers, terminal identifiers, etc. In an exemplary embodiment, the registration information of local tools reported by each terminal is obtained; wherein, the registration information includes account information, terminal information, and local tool information; based on the registration information, the association relationship between account information, terminal information, and local tool information is determined.

[0327] During the task planning phase, refer to Figure 10 Task planning in section 1003 includes methods such as: (refer to...) Figure 11 As shown, the core intelligent agent performs problem understanding and task planning on the user's query content. Step S203 involves performing task planning on the query content to obtain at least one task step. (Referring to...) Figure 11As shown, this application provides intelligent agents with various functions, forming a list of intelligent agents. Different intelligent agents are used to execute different task steps. In an exemplary embodiment, the task planning for the query content includes: parsing the query request and determining the target account information corresponding to the query request; determining at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information based on the association relationship between account information, terminal information, and local tool information; and performing task planning for the query content based at least on the at least one terminal information and the at least one local tool information. In an exemplary embodiment, the task planning for the query content to obtain at least one task step includes: decomposing the query content into tasks to obtain at least one task objective; for the task objective, determining a target local tool whose functional scope meets the task objective based at least on the tool information corresponding to the query request; and determining at least one task step based on the task objective, the tool identifier of the target local tool, and the terminal identifier of the terminal associated with the target local tool. In an exemplary embodiment, determining at least one task step based on the task objective, the tool identifier of the terminal local tool, the input parameters of the terminal local tool, and the corresponding parameter content includes: if the target local tool is a standard protocol tool, obtaining the identification information of the intelligent agent executing the standard protocol tool; and determining at least one task step based on the task objective, the tool identifier of the target local tool corresponding to the task objective, the terminal identifier corresponding to the target local tool, and the identification information of the intelligent agent executing the standard protocol tool.

[0328] During the task execution phase, refer to Figure 10The task execution 1005 method includes: step S205, for the task step, if the execution of the task step depends on a target local tool of an associated terminal, sending a request to invoke the target local tool to the associated terminal; the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal. In an exemplary embodiment, after performing task planning on the query content to obtain at least one task step, the method further includes: for each task step, determining a target execution agent from the agent resource pool whose functional scope matches the task step, and obtaining the target execution agent corresponding to the task step; when the execution of the task step depends on a target local tool of an associated terminal, sending a request to invoke the target local tool to the associated terminal includes: based on the target execution agent corresponding to the task step, determining target terminal information and target local tool information corresponding to the task step from at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information; wherein, the target account information is obtained by parsing the query request; when the target terminal information corresponds to an associated terminal and the target local tool information corresponds to a target local tool, sending a request to invoke the target local tool to the associated terminal.

[0329] In one embodiment, determining the target terminal information and target local tool information corresponding to the task step from at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information based on the execution agent corresponding to the task step includes: determining a target prompt based on at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information; inputting the target prompt to the target execution agent corresponding to the task step, and outputting the target terminal information and target local tool information corresponding to the task step. In one embodiment, determining the target terminal information and target local tool information corresponding to the task step includes: determining candidate local tools whose functional scope satisfies the task step, querying the online status information of the candidate local tools; determining the target local tool and the associated terminal where the target local tool is located from the candidate local tools based on the online status information; and determining the target local tool information corresponding to the target local tool and the target terminal information corresponding to the associated terminal.

[0330] In one embodiment, after performing task planning on the query content to obtain at least one task step, the method further includes: for the task step, if the task step satisfies preset conditions, performing task decomposition on the task step that satisfies the preset conditions to obtain at least one subtask; wherein the at least one subtask is matched with subtask tool information, and the subtask tool information is at least one type of tool information. In one embodiment, after performing task decomposition on the task step that satisfies the preset conditions to obtain at least one subtask, the method further includes: for the subtask, if the execution of the subtask depends on the target local tool of the corresponding associated terminal, sending a subtask invocation request of the target local tool to the associated terminal; determining the processing result of the subtask based on the invocation result corresponding to the subtask invocation request returned by the associated terminal; and determining the processing result of the task step that satisfies the preset conditions based on the processing result of the at least one subtask.

[0331] In one embodiment, when the execution of the task step depends on a target local tool on an associated terminal, sending a request to invoke the target local tool to the associated terminal includes: if, by an agent executing a standard protocol tool, it is determined that the execution of the task step depends on a target local tool on an associated terminal, then, for the terminal identifier of the associated terminal corresponding to the task step and the tool identifier corresponding to the target local tool, encapsulates them using a standard protocol data format to obtain a request corresponding to the task step; and sends the request to the associated terminal. In an exemplary embodiment, sending the request to invoke the target local tool to the associated terminal includes: obtaining the terminal identifier corresponding to the associated terminal and the account information corresponding to the query request; determining a long-connection communication interface corresponding to the terminal identifier and the account information; and sending the request to invoke the target local tool based on the long-connection communication interface.

[0332] In one embodiment, before sending the call request for the target local tool based on the long-connection communication interface, the method further includes: obtaining account information corresponding to the query request; determining terminal information corresponding to at least one terminal corresponding to the account information; and establishing long-connection communication interfaces with the account information and the terminals corresponding to the at least one terminal information, respectively. In an exemplary embodiment, determining the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal includes: if the call result corresponding to the call request returned by the associated terminal is a call failure, updating the execution of the target local tool to a backup local tool; and determining the processing result of the task step based on the processing result of the backup local tool.

[0333] During the response content generation stage, refer to Figure 10 In step S207, after returning the final result 1007, based on the call result corresponding to the call request returned by the associated terminal, the processing result of the task step is determined. Based on the processing results corresponding to the at least one task step, the response content for replying to the query content is determined. In an exemplary embodiment, after updating the execution of the target local tool to a backup local tool, the method further includes: if the backup local tool call fails, re-planning the task for the query content to obtain at least one updated task step; determining the response content for replying to the query content based on the processing results corresponding to the at least one task step includes: determining the response content for replying to the query content based on the processing results corresponding to the at least one updated task step.

[0334] In a specific application scenario, this application can be applied to complex task processing scenarios involving multi-terminal collaboration. In an exemplary embodiment, the query content can be "Please help me find recent football matches, book tickets for popular football matches, and send them to my communication friend F." The intelligent agent responsible for the task steps plans the task content, resulting in task steps A, B, and C. Task step A: Use the browser tool on the browser terminal to query recent football matches and determine the most popular matches. Task step B: Use the ticketing tool on the mobile terminal to purchase tickets for the most popular matches. Task step C: Send the purchased tickets to communication friend F via the instant messaging tool on the computer terminal. For task step A, the server calls the browser tool on the browser terminal to query football matches. Based on the query results, it determines the most popular matches. It calls the ticketing tool on the mobile terminal to purchase tickets for the most popular matches. It calls the instant messaging tool on the computer terminal to send the purchased tickets to communication friend F.

[0335] In another specific application scenario, in an exemplary embodiment, the query content could be: "Help me collect the latest viewpoints and data on 'Artificial Intelligence Trends in 2025' and create a simple PPT." The agent responsible for the task steps understands the above query content and formulates the following task steps: Task Step 1, use the browser tool on the browser terminal to search for information on the Internet; Task Step 2, use the slide creation tool on the computer terminal to convert the searched information into a slide format; Task Step 3, use the mobile terminal to send the slide to the communication account via an instant messaging tool; the processing result of the task steps is determined based on the call result corresponding to the call request returned by the associated terminal.

[0336] Optionally, the server call center schedules and executes each task step separately. Specifically, the browser tool on the browser terminal searches multiple authoritative technology media and research institution websites to extract key information and data, generating a structured content summary, thus obtaining the processing result of task step one. The computer terminal's slide generator is then used, taking the structured summary returned by the browser terminal as input, to automatically generate a PPT document containing a title, table of contents, core viewpoints, and data charts. This yields the processing result of task step two. Finally, the mobile terminal's instant messaging tool sends the PPT document generated on the computer terminal to a designated communication group, along with a message: "This is a newly compiled 2025 AI trend report; please refer to it." Optionally, the execution status and results of the entire task step can be notified via the mobile terminal or computer terminal.

[0337] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0338] Based on the same inventive concept, this application also provides a task processing apparatus for implementing the task processing method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more task processing apparatus embodiments provided below can be found in the limitations of the task processing method described above, and will not be repeated here.

[0339] In one embodiment, such as Figure 12 As shown, a task processing device 1200 is provided, comprising:

[0340] The first acquisition module 1201 is used to acquire the sent query request; wherein, the query request includes query content;

[0341] Planning module 1203 is used to perform task planning on the query content to obtain at least one task step;

[0342] The calling module 1205 is configured to send a calling request for the target local tool to the associated terminal when the execution of the task step depends on the target local tool of the associated terminal; wherein the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal.

[0343] The first determining module 1207 is used to determine the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal.

[0344] The second determining module 1209 is used to determine the response content to the query content based on the processing results corresponding to the at least one task step.

[0345] In one embodiment, the planning module is further configured to:

[0346] Parse the query request to determine the target account information corresponding to the query request;

[0347] Based on the association between account information, terminal information and local tool information, at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information are determined.

[0348] Task planning is performed on the query content based at least on the at least one terminal information and the at least one local tool information.

[0349] In one embodiment, the planning module is further configured to:

[0350] For each task step, a target execution agent whose functional scope matches the task step is determined from the agent resource pool, thus obtaining the target execution agent corresponding to the task step;

[0351] The calling module is also used for:

[0352] Based on the target execution agent corresponding to the task step, the target terminal information and target local tool information corresponding to the task step are determined from at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information; wherein, the target account information is obtained by parsing the query request;

[0353] If the target terminal information corresponds to an associated terminal and the target local tool information corresponds to a target local tool, a request to invoke the target local tool is sent to the associated terminal.

[0354] In one embodiment, the calling module is further configured to:

[0355] The target prompt is determined based on at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information.

[0356] The target prompt is input into the target execution agent corresponding to the task step, and the target terminal information and target local tool information corresponding to the task step are output.

[0357] In one embodiment, the calling module is further configured to:

[0358] Identify candidate local tools whose functional scope satisfies the task steps, and query the online status information of the candidate local tools;

[0359] Based on the online status information, the target local tool and the associated terminal where the target local tool is located are determined from the candidate local tools;

[0360] Determine the target local tool information corresponding to the target local tool and the target terminal information corresponding to the associated terminal.

[0361] In one embodiment, the calling module is further configured to:

[0362] Identify candidate local tools whose functional scope satisfies the task steps, and query the performance information and historical execution result accuracy of the terminal where the candidate local tools are located;

[0363] Based on the performance information and the accuracy of the execution results, a target local tool and the associated terminal where the target local tool is located are determined from the candidate local tools.

[0364] Determine the target local tool information corresponding to the target local tool and the target terminal information corresponding to the associated terminal.

[0365] In one embodiment, the planning module is further configured to:

[0366] The query content is decomposed into tasks to obtain at least one task objective;

[0367] For the task objective, at least based on the tool information corresponding to the query request, determine the target local tool whose functional scope meets the task objective;

[0368] Based on the task objective, the tool identifier of the target local tool, and the terminal identifier of the associated terminal where the target local tool is located, at least one task step is determined.

[0369] In one embodiment, the planning module is further configured to:

[0370] If the target local tool is a standard protocol tool, obtain the identification information of the agent executing the standard protocol tool;

[0371] Based on the task objective, the tool identifier of the target local tool corresponding to the task objective, the terminal identifier of the target local tool, and the identification information of the agent executing the standard protocol tool, at least one task step is determined.

[0372] In one embodiment, the calling module is further configured to:

[0373] If the target local tool is a standard protocol tool, the input parameters of the target local tool are queried from the target local tool information corresponding to the target local tool;

[0374] Based on the task steps corresponding to the target local tool, determine the parameter content of the input parameters;

[0375] Based on the input parameters of the target local tool and the corresponding parameter content, the invocation request of the target local tool is determined.

[0376] In one embodiment, the apparatus further includes a task decomposition module, the task decomposition module being used to:

[0377] For the task steps, if the task steps meet preset conditions, the task steps that meet the preset conditions are decomposed to obtain at least one subtask; wherein, the subtask tool information corresponding to the matching of the at least one subtask is at least one type of tool information.

[0378] In one embodiment, the task decomposition module is used to:

[0379] For the subtask, if the execution of the subtask depends on the target local tool of the corresponding associated terminal, a subtask invocation request of the target local tool is sent to the associated terminal;

[0380] Based on the call result corresponding to the call request of the subtask returned by the associated terminal, the processing result of the subtask is determined;

[0381] Based on the processing results of the at least one sub-task, the processing results of the task steps that meet the preset conditions are determined.

[0382] In one embodiment, the calling module is further configured to:

[0383] When the execution of the task step depends on the target local tool of the associated terminal by the intelligent agent executing the standard protocol tool, the terminal identifier of the associated terminal corresponding to the task step and the tool identifier corresponding to the target local tool are encapsulated in the data format of the standard protocol to obtain the invocation request corresponding to the task step.

[0384] The call request is sent to the associated terminal.

[0385] In one embodiment, the calling module is further configured to:

[0386] Obtain the terminal identifier corresponding to the associated terminal and the account information corresponding to the query request;

[0387] Determine the long-connection communication interface corresponding to the terminal identifier and the account information;

[0388] Based on the long-connection communication interface, a request to invoke the target local tool is sent.

[0389] In one embodiment, the device further includes a communication module, the communication module being used for:

[0390] Obtain the account information corresponding to the query request;

[0391] Determine the terminal information corresponding to at least one terminal associated with the account information;

[0392] Establish long-connection communication interfaces for the terminals corresponding to the account information and the at least one terminal information, respectively.

[0393] In one embodiment, the apparatus further includes a registration module, the registration module being used for:

[0394] Obtain the registration information of local tools reported by each terminal; wherein, the registration information includes account information, terminal information, and local tool information;

[0395] Based on the registration information, the relationships between account information, terminal information, and local tool information are determined.

[0396] In one embodiment, the first determining module is further configured to:

[0397] If the call result corresponding to the call request returned by the associated terminal is a call failure, the target local tool is updated to a backup local tool;

[0398] Based on the processing results of the backup local tool, the processing result of the task step is determined.

[0399] In one embodiment, the second determining module is further configured to:

[0400] If the backup local tool fails to be invoked, the query content is re-planned to obtain at least one updated task step;

[0401] The step of determining the response content based on the processing results corresponding to the at least one task step includes:

[0402] Based on the processing results corresponding to the at least one updated task step, the response content for replying to the query content is determined.

[0403] In one embodiment, the planning module is further configured to:

[0404] Retrieve historical query content that matches the query request from the experience knowledge base;

[0405] If a historical query that matches the query request is found, the historical task steps corresponding to the historical query are obtained; based on the historical task steps, at least one task step to reply to the query is determined.

[0406] If no historical query content matching the query request is found, task planning is performed on the query content to obtain at least one task step.

[0407] In one embodiment, the associated terminal includes at least one of a mobile terminal, a computer terminal, and a browser terminal, and the calling module is further configured to:

[0408] If the execution of the task step depends on the target local tool of the corresponding mobile terminal, a request to invoke the target local tool is sent to the mobile terminal.

[0409] If the execution of the task step depends on the target local tool of the corresponding computer terminal, a request to invoke the target local tool is sent to the computer terminal.

[0410] If the execution of the task step depends on the target local tool of the corresponding browser terminal, a request to invoke the target local tool is sent to the browser terminal.

[0411] In one embodiment, the second determining module is further configured to:

[0412] Based on the invocation result of the first local tool returned by the first terminal, the processing result of the first task step is determined; wherein, the first task step includes using the first local tool to search for information from the Internet;

[0413] Based on the invocation result of the second local tool returned by the second terminal, the processing result of the second task step is determined; wherein, the second task step includes using the second local tool to convert the searched information into a target format file;

[0414] Based on the invocation result of the third local tool returned by the third terminal, the processing result of the third task step is determined; wherein, the third task step includes using the third local tool to send the target format file to the communication account via an instant messaging tool.

[0415] In one embodiment, such as Figure 13 As shown, a task processing device 1300 is provided, comprising:

[0416] The second acquisition module 1301 is used to determine a query request based on the query content;

[0417] The sending module 1303 is used to send the query request to the server to instruct the server to perform task planning on the query content and obtain at least one task step;

[0418] The first display module 1305 is used to display the step nodes corresponding to the at least one task step; the server is also used to send a call request for the target local tool to the associated terminal when the execution of the task step depends on the target local tool of the associated terminal; and to determine the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal; wherein the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal;

[0419] The second display module 1307 is used to display the result nodes corresponding to the processing results of the task steps; the server is also used to determine the response content to the query content based on the processing results corresponding to the at least one task step.

[0420] The third display module 1309 is used to display the reply content.

[0421] In one embodiment, the apparatus further includes a configuration module, the configuration module being configured to:

[0422] Display the target page, on which configuration controls for the local tools are displayed;

[0423] In response to a triggered event on the configuration control, display a list of local tools;

[0424] In response to a selection event for a target local tool from the list of local tools, the target local tool is installed to at least one of the associated terminals.

[0425] In one embodiment, the device further includes a sharing module, the sharing module being used for:

[0426] In response to a share event for a target local tool from the list of local tools, the link information of the target local tool is sent to other accounts.

[0427] In one embodiment, the configuration module is further configured to:

[0428] Display the target page, on which a new control for local tools is displayed;

[0429] In response to the trigger event for the newly created control, display the local tool creation page;

[0430] The local tool import page receives an input event for the link information of a local tool, and downloads the local tool from a third-party platform based on the link information to obtain a newly created local tool.

[0431] In one embodiment, the sending module is further configured to send the query request to the server based on the associated terminal where the target local tool is located.

[0432] The device further includes a processing module, the processing module being used for:

[0433] Parse the call request to obtain the call instructions for the target local tool to execute the task steps;

[0434] Based on the invocation instruction and the target local tool, the task steps are processed to obtain the processing result, and the processing result is sent to the server.

[0435] Each module in the aforementioned task processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can invoke and execute the operations corresponding to each module.

[0436] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 14As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores task processing data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a task processing method.

[0437] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 15 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational 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 an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a task processing method. The display unit of the computer device is used to form a visually visible image. It can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0438] Those skilled in the art will understand that Figure 15The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0439] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0440] Those skilled in the art will understand that all or part of the processes in 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 described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, etc., and are not limited to these.

[0441] 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.

[0442] 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 task processing method, characterized in that, Applied to a server, the method includes: Obtain the sent query request; wherein the query request includes query content; The query content is used for task planning to obtain at least one task step; For the task step, if the execution of the task step depends on a target local tool on an associated terminal, a request to invoke the target local tool is sent to the associated terminal; wherein, the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal; Based on the call result corresponding to the call request returned by the associated terminal, the processing result of the task step is determined; Based on the processing results corresponding to the at least one task step, the response content for replying to the query content is determined.

2. The method according to claim 1, characterized in that, The task planning for the query content includes: Parse the query request to determine the target account information corresponding to the query request; Based on the association between account information, terminal information and local tool information, at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information are determined. Task planning is performed on the query content based at least on the at least one terminal information and the at least one local tool information.

3. The method according to claim 1, characterized in that, After performing task planning on the query content to obtain at least one task step, the method further includes: For each task step, a target execution agent whose functional scope matches the task step is determined from the agent resource pool, thus obtaining the target execution agent corresponding to the task step; In the case where the execution of the task step depends on a target local tool on the associated terminal, sending a request to invoke the target local tool to the associated terminal includes: Based on the target execution agent corresponding to the task step, the target terminal information and target local tool information corresponding to the task step are determined from at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information; wherein, the target account information is obtained by parsing the query request; If the target terminal information corresponds to an associated terminal and the target local tool information corresponds to a target local tool, a request to invoke the target local tool is sent to the associated terminal.

4. The method according to claim 3, characterized in that, The execution agent corresponding to the task step determines the target terminal information and target local tool information corresponding to the task step from at least one terminal information corresponding to the target account information and at least one local tool information corresponding to each terminal information. This includes: determining the target prompt based on at least one terminal information corresponding to the target account information, and at least one local tool information corresponding to each terminal information; The target prompt is input into the target execution agent corresponding to the task step, and the target terminal information and target local tool information corresponding to the task step are output.

5. The method according to claim 3, characterized in that, The determination of the target terminal information and target local tool information corresponding to the task steps includes: Identify candidate local tools whose functional scope satisfies the task steps, and query the online status information of the candidate local tools; Based on the online status information, the target local tool and the associated terminal where the target local tool is located are determined from the candidate local tools; Determine the target local tool information corresponding to the target local tool and the target terminal information corresponding to the associated terminal.

6. The method according to claim 3, characterized in that, The determination of the target terminal information and target local tool information corresponding to the task steps includes: Identify candidate local tools whose functional scope satisfies the task steps, and query the performance information and historical execution result accuracy of the terminal where the candidate local tools are located; Based on the performance information and the accuracy of the execution results, a target local tool and the associated terminal where the target local tool is located are determined from the candidate local tools. Determine the target local tool information corresponding to the target local tool and the target terminal information corresponding to the associated terminal.

7. The method according to claim 1, characterized in that, The task planning for the query content, resulting in at least one task step, includes: The query content is decomposed into tasks to obtain at least one task objective; For the task objective, at least based on the tool information corresponding to the query request, determine the target local tool whose functional scope meets the task objective; Based on the task objective, the tool identifier of the target local tool, and the terminal identifier of the associated terminal where the target local tool is located, at least one task step is determined.

8. The method according to claim 7, characterized in that, The step of determining at least one task step based on the task objective, the tool identifier of the terminal local tool, the input parameters of the terminal local tool, and the corresponding parameter content includes: If the target local tool is a standard protocol tool, obtain the identification information of the agent executing the standard protocol tool; Based on the task objective, the tool identifier of the target local tool corresponding to the task objective, the terminal identifier of the target local tool, and the identification information of the agent executing the standard protocol tool, at least one task step is determined.

9. The method according to claim 1, characterized in that, Before sending the invocation request for the target local tool to the associated terminal, the method further includes: If the target local tool is a standard protocol tool, the input parameters of the target local tool are queried from the target local tool information corresponding to the target local tool; Based on the task steps corresponding to the target local tool, determine the parameter content of the input parameters; Based on the input parameters of the target local tool and the corresponding parameter content, the invocation request of the target local tool is determined.

10. The method according to claim 1, characterized in that, After performing task planning on the query content to obtain at least one task step, the method further includes: For the task steps, if the task steps meet preset conditions, the task steps that meet the preset conditions are decomposed to obtain at least one subtask; wherein, the subtask tool information corresponding to the matching of the at least one subtask is at least one type of tool information.

11. The method according to claim 10, characterized in that, After decomposing the task steps that meet the preset conditions to obtain at least one sub-task, the method further includes: For the subtask, if the execution of the subtask depends on the target local tool of the corresponding associated terminal, a subtask invocation request of the target local tool is sent to the associated terminal; Based on the call result corresponding to the call request of the subtask returned by the associated terminal, the processing result of the subtask is determined; Based on the processing results of the at least one sub-task, the processing results of the task steps that meet the preset conditions are determined.

12. The method according to claim 1, characterized in that, In the case where the execution of the task step depends on a target local tool on the associated terminal, sending a request to invoke the target local tool to the associated terminal includes: When the execution of the task step depends on the target local tool of the associated terminal by the intelligent agent executing the standard protocol tool, the terminal identifier of the associated terminal corresponding to the task step and the tool identifier corresponding to the target local tool are encapsulated in the data format of the standard protocol to obtain the invocation request corresponding to the task step. The call request is sent to the associated terminal.

13. The method according to claim 1, characterized in that, Sending the request to invoke the target local tool to the associated terminal includes: Obtain the terminal identifier corresponding to the associated terminal and the account information corresponding to the query request; Determine the long-connection communication interface corresponding to the terminal identifier and the account information; Based on the long-connection communication interface, a request to invoke the target local tool is sent.

14. The method according to claim 13, characterized in that, Before sending the invocation request for the target local tool based on the long-connection communication interface, the method further includes: Obtain the account information corresponding to the query request; Determine the terminal information corresponding to at least one terminal associated with the account information; Establish long-connection communication interfaces for the terminals corresponding to the account information and at least one terminal information, respectively.

15. The method according to claim 1, characterized in that, Before performing task planning on the query content to obtain at least one task step, the method further includes: Obtain the registration information of local tools reported by each terminal; wherein, the registration information includes account information, terminal information, and local tool information; Based on the registration information, the relationships between account information, terminal information, and local tool information are determined.

16. The method according to claim 1, characterized in that, Based on the call result corresponding to the call request returned by the associated terminal, the processing result of the task step is determined, including: If the call result corresponding to the call request returned by the associated terminal is a call failure, the target local tool is updated to a backup local tool; Based on the processing results of the backup local tool, the processing result of the task step is determined.

17. The method according to claim 16, characterized in that, After updating the target local tool to an alternate local tool, the process also includes: If the backup local tool fails to be invoked, the query content is re-planned to obtain at least one updated task step; The step of determining the response content based on the processing results corresponding to the at least one task step includes: Based on the processing results corresponding to the at least one updated task step, the response content for replying to the query content is determined.

18. The method according to claim 1, characterized in that, The task planning for the query content, resulting in at least one task step, includes: Retrieve historical query content that matches the query request from the experience knowledge base; If a historical query that matches the query request is found, the historical task steps corresponding to the historical query are obtained; based on the historical task steps, at least one task step to reply to the query is determined. If no historical query content matching the query request is found, task planning is performed on the query content to obtain at least one task step.

19. The method according to claim 1, characterized in that, The associated terminal includes at least one of a mobile terminal, a computer terminal, and a browser terminal. When the execution of the task step depends on a target local tool on the corresponding associated terminal, sending a request to the associated terminal to invoke the target local tool includes: If the execution of the task step depends on the target local tool of the corresponding mobile terminal, a request to invoke the target local tool is sent to the mobile terminal. If the execution of the task step depends on the target local tool of the corresponding computer terminal, a request to invoke the target local tool is sent to the computer terminal. If the execution of the task step depends on the target local tool of the corresponding browser terminal, a request to invoke the target local tool is sent to the browser terminal.

20. The method according to claim 1, characterized in that, The process of determining the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal includes: Based on the invocation result of the first local tool returned by the first terminal, the processing result of the first task step is determined; wherein, the first task step includes using the first local tool to search for information from the Internet; Based on the invocation result of the second local tool returned by the second terminal, the processing result of the second task step is determined; wherein, the second task step includes using the second local tool to convert the searched information into a target format file; Based on the invocation result of the third local tool returned by the third terminal, the processing result of the third task step is determined; wherein, the third task step includes using the third local tool to send the target format file to the communication account via an instant messaging tool.

21. A task processing method, characterized in that, Applied to a terminal, the method includes: Obtain the query content, and determine the query request based on the query content; The query request is sent to the server to instruct the server to perform task planning on the query content and obtain at least one task step; The server displays the step nodes corresponding to the at least one task step; the server is also configured to send a call request for the target local tool to the associated terminal when the execution of the task step depends on the target local tool of the associated terminal; and determine the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal; wherein the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal. The server displays the result nodes corresponding to the processing results of the task steps; the server is also used to determine the response content to the query content based on the processing results corresponding to the at least one task step. The response content is displayed.

22. The method according to claim 21, characterized in that, Before obtaining the query content, the following is also included: Display the target page, on which configuration controls for the local tools are displayed; In response to a triggered event on the configuration control, display a list of local tools; In response to a selection event for a target local tool from the list of local tools, the target local tool is installed to at least one of the associated terminals.

23. The method according to claim 22, characterized in that, Following the display of the list of local tools, the following is also included: In response to a share event for a target local tool from the list of local tools, the link information of the target local tool is sent to other accounts.

24. The method according to claim 22, characterized in that, Before obtaining the query content, the following is also included: Display the target page, on which a new control for local tools is displayed; In response to the trigger event for the newly created control, display the local tool creation page; The local tool import page receives an input event for the link information of a local tool, and downloads the local tool from a third-party platform based on the link information to obtain a newly created local tool.

25. The method according to claim 22, characterized in that, Sending the query request to the server includes: The query request is sent to the server based on the associated terminal where the target local tool is located; The document further includes, after displaying the step nodes corresponding to the at least one task step; Parse the call request to obtain the call instructions for the target local tool to execute the task steps; Based on the invocation instruction and the target local tool, the task steps are processed to obtain the processing result, and the processing result is sent to the server.

26. A task processing device, characterized in that, The device includes: The first acquisition module is used to acquire the sent query request; wherein the query request includes query content; The planning module is used to plan tasks for the query content and obtain at least one task step; The invocation module is configured to, for the task step, send an invocation request to the associated terminal for the target local tool when the execution of the task step depends on the target local tool of the associated terminal; wherein the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal; The first determining module is used to determine the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal. The second determining module is used to determine the response content to the query content based on the processing results corresponding to the at least one task step.

27. A task processing device, characterized in that, The device includes: The second acquisition module is used to acquire query content and determine a query request based on the query content; The sending module is used to send the query request to the server to instruct the server to perform task planning on the query content and obtain at least one task step; The first display module is used to display the step nodes corresponding to the at least one task step; the server is also used to send a call request for the target local tool to the associated terminal when the execution of the task step depends on the target local tool of the associated terminal; and to determine the processing result of the task step based on the call result corresponding to the call request returned by the associated terminal; wherein the associated terminal is at least one of the terminals associated with the query request, and the target local tool is at least one of the tools installed on the associated terminal; The second display module is used to display the result nodes corresponding to the processing results of the task steps; the server is also used to determine the response content to the query content based on the processing results corresponding to the at least one task step. The third display module is used to display the response content.

28. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 20 or the steps of the method according to any one of claims 21 to 25.

29. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 20 or the steps of the method according to any one of claims 21 to 25.

30. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 20 or the steps of the method according to any one of claims 21 to 25.

Images