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

Abstract

The invention relates to a data processing method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring query content, and performing task planning on the query content based on unit information of a started processing unit to obtain initial planning information which corresponds to the query content and carries missing capability information; the missing capability information is generated under the condition that the processing capability of the started processing unit does not meet the query content, and is used for representing the missing capability for executing the query content; based on missing capability information in the initial planning information, determining a recommended processing unit corresponding to the missing capability information from the candidate processing unit set; and determining a newly-added enabled processing unit based on the recommendation processing unit, and updating the initial planning information according to the query content and the unit information of the newly-added enabled processing unit to obtain target planning information corresponding to the query content. By adopting the method, the planning information generation accuracy can be improved.

Description

Technical Field

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

[0002] With the development of artificial intelligence technology, users can use intelligent agents to search for content and quickly obtain the corresponding search results. However, before obtaining the search results, task planning is required to better execute the search.

[0003] In traditional technologies, when planning tasks for query content, complex tasks such as query content are usually broken down into a large number of atomic steps to obtain corresponding planning information. However, the planning information generated in this way is relatively lengthy, prone to errors during task planning, and cannot detect whether the generated planning information has any capability deficiencies, ultimately resulting in low accuracy of planning information generation. Summary of the Invention

[0004] Therefore, it is necessary to provide a data processing method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve the accuracy of planning information generation in response to the above-mentioned technical problems.

[0005] In a first aspect, this application provides a data processing method, including:

[0006] The query content is obtained, and task planning is performed on the query content based on the unit information of the enabled processing unit to obtain the initial planning information corresponding to the query content carrying missing capability information. The missing capability information is generated when the processing capability of the enabled processing unit does not meet the query content, and is used to characterize the missing capability for executing the query content.

[0007] Based on the missing capability information in the initial planning information, the recommended processing unit corresponding to the missing capability information is determined from the candidate processing unit set;

[0008] Based on the recommendation processing unit, a new activation processing unit is determined. According to the query content and the unit information of the new activation processing unit, the initial planning information is updated to obtain the target planning information corresponding to the query content.

[0009] Secondly, this application also provides another data processing method, including:

[0010] The search results are displayed on the search page;

[0011] The query page displays initial planning information corresponding to the query content, carrying missing capability information; the initial planning information is obtained by performing task planning on the query content based on the unit information of the enabled processing units; the missing capability information is generated when the processing capability of the enabled processing units does not meet the query content, and is used to characterize the missing capability for executing the query content.

[0012] The query page displays the target planning information corresponding to the query content; the target planning information is obtained by updating the initial planning information based on the query content and the unit information of the newly added activated processing unit; the newly added activated processing unit is determined based on the recommended processing unit; the recommended processing unit is used to characterize the recommended processing unit corresponding to the missing capability information determined from the candidate processing unit set based on the missing capability information in the initial planning information.

[0013] Thirdly, this application also provides a data processing apparatus, comprising:

[0014] The task planning module is used to obtain query content, perform task planning on the query content based on the unit information of the enabled processing units, and obtain initial planning information corresponding to the query content carrying missing capability information; the missing capability information is generated when the processing capability of the enabled processing units does not meet the query content, and is used to characterize the missing capability for executing the query content.

[0015] The unit determination module is used to determine the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information;

[0016] The information update module is used to determine the newly enabled processing unit based on the recommendation processing unit, and update the initial planning information according to the query content and the unit information of the newly enabled processing unit to obtain the target planning information corresponding to the query content.

[0017] Fourthly, this application also provides another data processing apparatus, including:

[0018] The first display module is used to display the query content on the query page;

[0019] The second display module is used to display initial planning information carrying missing capability information corresponding to the query content on the query page; the initial planning information is obtained by performing task planning on the query content based on the unit information of the enabled processing unit; the missing capability information is generated when the processing capability of the enabled processing unit does not meet the query content, and is used to characterize the missing capability for executing the query content.

[0020] The third display module is used to display the target planning information corresponding to the query content on the query page; the target planning information is obtained by updating the initial planning information based on the query content and the unit information of the newly added activated processing unit; the newly added activated processing unit is determined based on the recommended processing unit; the recommended processing unit is used to characterize the recommended processing unit corresponding to the missing capability information determined from the candidate processing unit set based on the missing capability information in the initial planning information.

[0021] Fifthly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0022] The query content is obtained, and task planning is performed on the query content based on the unit information of the enabled processing unit to obtain the initial planning information corresponding to the query content carrying missing capability information. The missing capability information is generated when the processing capability of the enabled processing unit does not meet the query content, and is used to characterize the missing capability for executing the query content.

[0023] Based on the missing capability information in the initial planning information, the recommended processing unit corresponding to the missing capability information is determined from the candidate processing unit set;

[0024] Based on the recommendation processing unit, a new activation processing unit is determined. According to the query content and the unit information of the new activation processing unit, the initial planning information is updated to obtain the target planning information corresponding to the query content.

[0025] Sixthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0026] The query content is obtained, and task planning is performed on the query content based on the unit information of the enabled processing unit to obtain the initial planning information corresponding to the query content carrying missing capability information. The missing capability information is generated when the processing capability of the enabled processing unit does not meet the query content, and is used to characterize the missing capability for executing the query content.

[0027] Based on the missing capability information in the initial planning information, the recommended processing unit corresponding to the missing capability information is determined from the candidate processing unit set;

[0028] Based on the recommendation processing unit, a new activation processing unit is determined. According to the query content and the unit information of the new activation processing unit, the initial planning information is updated to obtain the target planning information corresponding to the query content.

[0029] In a seventh aspect, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0030] The query content is obtained, and task planning is performed on the query content based on the unit information of the enabled processing unit to obtain the initial planning information corresponding to the query content carrying missing capability information. The missing capability information is generated when the processing capability of the enabled processing unit does not meet the query content, and is used to characterize the missing capability for executing the query content.

[0031] Based on the missing capability information in the initial planning information, the recommended processing unit corresponding to the missing capability information is determined from the candidate processing unit set;

[0032] Based on the recommendation processing unit, a new activation processing unit is determined. According to the query content and the unit information of the new activation processing unit, the initial planning information is updated to obtain the target planning information corresponding to the query content.

[0033] The aforementioned data processing method, apparatus, computer equipment, computer-readable storage medium, and computer program product first acquire the query content. Based on the unit information of the activated processing units, task planning is performed on the query content to obtain initial planning information corresponding to the query content, carrying missing capability information. The missing capability information is generated when the processing capabilities of the activated processing units do not meet the query content requirements, and is used to characterize the missing capabilities for executing the query content. Next, based on the missing capability information in the initial planning information, recommended processing units corresponding to the missing capability information are determined from the candidate processing unit set. Finally, new activated processing units are determined based on the recommended processing units. The initial planning information is updated according to the query content and the unit information of the new activated processing units to obtain the target planning information corresponding to the query content. In this way, when performing task planning, task planning is performed on the query content based on the unit information of the activated processing units, achieving the goal of task planning according to the activated processing units. This ensures that the generated initial planning information can be executed by the activated processing units, thereby improving the accuracy of planning information generation. Furthermore, during task planning, when the processing capacity of the activated processing units is insufficient to meet the query requirements, missing capability information is generated to characterize the missing capabilities for executing the query. Based on the query and the unit information of the recommended processing units corresponding to the missing capability information, the initial planning information is updated to obtain the target planning information corresponding to the query. This achieves the goal of timely detection of capability deficiencies and capability completion of the initial planning information, which helps improve the accuracy of planning information generation. Moreover, the recommended processing units are determined from the candidate processing unit set based on the missing capability information, ensuring the compatibility between the determined recommended processing units and the missing capability information. This makes the target planning information updated based on the unit information of the recommended processing units more accurate, further improving the accuracy of planning information generation. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a diagram illustrating the application environment of a data processing method in one embodiment.

[0036] Figure 2 This is a flowchart illustrating a data processing method in one embodiment;

[0037] Figure 3This is a flowchart illustrating a thought-action-observation-based task execution method based on the MCP / tool ​​market in one embodiment.

[0038] Figure 4 This is a flowchart illustrating a task execution method based on a single-agent market model in one embodiment;

[0039] Figure 5 This is a flowchart illustrating a method for intelligent planning, execution, and capability completion recommendation of complex tasks based on an Agent capability market in one embodiment.

[0040] Figure 6 This is a flowchart illustrating the steps of task planning for query content in one embodiment;

[0041] Figure 7 This is a flowchart illustrating the data processing method in another embodiment;

[0042] Figure 8 This is a schematic diagram illustrating the display of query content on a query page in one embodiment;

[0043] Figure 9 This is a schematic diagram illustrating the display of initial planning information on a query page in one embodiment.

[0044] Figure 10 This is a schematic diagram illustrating the display of a smart recommendation pop-up window on a query page in one embodiment;

[0045] Figure 11 This is a schematic diagram illustrating the display of an agent with recommendations enabled on a query page in one embodiment.

[0046] Figure 12 This is a schematic diagram illustrating the display of target planning information on a query page in one embodiment.

[0047] Figure 13 This is a schematic diagram of an agent operation page in one embodiment;

[0048] Figure 14 This is a schematic diagram of a smart recommendation pop-up window in one embodiment;

[0049] Figure 15 This is a schematic diagram of a query page in one embodiment;

[0050] Figure 16 This is a schematic diagram showing the recommended purchase results for small kitchen appliances in one embodiment;

[0051] Figure 17 This is a flowchart illustrating the data processing method in yet another embodiment;

[0052] Figure 18This is an architecture diagram of a complex task intelligent planning, execution, and capability completion recommendation system based on an Agent capability market in one embodiment.

[0053] Figure 19 This is a structural block diagram of a data processing device in one embodiment;

[0054] Figure 20 This is a structural block diagram of the data processing apparatus in another embodiment;

[0055] Figure 21 This is an internal structural diagram of a computer device in one embodiment;

[0056] Figure 22 This is a diagram of the internal structure of a computer device in another embodiment. Detailed Implementation

[0057] 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.

[0058] For task planning scenarios, current tool markets based on protocols such as MCP (Model Context Protocol) can only provide atomic-level API (Application Programming Interface) functions. This results in fine-grained task planning with numerous steps and complex logic, making the planning fragile and unable to handle scenarios requiring complex analysis and customized logic. Furthermore, existing agent markets mostly operate on a "single-selection" model, where users can only choose one agent to execute a task at a time, lacking a mechanism to integrate the capabilities of multiple agents into a coherent, automated planning process. Simultaneously, the system cannot intelligently detect missing capabilities during planning and recommend supplementary information to the user.

[0059] Currently, mainstream complex task processing systems are mainly based on the following two paradigms:

[0060] 1. Planning based on the MCP / tool ​​market (such as LangChain, AutoGPT): Its core is to use LLM to call a series of atomic tools (APIs) to complete the task step by step; LLM acts as the brain, and the tool library acts as the limbs; the planning process is linear or finitely cyclical (Think-Act-Observe), see details for more information. Figure 3 .

[0061] 2. Single Agent Market Model: This model provides a list of agents, each encapsulating the ability to solve a specific type of problem. Users need to manually determine and select one agent to execute the entire task. Agents are isolated and mutually exclusive. (See details...) Figure 4 .

[0062] Current mainstream complex task processing systems have the following main drawbacks:

[0063] 1. MCP tool market: (1) Complex and fragile planning: complex tasks need to be broken down into a large number of atomic steps, the planning is lengthy and the fault tolerance is poor; (2) Lack of high-level abstraction: the tool has a single function and cannot directly complete sub-tasks that require multi-step reasoning and complex logic.

[0064] 2. Single Agent Market: (1) Unable to collaborate: Agents are isolated and cannot automatically combine the capabilities of multiple Agents to solve a complex task; (2) Dependent on user selection: Users are required to accurately understand the capabilities of each Agent and make the correct choice, which has a high threshold.

[0065] 3. Common shortcomings: (1) Lack of ability perception and recommendation: The system cannot perceive whether the current ability set is insufficient, let alone actively recommend how users can complete their abilities to better complete the task.

[0066] Based on this, in order to solve the above problems, this application proposes a data processing method, specifically a method for intelligent planning, execution and capability completion recommendation of complex tasks based on the Agent capability market, which can improve the accuracy of planning information generation and task execution accuracy, and can be applied to various query scenarios, task planning scenarios and task execution scenarios.

[0067] The data processing method provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store data that server 104 needs to process, such as query content. The data storage system can be integrated on server 104 or located on the cloud or other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart in-vehicle devices, projection devices, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. 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. Furthermore, terminal 102 and server 104 can be directly or indirectly connected via wired or wireless communication, which is not limited herein.

[0068] It should be noted that both terminal 102 and server 104 can be used independently to execute the data processing methods provided in the embodiments of this application.

[0069] For example, refer to Figure 1 First, terminal 102 obtains the query content and performs task planning on the query content based on the unit information of the enabled processing units, obtaining initial planning information corresponding to the query content and carrying missing capability information. The missing capability information is generated when the processing capabilities of the enabled processing units do not meet the query content and is used to characterize the missing capabilities for executing the query content. Next, based on the missing capability information in the initial planning information, terminal 102 determines the recommended processing unit corresponding to the missing capability information from the candidate processing unit set. Finally, terminal 102 determines the newly enabled processing unit based on the recommended processing unit, updates the initial planning information according to the query content and the unit information of the newly enabled processing unit, obtains the target planning information corresponding to the query content, and displays the target planning information on the terminal interface.

[0070] For example, refer to Figure 1First, server 104 obtains the query content and performs task planning on the query content based on the unit information of the enabled processing units, obtaining initial planning information corresponding to the query content and carrying missing capability information. The missing capability information is generated when the processing capabilities of the enabled processing units do not meet the query content and is used to characterize the missing capabilities for executing the query content. Next, terminal 102 determines the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information. Finally, terminal 102 determines the newly enabled processing unit based on the recommended processing unit, updates the initial planning information according to the query content and the unit information of the newly enabled processing unit, and obtains the target planning information corresponding to the query content.

[0071] It should be noted that the terminal 102 and the server 104 can also be used in conjunction to execute the data processing method provided in the embodiments of this application.

[0072] For example, refer to Figure 1 Terminal 102 obtains the query content and sends it to server 104. Server 104 performs task planning on the received query content based on the unit information of the enabled processing units, obtaining initial planning information carrying missing capability information corresponding to the query content. This missing capability information is generated when the processing capabilities of the enabled processing units do not meet the query content requirements, and is used to characterize the missing capabilities required to execute the query content. Next, based on the missing capability information in the initial planning information, server 104 determines the recommended processing unit corresponding to the missing capability information from the candidate processing unit set. Finally, server 104 determines the newly enabled processing unit based on the recommended processing unit, updates the initial planning information according to the query content and the unit information of the newly enabled processing unit, obtains the target planning information corresponding to the query content, and sends the target planning information to terminal 102, which then displays the target planning information on the terminal interface.

[0073] Before introducing the specific embodiments of this application, the technical terms involved in this application will be explained:

[0074] LLM (Large Language Model): A large-scale AI (Artificial Intelligence) model with natural language understanding and generation capabilities. It is the core engine that drives intelligent agents to perform task planning and decision-making.

[0075] An agent is an autonomous software entity that can perceive its environment, make plans, invoke tools (or its own capabilities), and perform actions to achieve its goals. An agent typically encapsulates complex logic and capabilities for solving a specific type of problem.

[0076] Agent Capability Market: A centralized repository or directory for registering, discovering, and managing multiple heterogeneous agents with different specialized capabilities. Each agent in this market declares its functionality, input / output formats, and performance characteristics through structured capability metadata.

[0077] Planning: refers to the process by which an LLM (acting as a central planner) decomposes and orchestrates a sequence of subtasks (i.e., planning) based on the currently available set of agent capabilities, from complex, high-level task objectives input by the user into a sequence of subtasks to be executed collaboratively by multiple agents.

[0078] Capability Completion Recommendation: When the central planner finds that the existing enabled agent capabilities cannot fully meet the task requirements, it proactively retrieves and analyzes the most relevant unenabled agents from the full agent capability market, generates a recommendation list and reasons, and prompts the user to enable them.

[0079] Dynamic Replanning: After a user enables a new agent capability based on a recommendation, the system does not need to start planning from scratch. Instead, it makes partial adjustments and optimizations to the original plan based on the new agent capability set.

[0080] Multi-LLM Decision Mechanism: In different stages of the system (such as planning generation, feasibility assessment, recommendation reason generation, etc.), it automatically selects and calls the LLM that is better suited to the specific task (such as one LLM for complex planning and another LLM for long text analysis) to participate in the decision-making process, so as to improve the overall adaptability and result accuracy.

[0081] In one exemplary embodiment, such as Figure 2 As shown, a data processing method is provided, illustrated by applying it to a computer device, which can be a terminal or a server. It is understood that this method can be executed independently by the terminal or server, or it can be implemented through interaction between the terminal and the server. (See reference) Figure 2 The data processing method includes the following steps S201 to S203. Wherein:

[0082] Step S201: Obtain the query content. Based on the unit information of the enabled processing units, perform task planning on the query content to obtain the initial planning information corresponding to the query content, which carries missing capability information. The missing capability information is generated when the processing capability of the enabled processing units does not meet the query content, and is used to characterize the missing capabilities for executing the query content.

[0083] The query content refers to the information that needs to be queried to obtain results. Specifically, it refers to the content that needs to be queried through an intelligent agent, and can be from various fields, such as medicine, scientific research, industry, agriculture, gaming, education, and finance. The query content can be input in various ways, such as voice or text; for example, references... Figure 8 Users can enter query content A in the message sending area of ​​the query page. Furthermore, the computer device can rewrite the query content, especially colloquial queries, into text that the large language model or intelligent agent can understand—the rewritten query content. Additionally, the query content can be for information in different formats, such as text, images, videos, and audio. This query content is typically user-inputted and requires a response from a multi-agent collaborative system based on the large language model. It can be various types of questions, such as inquiry-based questions, execution-based questions, and search-based questions. For example, an inquiry-based question entered by the user during interaction with the large model, such as "Which is more suitable for college students: phone A, phone B, or phone C?" or "Which has a higher cost-performance ratio: computer E or computer F?", will be used as the query content. Similarly, an execution-based question entered by the user during interaction with the large model, such as "Please help me post a Weibo post Z" or "Please help me delete folder K on drive D of my computer," will also be used as the query content. For example, if a user enters a search question during a session with a large model, such as "Please provide a review paper on the application of artificial intelligence in medical diagnosis" or "Please provide weather information for city W for the last three months," then that search question will be considered the query content. In real-world scenarios, the query content can refer to complex, high-level tasks entered by the user, such as a query for a departmental team-building activity: "I want to organize a departmental team-building activity next month, with a budget of 500 yuan per person. I need a complete plan that includes transportation, activities, and catering, and I would like to analyze which plan offers the best cost-effectiveness."

[0084] In this context, a processing unit can refer to an agent or a tool. Each agent possesses the ability to solve specific types of problems and can execute corresponding sub-tasks, such as traffic plan queries, team-building activity generation, restaurant recommendations, and multi-plan comparative analysis. Each tool also possesses the ability to solve specific types of problems and can execute corresponding sub-tasks, such as weather queries, train ticket booking, and image rendering. Tools can be local or cloud-based, such as weather query tools, ticket booking tools, and drawing tools. An enabled processing unit refers to a processing unit that has been activated (i.e., turned on), specifically an agent that the user has enabled; for example, a reference... Figure 13 The user has already enabled three Agents: a Transportation Plan Query Agent, a Restaurant Recommendation Agent, and a Data Visualization Agent. The unit information for enabled processing units refers to relevant information about the enabled processing units, such as unit identifier, capability description information, capability domain, input data structure, output data structure, and performance metrics. The unit identifier is the unique identifier of the processing unit, specifically the unit name, such as the Transportation Plan Query Agent or the Restaurant Recommendation Agent. The capability description information describes the Agent's processing capabilities, used to quickly understand the Agent's core functions and applicable scenarios. For example, the capability description for the Transportation Plan Query Agent is "to query and compare the time and cost of different modes of transportation (flight, train, driving) and provide the optimal travel plan." The capability domain refers to the application area or tag information of the Agent's processing capabilities, such as travel and transportation, lifestyle services, and data analysis. The input data structure defines the input parameters and their data types that must be provided when calling the Agent. The output data structure defines the format and data type of the results returned after the Agent completes execution, facilitating understanding how to use the Agent's output as input for subsequent Agents and enabling data flow between multiple Agents. Performance metrics refer to the performance data of an agent, such as success rate, response time, and resource consumption. When multiple agents are available, they can serve as a reference for selecting the optimal agent; for example, referring to... Figure 13 The performance metrics for the traffic plan query agent are: success rate 98%, response time 1.2 seconds, and medium resource consumption.

[0085] Task planning for the query content refers to decomposing and arranging the query content into a sequence of subtasks executed collaboratively by multiple processing units (i.e., agents) based on the unit information of the enabled processing units (i.e., enabled agents). Specifically, the query content is decomposed into a sequence of subtasks based on a DAG (Directed Acyclic Graph). The subtask sequence includes multiple subtasks, each corresponding to an enabled processing unit. These subtasks have dependencies on each other; for example, some subtasks can be executed in parallel, while others have a specific execution order. For instance, a subtask sequence might include three subtasks: subtask 1, subtask 2, and subtask 3. Subtask 1 and subtask 2 can be executed in parallel, while subtask 3 executes after subtask 1 and subtask 2, depending on the results of subtask 1 and subtask 2. It's important to note that in the task planning process, enabled processing units (i.e., enabled agents) are used as planning units. Each enabled processing unit can handle a complex subtask, greatly simplifying the planning logic and improving the reliability and execution efficiency of task planning.

[0086] Initial planning information refers to a sequence of sub-tasks, composed of decomposed and arranged sub-tasks executed collaboratively by multiple processing units, based on the unit information of the activated processing units. For example, the sub-task sequence corresponding to the departmental team building query (i.e., the content entered to query departmental team building activities) includes: transportation plan query sub-task, team building activity generation sub-task, restaurant recommendation sub-task, and multi-plan comparison analysis sub-task. If the processing capacity of the activated processing units does not meet the query requirements, it means that the activated processing units cannot complete the query. Specifically, it means that the processing capacity of the activated processing units cannot cover the target capabilities required to execute the query; for example, the activated Agents cannot cover all Agents required to execute the query. It should be noted that if the processing capacity of the activated processing units meets the query requirements, no missing capability information will be generated; only multiple sub-tasks will be generated. In this case, the generated initial planning information only includes the sub-tasks. If the processing capacity of the activated processing units does not meet the query requirements, missing capability information and multiple sub-tasks will be generated. That is, the generated initial planning information carries missing capability information and includes both sub-tasks and missing capability information.

[0087] The missing capability information is used to characterize the capabilities lacking when executing the query content based on the processing capabilities of the already enabled processing units. This can refer to the type of missing capability, such as the ability to generate team-building activities or the ability to compare and analyze multiple options; or it can refer to the description of the missing capability, such as team-building activity planning (requiring the generation of multiple optional options based on budget and number of participants) or quantitative analysis of the cost-effectiveness of multiple options (requiring a comprehensive comparison of transportation, catering, and activity dimensions). For example, for a query about departmental team-building activities, the user's enabled Agents include a transportation plan query Agent and a catering recommendation Agent. However, the Agents required to execute the query include a transportation plan query Agent, a catering recommendation Agent, a team-building activity generation Agent, and a multi-option comparison and analysis Agent. This indicates that the enabled Agents are missing the team-building activity generation Agent and the multi-option comparison and analysis Agent, meaning that the processing capabilities of the enabled Agents do not meet the query requirements, and the missing capabilities are the team-building activity generation capability and the multi-option comparison and analysis capability.

[0088] For example, a computer device responds to a query request and parses the query request to obtain the query content carried in the query request; for example, a user in... Figure 8 In the message sending area shown, users input query content A via voice or text and click send to trigger a query request. The computer device parses the query request to obtain query content A. Next, the computer device obtains the unit information of the enabled processing units. For example, based on the target user identifier (such as a user account) carried in the query request, it queries the correspondence between the user identifier and the enabled processing units to obtain the enabled processing units corresponding to the target user identifier, and retrieves the unit information of the enabled processing units corresponding to the target user identifier from the local database. Then, the computer device rewrites the query content in terms of structure or format to obtain the rewritten query content, for example, rewriting the colloquial query content into input text that is easier for large models or intelligent agents to understand. Finally, based on the unit information of the enabled processing units, the computer device decomposes the rewritten query content into multiple sub-task objectives. These sub-task objectives are then matched with the unit information of the enabled processing units to identify the matched enabled processing units, and sub-tasks corresponding to the sub-task objectives of the matched enabled processing units are generated. If any sub-task objective does not match an enabled processing unit, it indicates that the processing capacity of the enabled processing units does not meet the query requirements. These sub-task objectives are then treated as remaining sub-task objectives, and missing capacity information corresponding to these remaining sub-task objectives is generated. The sub-tasks corresponding to the enabled processing units and the missing capacity information corresponding to the remaining sub-task objectives are then summarized to obtain the initial planning information corresponding to the query content, such as... Figure 9The example shows the initial planning information corresponding to query content A. For instance, for the query content related to departmental team building, the activated agents include the transportation plan query agent and the catering recommendation agent; the sub-task objectives include the transportation plan query sub-task objective, the team building activity generation sub-task objective, the catering recommendation sub-task objective, and the multi-scheme comparison analysis sub-task objective. This indicates that the transportation plan query sub-task objective matches the transportation plan query agent, and the catering recommendation sub-task objective matches the catering recommendation agent. However, the team building activity generation sub-task objective and the multi-scheme comparison analysis sub-task objective do not match any activated agents. Therefore, the following sub-tasks are generated: the transportation plan query sub-task corresponding to the transportation plan query sub-task objective, the catering recommendation sub-task corresponding to the catering recommendation sub-task objective, and the missing capability information (i.e., team building activity generation capability information) and the missing capability information (i.e., multi-scheme comparison analysis capability information) corresponding to the team building activity generation sub-task objective. Finally, the transportation plan query sub-task, catering recommendation sub-task, team building activity generation capability information, and multi-scheme comparison analysis capability information are summarized to obtain the initial planning information corresponding to the departmental team building query content. For a detailed explanation, please refer to [reference needed]. Figure 5 .

[0089] Further, after matching the subtask targets with the unit information of the enabled processing units, if each subtask target matches a corresponding enabled processing unit, it indicates that the processing capacity of the enabled processing unit meets the query content and there is no missing capacity. The computer device then generates the subtask corresponding to the subtask target of the enabled processing unit. Next, based on the query content, the computer device determines the sorting order of the subtasks and arranges them according to this order, obtaining the arranged subtasks. The arranged subtasks are then summarized to obtain the initial planning information for the query content (without carrying initial planning information). Then, if the obtained initial planning information does not carry missing capacity information, the computer device calls the enabled processing unit corresponding to the subtask in the obtained initial planning information to execute the subtask, obtaining the query result corresponding to the query content. For details, refer to [link to relevant documentation]. Figure 5For example, if the subtask objectives corresponding to query content A include subtask objective 1, subtask objective 2, and subtask objective 3, and subtask objective 1 matches an enabled Agent A, subtask objective 2 matches an enabled Agent B, and subtask objective 3 matches an enabled Agent C, then subtask 1 corresponding to subtask objective 1, subtask 2 corresponding to subtask objective 2, and subtask 3 corresponding to subtask objective 3 are generated. Subtask 1, subtask 2, and subtask 3 are then arranged and summarized to obtain the initial planning information for query content A: subtask 1 → subtask 2 → subtask 3. Finally, according to the initial planning information for query content A, the computer device first calls enabled Agent A to execute subtask 1, then calls enabled Agent B to execute subtask 2, and finally calls enabled Agent C to execute subtask 3, ultimately obtaining the task execution result corresponding to query content A.

[0090] Step S202: Based on the missing capability information in the initial planning information, determine the recommended processing unit corresponding to the missing capability information from the candidate processing unit set.

[0091] The candidate processing unit set refers to the set of candidate processing units corresponding to the missing capability information, specifically including multiple candidate processing units, such as multiple candidate agents. A candidate processing unit is a processing unit whose corresponding unit information and the missing capability information have a similarity that meets a preset similarity condition, such as the top K agents in similarity ranking (arranged from largest to smallest), or agents with similarity greater than a preset similarity.

[0092] Here, the recommendation processing unit refers to the recommendation agent, such as... Figure 14 The diagram shows the team-building activity generation agent and the multi-solution comparison analysis agent. Each missing capability information corresponds to a recommendation processing unit. For example, the team-building activity generation capability corresponds to a recommendation agent (e.g., Team-building Activity Generation Agent), and the multi-solution comparison analysis capability corresponds to a recommendation agent (e.g., Multi-solution Comparison Analysis Agent). The recommendation processing unit corresponding to the missing capability information refers to the candidate processing unit with the highest recommendation degree in the candidate processing unit set. Specifically, it refers to the optimal intelligent agent corresponding to that missing capability information. For example, the recommendation intelligent agent corresponding to the team-building activity generation capability is the Team-building Activity Generation Agent.

[0093] For example, the computer device acquires the unit information (such as capability description information) of the processing units in the initial processing unit set, and performs feature extraction processing on the missing capability information and the unit information respectively to obtain the information feature vector of the missing capability information and the information feature vector of the unit information of the processing unit; then, based on the information feature vector of the missing capability information and the information feature vector of the unit information of the processing unit, the similarity (such as cosine similarity) between the missing capability information and the unit information of the processing unit is calculated; next, the top K candidate processing units with the highest similarity (arranged from largest to smallest, and K is a positive integer greater than 0) are selected from the initial processing unit set to obtain the candidate processing unit set corresponding to the missing capability information; or, candidate processing units with a similarity greater than a preset similarity (such as 0.7) are selected from the initial processing unit set to obtain the candidate processing unit set corresponding to the missing capability information; finally, the query content, the missing capability information, and the metadata of the candidate processing units in the candidate processing unit set are input into the large language model to obtain the recommendation degree of each candidate processing unit, and the candidate processing unit with the highest recommendation degree is selected from the candidate processing unit set as the recommended processing unit corresponding to the missing capability information. For details, please refer to [reference needed]. Figure 5 For example, computer equipment determines the recommended agent corresponding to the team building activity generation capability from the candidate agent set corresponding to the team building activity generation capability, i.e., the team building activity generation agent; and determines the recommended agent corresponding to the multi-solution comparison analysis capability from the candidate agent set corresponding to the multi-solution comparison analysis capability, i.e., the multi-solution comparison analysis agent.

[0094] Step S203: Based on the recommended processing units, determine the newly added enabled processing units, and update the initial planning information according to the query content and the unit information of the newly enabled processing units to obtain the target planning information corresponding to the query content.

[0095] Among them, newly added enabled processing units refer to recommended processing units that have been confirmed by users; for example, reference Figure 14 When the intelligent recommendation pop-up displays the recommended team-building activity generation agent and the multi-solution comparison analysis agent, if the user clicks "Enable and Continue," the team-building activity generation agent and the multi-solution comparison analysis agent are confirmed to be enabled. The unit information for newly enabled processing units refers to relevant information such as capability descriptions.

[0096] Updating the initial planning information refers to partial adjustments and optimizations, rather than regenerating a new planning information. For example, adjusting the subtask in the initial planning information based on the task adjustment information of the subtask in the initial planning information; deleting missing capability information in the initial planning information; and adding the new subtask corresponding to the newly added activated processing unit to the initial planning information.

[0097] Among them, target planning information refers to the planning information obtained after updating the initial planning information based on the query content and the unit information of newly enabled processing units. Specifically, it refers to a sequence of sub-tasks executed collaboratively by multiple processing units. For example, the target planning information corresponding to the department team building query content includes: traffic plan query sub-task, team building activity generation sub-task, catering recommendation sub-task, and multi-plan comparison analysis sub-task.

[0098] For example, in response to a user's activation operation of the recommendation processing unit, the computer device confirms the recommendation processing unit as a newly enabled processing unit; for example, refer to Figure 10 When the user clicks "Start and Continue," it confirms the activation of agents M and N, which is equivalent to... Figure 11 The input shown is "Enable recommended agents M and N"; then, the computer device retrieves the unit information (such as capability description information) of the newly enabled processing unit from the local database, and determines the updated information of the initial planning information based on the query content, the unit information of the newly enabled processing unit, and the initial planning information, such as the task adjustment information of subtasks in the initial planning information, and the newly added subtasks corresponding to the newly enabled processing units (such as agents M and N); finally, the updated information of the initial planning information is used to update the initial planning information to obtain the updated planning information, which is used as the target planning information corresponding to the query content, such as... Figure 12 The query content A shown here corresponds to the target planning information.

[0099] In the above data processing method, the query content is first obtained. Based on the unit information of the enabled processing units, task planning is performed on the query content to obtain initial planning information corresponding to the query content, carrying missing capability information. Missing capability information is generated when the processing capabilities of the enabled processing units do not meet the query content requirements, and is used to characterize the missing capabilities for executing the query content. Next, based on the missing capability information in the initial planning information, recommended processing units corresponding to the missing capability information are determined from the candidate processing unit set. Finally, new enabled processing units are determined based on the recommended processing units. The initial planning information is updated according to the query content and the unit information of the new enabled processing units to obtain the target planning information corresponding to the query content. In this way, when performing task planning, task planning is performed on the query content based on the unit information of the enabled processing units, achieving the goal of task planning according to the enabled processing units. This ensures that the generated initial planning information can be executed by the enabled processing units, thereby improving the accuracy of planning information generation. Furthermore, during task planning, when the processing capacity of the activated processing units is insufficient to meet the query requirements, missing capability information is generated to characterize the missing capabilities for executing the query. Based on the query and the unit information of the recommended processing units corresponding to the missing capability information, the initial planning information is updated to obtain the target planning information corresponding to the query. This achieves the goal of timely detection of capability deficiencies and capability completion of the initial planning information, which helps improve the accuracy of planning information generation. Moreover, the recommended processing units are determined from the candidate processing unit set based on the missing capability information, ensuring the compatibility between the determined recommended processing units and the missing capability information. This makes the target planning information updated based on the unit information of the recommended processing units more accurate, further improving the accuracy of planning information generation.

[0100] In one exemplary embodiment, such as Figure 6 As shown, step S201 above, based on the unit information of the activated processing unit, performs task planning on the query content to obtain initial planning information carrying missing capability information corresponding to the query content, including the following steps S601 to S605. Wherein:

[0101] Step S601: Decompose the query content into tasks to obtain the sub-task objectives of the query content.

[0102] Step S602: Based on the unit information of the enabled processing units, determine the target enabled processing unit whose unit information satisfies the subtask objective from the enabled processing units.

[0103] Step S603: If the processing capacity of the target's enabled processing unit in the enabled processing unit does not meet the query content, determine the sub-task targets other than the sub-task targets of the target's enabled processing unit from each sub-task target of the query content, and obtain the remaining sub-task targets.

[0104] Step S604: Generate subtasks corresponding to subtasks of the target with enabled processing units, and generate missing capability information for the remaining subtask targets.

[0105] Step S605: Based on the subtask and missing capability information, generate the initial planning information corresponding to the query content.

[0106] Among them, the sub-task objectives of the query content refer to the sub-task objectives obtained after decomposing the query content into tasks based on the intent information of the query content, such as the sub-task objectives of transportation plan query, team building activity generation, restaurant recommendation, and multi-plan comparison analysis.

[0107] It should be noted that sub-task objectives are result-oriented sub-objectives obtained after decomposing the query content. They clearly define the specific results to be achieved in each decomposition step, such as transportation plan query, team building activity generation, restaurant recommendation, and multi-plan comparison analysis. They are the core basis for matching the processing capabilities of the enabled processing units. For example, based on the processing capabilities of the enabled processing units, those enabled processing units that can execute the sub-task objective are selected as the target enabled processing units corresponding to that sub-task objective. For instance, if a sub-task objective is a transportation plan query sub-task objective, and the enabled processing units include a transportation plan query agent, which is an agent specifically used to execute transportation plan query processing, it can execute the transportation plan query sub-task objective. Therefore, the target enabled processing unit corresponding to the transportation plan query sub-task objective is the transportation plan query agent. Subtasks are specific execution information generated based on subtask objectives and the processing capabilities of the corresponding enabled processing units. This includes information such as which enabled processing unit is selected, the corresponding target input parameters, and the corresponding target output. In other words, a subtask includes the specific execution information of the subtask objective, specifically the target input parameters, target output, and unit identifier of the enabled processing unit corresponding to the subtask objective. It specifies the steps, parameters, and execution flow that the enabled processing unit needs to complete to achieve the corresponding subtask objective. The two represent a correspondence between "objective" and "implementation path." The subtask objective determines "what result should be achieved," while the subtask clarifies "how to achieve it," jointly supporting the planning and execution of the query content.

[0108] Among them, the target enabled processing unit whose unit information meets the sub-task objective refers to the target enabled processing unit whose corresponding processing capability (such as capability description information) meets the sub-task objective, specifically the target enabled processing unit that can execute the sub-task objective.

[0109] The statement that the processing capacity of the target's enabled processing unit does not meet the query content can mean that the processing capacity of the target's enabled processing unit is only a portion of the target capacity required to execute the query content, or it can mean that the sub-task target of the target's enabled processing unit is only a portion of the sub-task targets of the query content.

[0110] Among them, the remaining subtask objectives refer to the subtask objectives other than those whose processing units have been activated, among the subtask objectives of the query content, such as the subtask objective of generating team building activities and the subtask objective of multi-scheme comparison analysis.

[0111] Among them, the subtasks corresponding to the target of the target's enabled processing unit specifically include the target input parameter information, target output information, and unit identifier (such as unit name) of the target's enabled processing unit.

[0112] The missing capability information corresponding to the remaining sub-task objectives refers to the capability descriptions or capability types required to execute the remaining sub-task objectives, such as the ability to generate team-building activities or the ability to compare and analyze multiple solutions. The initial planning information corresponding to the query content includes the sub-tasks corresponding to the sub-task objectives whose processing units have been activated, as well as the missing capability information corresponding to the remaining sub-task objectives.

[0113] For example, the computer device uses an intent recognition model to identify the intent of the query content, obtains the intent information of the query content, and decomposes the query content into multiple sub-task objectives based on the intent information. For example, for the query content of departmental team building input by the user, the computer device decomposes it into multiple sub-task objectives, namely, the sub-task objective of transportation plan query, the sub-task objective of team building activity generation, the sub-task objective of catering recommendation, and the sub-task objective of multi-plan comparison analysis. Next, the computer device matches the unit information (such as capability description information) of the enabled processing units with the sub-task target to obtain a matching result. Based on the matching result, it determines the enabled processing units whose unit information satisfies the sub-task target from the enabled processing units. That is, it determines the enabled processing units whose unit information successfully matches the sub-task target, and uses them as the target enabled processing units corresponding to the sub-task target. For example, the enabled agents corresponding to the user include a transportation plan query agent and a restaurant recommendation agent. The processing capability of the transportation plan query agent satisfies the transportation plan query sub-task, and the processing capability of the restaurant recommendation agent satisfies the restaurant recommendation sub-task. This means that the target enabled agent corresponding to the transportation plan query sub-task is the transportation plan query agent, and the target enabled agent corresponding to the restaurant recommendation sub-task is the restaurant recommendation agent. Next, the computer device determines whether the sub-task target of the target enabled processing unit is part of the sub-task targets of the query content. If so, it confirms that the processing capacity of the target enabled processing unit does not meet the query content, and removes the target enabled processing unit's sub-task target from the query content's sub-task targets, obtaining the remaining sub-task targets. Next, the computer device obtains the target input parameter information, target output information, and unit identifier of the target enabled processing unit, and generates sub-tasks corresponding to the target enabled processing unit's sub-task targets based on these parameters. Next, the computer device obtains the capability description information or capability type required to execute the remaining sub-task targets, as the missing capability information corresponding to the remaining sub-task targets. Finally, the computer device summarizes the generated sub-tasks and missing capability information to obtain the initial planning information corresponding to the query content.

[0114] In this embodiment, during the generation of initial planning information, if the processing capability of the target in the activated processing unit does not meet the query content, the remaining sub-task targets are determined from each sub-task target of the query content, and the missing capability information corresponding to the remaining sub-task targets is generated. Based on the sub-tasks and the missing capability information, the initial planning information corresponding to the query content is generated, so that the generated initial planning information carries the missing capability information, realizing the purpose of sensing the missing capability, which facilitates the subsequent updating of the initial planning information based on the missing capability information, thereby ensuring the accuracy of the final planning information and further improving the accuracy of planning information generation.

[0115] In an exemplary embodiment, step S604 above, generating a subtask corresponding to the subtask target of the target enabled processing unit, specifically includes the following: determining the target input parameter information of the target enabled processing unit based on the subtask target of the target enabled processing unit; determining the target output information of the target enabled processing unit based on the subtask target and the target input parameter information of the target enabled processing unit; and generating a subtask corresponding to the subtask target of the target enabled processing unit based on the target input parameter information, target output information, and unit identifier of the target enabled processing unit.

[0116] Among them, the target input parameter information of the target enabled processing unit refers to the input parameter information required to execute the sub-task target of the target enabled processing unit. Specifically, it includes the target input parameters and the parameter content of the target input parameters. For example, the target input parameters of the traffic plan query agent include the task scenario, budget per capita, core requirements, and constraints. The corresponding parameter content is "department team building", "500", "round-trip transportation options", "cost-effectiveness priority, matching team building time (next month), and compatible with activity venue location".

[0117] The target output information of the target enabled processing unit refers to the output information determined based on the sub-task target and target input parameter information of the target enabled processing unit. Specifically, it includes output connection information or output requirement information. For example, the output connection information of the transportation plan query agent is "synchronize the transportation mode, time, and cost details to the multi-plan comparison analysis agent", and the output requirement information of the multi-plan comparison analysis agent is "generate 2-3 complete team building plans, with a quantitative comparison table and the final recommendation reasons".

[0118] For example, the computer device determines the target input parameters and parameter content of the subtask target based on the subtask target of the target enabled processing unit and the unit information of the target enabled processing unit. It then summarizes the target input parameters and parameter content of the subtask target to obtain the target input parameter information of the target enabled processing unit. Next, the computer device determines the corresponding output information based on the subtask target and target input parameter information of the target enabled processing unit, which is used as the target output information of the target enabled processing unit. Finally, it obtains the unit identifier of the target enabled processing unit and summarizes the unit identifier, target input parameter information, and target output information of the target enabled processing unit in a preset order to obtain the subtask corresponding to the subtask target of the target enabled processing unit. For example, the subtasks corresponding to the transportation plan query subtask target include: "agent_id": "Transportation Plan Query Agent"; "input_params": {"Task Scenario": "Department Team Building"; "Budget per Person": "500"; "Core Requirements": "Round-trip transportation options"; "Constraints": "Prioritize cost-effectiveness, match team building time (next month), and be compatible with activity venue location"}; "Output Connection": "Synchronize transportation mode, time, and cost details to [Multi-Plan Comparison Analysis Agent]".

[0119] Furthermore, when determining the target input parameter information of the target enabled processing unit, if the subtask of the target enabled processing unit depends on the output results of the preceding subtasks, then the associated dependency parameter information of the subtask of the target enabled processing unit is determined based on the output results of the dependent subtasks. For example, the output results of the dependent subtasks are used as the associated dependency parameter information of the subtask of the target enabled processing unit. Next, the original input parameter information (such as task scenario, budget per capita, core requirements, and constraints) and the associated dependency parameter information of the subtask of the target enabled processing unit are summarized to obtain the target input parameter information of the target enabled processing unit. For example, the associated dependency parameter of the catering recommendation agent is: receiving the activity venue location output by the [team building activity generation agent].

[0120] Furthermore, when determining the target output information of the target enabled processing unit, if the output result of the subtask of the target enabled processing unit needs to be used as the input of the subsequent associated subtask, then based on the output result of the subtask of the target enabled processing unit and the enabled processing unit corresponding to the associated subtask, the output connection information of the target enabled processing unit is determined. For example, the output result of the subtask of the target enabled processing unit is synchronized to the enabled processing unit corresponding to the associated subtask, and this output connection information is used as the target output information of the target enabled processing unit. For example, the output connection information of the restaurant recommendation agent is: synchronize the restaurant name, address, average cost per person, and dish features to the [multi-solution comparison analysis agent].

[0121] In this embodiment, during the process of generating the subtask corresponding to the subtask target with the target's enabled processing unit, the target input parameter information, target output information, and unit identifier of the target's enabled processing unit are comprehensively considered, which helps to improve the accuracy of subtask generation.

[0122] In an exemplary embodiment, step S604 above, generating missing capability information corresponding to the remaining sub-task objectives, specifically includes the following: obtaining capability information required to execute the remaining sub-task objectives; and obtaining missing capability information corresponding to the remaining sub-task objectives based on the capability information required to execute the remaining sub-task objectives.

[0123] Among them, the capability information required to execute the remaining sub-task objectives refers to the capability type or capability description information required to execute the remaining sub-task objectives, such as the ability to generate team building activities, or the ability to plan team building activities (which needs to generate multiple optional plans in combination with budget and number of people).

[0124] For example, the computer device obtains the capability types required to execute the remaining sub-task objectives and uses these capability types as the missing capability information corresponding to the remaining sub-task objectives. For instance, if the capability type required to execute the team building activity generation sub-task objective is "team building activity generation capability," then the missing capability information corresponding to the team building activity generation sub-task objective is "team building activity generation capability." Alternatively, the computer device obtains the capability description information required to execute the remaining sub-task objectives and uses this capability description information as the missing capability information corresponding to the remaining sub-task objectives. For instance, if the capability description information required to execute the team building activity generation sub-task objective is "team building activity plan planning (requiring the generation of multiple optional plans based on budget and number of participants)," then the missing capability information corresponding to the team building activity generation sub-task objective is "team building activity plan planning (requiring the generation of multiple optional plans based on budget and number of participants)."

[0125] In this embodiment, the capability information required to execute the remaining sub-task objectives is obtained as the missing capability information corresponding to the remaining sub-task objectives. This allows for accurate perception of the missing capability information, facilitating subsequent capability recommendation and completion.

[0126] In an exemplary embodiment, step S605 above, which generates initial planning information corresponding to the query content based on subtasks and missing capability information, specifically includes the following: determining the order of subtasks based on the query content; arranging the subtasks according to the order to obtain the arranged subtasks; and summarizing the arranged subtasks and missing capability information to obtain the initial planning information corresponding to the query content.

[0127] The order of subtasks is determined by the intent information of the query content, specifically used to represent the dependencies between subtasks; for example, subtask 2 is executed after subtask 1. The ordered subtasks refer to the subtasks arranged in the specified order, such as subtask 1 → subtask 2 → subtask 3. It should be noted that the initial planning information includes both the ordered subtasks and missing capability information, with the missing capability information following the ordered subtasks.

[0128] For example, the computer determines the order of subtasks based on the intent information of the query content, and then arranges the subtasks according to the order to obtain the arranged subtasks, such as subtask 1, subtask 2 → subtask 3 → subtask 4; finally, the arranged subtasks and the missing capability information are summarized in the order of arranged subtasks first and missing capability information last to obtain the initial planning information corresponding to the query content. For example, the initial planning information for the department team building query is as follows: [{"agent_id": "Transportation Plan Query Agent", "input_params": {"Task Scenario": "Department Team Building", "Budget per Person": 500,"Core Requirements": "Round-trip Transportation Options", "Constraints": "Prioritize cost-effectiveness, match team building time (next month)"}}, {"agent_id": "Restaurant Recommendation Agent", "input_params": {"Task Scenario": "Department Team Building", "Budget per Person": 500, "Core Requirements": "Recommended Team Building Restaurant", "Constraints": "Accommodates department members, tastes suit most people's preferences"}}, {"missing_capability":["Team building activity plan planning (requires combining budget and number of people to generate multiple optional plans)", "Quantitative analysis of cost-effectiveness of multiple plans (requires comprehensive comparison of transportation, catering, and activity dimensions)"]}].

[0129] In this embodiment, the subtasks are arranged according to their order to obtain the arranged subtasks. The arranged subtasks and missing capability information are then summarized to obtain the initial planning information corresponding to the query content. In this way, the planning information includes the arranged subtasks and missing capability information, which facilitates subsequent capability recommendation and completion based on the missing capability information, thereby making the updated planning information more accurate. Moreover, arranging the subtasks ensures the accuracy of task execution, thereby ensuring the reliability of the task execution results.

[0130] In an exemplary embodiment, step S602, after determining the target enabled processing unit whose unit information satisfies the subtask objective from the enabled processing units based on the unit information of the enabled processing units, further includes a step of determining whether the processing capability of the target enabled processing unit satisfies the query content. Specifically, this includes: obtaining the target capability required to execute the query content; and confirming that the processing capability of the target enabled processing unit does not satisfy the query content if the processing capability of the target enabled processing unit is only a part of the target capability.

[0131] The target capabilities refer to all capabilities required to execute the query, such as the ability to query transportation options, generate team-building activities, recommend restaurants, and conduct multi-option comparative analysis. The fact that the processing capabilities of the target's enabled processing units do not meet the query requirements may mean that the processing capabilities of the target's enabled processing units are only a subset of the target capabilities.

[0132] For example, the computer device acquires all the capabilities required to execute the query content as target capabilities. For instance, the target capabilities required for executing the team building query content include transportation plan query capability, team building activity generation capability, restaurant recommendation capability, and multi-plan comparison analysis capability. Next, the computer device determines whether the processing capability of the target enabled processing unit is part of the target capabilities. If so, it confirms that the processing capability of the target enabled processing unit does not meet the query content. For example, if the target enabled processing unit includes a transportation plan query agent and a restaurant recommendation agent, it means that the processing capability of the target enabled processing unit only includes transportation plan query capability and restaurant recommendation capability, and lacks team building activity generation capability and multi-plan comparison analysis capability, thus confirming that the processing capability of the target enabled processing unit does not meet the query content.

[0133] Furthermore, if the processing capacity of the target's enabled processing unit is the same as the target's capacity required to execute the query content, the computer device confirms that the processing capacity of the target's enabled processing unit satisfies the query content.

[0134] In this embodiment, the target capabilities required to execute the query content are first obtained. Then, if the processing capability of the target enabled processing unit is only a part of the target capabilities, it is confirmed that the processing capability of the target enabled processing unit does not meet the query content. In this way, by comprehensively considering the processing capability of the target enabled processing unit and the target capabilities required to execute the query content, it is helpful to accurately determine whether there is a lack of capability.

[0135] In an exemplary embodiment, step S602, after determining the target enabled processing unit whose unit information satisfies the subtask objective from the enabled processing units based on the unit information of the enabled processing units, further includes a step of determining whether the processing capability of the target enabled processing unit satisfies the query content. Specifically, it includes the following: if the subtask objective of the target enabled processing unit is a part of the subtask objective of each subtask objective of the query content, it is confirmed that the processing capability of the target enabled processing unit does not satisfy the query content.

[0136] In this context, "the processing capacity of the target's enabled processing unit does not meet the query content" can refer to a subset of the sub-task targets among the sub-task targets of the target's enabled processing unit that are the query content.

[0137] For example, the computer device determines whether the sub-task target of the target's activated processing unit is a part of the sub-task targets of the query content. If so, it confirms that the processing capacity of the target's activated processing unit does not meet the query content. For example, the sub-task targets corresponding to the department team building query content include the transportation plan query sub-task target, the team building activity generation sub-task target, the catering recommendation sub-task target, and the multi-plan comparison analysis sub-task target. The target's activated processing unit includes the transportation plan query agent and the catering recommendation agent. The corresponding sub-task targets only include the transportation plan query sub-task target and the catering recommendation sub-task target, and lack the team building activity generation sub-task target and the multi-plan comparison analysis sub-task target. Therefore, it is confirmed that the processing capacity of the target's activated processing unit does not meet the query content.

[0138] Furthermore, if the sub-task targets of the target's activated processing unit are the same as the sub-task targets of the query content, the computer device confirms that the processing capacity of the target's activated processing unit satisfies the query content.

[0139] In this embodiment, when determining whether the processing capability of the target's enabled processing unit meets the query content, the sub-task objectives of the target's enabled processing unit and the sub-task objectives of the query content are comprehensively considered, which is beneficial to accurately determine whether there is a lack of capability.

[0140] In an exemplary embodiment, step S201 above, based on the unit information of the enabled processing units, performs task planning on the query content to obtain initial planning information corresponding to the query content. Specifically, this includes: determining a target planning information generation model corresponding to the query content; inputting the query content and the unit information of the enabled processing units into the target planning information generation model; decomposing the query content into tasks using the target planning information generation model to obtain sub-task targets for the query content; based on the unit information of the enabled processing units, identifying target enabled processing units whose unit information satisfies the sub-task targets from the enabled processing units; if the processing capacity of the target enabled processing units in the enabled processing units does not meet the query content's requirements, identifying sub-task targets other than the sub-task targets of the target enabled processing units from each sub-task target of the query content to obtain remaining sub-task targets; generating sub-tasks corresponding to the sub-task targets of the target enabled processing units; generating missing capability information corresponding to the remaining sub-task targets; and summarizing the sub-tasks and missing capability information to obtain the initial planning information corresponding to the query content.

[0141] Among them, the target planning information generation model refers to the planning information generation model (such as the large language model) that is adapted to the query content, and is used to generate planning information for the query content. For example, based on the query content and the unit information of the enabled processing units, the initial planning information for the query content is generated.

[0142] For example, the computer device selects a planning information generation model from multiple planning information generation models that corresponds to the content complexity of the query content, and uses this model as the target planning information generation model corresponding to the query content. Next, the query content and the unit information of the enabled processing units are input into the target planning information generation model. The target planning information generation model performs intent recognition on the query content to obtain intent information, and based on this intent information, it performs task decomposition on the query content to obtain multiple sub-task objectives. Then, the unit information of the enabled processing units (such as capability description information) is matched with the sub-task objectives to obtain matching results. Based on the matching results, the enabled processing units whose unit information satisfies the sub-task objectives are identified from the enabled processing units; that is, the enabled processing units whose unit information successfully matches the sub-task objectives are identified, and these are designated as the target enabled for that sub-task objective. The processing unit then determines whether the subtask target of the target's enabled processing unit is part of the subtask targets of the query content. If so, it confirms that the processing capability of the enabled processing unit does not meet the query content, and removes the subtask targets of the target's enabled processing unit from the subtask targets of the query content to obtain the remaining subtask targets. Next, it obtains the target input parameter information, target output information, and unit identifier of the target's enabled processing unit, and generates subtasks corresponding to the subtask targets of the target's enabled processing unit based on the target input parameter information, target output information, and unit identifier. Next, it obtains the capability description information or capability type required to execute the remaining subtask targets as the missing capability information corresponding to the remaining subtask targets. Finally, it summarizes the generated subtasks and missing capability information to obtain the initial planning information corresponding to the query content.

[0143] In this embodiment, by inputting the query content and the unit information of the enabled processing unit into the target planning information generation model corresponding to the query content, the initial planning information corresponding to the query content is obtained. In this way, by using the target planning information generation model corresponding to the query content to generate the initial planning information corresponding to the query content, the model advantages of the adapted model can be fully utilized, making the generated planning information more reliable and accurate.

[0144] In an exemplary embodiment, determining the target planning information generation model corresponding to the query content specifically includes the following: obtaining the content complexity of the query content; based on the content complexity of the query content and the correspondence between the query content complexity and the planning information generation model, obtaining the planning information generation model corresponding to the content complexity of the query content; and based on the planning information generation model corresponding to the content complexity of the query content, obtaining the target planning information generation model.

[0145] Content complexity refers to the execution complexity or task complexity of the query content, such as simple tasks and complex tasks. The planning information generation model refers to the model used to generate planning information, such as a large language model. There is a correspondence between content complexity and planning information generation models; different content complexities correspond to different planning information generation models. For example, if the content complexity of the query content is low, the corresponding planning information generation model is a low-cost, fast model; if the content complexity of the query content is high, the corresponding planning information generation model is a high-performance model.

[0146] Among them, the target planning information generation model refers to the planning information generation model corresponding to the content complexity of the query content.

[0147] For example, a computer device identifies the complexity of the query content to obtain its content complexity. This complexity can be determined, for instance, through a content complexity identification model (such as a deep learning model or a neural network model). Then, based on the content complexity of the query content and the correspondence between the query complexity and the planning information generation model, a planning information generation model corresponding to the content complexity of the query content is obtained, serving as the target planning information generation model. For example, if the content complexity of query content A is low, a low-cost, fast model is selected as the target planning information generation model for query content A; if the content complexity of query content A is high, a high-performance model is selected as the target planning information generation model for query content A.

[0148] In this embodiment, the content complexity of the query content is obtained, and the corresponding target planning information generation model is determined based on the content complexity of the query content. In this way, different target planning information generation models can be selected based on the content complexity of the query content, which is conducive to maximizing the use of the advantages of different models and improving the generation quality of planning information.

[0149] In an exemplary embodiment, the query content and the unit information of the enabled processing unit are input into the target planning information generation model, specifically including the following: obtaining a task planning prompt template corresponding to the query content; constructing a task planning prompt corresponding to the query content based on the task planning prompt template, the query content, and the unit information of the enabled processing unit; and inputting the task planning prompt into the target planning information generation model.

[0150] The query content is decomposed into sub-task objectives by using the goal planning information generation model. Specifically, the query content is decomposed into sub-task objectives based on task planning prompts.

[0151] Among them, the task planning prompt template refers to the template for generating task planning prompts; different query content corresponds to different task planning prompt templates; for example, query content A corresponds to task planning prompt template a; query content B corresponds to task planning prompt template b.

[0152] Among them, the task planning prompt refers to the prompt used to assist the target planning information generation model in generating initial planning information. Specifically, it includes role definition (such as "You are a senior task planner. Please generate an execution plan for the user task based on the list of currently available Agent capabilities"), unit information of the enabled processing units, query content, and output requirements (such as "Please output a JSON array listing the Agent IDs to be called and their brief input parameters in order; if the existing capabilities cannot fully meet the task, please indicate the missing capability types in the JSON").

[0153] For example, the computer device queries the correspondence between the query content and the preset query content and the task planning prompt template based on the query content, and obtains the task planning prompt template corresponding to the query content; then, it adds the query content and the unit information of the enabled processing unit to the corresponding position in the task planning prompt template, and obtains the task planning prompt corresponding to the query content; finally, it inputs the task planning prompt into the target planning information generation model, and through the target planning information generation model, it performs intent recognition on the query content based on the task planning prompt to obtain the intent information of the query content, and based on the intent information, it performs task decomposition on the query content to obtain multiple sub-task objectives of the query content.

[0154] For example, see reference. Figure 18The Central Planner is responsible for understanding the user task (i.e., the query content entered by the user) and generating an execution plan based on the currently available capability set. The implementation mechanism includes: 1. Multi-LLM routing: Automatically selecting an LLM (Large Language Model) based on task complexity; for example, choosing a low-cost, fast model for simple tasks and a high-performance model for complex planning / recommendation; 2. Plan generation: Using a carefully designed Prompt template to guide the LLM in DAG-based planning, ultimately outputting a structured plan (JSON) or a capability missing signal. The Prompt template content is as follows: Role: You are a senior task planner. Please generate an execution plan for the user task based on the currently available Agent capability list; Available Agent List: {enabled_agents_list}; User Task: {user_input}; Output Requirements: Please output a JSON array listing the Agent IDs to be invoked and their brief input parameters in order; If the existing capabilities cannot fully meet the task, please indicate the missing capability types in the JSON.

[0155] In this embodiment, based on the task planning prompt template corresponding to the query content, the query content, and the unit information of the enabled processing unit, a task planning prompt corresponding to the query content is constructed. The task planning prompt is then input into the target planning information generation model, which can help the target planning information generation model to better generate initial planning information, thereby making the planning information output by the model more accurate and reliable.

[0156] In an exemplary embodiment, step S202 above, before determining the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information, further includes a step of determining the candidate processing unit set, specifically including the following: obtaining the unit information of the processing units in the initial processing unit set; the processing unit is an intelligent agent; determining the similarity between the missing capability information and the unit information of the processing unit; determining the candidate processing units whose similarity meets the preset similarity conditions from the initial processing unit set, thereby obtaining the candidate processing unit set.

[0157] The initial processing unit set includes multiple processing units, specifically multiple intelligent agents, each with corresponding unit information. The initial processing unit set refers to the initial set of intelligent agents, such as... Figure 13 The Agent capability market is shown.

[0158] Among them, the similarity between missing capability information and unit information of processing unit is used to measure the semantic similarity between missing capability information and unit information of processing unit, such as cosine similarity.

[0159] The similarity condition meeting the preset similarity criteria means that the similarity ranks among the top K (e.g., 10, arranged in descending order) or the similarity is greater than the preset similarity (e.g., 0.7). The candidate processing unit set refers to the candidate agent set, including candidate processing units determined from the initial processing unit set whose similarity meets the preset similarity criteria. Specifically, it includes candidate agents that rank among the top K (e.g., 10, arranged in descending order) or candidate agents whose similarity is greater than the preset similarity (e.g., 0.7).

[0160] For example, the computer device acquires the unit information of the processing units in the initial set of processing units, and then inputs the unit information (such as capability description information) and missing capability information into an information feature extraction model (such as a text embedding model). The information feature extraction model performs information feature extraction processing on the unit information and missing capability information of the processing units to obtain the information feature vector of the missing capability information and the information feature vector of the unit information of the processing units. Next, using the cosine similarity calculation formula, based on the information feature vector of the missing capability information and the information feature vector of the unit information of the processing units, the cosine similarity between the missing capability information and the unit information of the processing units is calculated. Then, from the initial set of processing units, the top K candidate processing units with the highest cosine similarity (e.g., 10, arranged in descending order) are selected, or candidate processing units with a cosine similarity greater than a preset similarity (e.g., 0.7) are selected. Finally, the selected candidate processing units are summarized to obtain the candidate processing unit set corresponding to the missing capability information. For example, the set of candidate agents corresponding to the missing capability information "multi-solution comparison analysis capability" is: (candidate agent 1, candidate agent 2, ..., candidate agent 10).

[0161] In this embodiment, the similarity between the missing capability information and the unit information of the processing units in the initial processing unit set is first determined. Then, candidate processing units whose similarity meets the preset similarity conditions are determined from the initial processing unit set to obtain a candidate processing unit set. In this way, determining the candidate processing unit set first makes it easier to quickly determine the recommended processing unit corresponding to the missing capability information from the candidate processing unit set. This avoids the cumbersome process of determining the recommended processing unit from a massive number of processing units, and realizes the process from coarse selection to fine selection, which is conducive to improving the efficiency and accuracy of determining the recommended processing unit.

[0162] In an exemplary embodiment, step S201 above, before obtaining the query content, further includes a step of constructing an initial processing unit set, specifically including the following: obtaining the metadata of the processing unit; the metadata of the processing unit includes the unit information of the processing unit; according to the unit identifier of the processing unit, the metadata of the processing unit is associated and stored to obtain the initial processing unit set; the initial processing unit set includes the unit identifier and metadata of the processing unit.

[0163] The metadata of a processing unit includes relevant information such as unit identifier, unit name, capability description, capability domain, input data structure, output data structure, and performance metrics. In the initial set of processing units, each unit identifier is associated with corresponding metadata; for example, a traffic plan query agent is associated with corresponding metadata, and a restaurant recommendation agent is associated with corresponding metadata.

[0164] For example, a computer device acquires multiple processing units and constructs metadata for each processing unit, such as constructing metadata for a traffic plan query agent. Then, according to the unit identifier of the processing unit, the metadata of the processing units is associated and stored to obtain an initial set of processing units (unit identifier → metadata). For further illustration, see [reference]. Figure 18 The full-scale Agent capability marketplace serves as a core repository for registering, discovering, and managing all Agent capabilities. Its implementation mechanism includes: 1. Storing Agent metadata (Capability Metadata) using a database (such as MySQL); an example metadata structure is as follows: {"agent_id": "price_trend_analyzer_01", "name": "Historical Price Trend Analysis Agent", "description": "Analyzes historical price data of goods to predict future price trends", "capability_domain":["e-commerce", "price analysis"], "input_schema": {"type": "object", "properties": {"product_name": {"type": "string"}}}, "output_schema": {"type": "object", "properties": {"trend_prediction": {"type": "string"}}},"performance_metrics":{"avg_success_rate": 0.95, "avg_latency_ms": 1200}}; 2. Providing APIs for the central planner to query and retrieve, for example in... Figure 13On the Agent Capability Market interface shown, users can search for the Agent they need.

[0165] In this embodiment, the metadata of the processing unit is first obtained, and then the metadata of the processing unit is associated and stored according to the unit identifier of the processing unit to obtain an initial set of processing units. In this way, by pre-constructing the initial set of processing units, it is easier to determine the required candidate processing units from the initial set of processing units, which is beneficial to laying the foundation for accurate planning and recommendation in the future.

[0166] In an exemplary embodiment, step S202 above, based on the missing capability information in the initial planning information, determines the recommended processing unit corresponding to the missing capability information from the candidate processing unit set. Specifically, this includes: obtaining the metadata of the candidate processing units in the candidate processing unit set; inputting the query content, missing capability information, and the metadata of the candidate processing units into the target unit recommendation model; performing unit recommendation processing based on the query content, missing capability information, and the metadata of the candidate processing units through the target unit recommendation model to obtain the recommendation degree of the candidate processing units; determining the candidate processing unit with the highest recommendation degree from the candidate processing unit set, obtaining the recommended processing unit corresponding to the missing capability information, and generating the recommendation reason corresponding to the recommended processing unit.

[0167] Among them, the target unit recommendation model refers to the model used to perform unit recommendation processing. Specifically, it refers to the model used to output the recommendation degree of candidate processing units based on query content, missing ability information and metadata of candidate processing units, such as LLM which is good at analysis.

[0168] The recommendation score of a candidate processing unit is used to characterize the degree of recommendation of the candidate processing unit, such as 0.9, 0.7, 0.3, etc.

[0169] Among them, the recommendation processing unit corresponding to the missing capability information refers to the candidate processing unit with the highest recommendation degree in the set of candidate processing units corresponding to the missing capability information. Specifically, it refers to the recommendation agent corresponding to the missing capability information, such as the optimal agent corresponding to the missing capability information.

[0170] Here, the recommendation reason refers to the specific recommendation reason given by the recommendation processing unit. For example, reference Figure 14 The recommendation reason for the team building activity agent is: the current task explicitly requires providing an "activity" plan, and this agent specializes in this area. The recommendation reason for the multi-solution comparison analysis agent is: the task requires "analyzing the most cost-effective option," and this agent can provide data-driven decision support.

[0171] For example, a computer device acquires the metadata of candidate processing units from a set of candidate processing units. Then, from multiple large language models, it determines a target unit recommendation model that is adept at analysis. The query content, missing capability information, and the metadata of the candidate processing units are input into the target unit recommendation model. Based on the query content, missing capability information, and the metadata of the candidate processing units, the target unit recommendation model performs unit recommendation processing (such as similarity calculation) to obtain the similarity between the metadata of the candidate processing unit and the missing capability information, which is used as the recommendation score of the candidate processing unit. Next, the candidate processing units are re-ranked according to their recommendation scores to obtain ranked candidate processing units. Finally, from the ranked candidate processing units, the candidate processing unit with the highest recommendation score is selected as the recommended processing unit corresponding to the missing capability information, and a recommendation reason corresponding to the recommended processing unit is generated. For example, refer to... Figure 14 For the query content related to departmental team building activities, the recommended Agent for the missing capability information "team building activity generation capability" is the Team Building Activity Generation Agent, and the reason for recommending the Team Building Activity Generation Agent is: the current task clearly requires the provision of "activity" plans, and this Agent specializes in this.

[0172] In this embodiment, the query content, missing capability information, and metadata of candidate processing units are input into the target unit recommendation model to obtain the recommended processing unit corresponding to the missing capability information and the recommendation reason corresponding to the recommended processing unit. In this way, by using the model and referring to the metadata of the query content, missing capability information, and candidate processing units, the final determined recommended processing unit can be more accurate, thereby improving the accuracy of the determination of the recommended processing unit. At the same time, the recommendation reason corresponding to the recommended processing unit is also generated, providing a reference for subsequent selection of recommended processing units.

[0173] In an exemplary embodiment, the query content, missing capability information, and metadata of candidate processing units are input into the target unit recommendation model. Specifically, this includes: obtaining a unit recommendation prompt template for the target unit recommendation model; adding the query content, missing capability information, and metadata of candidate processing units to the unit recommendation prompt template to obtain a unit recommendation prompt; and inputting the unit recommendation prompt into the target unit recommendation model.

[0174] The target unit recommendation model performs unit recommendation processing based on query content, missing capability information, and metadata of candidate processing units to obtain the recommendation degree of candidate processing units. Specifically, it includes the following: The target unit recommendation model performs unit recommendation processing based on unit recommendation prompts to obtain the recommendation degree of candidate processing units.

[0175] Among them, the unit recommendation prompt template refers to the template for generating unit recommendation prompts, specifically the template for generating agent recommendation prompts.

[0176] Among them, the unit recommendation prompt refers to the prompt used to assist the target unit recommendation model in determining the recommendation processing unit and generating the recommendation reason for the recommendation processing unit. It can be generated by query content, missing capability information, metadata of candidate processing units, and unit recommendation prompt template. Specifically, it includes task requirements (such as "Recommend the most suitable supplementary Agent for the user task"), query content (such as user task {user_input}), missing capability information (such as missing capability description), metadata of candidate processing units (such as candidate Agent list {candidate_agents_list}), and output requirements (such as "Please output a JSON object containing: recommended_agent_id: the most recommended Agent ID; reasoning: detailed recommendation reason").

[0177] For example, the computer device retrieves the unit recommendation prompt template of the target unit recommendation model from the local database; then, it adds the query content, missing capability information, and metadata of the candidate processing unit to the corresponding positions in the unit recommendation prompt template to obtain the unit recommendation prompt; then, it inputs the unit recommendation prompt into the target unit recommendation model, and through the target unit recommendation model, based on the unit recommendation prompt, performs unit recommendation processing (such as similarity calculation) to obtain the similarity between the metadata of the candidate processing unit and the missing capability information, which is used as the recommendation degree of the candidate processing unit.

[0178] For example, see reference. Figure 18The Capability Matcher & Recommender engine functions as follows: it receives capability missing signals from the central planner, retrieves and recommends the most suitable Agent from the entire market. Its implementation mechanism is as follows: 1. Semantic Retrieval: Using a text embedding model (such as BGE), the missing capability description and the Agent descriptions in the Agent capability market are vectorized, and the matching degree is calculated using cosine similarity; 2. LLM Ranking and Reason Generation: The Top-K candidate Agents and their metadata, missing capability descriptions, and user tasks (i.e., the user-input query content) are input into an LLM algorithm skilled in analysis, performing final ranking and generating recommendation reasons. Finally, a list of recommended Agent IDs and their reasons are output and presented to the user through a UI. The LLM Prompt example is as follows: Task: Recommend the most suitable supplementary Agent for the user task; User task: {user_input}; Missing capability description: {missing_capability_description}; Candidate Agent list: {candidate_agents_list}; Output requirements: Please output a JSON object containing: recommended_agent_id: the most recommended Agent ID; reasoning: detailed reasons for the recommendation.

[0179] In this embodiment, the query content, missing capability information, and metadata of candidate processing units are added to the unit recommendation prompt template to obtain the unit recommendation prompt. The unit recommendation prompt is then input into the target unit recommendation model, which helps the target unit recommendation model to better determine the recommended processing units. This makes the recommended processing units determined by the model more accurate and reliable, and further improves the accuracy of the determination of recommended processing units.

[0180] In an exemplary embodiment, step S203 above, which determines the newly enabled processing unit based on the recommendation processing unit, specifically includes the following: obtaining the enable information for the recommendation processing unit; the enable information is used to characterize the enabling operation of the recommendation processing unit based on the unit identifier and recommendation reason corresponding to the recommendation processing unit; and confirming the recommendation processing unit as the newly enabled processing unit based on the enable information.

[0181] The activation information for the recommendation processing unit is used to characterize the activation operation for the recommendation processing unit, such as the generation of an Agent for a user to activate a recommended team building activity.

[0182] For example, after determining the unit identifier and recommendation reason corresponding to the recommendation processing unit, the computer device obtains the activation operation for the recommendation processing unit as activation information for the recommendation processing unit; then, upon detecting the activation information for the recommendation processing unit, the recommendation processing unit is confirmed as a newly activated processing unit. For example, refer to... Figure 14 When the intelligent recommendation pop-up displays the recommended team-building activity generation agent and its recommendation reasons, as well as the multi-scheme comparison analysis agent and its recommendation reasons, if the user clicks "Enable and Continue," they agree to enable the recommended team-building activity generation agent and multi-scheme comparison analysis agent. The computer then adds these two agents as newly enabled agents. If the user clicks "Do Not Enable," they disagree with enabling the recommended team-building activity generation agent and multi-scheme comparison analysis agent. The computer then executes the initial planning information according to the existing enabled agents, or terminates the query execution. See details in [link to relevant documentation]. Figure 5 .

[0183] In this embodiment, the activation information for the recommended processing unit is obtained, and based on the activation information, the recommended processing unit is confirmed as a newly activated processing unit, thus achieving the purpose of adding a new activated processing unit, that is, achieving the purpose of capability recommendation completion. This facilitates the subsequent updating of the initial planning information based on the unit information of the newly activated processing unit, which is beneficial to further improve the accuracy of planning information generation.

[0184] In an exemplary embodiment, step S203 above, which updates the initial planning information based on the query content and the unit information of the newly enabled processing unit to obtain the target planning information corresponding to the query content, specifically includes the following: confirming the updated information corresponding to the initial planning information based on the initial planning information, the query content, and the unit information of the newly enabled processing unit; the updated information includes at least the newly added subtask corresponding to the newly enabled processing unit, and the task adjustment information for the subtask in the initial planning information; updating the initial planning information based on the updated information to obtain the updated planning information corresponding to the query content; and obtaining the target planning information corresponding to the query content based on the updated planning information.

[0185] The updated information refers to information that updates the initial planning information. Specifically, this includes newly added subtasks corresponding to newly activated processing units, as well as task adjustment information for subtasks in the initial planning information. Newly added subtasks refer to the subtasks corresponding to the objectives of newly activated processing units, specifically including the unit identifier, target input parameters, and target output information of the newly activated processing unit. Task adjustment information refers to any adjustments made to the original subtasks in the initial planning information, such as adjusting target input parameters, adjusting target output information, or adjusting the order of subtasks.

[0186] Among them, target planning information refers to the updated planning information corresponding to the query content. Updated planning information refers to the planning information obtained after updating the initial planning information based on the updated information.

[0187] For example, the computer device determines the task adjustment information for the subtasks in the initial planning information and the subtask objectives for the newly activated processing unit based on the initial planning information, the query content, and the unit information of the newly activated processing unit. Next, based on the subtask objectives of the newly activated processing unit, it determines the target input parameter information for the newly activated processing unit. Based on the subtask objectives and target input parameter information of the newly activated processing unit, it determines the target output information for the newly activated processing unit. Based on the target input parameter information, target output information, and unit identifier of the newly activated processing unit, it generates the subtasks corresponding to the subtask objectives of the newly activated processing unit, which are then used as the newly added subtasks corresponding to the newly activated processing unit. Next, it summarizes the newly added subtasks corresponding to the newly activated processing unit and the task adjustment information for the subtasks in the initial planning information to obtain the updated information corresponding to the initial planning information. Finally, it uses the updated information to update the initial planning information to obtain the updated planning information corresponding to the query content, which is then used as the target planning information corresponding to the query content. For example, the computer device inputs the initial planning information, the query content, and the unit information of the newly added activated processing unit into the target planning information generation model. The target planning information generation model generates the updated information corresponding to the initial planning information, and uses the updated information to update the initial planning information to obtain the target planning information corresponding to the query content.

[0188] In this embodiment, based on the initial planning information, the query content, and the unit information of the newly added activated processing unit, the update information corresponding to the initial planning information is confirmed, and the initial planning information is updated based on the update information to obtain the target planning information corresponding to the query content. This achieves the purpose of updating the initial planning information without regenerating a new planning information, which helps to improve the efficiency of planning information generation and avoids wasting computer resources.

[0189] In an exemplary embodiment, the initial planning information is updated based on the updated information to obtain the updated planning information corresponding to the query content. Specifically, this includes: adjusting the subtasks in the initial planning information according to the task adjustment information to obtain the adjusted planning information; deleting the missing capability information in the adjusted planning information and adding the new subtasks to the adjusted planning information to obtain the updated planning information corresponding to the query content.

[0190] The adjusted planning information refers to the planning information obtained by adjusting the sub-tasks in the initial planning information based on the task adjustment information.

[0191] For example, the computer device adjusts the subtasks in the initial planning information based on the task adjustment information for the subtasks in the initial planning information to obtain the adjusted planning information; for example, it adds output connection information to the traffic plan query subtask; then, it deletes missing capability information in the adjusted planning information, such as deleting "missing_capability": [ "team building activity planning (requires generating multiple optional plans based on budget and number of people)", "multiple plan cost-effectiveness quantitative analysis (requires comprehensive comparison of transportation, catering, and activity dimensions)"; finally, it adds the new subtasks to the adjusted planning information to obtain the updated planning information corresponding to the query content, such as adding a team building activity generation subtask and a multiple plan comparison analysis subtask, and uses the updated planning information as the target planning information corresponding to the query content.For example, regarding the query for departmental team building activities, the target planning information is as follows: [{"agent_id": "Transportation Plan Query Agent", "input_params":{"Task Scenario": "Departmental Team Building","Budget Per Person": 500,"Core Requirements": "Round-trip Transportation Options", "Constraints": "Prioritize cost-effectiveness, match team building time (next month), and be compatible with activity venue location"}, "Output Connection": "Synchronize transportation mode, time, and cost details to

Multi-Plan Comparison Analysis Agent

Restaurant Recommendation Agent

Multi-Plan Comparison Analysis Agent

Team Building Activity Generation Agent

Multi-Option Comparison Analysis Agent

[0192] For example, see reference. Figure 18The Replan Optimizer's function is to make partial adjustments and optimizations to the original plan after a user enables a new Agent. Its implementation mechanism is as follows: 1. It receives information about the newly enabled Agent; 2. It inputs the original plan, the new Agent's capability description, and the task objective back into the central planner; the central planner generates a new, typically more optimized or concise, execution plan based on the enhanced capability set; the central planner, through context management, retains the results of successful steps to avoid duplicate execution, i.e., it avoids repeatedly generating subtasks that do not need modification.

[0193] In this embodiment, the subtasks in the initial planning information are adjusted according to the task adjustment information to obtain the adjusted planning information. Then, the missing capability information in the adjusted planning information is deleted, and the newly added subtasks are added to the adjusted planning information to obtain the updated planning information corresponding to the query content. This achieves the purpose of efficient local adjustment and optimization of the initial planning information, which is conducive to the rapid generation of target planning information and further improves the efficiency of planning information generation.

[0194] In an exemplary embodiment, step S203, after updating the initial planning information based on the query content and the unit information of the newly added enabled processing unit to obtain the target planning information corresponding to the query content, further includes a step of executing the target planning information, specifically including the following: using the newly added enabled processing unit to update the already enabled processing unit to obtain the updated enabled processing unit; if the processing capacity of the updated enabled processing unit meets the query content, calling the enabled processing unit corresponding to the subtask in the target planning information to execute the subtask to obtain the query result of the query content; if the processing capacity of the updated enabled processing unit does not meet the query content, using the target planning information as the updated initial planning information, and jumping to the step of determining the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information, until the processing capacity of the updated enabled processing unit meets the query content.

[0195] The updated enabled processing units include newly added enabled processing units and previously enabled processing units. The processing capacity of the updated enabled processing units satisfies the query content. This can mean that the processing capacity of the updated enabled processing units is the same as the target capacity required to execute the query content, or that the sub-task targets of the updated enabled processing units are the same as the sub-task targets of the query content. The processing capacity of the updated enabled processing units does not satisfy the query content. This can mean that the processing capacity of the updated enabled processing units is only a portion of the target capacity required to execute the query content, or that the sub-task targets of the updated enabled processing units are only a portion of the sub-task targets of the query content.

[0196] The query results refer to the task execution results, specifically the results obtained after executing the target planning information, such as specific departmental team building plans, mobile phone A being more suitable for college students, and computer E offering higher cost-effectiveness.

[0197] For example, the computer device summarizes the newly added and already enabled processing units to obtain an updated list of enabled processing units. For instance, if the already enabled processing units are a transportation plan query agent and a restaurant recommendation agent, and the newly added enabled processing units are a team building activity generation agent and a multi-plan comparison analysis agent, then the updated list of enabled processing units includes the transportation plan query agent, team building activity generation agent, restaurant recommendation agent, and multi-plan comparison analysis agent. Next, the computer device determines whether the processing capacity of the updated enabled processing units meets the query requirements. If so, it calls the enabled processing unit corresponding to the sub-task in the target planning information to execute the sub-task and obtain the query results. For example, it first calls the transportation plan query agent to execute the transportation plan query sub-task, then calls the team building activity generation agent to execute the team building activity generation sub-task, then calls the restaurant recommendation agent to execute the restaurant recommendation sub-task, and finally calls the multi-plan comparison analysis agent to execute the multi-plan comparison analysis sub-task, ultimately outputting a complete departmental team building plan. If not, the target planning information is used as the updated initial planning information, and the process jumps to the step of determining the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information. That is, the process jumps to step S202 to continuously update the generated planning information until the processing capability of the updated enabled processing unit meets the query content. Then, the loop ends, and the enabled processing unit corresponding to the subtask in the finally obtained target planning information is called to execute the subtask and obtain the query result of the query content. For details, please refer to [link to relevant documentation]. Figure 5 .

[0198] In this embodiment, newly added enabled processing units are used to update the already enabled processing units, resulting in updated enabled processing units. If the processing capacity of the updated enabled processing units meets the query requirements, the target planning information is executed. If the processing capacity of the updated enabled processing units does not meet the query requirements, the target planning information is continuously updated until the processing capacity of the updated enabled processing units meets the query requirements. This achieves the goal of a capability completion recommendation closed loop, ensuring that the final target planning information does not have any missing capabilities, further improving the accuracy of planning information generation. Simultaneously, executing the final target planning information makes the final query results more accurate, thereby improving the reliability and accuracy of task execution.

[0199] In one exemplary embodiment, such as Figure 7 As shown, another data processing method is provided. Taking the application of this method to a terminal as an example, it includes the following steps S701 to S703. Wherein:

[0200] Step S701: Display the query results on the query page.

[0201] Step S702: Display the initial planning information corresponding to the query content, which includes information on missing capabilities, on the query page.

[0202] The initial planning information is obtained by planning tasks for the query content based on the unit information of the enabled processing units; the missing capability information is generated when the processing capability of the enabled processing units does not meet the query content, and is used to characterize the missing capabilities for executing the query content.

[0203] Step S703: Display the target planning information corresponding to the query content on the query page.

[0204] Specifically, the target planning information is obtained by updating the initial planning information based on the query content and the unit information of the newly added activated processing unit; the newly added activated processing unit is determined based on the recommended processing unit; the recommended processing unit is used to represent the recommended processing unit corresponding to the missing capability information determined from the candidate processing unit set based on the missing capability information in the initial planning information.

[0205] The query page refers to the page where users can perform queries and obtain query results. See details for further information. Figure 8 On the query page, users can enter their query in the message sending area, for example, query A.

[0206] For example, the terminal responds to the user's input and displays the search results on the search page; for example, refer to Figure 8 The user enters query content A in the message sending area of ​​the query page and clicks "send". The terminal responds to the user's query operation and displays query content A on the query page. Next, based on the unit information of the activated processing unit, the terminal performs task planning on the obtained query content, obtaining initial planning information corresponding to the query content and carrying missing capability information, and displays this initial planning information on the query page; for example, referring to... Figure 9The terminal displays the initial planning information corresponding to query content A on the query page. Then, if the processing capacity of the enabled processing units does not meet the query requirements, the terminal, based on the missing capacity information in the initial planning information, determines the recommended processing unit corresponding to the missing capacity information from the candidate processing unit set. A smart recommendation pop-up window is then displayed on the query page, showing the unit identifier and recommendation reason for the recommended processing unit, allowing the user to select and enable the recommended processing unit; for example, referring to... Figure 10 The intelligent recommendation pop-up displays agent M and its recommendation reason, as well as agent N and its recommendation reason. When the user clicks "Enable and Continue," they agree to enable agents M and N, equivalent to the user typing "Enable recommended agents M and N" in the message sending area; for example, refer to... Figure 11 The query page displays "Enable recommended agents M and N". Finally, in response to the activation of the recommendation processing unit, the terminal confirms the recommendation processing unit as a newly enabled processing unit, and updates the initial planning information based on the query content and the unit information of the newly enabled processing unit, obtaining the target planning information corresponding to the query content, and displays the target planning information corresponding to the query content on the query page; for example, referring to... Figure 12 The terminal displays the target planning information corresponding to query content A on the query page.

[0207] For example, a user perceives a super AI assistant with "scalable capabilities and automated planning." The user simply describes a complex task naturally, and the system automatically calls upon multiple suitable AI experts (Agents) to collaborate on the task. If a missing expert is detected, it intelligently recommends a suitable Agent. After the user agrees with a single click, the system immediately re-optimizes the plan and integrates the new expert, seamlessly continuing the task. For instance, the user inputs the task: "I want to organize a departmental team-building event next month, with a budget of 500 yuan per person. I need a complete plan including transportation, activities, and catering, and analyze which plan offers the best cost-effectiveness." Initial state: The user has activated the [Transportation Plan Query Agent] and [Catering Recommendation Agent]. Next, the system executes the following steps: 1. Planning and Identifying Missing Agents: The central planner analyzes the task, planning to call the [Transportation Agent] and [Catering Agent], but identifies a missing Agent specifically for "Activity Planning" and "Cost-Effectiveness Analysis"; 2. Intelligent Recommendation: A pop-up interface appears (see reference). Figure 14The system detects that your task requires 'event planning' and 'cost-effectiveness analysis' capabilities. It recommends enabling the [Team Building Activity Generation Agent] and [Multi-Option Comparison Analysis Agent] to achieve better results. Do you want to enable them now? 3. User Confirmation and Replanning: The user clicks "Enable and Continue," and the system quickly replans based on the two newly added agents. 4. Collaborative Execution: The system automatically executes the new plan: [Transportation Agent] queries round-trip transportation options; [Dining Agent] recommends restaurants with an average cost of less than 500 yuan per person; [Team Building Activity Agent] generates multiple activity plans (such as parties, hiking, skiing); [Comparison Analysis Agent] integrates all information, generates 2-3 complete plans, compares their cost-effectiveness, and provides a final recommendation. In this way, the user transforms from a complex decision-maker into a simple decision approver, only needing to describe the goal and manage the AI ​​team to obtain professional and comprehensive automated services.

[0208] In the above data processing method, the query content is first displayed on the query page; then, the initial planning information corresponding to the query content, carrying missing capability information, is displayed on the query page; the initial planning information is obtained by performing task planning on the query content based on the unit information of the enabled processing units; the missing capability information is generated when the processing capabilities of the enabled processing units do not meet the query content, and is used to characterize the missing capabilities required to execute the query content; finally, the target planning information corresponding to the query content is displayed on the query page; the target planning information is obtained by updating the initial planning information based on the query content and the unit information of newly enabled processing units; the newly enabled processing units are determined based on recommended processing units; the recommended processing units are used to characterize the recommended processing units corresponding to the missing capability information determined from the candidate processing unit set based on the missing capability information in the initial planning information. In this way, when performing task planning, the query content is task planned based on the unit information of the enabled processing units, achieving the goal of task planning according to the enabled processing units, ensuring that the generated initial planning information can be executed by the enabled processing units, thereby improving the accuracy of planning information generation. Furthermore, during task planning, when the processing capacity of the activated processing units is insufficient to meet the query requirements, missing capability information is generated to characterize the missing capabilities for executing the query. Based on the query and the unit information of the recommended processing units corresponding to the missing capability information, the initial planning information is updated to obtain the target planning information corresponding to the query. This achieves the goal of timely detection of capability deficiencies and capability completion of the initial planning information, which helps improve the accuracy of planning information generation. Moreover, the recommended processing units are determined from the candidate processing unit set based on the missing capability information, ensuring the compatibility between the determined recommended processing units and the missing capability information. This makes the target planning information updated based on the unit information of the recommended processing units more accurate, further improving the accuracy of planning information generation.

[0209] In an exemplary embodiment, before displaying the query content on the query page, step S701 above further includes a step of displaying the target metadata of the enabled processing unit corresponding to the activation event. Specifically, this includes: displaying the target metadata of the processing unit on the processing unit operation page; and, in response to the activation event for the processing unit, displaying the target metadata of the enabled processing unit corresponding to the activation event in the target area of ​​the processing unit operation page.

[0210] The processing unit operation page refers to the agent operation page, where users can select to activate the agent, view agent details, configure agent data, etc. See details for further information. Figure 13 .

[0211] Among them, target metadata refers to the key metadata of the processing unit, such as the name of the agent, capability domain, capability description information, performance indicators, etc.

[0212] The "activation event" for the processing unit refers to the activation operation of the processing unit. The "target area" refers to the area displaying the activated processing unit, specifically the display area of ​​the activated intelligent agent, such as... Figure 13 The area showing enabled capabilities is indicated. Enabled processing units refer to the processing units corresponding to the enable event.

[0213] For example, in the first area of ​​the processing unit operation page, the terminal displays target metadata for multiple processing units; wherein, the first area refers to the display area for all processing units, specifically the area displaying all processing units, such as... Figure 13 The entire market capability display area is shown; then, in response to an activation operation on a processing unit in the first area, the target metadata of the enabled processing unit corresponding to the activation event is displayed in the target area of ​​the processing unit operation page, and the target metadata of the enabled processing unit corresponding to the activation event is de-displayed in the first area. For example, refer to... Figure 13 In the all-market capability display area of ​​the intelligent agent operation page, the target metadata of the team building activity generation agent is displayed, such as the team building activity generation agent, capability domain (activity planning), capability description information (generating a variety of interesting team building activity plans based on budget, number of people and preferences, and providing budget allocation suggestions), and performance indicators (success rate 95%, response time 2.5S, resource consumption is in progress). When the user clicks the "Enable Team Building Activity Generation Agent" button, it triggers the activation operation for the team building activity generation agent. The terminal responds to this activation operation by displaying the target metadata of the team building activity generation agent in the enabled capability display area and de-displaying the target metadata of the team building activity generation agent in the all-market capability display area.

[0214] In this embodiment, the target metadata of the processing unit is displayed on the processing unit operation page, which provides a reference for selecting and enabling the processing unit, thereby improving the convenience of selecting and enabling the processing unit.

[0215] In an exemplary embodiment, after displaying the target metadata of the processing unit on the processing unit operation page, the following content is further included: in response to a first target event for the processing unit, performing processing corresponding to the first target event on the processing unit; the first target event includes at least one of a details viewing event and a unit trial event.

[0216] The "Details View" event refers to the "Details View" operation, used to view detailed information about a processing unit, such as the agent's detailed information (e.g., detailed capability descriptions, input data structures, output data structures, etc.). The corresponding processing involves displaying the detailed information of the processing unit corresponding to the "Details View" operation on the details view page, such as displaying the agent's detailed information.

[0217] Here, a unit trial event refers to a unit trial operation, used to represent a trial processing unit, such as a trial agent. The corresponding processing for a unit trial operation refers to marking the processing unit corresponding to the unit trial operation as a trial state, such as marking an agent as a trial state.

[0218] For example, the first area of ​​the terminal in the processing unit operation page (such as...) Figure 13 In the area showing all market capabilities, target metadata for multiple processing units is displayed; then, in response to a details view operation for a processing unit in the first area, details of the processing unit corresponding to the details view operation are displayed on the details view page. For example, refer to... Figure 13 In the full market capability display area of ​​the intelligent agent operation page, the target metadata of the team building activity generated agent is displayed. When the user clicks the details button of the team building activity generated agent, it triggers the operation of viewing the details of the team building activity generated agent. The terminal responds to the details viewing operation and displays the detailed information of the team building activity generated agent, such as detailed capability description information, on the details viewing page.

[0219] For example, the first area of ​​the terminal in the processing unit operation page (such as...) Figure 13 In the area showing all market capabilities, target metadata for multiple processing units is displayed; then, in response to a trial operation for a processing unit in the first area, the processing unit corresponding to the trial operation is marked as trial status. For example, refer to... Figure 13 In the full market capability display area of ​​the intelligent agent operation page, the target metadata of the team building activity generation agent is displayed. When the user clicks the trial button of the team building activity generation agent, a trial operation for the team building activity generation agent is triggered; the terminal responds to the trial operation and marks the team building activity generation agent as a trial status.

[0220] In this embodiment, in response to a first target event for the processing unit, the processing unit performs processing corresponding to the first target event, thereby achieving the purpose of quickly viewing the detailed information of the selected processing unit and trying out the selected processing unit.

[0221] In an exemplary embodiment, after displaying the target metadata of the enabled processing unit corresponding to the enable event, the method further includes: in response to a second target event for the enabled processing unit, performing processing corresponding to the second target event on the enabled processing unit; the second target event includes at least one of an enable cancellation event, a details viewing event, and a data configuration event.

[0222] The "cancel enable" event refers to a cancellation operation, used to indicate the cancellation of an enabled processing unit, such as canceling the activation of an enabled agent. The corresponding processing involves removing the display of the target metadata of the enabled processing unit corresponding to the cancellation operation from the target area of ​​the processing unit operation page, and displaying the target metadata of the enabled processing unit corresponding to the cancellation operation in the first area.

[0223] The "Details View" event refers to the "Details View" operation, used to view the detailed information of an enabled processing unit, such as the detailed information of an enabled agent (e.g., detailed capability descriptions, input data structures, output data structures, etc.). The corresponding processing involves displaying the detailed information of the enabled processing unit on the "Details View" page, such as displaying the detailed information of an enabled agent.

[0224] Data configuration events refer to data configuration operations, used to configure relevant data for enabled processing units. This includes configuring data for enabled agents (such as input and output data structures), such as setting optional / required fields for input fields; input format adaptation (e.g., text / numeric / date type calibration); selecting output result field display dimensions (e.g., whether to display detailed parameters / simplified results); and adjusting output data formats (e.g., switching between JSON / text / table). The corresponding processing refers to configuring the relevant data for the enabled processing units on the data configuration page, such as configuring data for enabled agents.

[0225] For example, in response to a deactivation operation on an enabled processing unit in a target area, the terminal removes the display of target metadata for the enabled processing unit corresponding to the deactivation operation in the target area of ​​the processing unit operation page, and displays the target metadata for the enabled processing unit corresponding to the deactivation operation in the first area of ​​the processing unit operation page. For example, refer to... Figure 13In the enabled capability display area of ​​the intelligent agent operation page, the target metadata of the traffic plan query agent is displayed. When the user clicks the cancel enable button for the traffic plan query agent, a cancel enable operation is triggered for the traffic plan query agent. The terminal responds to the cancel enable operation by de-displaying the target metadata of the traffic plan query agent in the enabled capability display area and displaying the target metadata of the traffic plan query agent in the all market capability display area.

[0226] For example, in response to a details viewing operation for an enabled processing unit in a target area, the terminal displays details of the enabled processing unit corresponding to the details viewing operation on the details viewing page. For example, refer to... Figure 13 In the enabled capability display area of ​​the intelligent agent operation page, the target metadata of the traffic plan query agent is displayed. When the user clicks the details button of the traffic plan query agent, it triggers the operation of viewing the details of the traffic plan query agent. The terminal responds to the operation of viewing the details and displays the detailed information of the traffic plan query agent, such as detailed capability description information, on the details viewing page.

[0227] For example, in response to a data configuration operation for an enabled processing unit in a target area, the terminal displays relevant data for the enabled processing unit corresponding to the data configuration operation on the data configuration page for user configuration. For example, refer to... Figure 13 In the enabled capability display area of ​​the intelligent agent operation page, the target metadata of the traffic plan query agent is displayed. When the user clicks the configuration button of the traffic plan query agent, a data configuration operation for the traffic plan query agent is triggered. The terminal responds to the data configuration operation and displays the relevant data of the traffic plan query agent on the data configuration page for the user to configure, such as the optional / required settings of input fields and the selection of the field display dimensions of the output results.

[0228] In this embodiment, in response to a second target event for an enabled processing unit, the enabled processing unit is processed in accordance with the second target event, thereby achieving the purpose of canceling the enabled selected processing unit, quickly viewing the details of the selected enabled processing unit, and configuring the data of the selected enabled processing unit.

[0229] In an exemplary embodiment, after displaying the initial planning information carrying missing capability information corresponding to the query content on the query page in step S702, the following content is also included: displaying a processing unit recommendation pop-up window on the query page; the processing unit recommendation pop-up window displays the unit identifier, capability information and recommendation reason of the recommended processing unit.

[0230] In step S703 above, the target planning information corresponding to the query content is displayed on the query page. Specifically, this includes the following: in response to the activation event of the recommendation processing unit, the query page switches from displaying the recommendation pop-up of the processing unit to displaying the target planning information corresponding to the query content.

[0231] The processing unit recommendation pop-up refers to a pop-up on the query page that displays the unit identifier (e.g., a team building activity generation agent), capability information (e.g., generating various interesting team building activity plans based on budget, number of people, and preferences), and recommendation reasons (e.g., the current task explicitly requires providing "activity" plans, and this agent specializes in this). Figure 14 The intelligent recommendation pop-up window shown allows users to click and select a recommendation processing unit, such as clicking to select a team building activity to generate an agent or a multi-solution comparison analysis agent.

[0232] Among them, the start event for the recommendation processing unit refers to the start operation for the recommendation processing unit, which is used to indicate that the recommendation processing unit is started, such as starting the recommendation agent.

[0233] For example, after displaying the initial planning information containing missing capability information corresponding to the query content on the query page, if the terminal identifies that the processing capability of the enabled processing unit does not meet the query content, it determines the recommended processing unit corresponding to the missing capability information from the candidate processing unit set corresponding to the missing capability information based on the missing capability information in the initial planning information, and displays a processing unit recommendation pop-up window on the query page. This pop-up window displays the unit identifier, capability information, and recommendation reason of the recommended processing unit. Then, in response to the activation operation for the recommended processing unit, the terminal switches from displaying the processing unit recommendation pop-up window on the query page to displaying the target planning information corresponding to the query content. The terminal uses this activation operation to confirm the recommended processing unit as a newly enabled processing unit, and updates the initial planning information according to the query content and the unit information of the newly enabled processing unit to obtain the target planning information corresponding to the query content. For example, refer to... Figure 14The intelligent recommendation pop-up displays the team-building activity generation agent and its capability description: "Generates a variety of interesting team-building activity plans based on budget, number of participants, and preferences," with the recommendation reason: "The current task clearly requires providing activity plans, and this agent specializes in this." It also displays the multi-plan comparison analysis agent and its capability description: "Performs multi-dimensional (price, experience, cost-effectiveness) quantitative comparison analysis of multiple complex plans," with the recommendation reason: "The task requires the highest cost-effectiveness analysis; this agent can provide data-driven decision support." When the user clicks the "Start and Continue" button, it triggers the activation of the recommended team-building activity generation agent and multi-plan comparison analysis agent. In response to this activation, the terminal switches from displaying the intelligent recommendation pop-up to displaying the target planning information corresponding to the query content, such as... Figure 12 In the terminal, the target planning information corresponding to query content A is displayed on the query page.

[0234] In this embodiment, a pop-up window recommending processing units is displayed on the query page. Then, in response to the activation event for the recommended processing unit, the display of the pop-up window is switched to displaying the target planning information corresponding to the query content. This achieves the goal of quickly activating the selected recommended processing unit. At the same time, the pop-up window displays the unit identifier, capability information, and recommendation reasons of the recommended processing unit, providing a reference for selecting the recommended processing unit and avoiding the defect of selecting the recommended processing unit without any basis.

[0235] In one exemplary embodiment, the query content includes activity recommendations.

[0236] Step S703 above displays the target planning information corresponding to the query content on the query page, specifically including the following: displaying the target planning information corresponding to the activity recommendation content on the query page; the target planning information is obtained by updating the initial planning information based on the activity recommendation content and the agent information of the newly enabled agent; the newly enabled agent is determined based on the recommended agent; the recommended agent is used to represent the recommended agent corresponding to the missing capability information determined from the candidate agent set based on the missing capability information in the initial planning information.

[0237] The data processing method provided in this application also includes: in response to the execution operation on the target planning information, displaying the activity recommendation results corresponding to the activity recommendation content on the query page.

[0238] The activity recommendation content refers to the query content for activity recommendations, specifically departmental team building recommendations. For example, "I want to organize a departmental team building activity next month, with a budget of 500 yuan per person. I need a complete plan that includes transportation, activities, and catering, and analyze which plan has the best cost performance."

[0239] The agent information includes agent identifier, capability description information, capability domain, input data structure, output data structure, capability indicators, etc.

[0240] Among them, the activity recommendation results refer to the recommendation results corresponding to the activity recommendation content, specifically referring to complete team activity plans, such as complete departmental team building plans.

[0241] For example, in an activity recommendation scenario, such as a departmental team-building activity recommendation scenario, the terminal responds to the user's input and displays activity recommendations on the query page. For instance, a user pre-activates a transportation plan query agent and a catering recommendation agent, enters the activity recommendation content "I want to organize a departmental team-building activity next month, with a budget of 500 yuan per person. I need a complete plan including transportation, activities, and catering, and an analysis of which plan offers the best cost-effectiveness," in the message sending area of ​​the query page, and clicks "send." The terminal responds to the user's query and displays the activity recommendations on the query page. Next, based on the agent information of the activated agent, the terminal performs task planning on the obtained activity recommendations, obtaining initial planning information corresponding to the activity recommendations and including information on missing capabilities, and displays this initial planning information on the query page. Next, if the processing capacity of the enabled agent is insufficient to meet the activity recommendation requirements, the terminal, based on the missing capability information in the initial planning information, determines the recommended agent corresponding to the missing capability from the candidate agent set corresponding to the missing capability information. A smart recommendation pop-up is then displayed on the query page, showing the agent's identifier and recommendation reason, allowing the user to select and enable the recommended agent. For example, the pop-up might display a team-building activity generation agent and its recommendation reason, or a multi-solution comparison analysis agent and its recommendation reason. When the user clicks "Enable and Continue," they agree to enable the team-building activity generation agent and the multi-solution comparison analysis agent, equivalent to the user entering "Enable recommended team-building activity generation agent and multi-solution comparison analysis agent" in the message sending area. Following this, in response to the enabling operation of the recommended agent, the terminal confirms the recommended agent as a newly enabled agent and updates the initial planning information based on the activity recommendation content and the newly enabled agent's information, obtaining the target planning information corresponding to the activity recommendation content. This target planning information is then displayed on the query page. The user clicks the "Execute" button for the target planning information, triggering the execution operation for that target planning information. Finally, the terminal responds to the operation by calling the enabled agent corresponding to the subtask in the target planning information, executing the corresponding subtask, obtaining the activity recommendation results corresponding to the activity recommendation content, and displaying the activity recommendation results on the query page, such as displaying a complete department team building plan.

[0242] For example, in a product recommendation scenario, such as a computer recommendation scenario, the terminal responds to the user's input and displays product recommendation content (i.e., the query content for product recommendation) on the query page. For instance, the user pre-activates agents A and B, enters product recommendation content "Which of computers E, F, and G offers the best value for money, considering budget, performance, lifespan, and battery life?" in the message sending area of ​​the query page, and clicks "Send." The terminal responds to the user's query operation and displays the product recommendation content on the query page. Next, based on the agent information of the activated agents, the terminal performs task planning on the acquired product recommendation content, obtaining initial planning information corresponding to the product recommendation content and including information on missing capabilities, and displays this initial planning information on the query page. Next, if the processing capacity of the enabled agents is insufficient to meet the product recommendation requirements, the terminal, based on the missing capability information in the initial planning information, determines the recommended agent corresponding to the missing capability from the set of candidate agents corresponding to the missing capability information. A smart recommendation pop-up is then displayed on the query page, showing the agent's identifier and recommendation reason, allowing the user to select and enable the recommended agent. For example, the pop-up displays agent C and its recommendation reason, as well as agent D and its recommendation reason. When the user clicks "Enable and Continue," they agree to enable agents C and D, equivalent to the user entering "Enable recommended agents C and D" in the message sending area. Following this, in response to the enabling operation of the recommended agent, the terminal confirms the recommended agent as a newly enabled agent and updates the initial planning information based on the product recommendation content and the newly enabled agent's information, obtaining the target planning information corresponding to the product recommendation content. This target planning information is then displayed on the query page. The user clicks the "Execute" button for the target planning information, triggering the execution operation for that target planning information. Finally, the terminal responds to the operation by invoking the enabled agent corresponding to the subtask in the target planning information, executing the corresponding subtask, obtaining the product recommendation results corresponding to the product recommendation content, and displaying the product recommendation results on the query page, such as displaying the complete product recommendation plan. In this way, the process of obtaining the product recommendation results is carried out with reference to the target planning information corresponding to the product recommendation content, making the final product recommendation results more accurate and reliable, thereby improving the product recommendation accuracy rate.

[0243] For example, in a home appliance recommendation scenario, such as a kitchen small appliance selection recommendation scenario, the terminal responds to the user's input operation and displays recommended kitchen small appliances (i.e., the search content for kitchen small appliance recommendations) on the query page; for example, refer to Figure 15Users can pre-activate multiple smart agents, such as agent H and agent I, and enter kitchen appliance recommendations (e.g., "What kitchen appliances are worth buying?") in the message sending area of ​​the query page, then click "Send." The terminal responds to the user's query by displaying the kitchen appliance recommendations on the query page. In addition, the query page includes task record and new task options. Users can click the task record option to view relevant task information; clicking the new task option allows users to enter new query content on a newly created query page. Next, based on the agent information of the enabled agent, the terminal performs task planning on the acquired kitchen appliance recommendations, obtaining initial planning information corresponding to the recommendations. This initial planning information, along with the planning process, is displayed on the query page. The planning process includes multiple stages, each with a corresponding task description and execution time. For example, based on the input kitchen appliance recommendation content "What kitchen appliances are worth buying?", the terminal executes a "Kitchen Appliance Purchase Recommendation" plan, resulting in multiple sub-task objectives: identifying kitchen appliance candidates, comparing key parameters and scenarios, comparing prices and cost-effectiveness, and generating a comprehensive recommendation result, i.e., the kitchen appliance... Candidate selection → Comparison of key parameters and scenarios, price and cost-effectiveness → Generation of comprehensive recommendation results; For each sub-task objective, the terminal can further plan it. For example, the sub-task objective of selecting kitchen small appliances can be further divided into: kitchen small appliance brand screening, specific model screening, and intermediate result generation; Then, based on the sub-task objective and the relevant information of the corresponding activated intelligent agents, corresponding sub-tasks are generated, and all generated sub-tasks are summarized according to the execution order of the sub-tasks to obtain the initial planning information, namely the "Kitchen Small Appliance Purchase Recommendation" plan; Next, on the query page, the user clicks the execution button for the initial planning information to trigger the execution operation for the initial planning information. The terminal responds to the execution operation, calls the activated intelligent agents corresponding to the sub-tasks in the initial planning information, executes the corresponding sub-tasks, obtains the kitchen small appliance recommendation results corresponding to the recommended content, and displays the kitchen small appliance recommendation results on the query page, such as displaying a complete kitchen small appliance recommendation plan. See the specific reference. Figure 16 Each recommended small kitchen appliance is accompanied by corresponding price information, key advantages, target audience information, and purchase link information.

[0244] Furthermore, if the processing capacity of the enabled intelligent agent is insufficient to meet the recommended content for small kitchen appliances, the terminal, based on the missing capability information in the initial planning information, determines the recommended intelligent agent corresponding to the missing capability information from the candidate intelligent agent set corresponding to the missing capability information. A smart recommendation pop-up window is then displayed on the query page, showing the intelligent agent's identifier and recommendation reason, allowing the user to select and enable the recommended intelligent agent. For example, the smart recommendation pop-up window may display intelligent agent M and its recommendation reason, and intelligent agent N and its recommendation reason. When the user clicks "Enable and Continue," they agree to enable intelligent agents M and N, equivalent to the user entering "Enable recommended intelligent agents M and N" in the message sending area. Next, in response to the enabling operation of the recommended intelligent agent, the terminal confirms the recommended intelligent agent as a newly enabled intelligent agent and updates the initial planning information based on the recommended content for small kitchen appliances and the intelligent agent information of the newly enabled intelligent agent, obtaining the target planning information corresponding to the recommended content for small kitchen appliances. This target planning information is then displayed on the query page. The user clicks the execute button for the target planning information, triggering the execution operation for that target planning information. Finally, the terminal responds to the operation by invoking the enabled agent corresponding to the subtask in the target planning information, executing the corresponding subtask, obtaining the kitchen appliance recommendation results corresponding to the recommended content, and displaying the final kitchen appliance recommendation results on the query page, such as displaying the final complete kitchen appliance recommendation scheme. In this way, the process of obtaining the kitchen appliance recommendation results is carried out with reference to the target planning information corresponding to the recommended content, making the final kitchen appliance recommendation results more accurate and reliable, thereby improving the accuracy rate of kitchen appliance recommendations.

[0245] In this embodiment, the target planning information corresponding to the activity recommendation content is displayed on the query page. Then, in response to the execution operation on the target planning information, the activity recommendation result corresponding to the activity recommendation content is displayed on the query page. In this way, the process of obtaining the activity recommendation result corresponding to the activity recommendation content is carried out by referring to the target planning information corresponding to the activity recommendation content to execute the corresponding sub-tasks, which makes the final activity recommendation result more accurate and reliable, thereby improving the activity recommendation accuracy.

[0246] In one exemplary embodiment, such as Figure 17 As shown, another data processing method is provided, illustrated by its application to a computer device, which can be a terminal or a server. It can be understood that this method can be executed independently by the terminal or server, or it can be implemented through interaction between the terminal and the server. (See reference) Figure 17 The defect detection method includes the following steps S1701 to S1718. Wherein:

[0247] Step S1701: Obtain the metadata of the processing unit; the metadata of the processing unit includes the unit information of the processing unit; according to the unit identifier of the processing unit, the metadata of the processing unit is associated and stored to obtain an initial set of processing units; the initial set of processing units includes the unit identifier and metadata of the processing unit.

[0248] Step S1702: Obtain the query content; obtain the content complexity of the query content; based on the content complexity of the query content and the correspondence between the query content complexity and the planning information generation model, obtain the planning information generation model corresponding to the content complexity of the query content, and use it as the target planning information generation model corresponding to the query content.

[0249] Step S1703: Obtain the task planning prompt template corresponding to the query content; based on the task planning prompt template, the query content, and the unit information of the enabled processing units, construct the task planning prompt corresponding to the query content; input the task planning prompt into the target planning information generation model.

[0250] Step S1704: Using the target planning information generation model, based on the task planning prompts, the query content is decomposed into sub-task targets of the query content; based on the unit information of the enabled processing units, the enabled processing units whose unit information satisfies the sub-task targets are determined from the enabled processing units.

[0251] Step S1705: If the processing capability of the target enabled processing unit is only a portion of the target capability required to execute the query content, or if the subtask target of the target enabled processing unit is only a portion of the subtask targets of the query content, then it is confirmed that the processing capability of the target enabled processing unit in the enabled processing unit does not meet the query content.

[0252] Step S1706: From the sub-task targets of the query content, identify the sub-task targets other than those whose processing units have been enabled, and obtain the remaining sub-task targets; obtain the capability information required to execute the remaining sub-task targets; based on the capability information required to execute the remaining sub-task targets, obtain the missing capability information corresponding to the remaining sub-task targets.

[0253] Step S1707: Based on the sub-task target of the target enabled processing unit, determine the target input parameter information of the target enabled processing unit; based on the sub-task target and target input parameter information of the target enabled processing unit, determine the target output information of the target enabled processing unit.

[0254] Step S1708: Based on the target input parameter information, target output information and unit identifier of the target enabled processing unit, generate the subtask corresponding to the target's subtask target.

[0255] Step S1709: Based on the query content, determine the order of the subtasks; arrange the subtasks according to the order to obtain the arranged subtasks; summarize the arranged subtasks and missing capability information to obtain the initial planning information corresponding to the query content.

[0256] Step S1710: Obtain the unit information of the processing units in the initial processing unit set; the processing unit is an intelligent agent; determine the similarity between the missing capability information in the initial planning information and the unit information of the processing unit; from the initial processing unit set, determine the candidate processing units whose similarity meets the preset similarity conditions, and obtain the candidate processing unit set.

[0257] Step S1711: Obtain the metadata of the candidate processing units in the candidate processing unit set; obtain the unit recommendation prompt template of the target unit recommendation model; add the query content, missing capability information and the metadata of the candidate processing units to the unit recommendation prompt template to obtain the unit recommendation prompt; input the unit recommendation prompt into the target unit recommendation model.

[0258] Step S1712: Using the target unit recommendation model, based on the unit recommendation prompts, perform unit recommendation processing to obtain the recommendation degree of the candidate processing units; from the set of candidate processing units, determine the candidate processing unit with the highest recommendation degree, obtain the recommended processing unit corresponding to the missing capability information, and generate the recommendation reason corresponding to the recommended processing unit.

[0259] Step S1713: Obtain the activation information for the recommendation processing unit; the activation information is used to characterize the activation operation of the recommendation processing unit based on the unit identifier and recommendation reason corresponding to the recommendation processing unit; based on the activation information, confirm the recommendation processing unit as a newly activated processing unit.

[0260] Step S1714: Based on the initial planning information, the query content, and the unit information of the newly added activated processing unit, confirm the update information corresponding to the initial planning information; the update information includes at least the newly added sub-tasks corresponding to the newly activated processing unit, and the task adjustment information for the sub-tasks in the initial planning information.

[0261] Step S1715: Adjust the subtasks in the initial planning information according to the task adjustment information to obtain the adjusted planning information; delete the missing capability information in the adjusted planning information and add the new subtasks to the adjusted planning information to obtain the updated planning information corresponding to the query content; based on the updated planning information, obtain the target planning information corresponding to the query content.

[0262] Step S1716: Update the already enabled processing unit using the newly added enabled processing unit to obtain the updated enabled processing unit.

[0263] Step S1717: If the processing capacity of the updated enabled processing unit meets the query content, call the enabled processing unit corresponding to the subtask in the target planning information to execute the subtask and obtain the query result of the query content.

[0264] In step S1718, if the processing capacity of the updated enabled processing unit does not meet the query content, the target planning information is used as the initial planning information for the update, and the process jumps to step S1710 until the processing capacity of the updated enabled processing unit meets the query content.

[0265] In the aforementioned data processing method, during task planning, task planning is performed on the query content based on the unit information of the activated processing units. This achieves the goal of task planning according to the activated processing units, ensuring that the generated initial planning information can be executed by the activated processing units, thereby improving the accuracy of planning information generation. Furthermore, during task planning, if the processing capacity of the activated processing units is insufficient for the query content, missing capability information is generated to characterize the missing capabilities for executing the query content. Based on the query content and the unit information of the recommended processing units corresponding to the missing capability information, the initial planning information is updated to obtain the target planning information corresponding to the query content. This achieves the goal of timely detection of capability deficiencies and capability completion of the initial planning information, which helps improve the accuracy of planning information generation. Moreover, the recommended processing units are determined from the candidate processing unit set based on the missing capability information, ensuring the compatibility between the determined recommended processing units and the missing capability information. This makes the target planning information updated based on the unit information of the recommended processing units more accurate, further improving the accuracy of planning information generation.

[0266] To more clearly illustrate the data processing method provided in the embodiments of this application, the following describes the data processing method in detail with a specific embodiment. In an exemplary embodiment, such as Figure 5As shown, a method for intelligent planning, execution, and capability completion recommendation of complex tasks based on an agent capability market (i.e., intelligent planning and execution method based on an agent capability market) is provided. Its core idea is to construct a "capability market" with agents as capability units, where users can flexibly configure their "enabled sets." A central planner decomposes and orchestrates tasks based on this "enabled set." When capabilities are insufficient, the system can proactively analyze and recommend completion from the "full capability market." After user confirmation, the system can dynamically replan and integrate new capabilities. Key technical points include: 1. Construction and metadata description of the agent capability market: Defining a structured capability metadata schema allows each agent to self-describe its functional boundaries, interfaces, and characteristics, laying the foundation for accurate planning and recommendation. 2. Multi-Agent collaborative planning based on enabled sets: The central planner (LLM) decomposes user tasks into an execution plan based on a DAG (Directed Acyclic Graph), where each node is a call to a specific agent, realizing automated pipeline processing of complex tasks. 3. Capability Deficiency Analysis and Intelligent Recommendation: When the central planner identifies that the required capabilities exceed the user's current "enabled set," it automatically triggers the recommendation process. LLM analysis is used to determine the match between the task objective and the capabilities of agents across the entire market, generating a precise list of supplementary recommendations with justifications. 4. Dynamic Replanning and Capability Integration: After a user adopts a recommendation, the system doesn't need to completely rewrite the plan. Instead, based on the new, enhanced capability set, it efficiently adjusts and optimizes the original plan locally, quickly generating a new executable plan. 5. Multi-LLM Decision-Making Mechanism: In different stages such as planning, evaluation, and recommendation, intelligent routing is used to process the most suitable LLM model, maximizing the advantages of different models and improving the decision-making quality at each stage of the system.

[0267] The application makes the following: Figure 18The system shown extends the original single-agent architecture with a parallel reasoning and decision-making layer. The specific execution process is as follows: The user inputs a task (e.g., querying departmental team building activities, querying products) to the central planner (Multi-LLM) in the decision-making layer; after receiving the task, the central planner queries the available capabilities of the enabled capability set (Agent A, Agent B), and based on these available capabilities, performs task planning for the user-input task, obtaining initial planning information; if the enabled capability set does not meet the user-input task, i.e., when capabilities are insufficient, the capability matching and recommendation engine is triggered; the capability matching and recommendation engine, based on the missing capability information generated by the central planner, searches and analyzes the full agent capability market (Agent C, Agent D, Agent E, etc.). The optimal Agent matching the missing capability information is obtained, and an Agent recommendation list is generated and sent to the user. After the user confirms the Agent recommendation list, the enabled capability set is updated. The Replan optimizer obtains the new capability set from the enabled capability set and sends it to the central planner to request Replan execution. Based on the initial planning information, the capability description information of the new Agents in the new capability set, and the user-input task, the central planner makes local adjustments and optimizations to the initial planning information to obtain the target planning information (i.e., the final plan). Finally, based on the target planning information, multiple corresponding Agents are invoked to execute collaboratively to obtain the execution results corresponding to the user-input task, such as departmental team building query results or product query results.

[0268] Figure 5This describes the complete workflow of the system from task input to final completion, especially the closed loop of capability completion and recommendation. The specific steps are as follows: 1. Task Input and Parsing: The user submits a task, and the Central Planner (LLM) identifies the intent. 2. Initial Plan Generation: The planner generates an initial plan P0 based on the currently enabled Agent capability set. 3. Capability Satisfaction Judgment: The system determines whether the initial plan P0 can be satisfied by the current capability set, i.e., whether the current capability set satisfies the user's submitted task, thereby confirming whether any capabilities are missing. 4. Capability Execution and Result Summarization: If satisfied, the corresponding Agents are executed sequentially, and the results are summarized and output. 5. Capability Missing Analysis and Recommendation. If the requirements are not met, the central planner will identify the missing capability type; the capability recommendation engine will then start, using semantic retrieval and LLM ranking to find the most relevant Agent from the entire market and generate a recommendation reason; 6. User confirmation: The recommendation results are presented to the user through the UI interface, requesting confirmation to start; 7. Dynamic replan: After user confirmation, the new Agent is dynamically activated; the replan optimizer will make local adjustments to the original plan P0 based on the new capability set, generating an optimized new plan P1; 8. Iterative execution: The system, with the new plan P1, jumps back to step 3 to continue execution until the task is completed, that is, until the generated plan has no missing capabilities and then the plan is executed.

[0269] The above embodiments, by constructing an Agent capability market and introducing capability awareness, intelligent recommendation, and dynamic replanning mechanisms, produce the following significant beneficial effects compared to existing MCP tool markets or single Agent market solutions:

[0270] 1. Fundamentally enhances the ability to handle complex tasks: By using "Agent" as the planning unit, the abstraction level and robustness of the planning are greatly improved, enabling efficient and reliable handling of comprehensive tasks that require multiple steps of complex logic, which is unmatched by the atomic tool market.

[0271] 2. Create an "evolvable" user experience: The system has the evolutionary capability of "perception-recommendation-optimization". Users do not need to have all the Agent knowledge in advance. The system can guide users to gradually build the combination of capabilities that best suits their needs, which greatly reduces the threshold for use.

[0272] 3. Achieve intelligent capability scheduling and orchestration: Transform "multi-agent collaboration" from manual selection to automatic planning and execution, realize true automated collaboration of AI teams, and achieve a synergistic effect of "1+1>2".

[0273] 4. Provides cost-effective capability optimization paths: Users can flexibly start and stop agents according to actual needs and pay only as needed; the system's recommendation mechanism ensures that resources are accurately invested in the capabilities that generate the most value, avoiding resource waste.

[0274] 5. Enhanced abstraction level and planning robustness: Using "Agent" as the planning unit, each unit can handle a complex sub-task, which greatly simplifies the planning logic and improves the reliability and execution efficiency of the plan.

[0275] 6. Enables automatic multi-agent collaboration: The central planner automatically decomposes tasks and assigns them to multiple specialized agents to complete collaboratively, without requiring users to worry about the underlying call details.

[0276] 7. Intelligent capability completion: The system has the ability to perceive "capability gaps" and can intelligently recommend the most suitable supplementary agents from a global perspective, significantly reducing the user's learning curve and improving the task completion rate.

[0277] 8. Dynamic optimization and personalization: The user's capability set can be flexibly configured, and the system can dynamically adjust the planning strategy accordingly to achieve a truly personalized task processing pipeline.

[0278] 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 in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0279] Based on the same inventive concept, this application also provides a data processing apparatus for implementing the data processing method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more data processing apparatus embodiments provided below can be found in the limitations of the data processing method described above, and will not be repeated here.

[0280] In one exemplary embodiment, such as Figure 19 As shown, a data processing device 1900 is provided, including: a task planning module 1901, a unit determination module 1902, and an information update module 1903, wherein:

[0281] The task planning module 1901 is used to obtain query content and perform task planning on the query content based on the unit information of the enabled processing units to obtain initial planning information corresponding to the query content carrying missing capability information. The missing capability information is generated when the processing capability of the enabled processing units does not meet the query content and is used to characterize the missing capabilities for executing the query content.

[0282] The unit determination module 1902 is used to determine the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information.

[0283] The information update module 1903 is used to determine the newly added activated processing unit based on the recommended processing unit, and update the initial planning information according to the query content and the unit information of the newly added activated processing unit to obtain the target planning information corresponding to the query content.

[0284] In an exemplary embodiment, the task planning module 1901 is further configured to decompose the query content into sub-task objectives; based on the unit information of the enabled processing units, determine the target enabled processing units whose unit information satisfies the sub-task objectives from the enabled processing units; if the processing capacity of the target enabled processing unit in the enabled processing units does not meet the query content, determine the remaining sub-task objectives from each sub-task objective of the query content, excluding the sub-task objectives of the target enabled processing units; generate sub-tasks corresponding to the sub-task objectives of the target enabled processing units; generate missing capability information corresponding to the remaining sub-task objectives; and generate initial planning information corresponding to the query content based on the sub-tasks and missing capability information.

[0285] In an exemplary embodiment, the task planning module 1901 is further configured to: determine target input parameter information of the target enabled processing unit based on the sub-task target of the target enabled processing unit; determine target output information of the target enabled processing unit based on the sub-task target and the target input parameter information of the target enabled processing unit; and generate a sub-task corresponding to the sub-task target of the target enabled processing unit based on the target input parameter information, target output information and unit identifier of the target enabled processing unit.

[0286] In an exemplary embodiment, the task planning module 1901 is further configured to obtain capability information required to execute the remaining sub-task objectives; and based on the capability information required to execute the remaining sub-task objectives, obtain the missing capability information corresponding to the remaining sub-task objectives.

[0287] In an exemplary embodiment, the task planning module 1901 is further configured to determine the order of subtasks based on the query content; arrange the subtasks according to the order to obtain the arranged subtasks; and summarize the arranged subtasks and missing capability information to obtain the initial planning information corresponding to the query content.

[0288] In an exemplary embodiment, the task planning module 1901 is further configured to obtain the target capabilities required to execute the query content; and if the processing capability of the target enabled processing unit is only a portion of the target capabilities, confirm that the processing capability of the target enabled processing unit does not meet the query content.

[0289] In an exemplary embodiment, the task planning module 1901 is further configured to confirm that the processing capacity of the target's enabled processing unit does not meet the query content when the subtask target of the target's enabled processing unit is a subset of the subtask targets of the query content.

[0290] In an exemplary embodiment, the task planning module 1901 is further configured to determine a target planning information generation model corresponding to the query content, input the query content and the unit information of the enabled processing units into the target planning information generation model; decompose the query content into tasks through the target planning information generation model to obtain sub-task targets of the query content; based on the unit information of the enabled processing units, determine the target enabled processing units whose unit information satisfies the sub-task targets from the enabled processing units; if the processing capacity of the target enabled processing units in the enabled processing units does not meet the query content, determine the sub-task targets other than the sub-task targets of the target enabled processing units from each sub-task target of the query content to obtain the remaining sub-task targets; generate sub-tasks corresponding to the sub-task targets of the target enabled processing units, and generate missing capability information corresponding to the remaining sub-task targets; summarize the sub-tasks and missing capability information to obtain the initial planning information corresponding to the query content.

[0291] In an exemplary embodiment, the task planning module 1901 is further configured to obtain the content complexity of the query content; based on the content complexity of the query content and the correspondence between the query content complexity and the planning information generation model, obtain the planning information generation model corresponding to the content complexity of the query content; and based on the planning information generation model corresponding to the content complexity of the query content, obtain the target planning information generation model.

[0292] In an exemplary embodiment, the task planning module 1901 is further configured to obtain a task planning prompt template corresponding to the query content; construct a task planning prompt corresponding to the query content based on the task planning prompt template, the query content, and the unit information of the enabled processing unit; input the task planning prompt into the target planning information generation model; and, through the target planning information generation model, decompose the query content into tasks based on the task planning prompt to obtain the sub-task targets of the query content.

[0293] In an exemplary embodiment, the data processing device 1900 further includes a unit filtering module, used to obtain unit information of processing units in an initial set of processing units; the processing unit is an intelligent agent; determine the similarity between missing capability information and unit information of the processing unit; and determine candidate processing units from the initial set of processing units whose similarity meets a preset similarity condition to obtain a set of candidate processing units.

[0294] In an exemplary embodiment, the data processing apparatus 1900 further includes a set construction module for obtaining metadata of processing units; the metadata of processing units includes unit information of processing units; the metadata of processing units is associated and stored according to the unit identifier of the processing units to obtain an initial set of processing units; the initial set of processing units includes the unit identifier and metadata of the processing units.

[0295] In an exemplary embodiment, the unit determination module 1902 is further configured to obtain the metadata of candidate processing units in the candidate processing unit set; input the query content, missing capability information and the metadata of candidate processing units into the target unit recommendation model; perform unit recommendation processing based on the query content, missing capability information and the metadata of candidate processing units through the target unit recommendation model to obtain the recommendation degree of the candidate processing units; determine the candidate processing unit with the highest recommendation degree from the candidate processing unit set to obtain the recommended processing unit corresponding to the missing capability information, and generate the recommendation reason corresponding to the recommended processing unit.

[0296] In an exemplary embodiment, the unit determination module 1902 is further configured to obtain a unit recommendation prompt template of the target unit recommendation model; add the query content, missing capability information and metadata of the candidate processing unit to the unit recommendation prompt template to obtain a unit recommendation prompt; input the unit recommendation prompt into the target unit recommendation model; and perform unit recommendation processing based on the unit recommendation prompt through the target unit recommendation model to obtain the recommendation degree of the candidate processing unit.

[0297] In an exemplary embodiment, the information update module 1903 is further configured to obtain activation information for the recommendation processing unit; the activation information is used to characterize the activation operation of the recommendation processing unit based on the unit identifier and recommendation reason corresponding to the recommendation processing unit; based on the activation information, the recommendation processing unit is confirmed as a newly added activated processing unit.

[0298] In an exemplary embodiment, the information update module 1903 is further configured to confirm the update information corresponding to the initial planning information based on the initial planning information, the query content, and the unit information of the newly added activated processing unit; the update information includes at least the newly added sub-task corresponding to the newly added activated processing unit, and the task adjustment information for the sub-task in the initial planning information; based on the update information, the initial planning information is updated to obtain the updated planning information corresponding to the query content; based on the updated planning information, the target planning information corresponding to the query content is obtained.

[0299] In an exemplary embodiment, the information update module 1903 is further configured to adjust the subtasks in the initial planning information according to the task adjustment information to obtain the adjusted planning information; delete the missing capability information in the adjusted planning information and add the new subtasks to the adjusted planning information to obtain the updated planning information corresponding to the query content.

[0300] In an exemplary embodiment, the data processing apparatus 1900 further includes a task execution module, configured to update the already enabled processing units using newly added enabled processing units to obtain updated enabled processing units; if the processing capacity of the updated enabled processing units meets the query content, the module calls the enabled processing units corresponding to the subtasks in the target planning information to execute the subtasks and obtain the query results for the query content; if the processing capacity of the updated enabled processing units does not meet the query content, the module uses the target planning information as the updated initial planning information and jumps to the step of determining the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information, until the processing capacity of the updated enabled processing units meets the query content.

[0301] In one exemplary embodiment, such as Figure 20 As shown, another data processing device 2000 is provided, including: a first display module 2001, a second display module 2002, and a third display module 2003, wherein:

[0302] The first display module 2001 is used to display the query content on the query page.

[0303] The second display module 2002 is used to display the initial planning information corresponding to the query content and carrying missing capability information on the query page; the initial planning information is obtained by performing task planning on the query content based on the unit information of the enabled processing units; the missing capability information is generated when the processing capability of the enabled processing units does not meet the query content, and is used to characterize the missing capabilities for executing the query content.

[0304] The third display module 2003 is used to display the target planning information corresponding to the query content on the query page; the target planning information is obtained by updating the initial planning information based on the query content and the unit information of the newly added activated processing unit; the newly added activated processing unit is determined based on the recommended processing unit; the recommended processing unit is used to represent the recommended processing unit corresponding to the missing capability information determined from the candidate processing unit set based on the missing capability information in the initial planning information.

[0305] In an exemplary embodiment, the data processing apparatus 2000 further includes a fourth display module for displaying target metadata of the processing unit in the processing unit operation page; and for displaying the target metadata of the enabled processing unit corresponding to the enable event in the target area of ​​the processing unit operation page in response to an enable event for the processing unit.

[0306] In an exemplary embodiment, the data processing apparatus 2000 further includes a first processing module, configured to perform processing on the processing unit corresponding to a first target event in response to a first target event for the processing unit; the first target event includes at least one of a details viewing event and a unit trial event.

[0307] In an exemplary embodiment, the data processing apparatus 2000 further includes a second processing module, configured to perform processing corresponding to the second target event on the enabled processing unit in response to the second target event; the second target event includes at least one of a deactivation event, a details viewing event, and a data configuration event.

[0308] In an exemplary embodiment, the data processing device 2000 further includes a fifth display module for displaying a processing unit recommendation pop-up window on the query page; the processing unit recommendation pop-up window displays the unit identifier, capability information and recommendation reasons of the recommended processing unit.

[0309] In an exemplary embodiment, the third display module 2003 is further configured to, in response to an activation event for the recommendation processing unit, switch from displaying the recommendation pop-up of the display processing unit on the query page to displaying the target planning information corresponding to the query content.

[0310] In one exemplary embodiment, the query content includes activity recommendations.

[0311] The third display module 2003 is also used to display the target planning information corresponding to the activity recommendation content on the query page; the target planning information is obtained by updating the initial planning information based on the activity recommendation content and the agent information of the newly enabled agent; the newly enabled agent is determined based on the recommended agent; the recommended agent is used to represent the recommended agent corresponding to the missing capability information determined from the candidate agent set based on the missing capability information in the initial planning information.

[0312] The data processing device 2000 also includes a sixth display module, which displays the activity recommendation results corresponding to the activity recommendation content on the query page in response to the execution operation on the target planning information.

[0313] Each module in the aforementioned data processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0314] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 21 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 unit information of enabled processing units. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a data processing method.

[0315] In one exemplary embodiment, another computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 22As 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 computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides 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, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a data processing method. The display unit is used to form a visually visible image and 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.

[0316] Those skilled in the art will understand that Figure 21 or Figure 22 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.

[0317] In one exemplary 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-described method embodiments.

[0318] In one exemplary 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-described method embodiments.

[0319] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0320] 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 the relevant data must comply with relevant regulations.

[0321] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory 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, data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0322] 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 application.

[0323] 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 data processing method, characterized in that, The method includes: The query content is obtained, and task planning is performed on the query content based on the unit information of the enabled processing unit to obtain the initial planning information corresponding to the query content carrying missing capability information. The missing capability information is generated when the processing capability of the enabled processing unit does not meet the query content, and is used to characterize the missing capability for executing the query content. Based on the missing capability information in the initial planning information, the recommended processing unit corresponding to the missing capability information is determined from the candidate processing unit set; Based on the recommendation processing unit, a new activation processing unit is determined. According to the query content and the unit information of the new activation processing unit, the initial planning information is updated to obtain the target planning information corresponding to the query content.

2. The method according to claim 1, characterized in that, The step of performing task planning on the query content based on the unit information of the enabled processing unit to obtain initial planning information carrying missing capability information corresponding to the query content includes: The query content is decomposed into sub-task objectives to obtain the query content; Based on the unit information of the enabled processing units, the target enabled processing units whose unit information satisfies the subtask objective are determined from the enabled processing units. If the processing capacity of the target enabled processing unit in the enabled processing unit does not meet the query content, the remaining sub-task targets are obtained from each sub-task target of the query content, excluding the sub-task targets of the target enabled processing unit. Generate subtasks corresponding to the subtasks of the target whose processing units have been enabled, and generate the missing capability information corresponding to the remaining subtasks; Based on the subtask and the missing capability information, the initial planning information corresponding to the query content is generated.

3. The method according to claim 2, characterized in that, The subtasks corresponding to the subtask target for which the target processing unit has been enabled include: Based on the sub-task target of the target's enabled processing unit, determine the target input parameter information of the target's enabled processing unit; Based on the sub-task target and target input parameter information of the target enabled processing unit, determine the target output information of the target enabled processing unit; Based on the target input parameter information, target output information, and unit identifier of the target enabled processing unit, a subtask corresponding to the target enabled processing unit is generated.

4. The method according to claim 2, characterized in that, The generation of the missing capability information corresponding to the remaining sub-task objectives includes: Obtain the capability information required to execute the remaining sub-task objectives; Based on the capability information required to execute the remaining sub-task objectives, the missing capability information corresponding to the remaining sub-task objectives is obtained.

5. The method according to claim 2, characterized in that, The step of generating the initial planning information corresponding to the query content based on the subtask and the missing capability information includes: Based on the query content, determine the order of the subtasks; The subtasks are arranged according to the given order to obtain the arranged subtasks; The sorted subtasks and the missing capability information are summarized to obtain the initial planning information corresponding to the query content.

6. The method according to claim 2, characterized in that, After determining, based on the unit information of the enabled processing units, the target enabled processing unit whose unit information satisfies the sub-task objective from the enabled processing units, further includes: Obtain the target capabilities required to execute the query; If the processing capability of the target's enabled processing unit is only a portion of the target's capabilities, it is confirmed that the processing capability of the target's enabled processing unit does not satisfy the query content.

7. The method according to claim 2, characterized in that, After determining, based on the unit information of the enabled processing units, the target enabled processing unit whose unit information satisfies the sub-task objective from the enabled processing units, further includes: If the subtask target of the target's enabled processing unit is only a portion of the subtask targets of the query content, it is confirmed that the processing capacity of the target's enabled processing unit does not meet the query content.

8. The method according to claim 1, characterized in that, The step of performing task planning on the query content based on the unit information of the activated processing unit to obtain the initial planning information corresponding to the query content includes: Determine the target planning information generation model corresponding to the query content, and input the query content and the unit information of the enabled processing unit into the target planning information generation model; The target planning information generation model is used to decompose the query content into sub-task targets. Based on the unit information of the enabled processing units, the enabled processing units whose unit information satisfies the sub-task targets are determined from the enabled processing units. If the processing capacity of the target enabled processing unit in the enabled processing unit does not meet the query content, the remaining sub-task targets are obtained from each sub-task target of the query content, excluding the sub-task targets of the target enabled processing unit. Generate subtasks corresponding to the subtasks of the target whose processing units have been enabled, and generate the missing capability information corresponding to the remaining subtasks; The subtasks and the missing capability information are summarized to obtain the initial planning information corresponding to the query content.

9. The method according to claim 8, characterized in that, The step of determining the target planning information generation model corresponding to the query content includes: Obtain the content complexity of the query content; Based on the content complexity of the query content and the correspondence between the query content complexity and the planning information generation model, the planning information generation model corresponding to the content complexity of the query content is obtained. The target planning information generation model is obtained based on the planning information generation model corresponding to the content complexity of the query content.

10. The method according to claim 8, characterized in that, The step of inputting the query content and the unit information of the enabled processing unit into the target planning information generation model includes: Obtain the task planning prompt template corresponding to the query content; Based on the task planning prompt template, the query content, and the unit information of the enabled processing units, a task planning prompt corresponding to the query content is constructed; Input the task planning prompts into the target planning information generation model; The step of decomposing the query content into sub-task objectives using the target planning information generation model to obtain the query content includes: Using the target planning information generation model, and based on the task planning prompts, the query content is decomposed into sub-task targets to obtain the query content.

11. The method according to claim 1, characterized in that, Before determining the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information, the method further includes: Obtain the unit information of the processing units in the initial processing unit set; the processing unit is an intelligent agent; Determine the similarity between the missing capability information and the unit information of the processing unit; From the initial set of processing units, candidate processing units whose similarity satisfies the preset similarity condition are determined, thus obtaining the candidate processing unit set.

12. The method according to claim 11, characterized in that, Before retrieving the query results, it also includes: Obtain the metadata of the processing unit; the metadata of the processing unit includes the unit information of the processing unit; The metadata of the processing unit is associated and stored according to the unit identifier of the processing unit to obtain the initial processing unit set; the initial processing unit set includes the unit identifier and metadata of the processing unit.

13. The method according to claim 1, characterized in that, The step of determining the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information includes: Obtain the metadata of the candidate processing units in the candidate processing unit set; The query content, the missing capability information, and the metadata of the candidate processing unit are input into the target unit recommendation model; The target unit recommendation model performs unit recommendation processing based on the query content, the missing capability information, and the metadata of the candidate processing units to obtain the recommendation degree of the candidate processing units. From the set of candidate processing units, the candidate processing unit with the highest recommendation degree is determined, the recommended processing unit corresponding to the missing capability information is obtained, and the recommendation reason corresponding to the recommended processing unit is generated.

14. The method according to claim 13, characterized in that, The step of inputting the query content, the missing capability information, and the metadata of the candidate processing unit into the target unit recommendation model includes: Obtain the unit recommendation prompt template for the target unit recommendation model; The query content, the missing capability information, and the metadata of the candidate processing unit are added to the unit recommendation prompt template to obtain the unit recommendation prompt; Input the unit recommendation prompt into the target unit recommendation model; The step of performing unit recommendation processing based on the query content, the missing capability information, and the metadata of the candidate processing units through the target unit recommendation model to obtain the recommendation degree of the candidate processing units includes: Using the target unit recommendation model, and based on the unit recommendation prompts, unit recommendation processing is performed to obtain the recommendation degree of the candidate processing units.

15. The method according to claim 13, characterized in that, The step of determining the newly enabled processing unit based on the recommendation processing unit includes: Obtain the activation information for the recommendation processing unit; the activation information is used to characterize the activation operation of the recommendation processing unit based on the unit identifier and recommendation reason corresponding to the recommendation processing unit; Based on the activation information, the recommended processing unit is confirmed as a newly activated processing unit.

16. The method according to claim 1, characterized in that, The step of updating the initial planning information based on the query content and the unit information of the newly activated processing unit to obtain the target planning information corresponding to the query content includes: Based on the initial planning information, the query content, and the unit information of the newly activated processing unit, the update information corresponding to the initial planning information is confirmed; the update information includes at least the newly added sub-task corresponding to the newly activated processing unit, and task adjustment information for the sub-task in the initial planning information; Based on the updated information, the initial planning information is updated to obtain the updated planning information corresponding to the query content; Based on the updated planning information, the target planning information corresponding to the query content is obtained.

17. The method according to claim 16, characterized in that, The step of updating the initial planning information based on the updated information to obtain the updated planning information corresponding to the query content includes: Based on the task adjustment information, the subtasks in the initial planning information are adjusted to obtain the adjusted planning information; The missing capability information in the adjusted planning information is deleted, and the newly added subtask is added to the adjusted planning information to obtain the updated planning information corresponding to the query content.

18. The method according to claim 1, characterized in that, After updating the initial planning information based on the query content and the unit information of the newly enabled processing unit to obtain the target planning information corresponding to the query content, the method further includes: The newly added enabled processing unit is used to update the already enabled processing unit to obtain the updated enabled processing unit. If the processing capacity of the updated enabled processing unit meets the query content, the enabled processing unit corresponding to the subtask in the target planning information is invoked to execute the subtask and obtain the query result of the query content. If the processing capacity of the updated enabled processing unit does not meet the query content, the target planning information is used as the updated initial planning information, and the process jumps to the step of determining the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information, until the processing capacity of the updated enabled processing unit meets the query content.

19. A data processing method, characterized in that, The method includes: The search results are displayed on the search page; The query page displays initial planning information corresponding to the query content, carrying missing capability information; the initial planning information is obtained by performing task planning on the query content based on the unit information of the enabled processing units; the missing capability information is generated when the processing capability of the enabled processing units does not meet the query content, and is used to characterize the missing capability for executing the query content. The query page displays the target planning information corresponding to the query content; the target planning information is obtained by updating the initial planning information based on the query content and the unit information of the newly added activated processing unit; the newly added activated processing unit is determined based on the recommended processing unit; the recommended processing unit is used to characterize the recommended processing unit corresponding to the missing capability information determined from the candidate processing unit set based on the missing capability information in the initial planning information.

20. The method according to claim 19, characterized in that, Before the search results are displayed on the search page, it also includes: The target metadata of the processing unit is displayed on the processing unit operation page; In response to an activation event for the processing unit, target metadata of the enabled processing unit corresponding to the activation event is displayed in the target area of ​​the processing unit operation page.

21. The method according to claim 20, characterized in that, After displaying the target metadata of the processing unit on the processing unit operation page, it also includes: In response to a first target event for the processing unit, the processing unit performs processing corresponding to the first target event; the first target event includes at least one of a details viewing event and a unit trial event.

22. The method according to claim 20, characterized in that, After displaying the target metadata of the enabled processing unit corresponding to the activation event, the method further includes: In response to a second target event for the enabled processing unit, processing corresponding to the second target event is performed on the enabled processing unit; the second target event includes at least one of a deactivation event, a details viewing event, and a data configuration event.

23. The method according to claim 19, characterized in that, After displaying the initial planning information corresponding to the query content with missing capability information on the query page, the page also includes: A pop-up window recommending processing units is displayed on the query page; the pop-up window displays the unit identifier, capability information, and recommendation reasons of the recommended processing unit. The step of displaying the target planning information corresponding to the query content on the query page includes: In response to the activation event of the recommendation processing unit, the display of the recommendation pop-up window of the processing unit on the query page is switched to displaying the target planning information corresponding to the query content.

24. The method according to claim 19, characterized in that, The query content includes activity recommendations; The step of displaying the target planning information corresponding to the query content on the query page includes: The query page displays the target planning information corresponding to the recommended content of the activity; the target planning information is obtained by updating the initial planning information based on the recommended content of the activity and the information of the newly enabled intelligent agent; the newly enabled intelligent agent is determined based on the recommended intelligent agent; the recommended intelligent agent is used to represent the recommended intelligent agent corresponding to the missing capability information determined from the candidate intelligent agent set based on the missing capability information in the initial planning information; The method further includes: In response to the execution operation on the target planning information, the activity recommendation results corresponding to the activity recommendation content are displayed on the query page.

25. A data processing apparatus, characterized in that, The device includes: The task planning module is used to obtain query content, perform task planning on the query content based on the unit information of the enabled processing units, and obtain initial planning information corresponding to the query content carrying missing capability information; the missing capability information is generated when the processing capability of the enabled processing units does not meet the query content, and is used to characterize the missing capability for executing the query content. The unit determination module is used to determine the recommended processing unit corresponding to the missing capability information from the candidate processing unit set based on the missing capability information in the initial planning information; The information update module is used to determine the newly enabled processing unit based on the recommendation processing unit, and update the initial planning information according to the query content and the unit information of the newly enabled processing unit to obtain the target planning information corresponding to the query content.

26. A data processing apparatus, characterized in that, The device includes: The first display module is used to display the query content on the query page; The second display module is used to display initial planning information carrying missing capability information corresponding to the query content on the query page; the initial planning information is obtained by performing task planning on the query content based on the unit information of the enabled processing unit; the missing capability information is generated when the processing capability of the enabled processing unit does not meet the query content, and is used to characterize the missing capability for executing the query content. The third display module is used to display the target planning information corresponding to the query content on the query page; the target planning information is obtained by updating the initial planning information based on the query content and the unit information of the newly added activated processing unit; the newly added activated processing unit is determined based on the recommended processing unit; the recommended processing unit is used to characterize the recommended processing unit corresponding to the missing capability information determined from the candidate processing unit set based on the missing capability information in the initial planning information.

27. 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 24.

28. 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 24.

29. 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 24.

Images