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

Abstract

The invention relates to a task execution method and device, computer equipment, a storage medium and a computer program product. The method comprises the steps of obtaining a to-be-executed task corresponding to query content, wherein the to-be-executed task is obtained by performing task planning on the query content and is used for obtaining a query result; task execution is carried out based on the to-be-executed task, a task execution result corresponding to the to-be-executed task and reference information corresponding to the to-be-executed task are obtained, and the reference information comprises dependency data information of the obtained task execution result; and obtaining a query result of the query content based on the task execution result and the reference information. By adopting the method, the reliability of the query result can be improved.

Description

Technical Field

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

[0002] With the development of large-scale model technology, users can perform content queries in multi-agent collaborative systems based on large language models, thereby improving the accuracy and efficiency of the retrieved content. Currently, multi-agent collaborative systems based on large language models typically divide complex query tasks into multiple planned tasks, which are then assigned to different agents with specific functions for sequential or parallel processing to obtain the final query results.

[0003] However, this query method usually only yields the final query result and cannot reveal the reasoning information of the internal task, resulting in poor interpretability of the query result and consequently reduced reliability of the query result. Summary of the Invention

[0004] Therefore, it is necessary to provide a task execution method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve the interpretability of query results and thus improve the reliability of query results, in order to address the above-mentioned technical problems.

[0005] On one hand, this application provides a task execution method. The method includes:

[0006] Retrieve the tasks to be executed corresponding to the query content. The tasks to be executed are obtained by planning the query content and used to obtain the query results.

[0007] Based on the task to be executed, the task is executed to obtain the task execution result and the reference information corresponding to the task to be executed. The reference information includes the dependency data information for obtaining the task execution result.

[0008] The query results are obtained based on the task execution results and reference information.

[0009] On the other hand, this application also provides a task execution apparatus. The apparatus includes:

[0010] The task acquisition module is used to acquire the tasks to be executed corresponding to the query content. The tasks to be executed are obtained by planning the query content and used to obtain the query results.

[0011] The task execution module is used to execute tasks based on the tasks to be executed, and to obtain the task execution results and reference information corresponding to the tasks to be executed. The reference information includes the dependency data information for obtaining the task execution results.

[0012] The results module is used to obtain query results based on task execution results and reference information.

[0013] On the other hand, 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 perform the following steps:

[0014] Get the tasks to be executed corresponding to the query content. The tasks to be executed are obtained by task planning of the query content.

[0015] Based on the task to be executed, the task is executed to obtain the task execution result and the reference information corresponding to the task to be executed. The reference information includes the dependency data information for obtaining the task execution result.

[0016] The query results are obtained based on the task execution results and reference information.

[0017] On the other hand, 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, performs the following steps:

[0018] Retrieve the tasks to be executed corresponding to the query content. The tasks to be executed are obtained by planning the query content and used to obtain the query results.

[0019] Based on the task to be executed, the task is executed to obtain the task execution result and the reference information corresponding to the task to be executed. The reference information includes the dependency data information for obtaining the task execution result.

[0020] The query results are obtained based on the task execution results and reference information.

[0021] On the other hand, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0022] Retrieve the tasks to be executed corresponding to the query content. The tasks to be executed are obtained by planning the query content and used to obtain the query results.

[0023] Based on the task to be executed, the task is executed to obtain the task execution result and the reference information corresponding to the task to be executed. The reference information includes the dependency data information for obtaining the task execution result.

[0024] The query results are obtained based on the task execution results and reference information.

[0025] The aforementioned task execution method, apparatus, computer equipment, storage medium, and computer program product obtain the task to be executed corresponding to the query content. The task to be executed is obtained by task planning based on the query content and used to obtain the query results. The task to be executed is then executed to obtain the task execution result and reference information corresponding to the task to be executed. The reference information includes dependency data information for obtaining the task execution result. Then, the query result of the query content is obtained based on the task execution result and the reference information. This means that the obtained query result includes not only the task execution result corresponding to the task to be executed but also the dependency data information of the task execution result corresponding to the task to be executed. That is, interpreting the obtained task execution result through dependency data information can improve the interpretability of the obtained query result, thereby improving the reliability of the query result.

[0026] On one hand, this application provides a task execution method. The method includes:

[0027] The search results are displayed on the search page;

[0028] The system displays the task execution results and reference information corresponding to the query content. The reference information includes the dependent data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the task to be executed corresponding to the query content. The task to be executed is obtained by planning the query content and using it to obtain the query results.

[0029] On the other hand, this application also provides a task execution apparatus. The apparatus includes:

[0030] The content display module is used to display the query content on the query page;

[0031] The results display module is used to display the task execution results and reference information corresponding to the query content. The reference information includes the dependency data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the task to be executed corresponding to the query content. The task to be executed is obtained by planning the query content and using it to obtain the query results.

[0032] On the other hand, 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 perform the following steps:

[0033] The search results are displayed on the search page;

[0034] The system displays the task execution results and reference information corresponding to the query content. The reference information includes the dependent data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the task to be executed corresponding to the query content. The task to be executed is obtained by planning the query content and using it to obtain the query results.

[0035] On the other hand, 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, performs the following steps:

[0036] The search results are displayed on the search page;

[0037] The system displays the task execution results and reference information corresponding to the query content. The reference information includes the dependent data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the task to be executed corresponding to the query content. The task to be executed is obtained by planning the query content and using it to obtain the query results.

[0038] On the other hand, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0039] The search results are displayed on the search page;

[0040] The system displays the task execution results and reference information corresponding to the query content. The reference information includes the dependent data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the task to be executed corresponding to the query content. The task to be executed is obtained by planning the query content and using it to obtain the query results.

[0041] The aforementioned task execution method, apparatus, computer equipment, storage medium, and computer program product display query content on a query page. Then, it displays the task execution results and reference information corresponding to the query content. The reference information includes dependent data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the task to be executed corresponding to the query content. The task to be executed is obtained by planning the query content and using it to obtain the query results. Therefore, the obtained query results include not only the task execution results corresponding to the task to be executed, but also the dependent data information for obtaining the task execution results. That is, interpreting the obtained task execution results through dependent data information improves the interpretability of the obtained query results, thereby increasing the reliability of the query results. Attached Figure Description

[0042] Figure 1 This is an application environment diagram of the task execution method in one embodiment;

[0043] Figure 2 This is a flowchart illustrating a task execution method in one embodiment;

[0044] Figure 3 This is a schematic diagram of the framework for obtaining query results in a specific embodiment;

[0045] Figure 4 This is a schematic diagram of a query page in a specific embodiment;

[0046] Figure 5 This is a schematic diagram of selecting a task node in a specific embodiment;

[0047] Figure 6 This is a schematic diagram showing the parent-child connection edge in a specific embodiment;

[0048] Figure 7 This is a comparative schematic diagram of the task execution flowchart in a specific embodiment;

[0049] Figure 8 This is a schematic diagram of task node editing in a specific embodiment;

[0050] Figure 9 This is a flowchart illustrating the task execution process in a specific embodiment;

[0051] Figure 10 This is a schematic diagram of task execution results and reference information in a specific embodiment.

[0052] Figure 11 This is a schematic diagram of a data source viewing control in a specific embodiment;

[0053] Figure 12 This is a flowchart illustrating a task execution method in a specific embodiment;

[0054] Figure 13 This is a schematic diagram of the task execution framework in a specific embodiment;

[0055] Figure 14 This is a structural block diagram of a task execution device in one embodiment;

[0056] Figure 15 This is a structural block diagram of the task execution device in another embodiment;

[0057] Figure 16 This is an internal structural diagram of a computer device in one embodiment;

[0058] Figure 17 This is a diagram of the internal structure of a computer device in another embodiment. Detailed Implementation

[0059] 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.

[0060] The task execution method provided in this application embodiment can be applied to, for example, Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on another server. Terminal 102 sends a query request to server 104, carrying the query content. Server 104 receives the query request, parses it to obtain the query content, and then performs task planning based on the query content to obtain tasks to be executed. At this time, server 104 executes the tasks based on the tasks to be executed, obtaining the task execution result and reference information corresponding to the tasks. The reference information includes the dependency data information for obtaining the task execution result. Server 104 obtains the query result of the query content based on the task execution result and the reference information. Server 104 returns the query result of the query content to terminal 102, which can display the query result. Simultaneously, server 104 can also store the query result of the query content in the data storage system. The terminal 102 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, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal and server can be directly or indirectly connected via wired or wireless communication; this application does not impose any restrictions on this connection.

[0061] In one embodiment, such as Figure 2 As shown, a task execution method is provided, which can be applied to Figure 1 Taking a server as an example, it can be understood that this method can also be applied to a terminal, and also to a system that includes both a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0062] S202, obtain the task to be executed corresponding to the query content. The task to be executed is obtained by task planning of the query content and used to obtain the query results.

[0063] The query content refers to the information that needs to be queried by the intelligent agent. This can be from various fields, such as gaming, education, finance, industry, agriculture, healthcare, and scientific research. The query content can be for information in different formats, such as text, images, video, and audio. This query content is typically user-inputted and requires a response from a multi-agent collaborative system based on a large language model. It can be a question, a search query, etc. For example, if a user enters a question like, "I recently bought a phone, what are some cost-effective headphones you can recommend? Give me a comprehensive recommendation analysis report," then this question is considered the query content. The pending task refers to the task planned based on the query content. For example, it could be a task planned using LLM Planning (Large Language Model Planning). Executing this pending task will retrieve the query results.

[0064] In this embodiment, the server obtains the query content sent by the terminal, and then performs task planning on the query content to obtain the tasks to be executed. During task planning, each task to be executed includes the corresponding execution agent information.

[0065] Understandably, users typically interact with multi-agent collaborative systems to obtain query results. In this scenario, users can enter their queries on the system's query page and receive the corresponding results. The agents can be built upon large language models and possess the ability to autonomously understand, perceive, plan, remember, and use tools. Agents can communicate based on a single large language model or through the collaboration of multiple large language models. Different agents typically perform different functions.

[0066] Understandably, servers typically plan the query content to obtain a sequence of tasks to be executed. This sequence includes individual tasks and their execution order. The server then executes the tasks in the sequence sequentially, either in parallel or sequentially. For example, planning the query "Analyze the correlation between A and B over the past three months and generate a report" might result in a sequence of five tasks: [Search "A in the last 3 months"] and [Search "B in the last 3 months"] -> [Data preprocessing] -> [Calculate correlation] -> [Generate report], which are then executed sequentially according to the arrow direction. Each task typically refers to a specific, executable step. Tasks usually include corresponding action information and execution parameters; different tasks typically have different action information and parameters. Action information refers to the specific actions the task requires, such as web search, comparative analysis, or chart generation. Execution parameters refer to the parameters required to execute an operation based on the operation action information. These parameters typically include information carried in the query content; for example, the execution parameters for a web search action might be the search time range, the range of web pages to be searched, etc., and may also include the execution results of other tasks to be executed. The steps to be executed refer to the steps that need to be performed currently. The execution tools that need to be called when an operation action is executed are usually pre-set; for example, the name, identifier, and specific content of the execution tools that can be called when the operation action is executed are pre-defined.

[0067] S204, Execute the task based on the task to be executed, and obtain the task execution result and reference information corresponding to the task to be executed. The reference information includes the dependency data information for obtaining the task execution result.

[0068] The task execution result is used to characterize the outcome of the task to be executed. For example, the execution result of a task searching for "A in the last 3 months" is the information about A found within the last 3 months. Reference information refers to information used to trace and interpret the obtained task execution result. Dependency data information refers to the data information that the agent relies on when generating the task execution result. This can include dependent input information, dependent reference information, dependent task information, etc. Dependency data information is used to trace the source of the obtained task execution result, thereby improving the interpretability of the task execution result.

[0069] In this embodiment, the server can obtain the execution agent corresponding to the task to be executed from a plurality of pre-set execution agents. Specifically, the execution agent corresponding to the task to be executed can be determined from the plurality of execution agents based on the calling tool information corresponding to the task. Then, the task to be executed is input into the execution agent for execution. The execution agent obtains the input data according to the execution parameters in the task to be executed, and performs the task execution according to the corresponding operation action information, obtaining the output task execution result. Simultaneously, the execution agent generates reference information corresponding to the task to be executed. Each execution agent has a pre-set output reference information format, thus enabling the execution agent to output the corresponding reference information according to the pre-set reference information format.

[0070] Understandably, the server can obtain different tasks to be executed, and then execute these different tasks simultaneously to obtain the task execution results and reference information corresponding to each task, thereby improving the efficiency of task execution.

[0071] S206, Query results are obtained based on task execution results and reference information.

[0072] The query results refer to the results obtained by querying according to the query content. The query results include reference information for tracing and interpreting the results of task execution.

[0073] In this embodiment, the server uses the task execution result and reference information as the query result of the query content, and returns the query result of the query content to the terminal that sent the query content for display.

[0074] Understandably, the server can use the execution results and reference information of all pending tasks corresponding to the query content as the query result, and display these results on the terminal that sent the query content. In some embodiments, the server can also use the execution result and reference information of the last pending task as the query result, and display this information on the terminal that sent the query content. In this case, the user can also trigger an action to view the execution results and reference information of other pending tasks. When the server receives a viewing request, it will display the viewed task's execution results and reference information on the terminal that sent the query content.

[0075] Understandably, users can understand the reasons for generating the corresponding task execution results by referring to the information in the query results, thereby increasing the transparency of the query content in the task execution process.

[0076] The above task execution method obtains the tasks to be executed corresponding to the query content. These tasks are generated by planning the query content and used to obtain the query results. The tasks are then executed to obtain the execution results and reference information, including dependency data that is used to obtain the execution results. Finally, the query results are obtained based on the execution results and reference information. This ensures that the query results include not only the execution results of the tasks to be executed but also the dependency data that is used to obtain those results. By interpreting the execution results using dependency data, the interpretability and reliability of the query results are improved.

[0077] In some embodiments, S204 involves executing a task based on the task to be executed, obtaining the task execution result corresponding to the task to be executed and the reference information corresponding to the task to be executed. The reference information includes dependency data information for obtaining the task execution result, and includes the following steps:

[0078] Obtain the execution agent corresponding to the task to be executed; input the task to be executed into the execution agent to execute the task, and obtain the task execution result output according to the preset execution result format and the reference information output according to the preset reference information format.

[0079] In this context, an execution agent refers to the agent used to perform a task. Different tasks perform different functions, such as searching, comparing, and summarizing. Different functions can be executed using corresponding execution agents; for example, there's a search execution agent, a comparison execution agent, and a summary execution agent. Information search tasks can be executed using corresponding information search agents, code execution tasks using corresponding code execution agents, and information summarizing tasks using corresponding information summarizing agents. The preset execution result format refers to the pre-defined format of the execution results included in the output of the execution agent. The preset reference information format refers to the pre-defined format of the reference information included in the output of the execution agent. Essentially, each execution agent's output includes both the task execution result and corresponding reference information.

[0080] In this embodiment, the return value of each executing agent includes the task execution result and reference information. It is understood that all executing agents on the server have a corresponding calling interface format, which includes a predefined execution result parameter format and an additional reference information parameter format.

[0081] Understandably, when the task planner schedules an agent to execute a task, it doesn't directly call its original function. Instead, it calls a pre-defined encapsulated executor. Before invoking the agent's core logic, this encapsulated executor records the call context, such as input parameters and timestamps. After the agent's core logic completes, it intercepts its original output and, based on a predefined reference information parameter format, causes the agent to generate a reference information object containing fields such as dependency data. Finally, the encapsulated executor combines the task execution result and the reference information object into a complete response, which becomes the agent's output.

[0082] In some embodiments, the server can obtain a pre-set prompt for output reference information, and input the task to be executed and the prompt for output reference information into the execution agent. During the execution process, the execution agent performs the task according to the pre-set prompt for the task to be executed, and at the same time collects and generates reference information according to the prompt for output reference information, so as to obtain the output task execution result and reference information.

[0083] Understandably, when the server receives the task execution results and reference information for each task to be executed, it can associate and save the task to be executed with the corresponding task results and reference information for easy querying and use by users later.

[0084] In the above embodiments, by inputting the task to be executed into the execution agent for task execution, the execution result is output according to a preset execution result format and the reference information is output according to a preset reference information format. That is, the execution agent can obtain the task execution result and the reference information simultaneously, without needing to collect the reference information after obtaining the task execution result, thereby improving the resource utilization rate during task execution and saving resources.

[0085] In some embodiments, the tasks to be performed include at least two, and the dependency data information includes dependency task information;

[0086] S206, which refers to obtaining query results based on task execution results and reference information, includes the following steps:

[0087] Obtain current reference information corresponding to the currently executed task from at least two pending tasks, and obtain historical reference information corresponding to the previously executed tasks; generate task execution chain data based on historical dependent task information in the historical reference information and current dependent task information in the current reference information to obtain current task execution chain data; obtain query results for the query content based on the current task execution result corresponding to the currently executed task, the historical execution result corresponding to the previously executed task, and the current task execution chain data.

[0088] The dependency task information describes the information of completed tasks that the task to be executed depends on during its execution. This information can be the identifier or name of a completed task, etc. The dependent completed task refers to the task whose execution result is required as input for the task to be executed. It can be understood that the dependency task information can include one or more completed tasks; for example, the dependency task information can be a list of dependency task identifiers. The currently executed task refers to the task that has already been completed. This task is obtained by planning the query content, i.e., the task whose execution result and reference information have been obtained. The current reference information refers to the reference information obtained when the currently executed task completes its execution. The historical executed task refers to the task that was completed before the currently executed task. The historical reference information refers to the reference information obtained when the historically executed task completes its execution.

[0089] The current task execution chain data refers to the currently generated task execution chain data. This data is formed based on the dependencies between each executed task and the query content, representing the information required for execution. Task execution dependencies represent the completed tasks required for a task to execute. The task execution chain data includes each task and the links between them, which can be determined based on the task execution dependencies. This task execution chain data is graph-structured data, such as a DAG (Directed Acyclic Graph).

[0090] In this embodiment, when the server obtains the task execution results and reference information for each task, it can determine the currently executing task and the historically executed tasks from each task. The server can select any task other than the starting task as the currently executing task, and then select tasks completed before the currently executing task as historically executed tasks. At this time, the server retrieves the current reference information corresponding to the currently executing task and the historical reference information corresponding to the historically executed tasks from the database. Then, based on the historical dependent task information in the historical reference information and the current dependent task information in the current reference information, task execution chain data is generated to obtain the current task execution chain data. Specifically, a link relationship can be established between the currently executing task and the historically executed tasks based on the current dependent task information, and a link relationship can be established between the historically executed tasks based on the historical dependent task information. For example, if the current dependent task information records that the historically executed tasks M and N depend on the current executing task Q, then the link relationship between historically executed tasks M and N and the current executing task Q is saved. If the historical dependent task information records that the historically executed task N depends on historically executed tasks M and H, then the link relationship between historically executed tasks M and H and the historically executed task N is saved. The server iterates through all executed tasks to obtain the link relationships between them, thus retrieving the current task execution chain data. Finally, the server can use the current task execution result, the historical execution results of previously executed tasks, and the current task execution chain data as the query results.

[0091] Understandably, the reference information for the initial task does not include information about dependent tasks. Typically, there are no previously completed tasks preceding the initial task, so this initial task will not have dependent task information, and its input usually only includes the query content.

[0092] It is understandable that tasks that have not been completed, that is, tasks that have not received reference information, cannot establish a link with other completed tasks.

[0093] In the above embodiments, the accuracy of the generated task execution chain data can be guaranteed by generating task execution chain data through the dependent task information in different reference information. Then, the task execution chain data is used as the query result. The source of the query result can be clearly identified through the task execution chain data, thereby enhancing the interpretability of the query result and improving its reliability.

[0094] In some embodiments, the current dependent task information includes the identification information of dependent historical tasks, where the dependent historical tasks are executed tasks that have a dependency relationship with the currently executed task.

[0095] Task execution chain data is generated based on historical dependent task information in historical reference information and current dependent task information in current reference information, resulting in current task execution chain data, including:

[0096] The currently executed task is designated as the current task node, and the dependent historical tasks are designated as historical task nodes. A link relationship is established between the historical task nodes and the current task node based on the identification information of the dependent historical tasks included in the current dependent task information. The current task execution chain data is obtained based on the current task node, historical task nodes, and the link relationship.

[0097] The identification information for dependent historical tasks is used to uniquely identify the corresponding dependent historical task, and can be a number, sequence number, code, etc. A dependent historical task refers to a task that a currently executed task needs to perform; that is, the currently executed task requires the execution result of a dependent historical task to execute. The current task node represents the node of the currently executed task in the task execution chain. The historical task node represents the node of the dependent historical task in the task execution chain. The link relationship between the historical task node and the current task node indicates that the execution result of the historical task node is used as input information for the current task node during task execution.

[0098] In this embodiment, the server can use the currently executed task as the current task node and the dependent historical tasks as historical task nodes. Then, the server determines the identifier information of the dependent historical tasks from the current dependent task information. Finally, the server establishes the dependency relationship between the dependent historical tasks and the currently executed task, and then establishes a link relationship between the historical task nodes and the current task node based on the dependency relationship, obtaining the current task execution chain data. This current task execution chain data is used to represent the task execution chain data obtained when the query content completes the currently pending task.

[0099] It's understandable that the links established by the server pointing to the current task node can be either by associating and saving the historical task node with the current task node, or by saving pointers from the historical task node to the current task node. For example, the server can determine that task X depends on the execution of tasks Y1 and Y2 based on reference information. Then, the server can create pointers or associations from the records of tasks Y1 and Y2 to the record X, thereby obtaining a queryable and traversable graph data structure.

[0100] In the above embodiments, current task execution chain data of currently executed tasks and dependent historical tasks is established by using dependent task information. Since the dependent task information is output by the execution agent corresponding to the currently to be executed task, the accuracy of the obtained dependent task information is ensured, thereby guaranteeing the accuracy of the obtained current task execution chain data.

[0101] In some embodiments, the reference information includes dependency data information for obtaining the task execution result and explanatory data information for the task execution result. The dependency data information includes at least one of the following: dependency task information, dependency input information, and dependency reference information. The explanatory data information includes at least one of the following: task execution agent information, task execution inference information, task execution output information, and result confidence information. The dependency input information includes the execution result corresponding to the dependency task in the dependency task information. The dependency reference information includes the inference reference data source information and the reference data source information of the task execution result in the task execution inference information. The task execution output information includes the task execution result.

[0102] S206, which is the query result obtained based on the task execution result and reference information, also includes:

[0103] The query results are returned to the terminal corresponding to the query content. The terminal is used to display the task execution results in the query results and to display the reference information in response to the viewing event of the reference information in the query results.

[0104] The dependency task information includes the identifier of the dependency task. A dependency task is a task that requires a corresponding execution result for the current task to execute. The execution of the current task depends on the execution result of the dependency task. If there is no execution result for the dependency task, that is, if the dependency task has not yet completed its execution, the current task cannot be executed. The dependency input information refers to the input information that the task depends on for execution. This input information can include query parameters from the query content, or it can include the execution results of the dependency tasks.

[0105] Dependency reference information is used to characterize the information referenced during task execution, including the data source information referenced in the task execution inference information and the data source information referenced in the task execution result. The inference reference data source information characterizes the data source information corresponding to the information referenced during inference. This data source information can be information such as the data source link, address, and name; for example, the data source information can be a web address or a database name. The data source information referenced in the task execution result characterizes the data source information corresponding to the obtained task execution result. Explanation data information is used to interpret the task execution result. This task execution agent information refers to the information used to characterize the agent used during task execution. Task execution inference information refers to the information used by the agent for inference during task execution.

[0106] Task execution output information is used to characterize the task execution result. It can be the actual task execution result, or a result formatted according to the reference information. For example, converting the agent's output to JSON (an open standard file and data exchange format) yields the task execution output information from the reference information. Result confidence information characterizes the reliability of the task execution result. A higher confidence level indicates higher accuracy of the obtained result, and thus a more reliable query result.

[0107] Understandably, each query corresponds to a task with relevant reference information, including dependency and explanatory data. Finally, the server returns this reference information to the terminal corresponding to the query. Users can then use the dependency and explanatory data in the displayed reference information to trace the task execution results and provide explanations.

[0108] Understandably, the terminal can simultaneously display the task execution results and reference information from the query results, or it can display only the task execution results. When the terminal detects a viewing event for the reference information, it responds by displaying the reference information. This reference information includes dependency data information for obtaining the task execution results and explanatory data information for the results. The dependency data information can include at least one of the following: dependency task information, dependency input information, and dependency reference information. Therefore, the terminal can simultaneously display dependency task information, dependency input information, and dependency reference information, or it can display only what the user has set up. The reference information also includes explanatory data information for the task execution results. This explanatory data information can include at least one of the following: task execution agent information, task execution inference information, task execution output information, and result confidence information. Therefore, the terminal can simultaneously display task execution agent information, task execution inference information, task execution output information, and result confidence information, or it can display only what the user has set up. For example, if the user has set up to display result confidence information, then only the result confidence information can be displayed.

[0109] In the above embodiments, by displaying parameter information including dependency data and explanatory data on the terminal, the task execution results can be interpreted using both dependency and explanatory data. Specifically, dependency data can be used to trace the source of the task execution results. For example, dependency task information allows viewing the tasks that the current task depends on during execution; dependency input information allows understanding the execution results of upstream tasks; and dependency reference information allows understanding the data sources used during inference and the data sources of the obtained task execution results. Furthermore, explanatory data can be used to explain the task execution results, such as explaining the confidence level of the obtained task execution results, explaining the agent information used during execution, and explaining the inference process during agent execution. This improves the interpretability of the obtained task execution results and, consequently, enhances their reliability.

[0110] In some embodiments, S206, namely obtaining the query results based on the task execution results and reference information, includes the following steps:

[0111] The system retrieves historical task execution chain data corresponding to the query content. This historical task execution chain data is generated based on the reference information corresponding to each historically executed task. It then retrieves new reference information for newly executed tasks. Based on the new dependent task information in the new reference information, the system updates the historical task execution chain data to obtain updated task execution chain data. Finally, it obtains the query results based on the new task execution results corresponding to the newly executed tasks, the historical task execution results corresponding to each historically executed task, and the updated task execution chain data.

[0112] Historical task execution chain data refers to task execution chain data generated using the dependency task information corresponding to historically executed tasks. These historically executed tasks are those that were completed before the newly added task was completed. Newly added executed tasks refer to newly added tasks that have already been completed; that is, the query content can be executed in real-time, and the task execution chain data can also be updated based on tasks that have been completed in real-time. Newly added reference information refers to the reference information obtained when a newly added task is completed. Updated task execution chain data refers to updating historical task execution chain data using the newly added reference information.

[0113] In this embodiment, when the server detects that the task execution of the queried content has been completed, it obtains a newly executed task. At this time, the server can use the reference information of the newly executed task to incrementally update the historical task execution chain data. That is, the server determines the dependent tasks corresponding to the newly executed task from the newly added reference information, then establishes the link relationship between the newly executed task and the dependent tasks, and adds the link relationship and the newly executed task to the historical task execution chain data to obtain updated task execution chain data. For example, the server can obtain from the reference information that task B depends on the execution of tasks A1 and A2. Then the server can create pointers or associations from the records of tasks A1 and A2 to the record of task B, thereby obtaining the task execution chain data of tasks A1, A2, and task B. At this time, a newly executed task C is added, and the dependent tasks corresponding to the executed task C are B and A1. Then, a link relationship is established from task B and A1 to task C. At this time, updated task execution chain data is obtained, which includes A1, A2, B, and C, and also includes the link relationship from task A1 and A2 to task B and the link relationship from task B and A1 to task C.

[0114] Understandably, different tasks will complete at different times. The server can then obtain the latest task execution chain data, and when the last task in the query completes, it will obtain the overall task execution chain data corresponding to the query content. At this point, the server can persistently save the overall task execution chain data corresponding to the query content.

[0115] In the above embodiments, when a new executed task is obtained, the new reference information of the new executed task can be used to update the historical task execution chain data. That is, the task execution chain data can be incrementally generated according to the execution process of the query content, so as to avoid missing completed tasks and ensure the accuracy of the obtained task execution chain data.

[0116] In some embodiments, historical task execution chain data is updated based on newly added dependent task information in the newly added reference information to obtain updated task execution chain data, including:

[0117] The newly added executed tasks are treated as new task nodes in the historical task execution chain data; the executed tasks that the newly added executed tasks depend on are determined from the information of the newly added dependent tasks, and a new link relationship is established between the existing task nodes corresponding to the dependent executed tasks and the newly added task nodes; based on the newly added task nodes and the new link relationship, the updated task execution chain data corresponding to the historical task execution chain data is obtained.

[0118] The newly added task node represents a newly added executed task. The newly added link relationship represents that the execution result of the newly added executed task depends on the output of the corresponding executed task.

[0119] In this embodiment, the server adds newly executed tasks and new links to the historical task execution chain data to obtain updated task execution chain data. For example, the server can add corresponding new task nodes and corresponding new links to the saved graph structure data to obtain updated task execution chain data.

[0120] It is understandable that a newly added executed task can depend on at least two other executed tasks. Each dependent executed task establishes a corresponding new link relationship with the newly added executed task. The server then adds the newly added executed task and all corresponding new link relationships to the current task execution chain data, thus obtaining updated task execution chain data.

[0121] In the above embodiments, by adding task nodes and new link relationships to the task execution link data, updated task execution link data is obtained. This ensures that the obtained task execution link data is consistent with the task execution process of the query content, thereby guaranteeing the consistency of the obtained query results.

[0122] In some embodiments, the tasks to be performed include at least two, and the dependency data information includes dependency task information;

[0123] S206, which refers to obtaining query results based on task execution results and reference information, includes the following steps:

[0124] Retrieve the task execution chain data corresponding to the query content. The task execution chain data includes each task node and the corresponding link relationship between each task node. The link relationship is determined based on the dependent task information of each task node. Retrieve the query page layout information corresponding to the query content. Based on the query page layout information and the task execution chain data, perform page layout calculations to obtain the task execution flowchart information corresponding to the task execution chain data. Obtain the query results of the query content based on the task execution results and task execution flowchart information corresponding to each task node. Return the query results of the query content to the query page corresponding to the query content. The query page is used to display the task execution results and task execution flowchart information corresponding to each task node.

[0125] The task execution chain data refers to the task execution chain data generated by the executed tasks corresponding to the query content. The query page layout information describes the layout and size of the query page, which can include the page dimensions and formatting information such as fonts and graphics. The task execution flowchart information represents the flowchart of the execution process corresponding to the query content; for example, the task execution flowchart can be a mind map of the query content during task execution. This flowchart information can include information about each task node and the connections between them. The task node information can also include the display position of the task node on the page.

[0126] In this embodiment, the server can perform layout calculations based on the task execution chain data to obtain task execution flowchart information. At this point, the server can return the task execution flowchart information as a query result to the query page corresponding to the query content. The queried task can be rendered on the page according to the task execution flowchart information, resulting in the task execution flowchart displayed on the query page. That is, the query page can simultaneously display the task execution flowchart corresponding to the query content and the task execution result.

[0127] Understandably, the query page displays the task execution flowchart within a pre-defined flowchart display area. The server then calculates the flowchart layout based on the layout information of the flowchart display area to obtain the task execution flowchart information. The layout information of the flowchart display area may include information such as area size, area font, and graphic or table parameters.

[0128] In a specific embodiment, such as Figure 3The diagram illustrates the framework for obtaining query results. The server receives the query content, plans it to obtain tasks to be executed, and then executes each task through its corresponding execution agent, obtaining the execution results and reference information for each task. Next, the server uses the dependent task information from the reference information to generate task execution chain data, obtaining the task execution chain data corresponding to the query content. Then, the server uses the task execution chain data corresponding to the query content to perform page layout calculations, obtaining the task execution flowchart information corresponding to the query content. Finally, the server returns the task execution flowchart information and task execution results to the terminal. The terminal then renders the page content according to the task execution flowchart information and task execution results, resulting in a query page displaying the task execution flowchart and task execution results. For example, as shown... Figure 4 The image shows a schematic of the query page. This page includes the query content and the corresponding task execution flowchart on the left, and the task execution results on the right. The task execution results displayed on the right are typically the results of the last pending task node. For example, if the last pending task node is a summary task node, then the summary task execution result will be displayed. Users can view the task execution flowchart and task execution results on the query page.

[0129] In the above embodiments, the task execution flowchart information corresponding to the task execution chain data is obtained by calculating the page layout according to each task node and the corresponding link relationship. Using the link relationship determined by the dependency relationship generated by the executing agent for page layout calculation ensures the accuracy of the obtained task execution flowchart information.

[0130] In some embodiments, page layout calculation is performed based on query page layout information and task execution chain data to obtain task execution flowchart information corresponding to the task execution chain data, including the following steps:

[0131] The task nodes in the task execution chain data are layered to obtain the task nodes in each layer; the task nodes in each layer are sorted to obtain the sorted task nodes in each layer; and the page position of the sorted task nodes in each layer is calculated based on the query page layout information to obtain the page position information of the task nodes in the task execution chain data; the connection edge information of the task nodes is determined according to the link relationship of the task nodes in the task execution chain data, and the task execution flowchart information is obtained based on the page position information and the connection edge information of the task nodes.

[0132] The page location information is used to represent the specific location of the task node in the query page. It can be represented by page location coordinates, that is, the location coordinates are planar coordinates, and each task node is assigned a corresponding planar coordinate.

[0133] In this embodiment, the server can automatically perform layout calculations using graph structure data, i.e., task execution chain data, through a hierarchical layout algorithm, generating aesthetically pleasing and highly readable task execution flowchart information. Specifically, the server can detect whether there are circular dependencies in the graph structure data represented by the task execution chain data. If a circular dependency is found, the direction of the edges corresponding to the link relationships can be temporarily reversed to obtain a directed acyclic graph (DAG) and the reversed edges are recorded. Finally, the reversed edges are restored when the layout calculation is complete. If no circular dependencies are found, the server assigns task nodes to different levels according to the direction of the edges corresponding to the link relationships, aiming to minimize the number of levels crossed by an edge. At this time, the server sorts nodes within the same level, i.e., the server adjusts the order of task nodes within the same level to reduce the intersection between edges and improve readability. Finally, the server calculates the coordinates of the task nodes, i.e., assigns specific planar coordinates to each task node and draws the connecting edges. The purpose of the entire process is to transform abstract graph structure data into a clearly laid out and easily understood visual graph.

[0134] In the above embodiments, by layering and sorting the task nodes in the task execution chain data, and then calculating the page position and determining the connection edge information, the layout accuracy and visibility of the obtained task execution flowchart can be improved because layering and sorting can reduce the number of layers crossed by the edges and reduce the intersection between the edges.

[0135] In some embodiments, page layout calculation is performed based on query page layout information and task execution chain data to obtain task execution flowchart information corresponding to the task execution chain data, including the following steps:

[0136] Based on the query page layout information and the task nodes in the task execution chain data, the repulsive and attractive forces between nodes are simulated and calculated to obtain the simulated repulsive and attractive forces of the task nodes.

[0137] The page position corresponding to the task node is updated based on the simulated repulsive and simulated attractive forces of the task node to obtain the updated page position of the task node. Iterative calculations are then performed based on the updated page position of the task node until the force-directed layout completion condition is met, at which point the task execution flowchart information corresponding to the task execution chain data is obtained.

[0138] The simulated repulsive force of task nodes is used to characterize the simulated repulsive force between all task nodes, preventing excessive clustering and reducing overlap. The simulated attractive force of task nodes is used to characterize the simulated attractive force between linked tasks, maintaining the structural integrity of the graph.

[0139] In this embodiment, the server can simulate task nodes in the task execution chain data, i.e., the graph data structure of task execution, as charged particles that repel each other, and simulate edges as springs connecting dependent task nodes. Then, by simulating repulsive and attractive forces in a physical system, and through multiple iterative calculations, the task nodes gradually move to a position of force equilibrium, ultimately forming a graphical structure with evenly distributed task nodes, few edge intersections, and high readability, thus obtaining the task execution flowchart information. That is, the server can use a force-directed layout algorithm for layout calculations. This algorithm optimizes the position of graphical nodes by simulating the mechanical system in the physical world, generating a clear and balanced graph layout. For example, the server can randomly or heuristically set the initial coordinates of all task nodes based on the page size. Then, it iterates through each pair of task nodes, calculates the repulsive force between task nodes, and determines the net repulsive force of each task node. Then, it iterates through each edge, calculates the attractive force connecting task nodes, and updates the net attractive force of each task node. Finally, it updates the node positions: adjusting the node positions based on the net force (the resultant force of repulsive and attractive forces), typically combined with physical simulations of velocity and acceleration. Finally, the server performs iterative optimization, repeating the above steps until the force-directed layout completion conditions are met, such as reaching the maximum number of iterations or the energy change being less than a preset threshold. That is, the energy change being less than the preset threshold can be used to characterize that the position error of the task node before and after the iteration is less than a preset difference.

[0140] In the above embodiments, by calculating the simulated repulsive force and simulated attractive force of the task nodes and iteratively updating the page position corresponding to the task node, the task execution flowchart information corresponding to the task execution chain data is obtained. This can make the task nodes in the obtained task execution flowchart evenly distributed, with fewer edge intersections and high readability, thus improving the accuracy and visibility of the obtained task execution flowchart.

[0141] In some embodiments, after returning the query results of the query content to the query page corresponding to the query content, and the query page is used to display the task execution results corresponding to each task node and the task execution flowchart corresponding to the task execution flowchart information, the method further includes the following steps:

[0142] The system retrieves a view request for a selected task node in the task execution flowchart. In response, it obtains the selection reference information corresponding to the selected task node. This selection reference information includes the execution agent information of the selected task node, the input information of the selected task node, the execution inference information of the selected task node, the output information of the selected task node, the dependent task information of the selected task node, the reference information of the selected task node, and the result confidence information of the selected task node. The system then returns the selection reference information corresponding to the selected task node to the query page corresponding to the query content. The query page displays the execution agent information, input information, execution inference information, output information, dependent task information, reference information, and confidence information.

[0143] Here, "Selected Task Node" refers to a task node selected through a selection operation. The selection operation is the trigger action that selects a task node displayed on the query page. Trigger actions can include clicking, pressing, swiping, voice commands, gestures, etc. "View Request" refers to a request to find the reference information corresponding to the selected task node. "Selected Reference Information" refers to the reference information corresponding to the task represented by the selected task node. "Executing Agent Information" refers to the information of the executing agent used when the selected task node performs the task, which may include the agent's identifier, role, and functional description. "Input Information" refers to the information input to the executing agent when the selected task node performs the task, which may include the output of dependent tasks and pre-set inputs for task execution. These pre-set inputs may include prompts for the actions the executing agent should perform, such as a prompt to summarize the output of dependent tasks. "Execution Inference Information" describes the reasoning process of the executing agent during task execution, including the decision basis for the reasoning process. This decision basis can be the source of the data referenced in the reasoning, such as web page links, document names, database identifiers, etc. "Selected Task Execution Result" refers to the execution result of the task represented by the selected task node. Dependency task information refers to information about completed tasks that the task depends on. This can be an identifier of a completed task and is used to construct the dependency relationships in the task execution chain. Referenced data source information refers to the source information of the data referenced by the task represented by the selected task node during execution. This can include source information such as webpage links, document names, and database identifiers. Confidence information is used to characterize the reliability of the selected task's execution result. For example, it can be a confidence score, a confidence quantification value, a good confidence level, etc.

[0144] In this embodiment, the user can select a task node in the task execution flowchart through the query page on the terminal. At this time, the server receives the viewing request sent by the terminal, parses the query request to obtain the task identifier of the selected task node, and locates the pre-saved reference information based on the task identifier to obtain the selection reference information corresponding to the selected task node. Finally, the server can return the found selection reference information to the user's terminal. Upon receiving the selection reference information returned by the server, the terminal can display the selection reference information corresponding to the selected task node on the query page. This selection reference information includes the executing agent information, input information, execution inference information, output information, dependent task information, reference information, and confidence information corresponding to the selected task node.

[0145] Understandably, the reference information is the metadata object corresponding to the task result. This metadata object contains at least the following fields: agent_id (agent identifier): A unique identifier for the agent. Description: A description of the agent's role and function. Input: The specific input content for this call. Output: The result output for this call. Rationale: A textual description of the decision basis or reasoning process (generated by the agent). data_sources: A list of referenced data sources. Confidence: A confidence score (a quantified value or level) for the output result. parent_id: The identifier of the parent task node. depends_on: A list of identifiers of upstream tasks to be executed that the task depends on.

[0146] In a specific embodiment, such as Figure 5 The diagram illustrates the selection of a task node. Users can view any task node in the task execution flowchart on the query page. For example, task node B depends on task node A, the connection edge of task node A points to task node B, and the user can click on task node B. In this case, the terminal responds to the view event for task node B by sending a view request to the server on the query page. The server responds to the view request by finding the associated reference information B based on task node B. This reference information B includes the specific information of each field in the metadata object. Finally, the server returns the found reference information B to the query page. The terminal can then display the reference information B in a pop-up window on the query page.

[0147] In the above embodiments, by obtaining a viewing request, users can view the selected reference information corresponding to the selected task node on the query page. Since the selected reference information includes information about the executing agent, input information, execution inference information, output information, dependent task information, reference information, and confidence information, it can clearly and completely display the detailed input, output, and decision basis of the task node, including the basis for the result generation, the data source it depends on, the inference logic, and the confidence assessment, thereby solving the problem of opaque decision-making within complex intelligent agent systems, improving the interpretability of query results, and thus improving the reliability of query results.

[0148] In some embodiments, the dependency task information includes parent dependency task information and non-parent dependency task information, and the link relationship includes parent-child link relationship and non-parent-child link relationship;

[0149] The query results will be returned to the corresponding query page. This page displays the task execution results for each task node and the corresponding task execution flowchart, including:

[0150] The query results are returned to the corresponding query page. The query page displays the task execution results for each task node, as well as the target task execution flowchart that shows and hides the connection edges corresponding to non-parent-child links. The target task execution flowchart responds to the view event for the parent-child connection edge corresponding to the parent-child link, highlighting the parent-child connection edge, the parent task node and child task node corresponding to the parent-child connection edge, and the non-parent task node corresponding to the child task node and the non-parent task node corresponding to the non-parent-child connection edge.

[0151] Among them, parent dependency task information refers to the information about the parent tasks that the current task depends on during execution, which can be the identifier of the parent task. This parent relationship can be determined according to the execution order of the tasks to be executed corresponding to the query content. Non-parent dependency task information refers to the information about the non-parent tasks that the current task depends on during execution. These non-parent tasks can be determined based on the tasks corresponding to the inputs that the current task depends on during execution. For example, if the input includes the output of a non-parent task, then that non-parent task is considered the non-parent task corresponding to the current task. Parent-child link relationship refers to the link relationship determined according to parent dependency task information. Non-parent-child link relationship refers to the link relationship determined according to non-parent dependency task information. Parent-child connection edge refers to the connection edge drawn according to the parent-child link relationship. The parent-child connection edge viewing event refers to the trigger event for finding relevant information corresponding to the parent-child link relationship, which can be triggered by a trigger operation. Non-parent-child connection edge refers to the connection edge drawn according to the non-parent-child link relationship. Parent task node refers to the flowchart node corresponding to the parent task connected by the parent-child connection edge. A child task node is a flowchart node corresponding to a child task connected by a parent-child connection edge. A non-parent task node is a flowchart node corresponding to a child task that is not its parent task.

[0152] In this embodiment, the dependent task information includes parent dependent task information and non-parent dependent task information. Therefore, the link relationships determined when generating the task execution chain based on the dependent task information include parent-child link relationships corresponding to the parent dependent task information and non-parent-child link relationships corresponding to the non-parent dependent task information. At this time, when the server returns the query results of the query content to the query page corresponding to the query content, the terminal where the query page is located can draw the target task execution flowchart by hiding the connection edges corresponding to the non-parent-child link relationships based on the parent-child link relationships. Then, the terminal detects a viewing event for the parent-child connection edge in the target task execution flowchart and highlights the nodes and connection edges related to the parent-child connection edge. These related nodes and connection edges include parent-child connection edges, parent task nodes, child task nodes, non-parent-child connection edges, and non-parent task nodes. Highlighting refers to distinguishing the node from other task nodes and connection edges; for example, compared to other task nodes and connection edges, highlighting can be achieved through methods such as highlighting, magnification, or different color display.

[0153] In a specific embodiment, such as Figure 6 The diagram shows a parent-child connection edge. When the user clicks on the connection edge 605 between task node 601 and task node 602, task node 601, task node 602, and the clicked connection edge 605 are highlighted. At the same time, the non-parent-child connection edge 604 between task node 601 and task node 603, as well as task node 603, are highlighted.

[0154] In the above embodiments, by first displaying the target task execution flowchart with hidden connection edges corresponding to non-parent-child links, the number of edges in the displayed flowchart can be reduced, saving display resources and improving flowchart visibility. Then, in response to a view event for a parent-child connection edge corresponding to a parent-child link, the parent-child connection edge, the parent task node and child task node corresponding to the parent-child connection edge, and the non-parent task node corresponding to the child task node and the non-parent task node corresponding to the non-parent-child connection edge are highlighted, ensuring the completeness and detail of the displayed reference information, thereby improving the efficiency of tracing.

[0155] In some embodiments, S204, namely, performing task execution based on the task to be executed, and obtaining the task execution result and reference information corresponding to the task to be executed, includes the following steps:

[0156] Obtain at least two execution agents corresponding to the task to be executed; input the task to be executed into the at least two execution agents to perform task execution, and obtain at least two task execution results corresponding to the task to be executed and reference information corresponding to the at least two task execution results respectively;

[0157] S206, which refers to obtaining query results based on task execution results and reference information, includes the following steps:

[0158] At least two query results corresponding to the query content are obtained based on at least two task execution results and the reference information corresponding to each of the at least two task execution results.

[0159] In this embodiment, different execution agents may produce different execution results for the same task. For example, different inference processes or differences in the data sources used may produce different execution results. In this case, the server can simultaneously obtain the execution results generated by different execution agents and perform branch execution according to the execution results generated by different execution agents, thereby obtaining query results corresponding to different branches of the query content. For example, if two execution agents generate corresponding execution results for task A, these two execution results can be used to execute subsequent tasks, resulting in query results divided into two branches starting from task A.

[0160] It is understandable that the same execution agent may produce different execution results for the same task. In this case, the execution agent can also output all execution results and corresponding reference information.

[0161] In the above embodiments, by obtaining at least two task execution results corresponding to the task to be executed and the reference information corresponding to the at least two task execution results respectively, at least two query results corresponding to the query content can be obtained. That is, it is not necessary to select a final answer from them through a simple voting mechanism or rules, so it is easy to locate the root cause of the disagreement, such as whether the data source is different or the reasoning logic is different. This greatly facilitates result reproduction, problem investigation and auditing, and makes it easier to evaluate the reliability of different answers, further improving the reliability of the obtained query results.

[0162] In some embodiments, at least two query results corresponding to the query content are obtained based on at least two task execution results and reference information corresponding to the at least two task execution results, including:

[0163] Based on the reference information corresponding to at least two task execution results, a task execution chain is generated to obtain at least two current task execution chain data corresponding to the query content; the query page layout information corresponding to the query content is obtained, and the page layout is calculated based on the query page layout information and at least two current task execution chain data to obtain at least two task execution flowchart information corresponding to the query content; the at least two task execution results and the corresponding task execution flowchart information are returned to the query page corresponding to the query content, and the query page is used to compare and display at least two task execution results and the corresponding task execution flowchart.

[0164] In this embodiment, the server can generate task execution flowcharts for different branches of the same executed task. Specifically, it uses reference information corresponding to different task execution results of the executed task to generate task execution chain data for each branch, obtaining graph structure data for each branch. Then, it uses the generated different task execution chain data to perform page layout calculations, obtaining task execution flowchart information for each branch. This allows for parallel page layout calculations, improving computational efficiency. Finally, the server returns the task execution flowchart information for each branch and the corresponding task execution result to the terminal. The terminal then compares and displays each task execution result and its corresponding task execution flowchart on the query page corresponding to the query content.

[0165] In a specific embodiment, such as Figure 7The diagram shows a comparison of task execution flowcharts. The query page displays two task execution flowcharts, representing different execution results obtained by executing task node 702 through two different execution agents. Subsequent tasks are then executed according to these different results. The upper area displays the task execution flowchart following the first task's execution result, and the corresponding task execution result is displayed in the task execution result display area. Reference information for the task node to be viewed is also displayed in the reference information display area. Similarly, the lower area displays the task execution flowchart following the second task's execution result, and the corresponding task execution result is displayed in the task execution result display area. Reference information for the task node to be viewed is also displayed in the reference information display area.

[0166] In the above embodiments, by generating task execution flowchart information with different task execution results, at least two task execution results and their corresponding task execution flowcharts can be compared and displayed, which facilitates the source analysis of different execution results and enhances the reliability of task execution results and the scientific nature of decision-making.

[0167] In some embodiments, after returning at least two task execution results and corresponding task execution flowchart information to the query page corresponding to the query content, and the query page being used to compare and display at least two task execution results and corresponding task execution flowcharts, the method further includes:

[0168] Obtain the edit request for the target task node in the target task execution flowchart, respond to the edit request to obtain the target edit task, which is obtained by editing the target task of the target task node according to the reference information corresponding to the target task node; restart execution from the target edit task to obtain the update task execution flowchart information and the edit task execution result corresponding to the target task execution flowchart information; return the update task execution flowchart information and the edit task execution result to the query page corresponding to the query content, and the query page is used to display the update task execution flowchart and the edit task execution result corresponding to the target task execution flowchart.

[0169] The target task execution flowchart refers to the task execution flowcharts displayed on the query page that require editing. A target task node is a task node in the target task execution flowchart that needs editing. This editing can involve operations such as adjusting, deleting, inserting, adding, and replacing the target task; these operations can be determined by the user's editing requirements. An edit request is a request to re-edit the task executed by the target task node according to the user's editing requirements. The target edit task is the task obtained after editing the task of the target task node according to the edit request. Updating the task execution flowchart information refers to the layout calculation obtained using the task execution chain data corresponding to the target edit task; this task execution chain data is generated using the reference information corresponding to the target edit task. The edit task execution result refers to the result obtained by executing the target edit task through the corresponding execution agent.

[0170] In this embodiment, users can edit the task nodes displayed on the query page. Users submit their editing requests in natural language, based on the reference information corresponding to the target task node. These node editing requests instruct how the task node should be changed to meet the requirements. For example, if the data source for the target task node's reference information is platform A, and the user discovers an anomaly in the data from platform A, the user needs to change the data source of the target task node. The editing request could be "obtain data directly from platform B." In this case, the task to be performed by this task node is to obtain data from platform B. The server can generate a target editing task to obtain data from platform B, execute the target editing task, and perform subsequent tasks based on the execution result. When a user edits a target task node, triggering the node editing operation, the terminal displays a session area corresponding to the target task node for interacting with the large language model. For example, the target task node in the flowchart may display an edit button. When the user clicks the edit control in the target task node, the node editing operation is triggered, and the terminal displays a session area corresponding to that target task node for interacting with the large language model. Users can input their node editing needs in the chat area and then click the submit control displayed in the chat area. At this time, the terminal detects an edit event for the target task node, responds to the edit event by sending an edit request to the server, and the server responds to the edit request by generating the corresponding target edit task based on the node editing needs information input by the user in the edit event and using the large language model. At this point, the server restarts the execution of the target edit task according to the pending tasks and execution order of the query content. When the target edit task is completed, the server generates the corresponding update task execution flowchart information based on the reference information corresponding to the execution result of the target edit task. Finally, the server returns the update task execution flowchart information and the edit task execution result to the query page corresponding to the query content. The terminal updates the target task execution flowchart to the update task execution flowchart on the query page and updates the task execution result corresponding to the target task execution flowchart to the edit task execution result.

[0171] In a specific embodiment, such as Figure 8The diagram illustrates task node editing, with an edit button 801 displayed on the query page. When the user clicks the edit button 801, all task nodes in the task execution flowchart become editable. The user can then select any task node to edit, such as task node 802. The terminal then displays a conversation area 803 for interacting with task node 802. The user can then input their editing needs for task node 802 in the conversation area and click the send button. The terminal will then re-execute task node 802 and display the updated task execution flowchart and the edited task execution result.

[0172] In the above embodiments, by editing the target task of the target task node according to the reference information corresponding to the target task node, the target task node can be updated, avoiding the use of execution results corresponding to biased reference information for subsequent task execution, thereby obtaining more accurate node task execution results, and thus improving the accuracy of the query results corresponding to the query content.

[0173] In one embodiment, such as Figure 9 As shown, a task execution method is provided, which can be applied to Figure 1 Taking a terminal as an example, it can be understood that this method can also be applied to a server, and to a system that includes both a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0174] S902, Display the query results on the query page;

[0175] S904, Display the task execution results and reference information corresponding to the query content. The reference information includes the dependency data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the task to be executed corresponding to the query content. The task to be executed is obtained by planning the query content and using it to obtain the query results.

[0176] The query page can be a page where users interact with the multi-agent collaborative system to query content. That is, the query page can be a conversational interaction for content querying. Users input a conversational message through the query page, which becomes the query content. The multi-agent collaborative system then responds to the user's input, including the task execution results and reference information corresponding to the query content.

[0177] In this embodiment, the terminal detects a user's access to the query page and displays the query page in response. The user can log in to the query page using a pre-registered account. Then, the user enters the query content in the session message input area of ​​the query page and clicks the send button. At this time, the terminal displays the query content on the query page and sends a query request including the query content to the server. The server receives the query request and, in response, executes the query content according to the method steps of any of the above-described task execution methods to obtain the query result, which includes the task execution result and reference information. The server returns the query result to the terminal. The terminal displays the corresponding task execution result and reference information on the query page as a reply message.

[0178] Understandably, users can perform multiple queries on the query page. In this case, the query page will display each query result, the corresponding task execution result, and reference information in chronological order of the query time.

[0179] In a specific embodiment, such as Figure 10 The diagram shows the task execution results and reference information. The query page displays the session message sent by the user through the message sending area, i.e., query content A, and the reply session message, i.e., the task execution results and reference information corresponding to query content A. It can be the task execution results and reference information corresponding to the last task to be executed, or it can display the task execution results and reference information corresponding to each task to be executed.

[0180] The above task execution method displays the query content on the query page. Then, it displays the task execution results and reference information corresponding to the query content. The reference information includes dependency data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the tasks to be executed corresponding to the query content. The tasks to be executed are obtained by planning the query content and using it to obtain the query results. Therefore, the obtained query results include not only the task execution results corresponding to the tasks to be executed, but also the dependency data information for obtaining the task execution results. That is, interpreting the obtained task execution results through dependency data information improves the interpretability of the obtained query results, thereby improving the reliability of the query results.

[0181] In some embodiments, S904, which involves displaying the task execution results and reference information corresponding to the query content, includes the following steps:

[0182] The system displays the task execution results and a task execution flowchart corresponding to the query content. The task execution flowchart includes each task node and the connecting edges between each task node. The connecting edges are determined based on the linking relationships between each task node, which are determined based on the dependent task information corresponding to each task node. In response to a view event for a selected task node in the task execution flowchart, the system displays the selected reference information corresponding to the selected task node. The selected reference information includes the executing agent information of the selected task node, the input information of the selected task node, the execution inference information of the selected task node, the output information of the selected task node, the dependent task information of the selected task node, the reference information of the selected task node, and the result confidence information of the selected task node.

[0183] In this embodiment, the terminal can also display the task execution results and task execution flowchart corresponding to the query content on the query page. This task execution flowchart can be obtained by executing the method steps described in any of the above-described task execution methods. The user can then view the task nodes in the displayed task execution flowchart. At this time, the terminal displays selection reference information for the selected task node on the query page. This selection reference information can be displayed through a pop-up window or in a pre-set reference information display area on the query page.

[0184] In some embodiments, a user can perform multiple session interactions through the query page. The query page can then sequentially display the query content of each session interaction, as well as the corresponding task execution results and task execution flowchart.

[0185] It's understandable that when displaying the selection reference information for a selected task node, the terminal can show the referenced data source information of the selected task's execution result via a trigger button within the selected task's execution result. Then, when the user triggers the data source view control for the corresponding task's execution result, the referenced data source information is displayed. For example, a search page might display attribute A of the searched mobile phone A, along with a data source view control for attribute A. When the user clicks or moves to the data source view control, the data source for attribute A is displayed; this could be a source webpage, a source database, etc. For example... Figure 11 The diagram illustrates the data source viewing control. When the user clicks task node 1101, reference information corresponding to task node 1101 is displayed on the left. This reference information includes attribute A of mobile phone A and the data source viewing control 1102 for attribute A. The user can view the data source information for attribute A by clicking the data source viewing control 1102.

[0186] Understandably, the entire decision-making chain corresponding to the query content (including metadata of all intermediate steps) can be fully recorded and persistently stored to form an audit log, that is, the task execution chain data corresponding to the query content is stored in the database. Users can review the complete execution chain of queries in historical sessions at any time. The task execution flowchart corresponding to each query content can be exported in image formats such as PNG and SVG or structured formats such as Mermaid and JSON, which is convenient for inclusion in reports or for offline analysis.

[0187] In the above embodiments, by displaying the task execution results and task execution flowchart corresponding to the query content, and then responding to the viewing event of the selected task node in the task execution flowchart, the selected reference information corresponding to the selected task node is displayed. That is, by only displaying the task execution flowchart and not the reference information, page display resources and visibility can be saved. Then, when viewing the selected task node, the viewed reference information is displayed to facilitate tracing, thereby improving the interpretability and reliability of the task execution results.

[0188] In some embodiments, the dependency task information includes parent dependency task information and non-parent dependency task information, the link relationship includes parent-child link relationship and non-parent-child link relationship, and the task execution flowchart includes the target task execution flowchart that hides the connection edges corresponding to non-parent-child link relationships.

[0189] The task execution method also includes the following steps:

[0190] In response to a view event on the parent-child connection edge of the target task execution flowchart, highlight the parent-child connection edge, the parent task node and child task node corresponding to the parent-child connection edge, and highlight the non-parent-child connection edge corresponding to the child task node and the non-parent task node corresponding to the non-parent-child connection edge.

[0191] In this embodiment, the task execution flowchart displayed by the terminal is the target task execution flowchart that hides the connection edges corresponding to non-parent-child link relationships. When the user needs to trace the source, he / she can click on the connection edge in the target task execution flowchart to view it. The clicked connection edge is the parent-child connection edge. At this time, the terminal responds to the viewing event for the clicked parent-child connection edge and highlights all information related to the clicked connection edge, including the hidden connection edges corresponding to non-parent-child link relationships.

[0192] Understandably, in response to a "cancel viewing" event for a viewed parent-child connection edge, the terminal removes the highlighted parent-child connection edge, its corresponding parent task node, and child task node; hides the non-parent-child connection edges corresponding to the child task nodes; and removes the non-parent task nodes corresponding to the unhighlighted non-parent-child connection edges. This "cancel viewing" event refers to the trigger event for canceling the viewing of parent-child connection edges, which is triggered by a trigger operation. "Cancel highlighting" can mean restoring the display to its previous state, i.e., the terminal displays the restored target task execution flowchart.

[0193] Understandably, when users view task nodes in the target task execution flowchart, they can also simultaneously highlight the hidden non-parent-child link edges and view reference information for the task nodes. This highlighting of hidden non-parent-child link relationships refers to the non-parent-child link relationships corresponding to the viewed task nodes.

[0194] In the above embodiments, by viewing events for parent-child connection edges, the parent-child connection edge, the parent task node and child task node corresponding to the parent-child connection edge can be highlighted, as well as the non-parent-child connection edge corresponding to the child task node and the non-parent task node corresponding to the non-parent-child connection edge can be highlighted. All information related to the parent-child connection edge can be clearly viewed, which facilitates tracing and improves the interpretability and reliability of task execution results.

[0195] In some embodiments, the task execution results and task execution flowchart corresponding to the query content are displayed, including the following steps:

[0196] The comparison shows at least two task execution flowcharts corresponding to the query content and the task execution results corresponding to the at least two task execution flowcharts. The at least two task execution flowcharts are obtained by page layout calculation based on the at least two current task execution chain data corresponding to the query content. The at least two current task execution chain data are generated by task execution chain generation based on the reference information corresponding to the at least two task execution results. The reference information corresponding to the at least two task execution results is obtained by inputting the task to be executed corresponding to the query content into at least two execution agents for task execution.

[0197] In this embodiment, the terminal can simultaneously compare and display at least two task execution flowcharts corresponding to the unified query content and the task execution results corresponding to each task execution flowchart. For example, the terminal can display each task execution flowchart sequentially from the beginning to the end of the query page, i.e., each task execution flowchart is on a separate line. Alternatively, the terminal can display each task consultation flowchart side by side on the query page, i.e., each task execution flowchart is on a separate column.

[0198] Understandably, users can compare the differences between each task execution flowchart during execution by viewing the reference information of each task node. These differences might include variations in the data sources used, the inference process, and so on. Users can then select the more reliable task execution flowcharts to keep and delete the others. The terminal will then display the final retained task execution flowcharts and their corresponding execution results on the query page.

[0199] In the above embodiments, by comparing and displaying at least two task execution flowcharts corresponding to the query content and the task execution results corresponding to at least two task execution flowcharts, users can intuitively analyze the root cause of the divergence, which facilitates tracing the source and thus increases the reliability of the obtained task execution results.

[0200] In some embodiments, after comparing and displaying at least two task execution flowcharts corresponding to the query content and the task execution results corresponding to the at least two task execution flowcharts, the method further includes the following steps:

[0201] In response to an edit event for a target task node in the target task execution flowchart, the updated task execution flowchart and the edited task execution result are displayed. The updated task execution flowchart and the edited task execution result are obtained by restarting the target edited task corresponding to the target task node.

[0202] In this embodiment, the user can edit the task nodes in the compared task execution flowchart. This can be done by viewing the reference information of the task node and determining whether editing is necessary. If editing is required, such as modifying the referenced data source or pointing out errors in the inference process, an edit event can be triggered for the target task node. At this time, the terminal re-executes the task according to the edited task node, resulting in an updated task execution flowchart and the edited task execution result. The updated task execution flowchart includes the task nodes that remained unchanged after being executed before the edited task node was executed, the task nodes obtained by re-executing the edited task node, and the task nodes obtained by re-executing the task nodes that can only be executed after the edited task node was executed.

[0203] Understandably, users can also edit the entire task execution flowchart. For example, if a user finds that multiple task nodes in the task execution process are inconsistent with the user's needs by viewing the reference information of each task node in the task execution flowchart, and the task execution results represented by other task execution flowcharts are more reliable, then the user can delete the entire task execution flowchart. At this time, the terminal displays the page content after deleting the task execution flowchart on the query page.

[0204] Understandably, if the query page only displays a single task execution flowchart, users can edit the task nodes within that flowchart, such as modifying the data source referenced by the task node. In this case, the terminal responds to the edit event and displays the updated task execution flowchart obtained by restarting execution from the edited task node, along with the corresponding updated task execution results.

[0205] In the above embodiments, by triggering an edit event, the task nodes in the task execution flowchart can be edited, thereby obtaining a displayed updated task execution flowchart and edited task execution results. This allows users to edit and modify task nodes in real time based on reference information, further improving the reliability of the obtained task execution results.

[0206] In some embodiments, the query content includes product recommendations;

[0207] Displays the task execution results and reference information corresponding to the query content, including:

[0208] Displays the product recommendation results and the product recommendation task execution flowchart corresponding to the product recommendation content; in response to the view event of the product search task node in the product recommendation task execution flowchart, displays the product search reference information corresponding to the product search task node. The product search reference information includes the executing agent information of the product search task node, the input information of the product search task node, the product search output information of the product search task node, the dependent task information of the product search task node, the reference information of the product search task node, and the result confidence information of the product search node.

[0209] Here, "Product Recommendation Content" refers to the query content for recommending products. For example, it could be, "I recently bought a phone, what cost-effective headphones can you recommend? Give me a comprehensive recommendation analysis report." "Product Recommendation Results" refers to the results obtained from searching and recommending products based on the product recommendation content. For example, it could be "Recommendation analysis report for headphones A, B, and C." "Product Recommendation Task Execution Flowchart" is the flowchart for executing the task corresponding to the product recommendation content. "Product Search Task Nodes" represent the nodes that perform product search tasks. A product search task is a task to be executed, searching for product information. "Product Search Task Execution Results" refers to the results obtained from executing the product search task, which can include the searched product information. For example, the product search task execution results could be detailed information about headphones A, B, and C, including headphone attributes, performance information, etc. "Product Search Reference Information" refers to the reference information corresponding to the product search task execution results, such as the source webpage corresponding to the attribute information data of headphone A.

[0210] In this embodiment, the task execution method is applied to a scenario involving product recommendations, where a user can input product recommendations on a query page. The terminal detects a query event for product recommendations and sends the corresponding query request to the server. The server receives the query request and parses it to obtain the product recommendations carried within. Then, the server executes the steps in any of the embodiments described above to obtain the product recommendation results corresponding to the product recommendations, as well as a flowchart of the product recommendation task execution. Finally, the server returns the product recommendation results and the flowchart to the terminal. The terminal displays the product recommendation results and the flowchart on the query page.

[0211] In the above embodiments, by performing tasks on the product recommendation content, the product recommendation results and the product recommendation task execution flowchart corresponding to the product recommendation content are displayed. That is, the reasoning process of obtaining the product recommendation results can be shown through the product recommendation task execution flowchart, thereby improving the interpretability and reliability of the obtained product recommendation results.

[0212] In a specific embodiment, such as Figure 12 The diagram shows a flowchart of the task execution method, which is performed by a computer device, which can be a server or a terminal. The following steps are illustrated using a server as an example:

[0213] 1202, retrieve the query content, plan the query content, and obtain at least two tasks to be executed.

[0214] 1204. Obtain at least two execution agents corresponding to the tasks to be executed. Input the tasks to be executed into the corresponding execution agents in sequence for task execution. Obtain the task execution result output according to the preset execution result format and the reference information output according to the preset reference information format. The reference information includes the dependency data information for obtaining the task execution result and the interpretation data information for the task execution result. The dependency data information includes the dependency task information.

[0215] 1206. Obtain the current dependent task information corresponding to the currently executed task from at least two pending tasks. The current dependent task information includes the identification information of the dependent historical tasks.

[0216] 1208. The currently executed task is taken as the current task node, and the dependent historical tasks are taken as historical task nodes. A link relationship is established between the historical task nodes and the current task node. Based on the current task node, historical task nodes and the link relationship, the execution chain data of the current task is obtained.

[0217] 1210. Obtain the query page layout information corresponding to the query content, divide the task nodes in the current task execution chain data into layers to obtain the task nodes in each layer. Sort the task nodes in each layer to obtain the sorted task nodes in each layer, and calculate the page position of the sorted task nodes in each layer based on the query page layout information to obtain the page position information of the task nodes in the current task execution chain data.

[0218] 1212. Based on the link relationships between task nodes in the current task execution chain data, determine the connection edge information of the task nodes, and obtain the task execution flowchart information based on the page position information and connection edge information of the task nodes. Query results for the query content are obtained based on the task execution results and task execution flowchart information corresponding to each task node.

[0219] The dependency task information includes parent dependency task information and non-parent dependency task information, and the link relationship includes parent-child link relationship and non-parent-child link relationship.

[0220] 1214. Returns the query results to the corresponding query page. The query page displays the task execution results for each task node, as well as the task execution flowchart for showing and hiding the connection edges corresponding to non-parent-child links. The task execution flowchart responds to the view event for the parent-child connection edge corresponding to the parent-child link, highlighting the parent-child connection edge, the parent task node and child task node corresponding to the parent-child connection edge, and the non-parent task node corresponding to the child task node and the non-parent task node corresponding to the non-parent-child connection edge.

[0221] 1216. Obtain a view request for the selected task node in the task execution flowchart. Respond to the view request to obtain the selection reference information corresponding to the selected task node. The selection reference information includes the execution agent information of the selected task node, the input information of the selected task node, the output information of the selected task node, the dependent task information of the selected task node, the reference information of the selected task node, and the result confidence information of the selected task node.

[0222] 1218. The selected reference information corresponding to the selected task node will be returned to the query page corresponding to the query content. The query page is used to display the execution agent information, input information, output information, dependent task information, reference information and result confidence information corresponding to the selected task node.

[0223] In the above embodiments, by introducing a task execution chain mechanism, the decision-making footprint is determined through the task execution chain, and then the task execution flowchart is visualized as a traceability map. This effectively solves the problem of opaque decision-making processes in multi-agent systems, and realizes the full recording and traceability of the reasoning basis, data source, and confidence level of each task to be executed. This significantly improves the efficiency of error detection and attribution of discrepancies, enhances the interpretability, auditability, and user trust of the system, and improves the reliability of the obtained task execution results, thereby supporting its reliable application in scenarios with high reliability requirements such as finance and healthcare.

[0224] In one specific embodiment, this task execution method is applied in a search platform that responds to user queries through multi-agent collaboration. For example... Figure 13The diagram illustrates the framework for task execution of query content in a search platform. Specifically: A user enters a query on the search platform's query page; this query can be a question or inquiry, triggering a submission event. The server then receives the query request from the user terminal via the user interface layer, which carries the query content. The user interface layer then passes the query content to the task planning layer. The task planning layer performs task planning on the query content, identifying the intent of the query and generating an initial execution plan, which includes various tasks to be executed. The task planning layer then distributes these tasks to the agent execution layer. The agent execution layer inputs each task into the corresponding agent for execution, obtaining the output task results and structured metadata, which serves as reference information. The agent execution layer then passes the task results and structured metadata of each task to the decision footprint chain service layer. In the decision footprint chain service layer, the link generator obtains the task results and structured metadata, constructs task execution link data according to the structured metadata, and stores the constructed task execution link data. The task execution chain data is also known as the decision footprint graph structure data. The decision footprint chain service layer then persistently stores the structured metadata corresponding to the task execution chain data. Next, the decision footprint chain service layer uses the task execution chain data to perform layout calculations through the source graph engine, generating a visual graph. Finally, the decision footprint chain service layer returns the complete task execution chain data and source graph to the user interface layer; this source graph is the task execution flowchart. The user interface layer displays the final result and the source graph.

[0225] The task execution method proposed in this application brings the following significant technical effects:

[0226] 1. Completely break the "black box" effect and achieve end-to-end transparency and interpretability of the decision-making process: By attaching structured meta-information to the output of each agent, the implicit decision-making process in traditional multi-agent systems can be explicitly recorded and presented. Users and developers can no longer just obtain the final answer, but can clearly and completely understand the basis for each conclusion, the data sources it relies on, the reasoning logic, and the confidence assessment. This fundamentally solves the problem of opaque decision-making within complex agent systems, transforming the decision-making process from an incomprehensible "black box" into an auditable and understandable "glass box."

[0227] 2. Significantly improves the efficiency and accuracy of problem diagnosis and root cause tracing: When errors or deviations exist in the output, the source tracing diagram and interactive functions provided in this application enable users to quickly and accurately locate the specific task node where the problem occurred. Without the need for manual sifting through massive amounts of logs, users can directly find the task node leading to the error and view its detailed inputs, outputs, and decision-making basis through interactive operations via a visual interface. This greatly reduces the difficulty and time cost of operation and maintenance debugging, and improves the efficiency of system maintenance.

[0228] 3. Provides powerful attribution and comparative analysis capabilities for divergent results: When faced with scenarios where multiple agents output competing answers, this application compares and displays the complete decision-making process of each answer, enabling users to intuitively analyze the root causes of discrepancies. For example, is it due to inconsistent data sources relied upon by different agents, or differences in reasoning logic? This provides comprehensive and in-depth auxiliary information for final decision-making in high-risk fields such as finance and healthcare, enhancing the reliability of the output and the scientific nature of the decision.

[0229] 4. Enhancing user trust and system usability, and expanding the application boundaries of multi-agent systems: By providing an intuitive visual interface and one-step traceability capabilities, this application greatly improves the user experience of interacting with complex intelligent agent systems. Users can easily understand and verify the system's output, thereby establishing trust in the system. This enables multi-agent systems to overcome previous limitations and be reliably applied in scenarios with stringent requirements for interpretability, reliability, and auditability, such as financial risk control, medical diagnosis, legal consulting, and academic research.

[0230] 5. Establishing standardized and structured audit logs to provide a data foundation for system optimization and compliance review: This application generates and persistently stores metadata and task execution chain data, creating a complete, structured, and machine-readable audit log. This not only meets the data needs of internal system optimization but also provides tamper-proof and detailed evidence chains for external compliance reviews, quality certification, and liability determination, helping to meet increasingly stringent data supervision and industry compliance requirements.

[0231] 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.

[0232] Based on the same inventive concept, this application also provides a task execution apparatus for implementing the task execution 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 execution apparatus embodiments provided below can be found in the limitations of the task execution method described above, and will not be repeated here.

[0233] In one embodiment, such as Figure 14 As shown, a task execution device 1400 is provided, including: a task acquisition module 1402, a task execution module 1404, and a result acquisition module 1406, wherein:

[0234] The task acquisition module 1402 is used to acquire the tasks to be executed corresponding to the query content. The tasks to be executed are obtained by planning the query content and used to obtain the query results.

[0235] The task execution module 1404 is used to execute tasks based on tasks to be executed, and to obtain the task execution result and reference information corresponding to the task to be executed. The reference information includes the dependency data information for obtaining the task execution result.

[0236] The result module 1406 is used to obtain query results for the query content based on the task execution results and reference information.

[0237] In some embodiments, the task execution module 1404 is further configured to obtain the execution agent corresponding to the task to be executed; input the task to be executed into the execution agent for task execution, and obtain the task execution result output according to the preset execution result format and the reference information output according to the preset reference information format.

[0238] In some embodiments, the tasks to be executed include at least two, and the dependency data information includes dependency task information; the result obtaining module 1406 is further configured to obtain the current reference information corresponding to the currently executed task from the at least two tasks to be executed, and obtain the historical reference information corresponding to the historically executed task; generate task execution chain data based on the historical dependency task information in the historical reference information and the current dependency task information in the current reference information to obtain the current task execution chain data; and obtain the query result of the query content based on the current task execution result corresponding to the currently executed task, the historical execution result corresponding to the historically executed task, and the current task execution chain data.

[0239] In some embodiments, the current dependent task information includes the identification information of dependent historical tasks, where the dependent historical tasks are executed tasks that have a dependency relationship with the currently executed task; the result obtaining module 1406 is further configured to take the currently executed task as the current task node and the dependent historical task as the historical task node; establish a link relationship between the historical task node and the current task node based on the identification information of the dependent historical tasks included in the current dependent task information; and obtain the current task execution chain data based on the current task node, the historical task node, and the link relationship.

[0240] In some embodiments, the reference information includes dependency data information for obtaining the task execution result and explanatory data information for the task execution result. The dependency data information includes at least one of the following: dependency task information, dependency input information, and dependency reference information. The explanatory data information includes at least one of the following: task execution agent information, task execution inference information, task execution output information, and result confidence information. The dependency input information includes the execution result corresponding to the dependency task in the dependency task information. The dependency reference information includes the inference reference data source information and the reference data source information of the task execution result in the task execution inference information. The task execution output information includes the task execution result.

[0241] The task execution device 1400 also includes:

[0242] The results display module is used to return the query results to the terminal corresponding to the query content. The terminal is used to display the task execution results in the query results and to display the reference information in response to the viewing event of the reference information in the query results.

[0243] In some embodiments, the result obtaining module 1406 is further configured to obtain historical task execution chain data corresponding to the query content, wherein the historical task execution chain data is generated based on the reference information corresponding to each historically executed task; obtain new reference information for newly executed tasks; update the historical task execution chain data based on the new dependent task information in the new reference information to obtain updated task execution chain data; and obtain the query result of the query content based on the new task execution result corresponding to the newly executed task, the historical task execution result corresponding to each historically executed task, and the updated task execution chain data.

[0244] In some embodiments, the result obtaining module 1406 is further configured to treat the newly added executed task as a new task node in the historical task execution chain data; determine the executed task that the newly added executed task depends on from the new dependent task information, and establish a new link relationship between the existing task node corresponding to the dependent executed task and the new task node; and obtain the updated task execution chain data corresponding to the historical task execution chain data based on the new task node and the new link relationship.

[0245] In some embodiments, the tasks to be performed include at least two, and the dependency data information includes dependency task information;

[0246] Module 1406 also retrieves task execution chain data corresponding to the query content. The task execution chain data includes each task node and the corresponding link relationship, which is determined based on the dependent task information corresponding to each task node. It also retrieves the query page layout information corresponding to the query content, performs page layout calculations based on the query page layout information and the task execution chain data, and obtains the task execution flowchart information corresponding to the task execution chain data. Based on the task execution results and task execution flowchart information corresponding to each task node, it obtains the query results for the query content. Finally, it returns the query results to the query page corresponding to the query content, which displays the task execution results and task execution flowchart information corresponding to each task node.

[0247] In some embodiments, the result obtaining module 1406 is further configured to layer the task nodes in the task execution chain data to obtain the task nodes in each layer; sort the task nodes in each layer to obtain the sorted task nodes in each layer; calculate the page position of the sorted task nodes in each layer based on the query page layout information to obtain the page position information of the task nodes in the task execution chain data; determine the connection edge information of the task nodes according to the link relationship of the task nodes in the task execution chain data; and obtain the task execution flowchart information based on the page position information and the connection edge information of the task nodes.

[0248] In some embodiments, the result obtaining module 1406 is further configured to perform simulation calculations of repulsive and attractive forces between nodes based on the query page layout information and the task nodes in the task execution chain data, to obtain the simulated repulsive and attractive forces of the task nodes; update the page position corresponding to the task node based on the simulated repulsive and attractive forces of the task node, to obtain the updated page position of the task node, and perform iterative calculations based on the updated page position of the task node until the force-directed layout completion condition is met, to obtain the task execution flowchart information corresponding to the task execution chain data.

[0249] In some embodiments, the task execution device 1400 further includes:

[0250] The query request acquisition module is used to obtain viewing requests for selected task nodes in the task execution flowchart. It responds to these requests by retrieving the corresponding selection reference information for each selected task node. This selection reference information includes the executing agent information, input information, output information, dependent task information, reference information, and result confidence information for each selected task node. The module then returns this selection reference information to the query page corresponding to the query content. The query page displays the executing agent information, input information, output information, dependent task information, reference information, and result confidence information.

[0251] In some embodiments, the dependency task information includes parent dependency task information and non-parent dependency task information, and the link relationship includes parent-child link relationship and non-parent-child link relationship;

[0252] The search request retrieval module is also used to return the query results of the query content to the query page corresponding to the query content. The query page is used to display the task execution results corresponding to each task node, as well as to show and hide the target task execution flowchart corresponding to the connection edge of the non-parent-child link relationship. The target task execution flowchart is used to respond to the viewing event of the parent-child connection edge corresponding to the parent-child link relationship, highlighting the parent-child connection edge, the parent task node and child task node corresponding to the parent-child connection edge, and highlighting the non-parent task node corresponding to the child task node and the non-parent task node corresponding to the non-parent-child connection edge.

[0253] In some embodiments, the task execution module 1404 is further configured to obtain at least two execution agents corresponding to the task to be executed; input the task to be executed into the at least two execution agents for task execution, and obtain at least two task execution results corresponding to the task to be executed and reference information corresponding to the at least two task execution results respectively;

[0254] The result module 1406 is also used to obtain at least two query results corresponding to the query content based on at least two task execution results and the reference information corresponding to the at least two task execution results respectively.

[0255] In some embodiments, the result obtaining module 1406 is further configured to generate a task execution chain based on the reference information corresponding to at least two task execution results, and obtain at least two current task execution chain data corresponding to the query content; obtain the query page layout information corresponding to the query content, perform page layout calculation based on the query page layout information and at least two current task execution chain data, and obtain at least two task execution flowchart information corresponding to the query content; return at least two task execution results and corresponding task execution flowchart information to the query page corresponding to the query content, and the query page is used to compare and display at least two task execution results and corresponding task execution flowcharts.

[0256] In some embodiments, the task execution device 1400 further includes:

[0257] The edit request retrieval module is used to obtain edit requests for target task nodes in the target task execution flowchart, respond to edit requests to obtain target editing tasks that are edited according to the reference information corresponding to the target task nodes, restart execution from the target editing task corresponding to the target task node, obtain the update task execution flowchart information and the editing task execution result corresponding to the target task execution flowchart information, and return the update task execution flowchart information and the editing task execution result to the query page corresponding to the query content. The query page is used to display the update task execution flowchart and the editing task execution result corresponding to the target task execution flowchart information.

[0258] In one embodiment, such as Figure 15 As shown, a task execution device 1500 is provided, including: a content display module 1502 and a result display module 1504, wherein:

[0259] The content display module 1502 is used to display the query content on the query page;

[0260] The results display module 1504 is used to display the task execution results and reference information corresponding to the query content. The reference information includes the dependency data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the task to be executed corresponding to the query content. The task to be executed is obtained by planning the query content and using it to obtain the query results.

[0261] In some embodiments, the result display module 1504 is further configured to display the task execution result and task execution flowchart corresponding to the query content. The task execution flowchart includes each task node and the connection edges between each task node. The connection edges are determined based on the link relationships between each task node, and the link relationships are determined based on the dependent task information corresponding to each task node. In response to a viewing event for a selected task node in the task execution flowchart, the selected reference information corresponding to the selected task node is displayed. The selected reference information includes the execution agent information of the selected task node, the input information of the selected task node, the output information of the selected task node, the dependent task information of the selected task node, the reference information of the selected task node, and the result confidence information of the selected task node.

[0262] In some embodiments, the dependency task information includes parent dependency task information and non-parent dependency task information, the link relationship includes parent-child link relationship and non-parent-child link relationship, and the task execution flowchart includes the target task execution flowchart that hides the connection edges corresponding to non-parent-child link relationships.

[0263] The task execution device 1500 also includes:

[0264] The viewing module is used to respond to viewing events for parent-child connection edges in the target task execution flowchart, highlighting the parent-child connection edge, the parent task node and child task node corresponding to the parent-child connection edge, and highlighting the non-parent-child connection edge corresponding to the child task node and the non-parent task node corresponding to the non-parent-child connection edge.

[0265] In some embodiments, the result display module 1504 is further configured to compare and display at least two task execution flowcharts corresponding to the query content and the task execution results corresponding to the at least two task execution flowcharts. The at least two task execution flowcharts are obtained by page layout calculation based on the at least two current task execution chain data corresponding to the query content. The at least two current task execution chain data are obtained by generating task execution chains based on the reference information corresponding to the at least two task execution results. The reference information corresponding to the at least two task execution results is obtained by inputting the task to be executed corresponding to the query content into at least two execution agents for task execution.

[0266] In some embodiments, the task execution device 1500 further includes:

[0267] The editing module is used to respond to editing events for target task nodes in the target task execution flowchart, display the updated task execution flowchart information and the edited task execution result corresponding to the target task execution flowchart. The updated task execution flowchart and the edited task execution result are obtained by restarting the target editing task corresponding to the target task node.

[0268] In some embodiments, the query content includes product recommendations;

[0269] The results display module 1504 is also used to display the product recommendation results and the product recommendation task execution flowchart corresponding to the product recommendation content; in response to the viewing event of the product search task node in the product recommendation task execution flowchart, it displays the product search reference information corresponding to the product search task node. The product search reference information includes the execution agent information of the product search task node, the input information of the product search task node, the product search task execution result of the product search task node, the dependent task information of the product search task execution result, the referenced data source information of the product search task execution result, and the confidence information of the product search task execution result.

[0270] Each module in the aforementioned task execution device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0271] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 16 As 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 data such as task execution results, reference information, task execution chain data, and task execution flowchart information. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a task execution method.

[0272] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 17As 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 execution 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.

[0273] Those skilled in the art will understand that Figure 16 or Figure 17 The 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.

[0274] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0275] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0276] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0277] 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.

[0278] 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.

[0279] 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.

[0280] 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 execution method, characterized in that, The method includes: Obtain the task to be executed corresponding to the query content, wherein the task to be executed is obtained by task planning of the query content and used to obtain the query result; Based on the task to be executed, the task is executed to obtain the task execution result and the reference information corresponding to the task to be executed. The reference information includes the dependency data information for obtaining the task execution result. The query results for the query content are obtained based on the task execution results and the reference information.

2. The method according to claim 1, characterized in that, The step involves executing a task based on the task to be executed, obtaining the task execution result corresponding to the task to be executed and reference information corresponding to the task to be executed. The reference information includes dependency data information for obtaining the task execution result, including: Obtain the execution agent corresponding to the task to be executed; The task to be executed is input into the execution agent for task execution, and the task execution result and reference information are output according to the preset execution result format and the preset reference information format, respectively.

3. The method according to claim 1, characterized in that, The tasks to be executed include at least two, and the dependency data information includes dependency task information; The query results obtained based on the task execution results and the reference information include: Obtain current reference information corresponding to the currently executed task from the at least two tasks to be executed, and obtain historical reference information corresponding to the historically executed task; Based on the historical dependent task information in the historical reference information and the current dependent task information in the current reference information, task execution chain data is generated to obtain the current task execution chain data. The query result for the query content is obtained based on the current task execution result corresponding to the currently executed task, the historical execution result corresponding to the historical executed task, and the current task execution chain data.

4. The method according to claim 3, characterized in that, The current dependent task information includes the identification information of dependent historical tasks, whereby the dependent historical tasks are executed tasks that have a dependency relationship with the currently executed task. Based on the historical dependent task information in the historical reference information and the current dependent task information in the current reference information, task execution chain data is generated to obtain the current task execution chain data, including: The currently executed task is designated as the current task node, and the dependent historical task is designated as the historical task node; Based on the identification information of the dependent historical tasks included in the current dependent task information, a link relationship is established between the historical task node and the current task node; The current task execution chain data is obtained based on the current task node, the historical task nodes, and the link relationships.

5. The method according to claim 1, characterized in that, The reference information includes dependency data information for obtaining the task execution result and explanatory data information for the task execution result. The dependency data information includes at least one of the following: dependency task information, dependency input information, and dependency reference information. The explanatory data information includes at least one of the following: task execution agent information, task execution inference information, task execution output information, and result confidence information. The dependency input information includes the execution result corresponding to the dependency task in the dependency task information. The dependency reference information includes the inference reference data source information and the reference data source information of the task execution result in the task execution inference information. The task execution output information includes the task execution result. After obtaining the query result of the query content based on the task execution result and the reference information, the method further includes: The query results are returned to the terminal corresponding to the query content. The terminal is used to display the task execution results in the query results and to display the reference information in response to the viewing event of the reference information in the query results.

6. The method according to claim 1, characterized in that, The query results obtained based on the task execution results and the reference information include: Obtain the historical task execution chain data corresponding to the query content. The historical task execution chain data is generated based on the reference information corresponding to each historically executed task. Retrieve new reference information for newly executed tasks; The historical task execution chain data is updated based on the newly added dependent task information in the newly added reference information to obtain updated task execution chain data; The query results for the query content are obtained based on the execution results of the newly added executed tasks, the execution results of the historical tasks corresponding to each historical executed task, and the updated task execution chain data.

7. The method according to claim 6, characterized in that, The step of updating the historical task execution chain data based on the newly added dependent task information in the newly added reference information to obtain updated task execution chain data includes: The newly added executed task is treated as a new task node in the historical task execution chain data; From the newly added dependent task information, determine the already executed tasks that the newly added executed task depends on, and establish a new link relationship between the existing task node corresponding to the dependent executed task and the newly added task node. Based on the newly added task nodes and the newly added link relationships, the updated task execution chain data corresponding to the historical task execution chain data is obtained.

8. The method according to claim 1, characterized in that, The tasks to be executed include at least two, and the dependency data information includes dependency task information. The query results obtained based on the task execution results and the reference information include: Obtain the task execution chain data corresponding to the query content. The task execution chain data includes each task node and the link relationship corresponding to each task node. The link relationship is determined based on the dependent task information corresponding to each task node. Obtain the query page layout information corresponding to the query content, perform page layout calculation based on the query page layout information and the task execution chain data, and obtain the task execution flowchart information corresponding to the task execution chain data; The query results for the query content are obtained based on the task execution results corresponding to each task node and the task execution flowchart information. The query results of the query content are returned to the query page corresponding to the query content. The query page is used to display the task execution results corresponding to each task node and the task execution flowchart corresponding to the task execution flowchart information.

9. The method according to claim 8, characterized in that, The step of calculating the page layout based on the query page layout information and the task execution chain data to obtain the task execution flowchart information corresponding to the task execution chain data includes: The task nodes in the task execution chain data are divided into layers to obtain the task nodes in each layer. The task nodes in each level are sorted to obtain the sorted task nodes in each level. Based on the query page layout information, the page position of the sorted task nodes in each level is calculated to obtain the page position information of the task nodes in the task execution chain data. The connection edge information of the task nodes is determined based on the link relationship of the task nodes in the task execution chain data, and the task execution flowchart information is obtained based on the page position information of the task nodes and the connection edge information of the task nodes.

10. The method according to claim 8, characterized in that, The step of calculating the page layout based on the query page layout information and the task execution chain data to obtain the task execution flowchart information corresponding to the task execution chain data includes: Based on the query page layout information and the task nodes in the task execution chain data, simulated calculations of repulsive and attractive forces between nodes are performed to obtain simulated repulsive and attractive forces between task nodes. The page position corresponding to the task node is updated based on the simulated repulsive and simulated attractive forces of the task node to obtain the updated page position of the task node. Iterative calculations are then performed based on the updated page position of the task node until the force-directed layout completion condition is met, at which point the task execution flowchart information corresponding to the task execution chain data is obtained.

11. The method according to claim 8, characterized in that, After returning the query results of the query content to the query page corresponding to the query content, and the query page is used to display the task execution results corresponding to each task node and the task execution flowchart corresponding to the task execution flowchart information, the method further includes: Obtain a viewing request for the selected task node in the task execution flowchart, and respond to the viewing request to obtain the selection reference information corresponding to the selected task node. The selection reference information includes the execution agent information of the selected task node, the input information of the selected task node, the execution inference information of the selected task node, the output information of the selected task node, the dependent task information of the selected task node, the reference information of the selected task node, and the result confidence information of the selected task node. The selected reference information corresponding to the selected task node is returned to the query page corresponding to the query content. The query page is used to display the information of the executing agent, the input information, the execution inference information, the output information, the dependent task information, the reference information, and the result confidence information.

12. The method according to claim 8, characterized in that, The dependent task information includes parent dependent task information and non-parent dependent task information, and the link relationship includes parent-child link relationship and non-parent-child link relationship; The step of returning the query results of the query content to the query page corresponding to the query content, the query page being used to display the task execution results corresponding to each task node and the task execution flowchart corresponding to the task execution flowchart information, including: The query results of the query content are returned to the query page corresponding to the query content. The query page is used to display the task execution results corresponding to each task node, and to display and hide the target task execution flowchart corresponding to the non-parent-child link relationship. The target task execution flowchart is used to respond to the viewing event of the parent-child link relationship corresponding to the parent-child link edge, highlight the parent-child link edge, the parent task node and child task node corresponding to the parent-child link edge, and highlight the non-parent-child link edge corresponding to the child task node and the non-parent task node corresponding to the non-parent-child link edge.

13. The method according to claim 1, characterized in that, The step of executing the task based on the task to be executed, and obtaining the task execution result and reference information corresponding to the task to be executed, includes: Obtain at least two execution agents corresponding to the task to be executed; The task to be executed is input into the at least two execution agents for task execution, and at least two task execution results corresponding to the task to be executed and reference information corresponding to the at least two task execution results are obtained respectively; The query results obtained based on the task execution results and the reference information include: At least two query results corresponding to the query content are obtained based on the execution results of the at least two tasks and the reference information corresponding to the execution results of the at least two tasks.

14. The method according to claim 13, characterized in that, The process of obtaining at least two query results corresponding to the query content based on the execution results of the at least two tasks and the reference information corresponding to the execution results of the at least two tasks includes: Based on the reference information corresponding to the execution results of the at least two tasks, a task execution chain is generated to obtain at least two current task execution chain data corresponding to the query content; Obtain the query page layout information corresponding to the query content, and perform page layout calculation based on the query page layout information and the data of at least two current task execution chains to obtain the execution flowchart information of at least two tasks corresponding to the query content. The execution results of the at least two tasks and the corresponding task execution flowchart information are returned to the query page corresponding to the query content. The query page is used to compare and display the execution results of the at least two tasks and the corresponding task execution flowchart.

15. The method according to claim 14, characterized in that, After returning the execution results of the at least two tasks and the corresponding task execution flowchart information to the query page corresponding to the query content, and the query page being used to compare and display the execution results of the at least two tasks and the corresponding task execution flowchart, the method further includes: Obtain an edit request for a target task node in the target task execution flowchart, respond to the edit request to obtain a target edit task, wherein the target edit task is obtained by editing the target task of the target task node according to the reference information corresponding to the target task node; Restart execution from the target editing task to obtain the update task execution flowchart information and the editing task execution result corresponding to the target task execution flowchart information; The update task execution flowchart information and the edit task execution result are returned to the query page corresponding to the query content. The query page is used to display the update task execution flowchart and the edit task execution result corresponding to the target task execution flowchart.

16. A task execution method, characterized in that, The method includes: The search results are displayed on the search page; The task execution results and reference information corresponding to the query content are displayed. The reference information includes the dependency data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the task to be executed corresponding to the query content. The task to be executed is obtained by planning the query content and using it to obtain the query results.

17. The method according to claim 16, characterized in that, The display of the task execution results and reference information corresponding to the query content includes: The task execution results and task execution flowchart corresponding to the query content are displayed. The task execution flowchart includes each task node and the connection edges of each task node. The connection edges are determined based on the link relationship of each task node. The link relationship is determined according to the dependent task information corresponding to each task node. In response to a view event for a selected task node in the task execution flowchart, the selected reference information corresponding to the selected task node is displayed. The selected reference information includes the execution agent information of the selected task node, the input information of the selected task node, the execution inference information of the selected task node, the output information of the selected task node, the dependent task information of the selected task node, the reference information of the selected task node, and the result confidence information of the selected task node.

18. The method according to claim 17, characterized in that, The dependent task information includes parent dependent task information and non-parent dependent task information; the link relationship includes parent-child link relationship and non-parent-child link relationship; the task execution flowchart includes the target task execution flowchart that hides the connection edges corresponding to non-parent-child link relationships. The method further includes: In response to a view event of the parent-child connection edge of the execution flowchart of the target task, the parent-child connection edge, the parent task node and child task node corresponding to the parent-child connection edge are highlighted, and the non-parent-child connection edge corresponding to the child task node and the non-parent task node corresponding to the non-parent-child connection edge are highlighted.

19. The method according to claim 17, characterized in that, The process of displaying the task execution results and task execution flowchart corresponding to the query content includes: The comparison displays at least two task execution flowcharts corresponding to the query content and the task execution results corresponding to the at least two task execution flowcharts. The at least two task execution flowcharts are obtained by page layout calculation based on at least two current task execution chain data corresponding to the query content. The at least two current task execution chain data are generated by task execution chain generation based on reference information corresponding to at least two task execution results. The reference information corresponding to at least two task execution results is obtained by inputting the task to be executed corresponding to the query content into at least two execution agents for task execution.

20. The method according to claim 19, characterized in that, After comparing and displaying at least two task execution flowcharts corresponding to the query content and the task execution results corresponding to the at least two task execution flowcharts, the method further includes: In response to an edit event for a target task node in the target task execution flowchart, the updated task execution flowchart and the edit task execution result corresponding to the target task execution flowchart are displayed. The updated task execution flowchart and the edit task execution result are obtained by restarting the target edit task corresponding to the target task node.

21. The method according to claim 16, characterized in that, The query content includes product recommendations; The display of the task execution results and reference information corresponding to the query content includes: Displays the product recommendation results and the product recommendation task execution flowchart corresponding to the product recommendation content; In response to a view event for a product search task node in the product recommendation task execution flowchart, product search reference information corresponding to the product search task node is displayed. The product search reference information includes the execution agent information of the product search task node, the input information of the product search task node, the product search output information of the product search task node, the dependent task information of the product search task node, the reference information of the product search task node, and the result confidence information of the product search task node.

22. A task execution device, characterized in that, The device includes: The task acquisition module is used to acquire the tasks to be executed corresponding to the query content. The tasks to be executed are obtained by performing task planning on the query content and used to obtain the query results. The task execution module is used to perform task execution based on the task to be executed, and obtain the task execution result corresponding to the task to be executed and the reference information corresponding to the task to be executed. The reference information includes the dependency data information for obtaining the task execution result. The result acquisition module is used to obtain the query result of the query content based on the task execution result and the reference information.

23. A task execution device, characterized in that, The device includes: The content display module is used to display the query content on the query page; The results display module is used to display the task execution results and reference information corresponding to the query content. The reference information includes the dependency data information for obtaining the task execution results. The task execution results and reference information are obtained by executing the task to be executed corresponding to the query content. The task to be executed is obtained by planning the query content.

24. 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 21.

25. 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 21.

26. A computer program product, comprising a computer program, 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 21.

Images