Task request processing method, computing device, electronic device and storage medium

Abstract

The invention discloses a task request processing method, computing equipment, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence. The method comprises the following steps: in response to a received task request sent by a target client, constructing a cue word based on the task request; the cue word is input into an intelligent decision model, decision reasoning is conducted on the cue word through the intelligent decision model, an initial processing result and a calling decision result of the task request are obtained, and the calling decision result is used for representing whether the task request needs to be processed through a calling tool or not; and processing the task request based on the initial processing result and the calling decision result to obtain a target processing result. The technical problem of relatively low task processing efficiency in related technologies is solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology. Specifically, it relates to a task request processing method, a computing device, an electronic device, and a storage medium. Background Art

[0002] With the development of the e-commerce industry, large online shopping platforms are facing increasingly complex after-sales problems. In particular, the volume of offline dispute processing has increased sharply, with the number of disputes to be processed exceeding one hundred thousand per day, covering various task scenarios such as returns, exchanges, and repairs. Manual customer service for handling disputes is time-consuming and costly, affecting customer satisfaction and the overall shopping experience, posing a significant pressure on platform operations. That is, the efficiency of task processing in related technologies is relatively low.

[0003] For the above problems, no effective solutions have been proposed yet. Summary of the Invention

[0004] Embodiments of this application provide a task request processing method, a computing device, an electronic device, and a storage medium to at least solve the technical problem of relatively low task processing efficiency in related technologies.

[0005] According to one aspect of the embodiments of this application, a task request processing method is provided, including: in response to receiving a task request sent by a target client, constructing a prompt word based on the task request; inputting the prompt word into an intelligent decision-making model, and using the intelligent decision-making model to perform decision-making reasoning on the prompt word to obtain an initial processing result of the task request and a call decision result, where the call decision result is used to indicate whether it is necessary to process the task request by calling a tool; and processing the task request based on the initial processing result and the call decision result to obtain a target processing result.

[0006] According to another aspect of the embodiments of this application, a task request processing method is provided, including: in response to an input instruction acting on an operation interface, displaying a task request sent by a target client on the operation interface and constructing a prompt word based on the task request; in response to a processing instruction acting on the operation interface, displaying a target processing result of the task request on the operation interface, where the target processing result is obtained by processing the task request based on the initial processing result and the call decision result, and the initial processing result and the call decision result are obtained by inputting the prompt word into an intelligent decision-making model and using the intelligent decision-making model to perform decision-making reasoning on the prompt word, and the call decision result is used to indicate whether it is necessary to process the task request by calling a tool.

[0007] According to another aspect of the embodiments of the present application, a task request processing method is provided, including: obtaining a task request by calling a first interface, and constructing a prompt word based on the task request, where the first interface includes a first parameter, and the parameter value of the first parameter includes the task request; inputting the prompt word into an intelligent decision-making model, and using the intelligent decision-making model to perform decision-making reasoning on the prompt word to obtain an initial processing result of the task request and a call decision result, where the call decision result is used to indicate whether it is necessary to process the task request by calling a tool; processing the task request based on the initial processing result and the call decision result to obtain a target processing result; and outputting the target processing result by calling a second interface, where the second interface includes a second parameter, and the parameter value of the second parameter includes the target processing result.

[0008] According to another aspect of the embodiments of the present application, a computing device is further provided, including: a memory storing an executable program; and a processor for running the program, where when the program runs, it executes the methods in the various embodiments of the present application.

[0009] According to another aspect of the embodiments of the present application, an electronic device is further provided, including: a memory storing an executable program; and a processor connected to the memory through a bus for running the program, where when the program runs, it executes the methods in the various embodiments of the present application.

[0010] According to another aspect of the embodiments of the present application, a computer-readable storage medium is further provided, where the computer-readable storage medium includes a stored executable program, and when the executable program runs, it controls the device where the computer-readable storage medium is located to execute the methods in the various embodiments of the present application.

[0011] According to another aspect of the embodiments of the present application, a computer program product is further provided, including a computer program, where when the computer program is executed by a processor, it implements the methods in the various embodiments of the present application.

[0012] According to another aspect of the embodiments of the present application, a computer program product is further provided, including a non-volatile computer-readable storage medium storing a computer program, where when the computer program is executed by a processor, it implements the methods in the various embodiments of the present application.

[0013] According to another aspect of the embodiments of the present application, a computer program is further provided, where when the computer program is executed by a processor, it implements the methods in the various embodiments of the present application.

[0014] In an embodiment of the present application, in response to receiving a task request sent by a target client, a prompt word is constructed based on the task request; the prompt word is input into an intelligent decision-making model, and the intelligent decision-making model is used to perform decision-making reasoning on the prompt word to obtain an initial processing result of the task request and a call decision result, where the call decision result is used to indicate whether it is necessary to process the task request by calling a tool; the task request is processed based on the initial processing result and the call decision result to obtain a target processing result, achieving an improvement in task processing efficiency; it is easy to notice that through decision-making reasoning by the intelligent decision-making model and directly obtaining the initial processing result of the task and the judgment of whether a tool needs to be called, this process significantly reduces the time for manual judgment and decision-making. The intelligent decision-making model can quickly understand the context and requirements of the task request, and based on its powerful language understanding and reasoning capabilities, directly give a processing strategy and tool call suggestion, greatly shortening the preliminary preparation time of task processing. It can combine the initial processing result and the call decision result to automatically call the tool, further accelerating the execution stage of task processing, reducing manual intervention, and improving the degree of automation of the process, thereby effectively improving task processing efficiency, and further solving the technical problem of low task processing efficiency in the related art.

[0015] It is easy to notice that the above general description and the following detailed description are only for exemplifying and explaining the present application, and do not constitute a limitation to the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation to the present application. In the drawings:

[0017] Figure 1 is a schematic diagram of an application scenario for task request processing according to an embodiment of the present application;

[0018] Figure 2 is a flowchart of a task request processing method according to an embodiment of the present application;

[0019] Figure 3 is a schematic diagram of a model processing flow according to an embodiment of the present application;

[0020] Figure 4 is a decision link architecture diagram according to an embodiment of the present application;

[0021] Figure 5 is a flowchart of a task request processing method according to an embodiment of the present application;

[0022] Figure 6 is a flowchart of a task request processing method according to an embodiment of the present application;

[0023] Figure 7 It is a schematic diagram of a task request processing device according to an embodiment of the present application;

[0024] Figure 8 It is a schematic diagram of a task request processing device according to an embodiment of the present application;

[0025] Figure 9 It is a schematic diagram of a task request processing device according to an embodiment of the present application;

[0026] Figure 10 It is a structural block diagram of a computing device according to an embodiment of the present application;

[0027] Figure 11 It is a structural block diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0028] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application described here can be implemented in an order different from those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0030] The technical solution provided by this application is mainly implemented using large language model technology. Here, the large language model refers to a deep learning model with a large number of model parameters, usually including hundreds of millions, tens of billions, hundreds of billions, trillions or even more than one quadrillion model parameters. The large language model can also be called the Foundation Model. Through pre-training of the large language model with a large amount of unlabeled corpus, a pre-trained model with more than one hundred million parameters is produced. This model can adapt to a wide range of downstream tasks and has good generalization ability, such as the Large Language Model (LLM), multi-modal pre-training model, etc.

[0031] It should be noted that in actual applications, the large language model can be fine-tuned with a small number of samples for the pre-trained model, so that the large language model can be applied to different tasks. For example, the large language model can be widely applied to fields such as Natural Language Processing (NLP), computer vision, and speech processing. Specifically, it can be applied to tasks in the field of computer vision such as Visual Question Answering (VQA), Image Caption (IC), and image generation. It can also be widely applied to tasks in the field of natural language processing such as text-based sentiment classification, text summary generation, and machine translation. Therefore, the main application scenarios of the large language model include but are not limited to digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc. In the embodiments of this application, the data processing by the intelligent decision-making model in the task processing scenario is used as an example for explanation.

[0032] First, some nouns or terms that appear in the process of describing the embodiments of this application are applicable to the following explanations:

[0033] Large Language Model (LLM): A super-large deep learning model pre-trained based on a large amount of data.

[0034] After-Sales Service (ASS): In e-commerce, various services and supports provided by merchants or platforms after consumers purchase and receive goods.

[0035] Dispute: After purchasing goods or services, the disputes or conflicts arising between consumers and merchants or service providers due to after-sales service-related issues.

[0036] Ruling: When a dispute occurs between a consumer and a merchant, the platform is applied to intervene. The platform determines and divides the responsibilities based on the consumer's requests and the information provided by the merchant.

[0037] Agent Manager (abbreviated as AM): Responsible for managing the access policies of Agents, controlling that task work orders are not directly transferred to manual processing, and when an Agent is unable to handle or manual intervention is initiated, transferring the work order to the customer service staff for processing.

[0038] Agent Executor (abbreviated as AE): Responsible for scheduling requests to the LLM, including constructing Prompts, recalling relevant information through Retrieval-Augmented Generation (abbreviated as RAG), invoking tools, processing Memory, and managing the multi-round decision-making process.

[0039] With the rapid development of e-commerce, there are increasingly complex after-sales problems, especially a significant increase in the number of offline disputes generated in after-sales services. Currently, the magnitude of offline disputes in after-sales services is extremely large, with more than 100,000 offline disputes needing to be processed every day. These disputes not only involve various situations such as order returns, exchanges, and repairs, but also consume a large amount of time and human resources during the manual processing.

[0040] Each offline dispute requires dedicated customer service staff to conduct in-depth investigations and handling. These staff members need to carefully read the customer's complaint content, communicate with the merchant, and even intervene in mediation in some cases. This process is often cumbersome, not only time-consuming, but also likely to lead to a decrease in customer satisfaction, thereby affecting the entire shopping experience.

[0041] In response to the current situation, to address the problem of consumers' offline dispute demands. Existing technologies adopt long-process rulings, which rely more on operationally configured fixed processes, rule strategies to solve problems, and at the same time are too dependent on communication and coordination with consumers and merchants. Currently, it can only solve a limited number of problems in fixed scenarios, with a relatively low upper limit, and the operation effort and cost are relatively high.

[0042] This application introduces intelligent technologies (such as artificial intelligence chatbots, automated complaint handling systems, etc.) to improve the processing efficiency of offline customer service. In addition, this application considers establishing a perfect customer self-service system to enable consumers to solve some simple problems by themselves, thereby effectively reducing the burden on customer service staff.

[0043] According to an embodiment of the present application, a method for processing task requests is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0044] Considering that the number of model parameters of the large language model is huge and the computing resources of the mobile terminal are limited, the above method provided by the embodiment of the present application can be applied to Figure 1 the application scenarios shown, but not limited thereto. Figure 1 FIG. is a schematic diagram of an application scenario for processing task requests according to an embodiment of the present application. In Figure 1 the application scenario shown, the large language model is deployed in the server 10. The server 10 can be connected to one or more client devices 20 through a local area network connection, a wide area network connection, an Internet connection, or other types of data networks. Here, the client devices 20 can include but are not limited to: smart phones, tablet computers, laptop computers, palmtop computers, personal computers, smart home devices, in-vehicle devices, etc. The client device 20 can interact with the user through a graphical user interface to implement the invocation of the large language model, and further implement the method provided by the embodiment of the present application.

[0045] In the embodiment of the present application, the system composed of the client device and the server can execute the following steps: The client device executes the step of generating a task request. The server executes the steps of receiving the task request, constructing a prompt word based on the task request; inputting the prompt word into the intelligent decision-making model, and using the intelligent decision-making model to perform decision-making reasoning on the prompt word to obtain an initial processing result and an invocation decision result of the task request; processing the task request based on the initial processing result and the invocation decision result to obtain a target processing result.

[0046] It should be noted that with the rapid development of high-performance computing units, in other application scenarios, the above method provided by the embodiment of the present application can also be applied to a model all-in-one machine. In an optional embodiment, multiple models are built into the model all-in-one machine, and the user can select and adjust one model according to needs to obtain the user's own model. Thus, the high-performance computing unit built into the model all-in-one machine can directly call the adjusted model to execute the above method provided by the embodiment of the present application. In another optional embodiment, a trained model is built into the large language model all-in-one machine. Thus, the high-performance computing unit built into the model all-in-one machine can directly call the model to execute the above method provided by the embodiment of the present application.

[0047] Furthermore, when users need to train their own models, they can also upload their own data sets through the client, which are sent by the client to the server, so that the server can adjust the pre-trained model with the data set to obtain the user's own model and then deploy it to the production environment. In order to facilitate users' needs for model adjustment, the server can provide complete adjustment tools, development frameworks and processes, and can support multiple adjustment strategies, so that the adjusted model can better adapt to applications in different fields and achieve a high degree of customization.

[0048] Under the above operating environment, this application provides Figure 2 The task request processing method shown. Figure 2 is a flowchart of a task request processing method according to an embodiment of the present application. Figure 2 As shown, the method may include the following steps:

[0049] Step S202, in response to receiving a task request sent by the target client, constructing a prompt word based on the task request;

[0050] The target client mentioned above is the client that initiates the task request, which can be a consumer, a merchant, or other service requester. In the after-sales service scenario, the target client can be the client of a consumer who encounters a problem and seeks platform intervention to solve it, but is not limited to this. As the input end of the intelligent decision-making system, the target client provides the original request that triggers the automated processing flow. Its role is to start the entire technical solution and is the starting point for subsequent steps.

[0051] The above-mentioned task request is a request received from the target client for a specific service or problem solving, such as dispute resolution, order modification, etc. The task request is sent by the target client to the system to solve a specific problem or perform a specific service. In the after-sales scenario, this can be a consumer's request for return, exchange, refund, repair or resolution of any order-related dispute, but not limited to this. The task request is the core object processed by the intelligent decision-making system. It provides information about the specific problem that the system needs to solve. It contains key data such as the user's basic demands, order information, and dispute points. It is the basis for constructing prompt words.

[0052] The above-mentioned prompt words are structured text information that contains the key elements and expected goals of the task request, and are used to guide the intelligent decision model to make decision reasoning. Prompt words allow the intelligent decision model to understand the context and details of the task request, so as to give more accurate decision suggestions. The step of constructing prompt words is to convert the task request into a format that the model can understand and process, ensuring that the model can accurately grasp the user's intention and problem situation, so as to make decisions that are more in line with actual needs. The quality of prompt words directly affects the decision-making effect of the model.

[0053] In an alternative embodiment, after receiving a task request, the system first analyzes the key information in the request, including but not limited to the request type, involved transaction details, user requirements, historical interaction data, etc., structures this information and integrates it into a prompt. By converting the unstructured user request into a format that the intelligent decision-making model can understand, it facilitates model reasoning. Through precise extraction and structuring of the request information, the prompt construction step is expected to reduce the model's processing of irrelevant information, speed up the decision-making process, and improve the pertinence and efficiency of task processing.

[0054] Step S204: Input the prompt into the intelligent decision-making model, and use the intelligent decision-making model to perform decision-making reasoning on the prompt to obtain the initial processing result of the task request and the call decision result.

[0055] Among them, the call decision result is used to indicate whether it is necessary to process the task request by calling a tool.

[0056] The above intelligent decision-making model can be a model based on deep learning, natural language processing, and artificial intelligence technologies, and is trained to understand the prompt and output corresponding decisions. The above intelligent decision-making model can receive the prompt as input and, through reasoning, analysis, and decision-making mechanisms, output processing suggestions or solutions for specific tasks. Exemplarily, the intelligent decision-making model can be used to understand and solve complex problems encountered in after-sales services, such as handling disputes over returns, exchanges, repairs, etc.

[0057] The intelligent decision-making model is the core component of the automated processing flow. After receiving the constructed prompt, through the internal reasoning algorithm of the model, it analyzes the context of the task request, user requirements, order information, etc. to generate preliminary processing suggestions or solutions. The decision-making and reasoning ability of the model determines the accuracy and efficiency of the processing result.

[0058] The above initial processing result is the preliminary suggestion or solution given by the intelligent decision-making model based on the analysis of the prompt on how to solve the task request. It can include direct solution strategies, mediation suggestions, information supplementation requirements, etc. The initial processing result is based on the result of model reasoning, provides a guiding direction for subsequent task processing, includes a preliminary solution based on the task request, and is the basis for subsequent specific processing steps.

[0059] The above call decision result is the decision made by the intelligent decision-making model based on the complexity and specific requirements of the task request on whether it is necessary to call external tools to assist in task processing. External tools can be automated tools or services specifically used for communication and coordination, data query, etc. The call decision result is used to indicate the tool usage requirements during task processing, ensuring that appropriate tools can be called to collect additional information, perform specific operations, or verify data accuracy when dealing with complex tasks, thereby improving the accuracy and efficiency of processing.

[0060] Input the constructed prompt into the intelligent decision-making model. Utilize the model's reasoning ability to obtain the preliminary processing strategy and tool invocation indication for the task request. The intelligent decision-making model analyzes the information in the prompt, combines historical data and preset rules, and outputs a decision plan that includes the initial processing result and the invocation decision result. This decision plan can determine the direction of subsequent task processing and the selection of tools. Through the rapid reasoning of the intelligent decision-making model, the decision-making cycle can be significantly shortened, manual intervention can be reduced, the automation and intelligence of task processing can be achieved, and the processing efficiency and user satisfaction can be improved.

[0061] The intelligent decision-making model receives the constructed prompt and starts its decision-making reasoning process to generate the initial processing result and the invocation decision result for the task request. The model will conduct in-depth analysis and judgment based on the nature and requirements of the task request to determine the better processing path. If additional information verification or operation execution is required for task processing, the model will output the invocation decision result, instructing the system to invoke the corresponding tool to assist in completing the task. The technical effect of this stage is that through the intelligent decision-making of the model, the task processing is automated, the need for manual intervention is reduced, the processing efficiency and accuracy are improved, and at the same time, the flexibility and adaptability of the processing flow are ensured, and personalized processing can be carried out according to the characteristics of different tasks.

[0062] The mechanism of decision-making reasoning provides a feasible solution for subsequent automated processing, and through the output of the invocation decision result, the integrity of the task processing flow is ensured. Even in the face of complex scenarios, appropriate tools can be invoked to handle them. This step is the key to realizing the automated processing flow, which can convert unstructured user requirements into structured and executable solutions and auxiliary tools.

[0063] Step S206, process the task request based on the initial processing result and the invocation decision result to obtain the target processing result.

[0064] The above-mentioned target processing result is the final solution or processing result obtained after completing the relevant processing steps, which meets the solution requirements of the task request and provides a clear and reasonable conclusion or action plan for the user. The target processing result is the final output of the entire task processing flow. Based on the actual effects of the initial processing result and the invocation decision result, it provides a comprehensive solution based on the analysis of the intelligent decision-making model and the assistance of external tools. This result is directly related to whether the user's problem is effectively solved, as well as the user satisfaction and the quality of the experience.

[0065] The above steps are the execution and termination phases of the task processing flow. Based on the initial processing results and the invoked decision results output by the intelligent decision-making model, specific operations or service invocations can be carried out to solve the task request. This phase covers the execution of the preliminary solution and possible tool invocations to assist in task resolution, and finally generates and feedbacks the target processing results to the user. In this phase, the system will take corresponding actions according to the instructions of the initial processing results, such as directly solving the user's problem, conducting mediation, or collecting more evidence. At the same time, if the intelligent decision-making model determines that the assistance of specific tools is required, the system will invoke the corresponding tools or services to ensure the comprehensiveness and accuracy of the processing process. The invocations of these tools may include, but are not limited to, data query, automated task execution, etc., which can help the system collect more information, perform specific operations, or verify the accuracy of data when processing tasks.

[0066] Through the automated execution and the invocation of auxiliary tools, the efficient and accurate processing of task requests is achieved, ensuring that the user's problems can be solved in a timely and proper manner, thereby improving the user experience and satisfaction. At the same time, it also reduces the manual service cost of the platform and enhances the operational efficiency of the platform. By transforming the results of model reasoning and tool invocation into actual operations, a satisfactory solution is provided for the user, and at the same time, the advantages of automated services in improving efficiency and reducing costs are demonstrated.

[0067] According to the decision of the intelligent decision-making model, the system will execute a series of automated tasks, including but not limited to directly applying the solution, invoking tools for evidence review, or communicating with relevant parties. If the intelligent decision-making model determines that no tools need to be invoked, the system will directly generate the final processing solution based on the initial processing results; if tools need to be invoked, the tool invocation will be executed first. After obtaining the required information or completing specific tasks, the target processing results will be generated in combination with this information. This step ensures the accuracy and feasibility of the processing results. Through automated tool invocation and task execution, possible errors or oversights in manual processing are avoided, and the user's trust in the processing results is improved. At the same time, the automated process can significantly reduce the service response time, improve the user experience, and reduce the operating cost.

[0068] Exemplarily, if the target client is a consumer's client and submits a return application to the platform due to product quality issues. After receiving this task request, the system can first analyze the problem description, product information, order status, etc. provided by the consumer to construct a prompt containing the return request, product situation, and order details. The prompt is input into the intelligent decision-making model. After analysis, the model obtains a preliminary processing result: the return application meets the conditions, and it is necessary to further confirm the degree of product damage, and gives a call decision result: call the evidence review tool to verify the product status. The system calls the review tool (such as image recognition technology) to review the product damage photos uploaded by the consumer. After passing the review, the return process is automatically initiated, including generating a return form, notifying the seller, estimating the refund time, etc., and the processing result is fed back to the consumer, and the entire process does not require manual intervention.

[0069] From the reception and parsing of the task request, to the decision-making reasoning based on the intelligent decision-making model, and then to the automated task execution and result generation, the system ensures the efficiency, accuracy, and user-friendliness of the processing process. The closed-loop improvement mechanism is reflected in the continuous learning and adjustment of the model. Different processing results will be recorded and analyzed to improve the model decision-making and prompt construction strategies, thereby continuously improving the overall processing effect. This application realizes the rapid response, automated processing, and decision-making accuracy of task requests, improves user satisfaction and the operation efficiency of the platform, and reduces service costs.

[0070] Through the above steps, in response to receiving a task request sent by the target client, a prompt is constructed based on the task request; the prompt is input into the intelligent decision-making model, and the intelligent decision-making model is used to perform decision-making reasoning on the prompt to obtain the initial processing result and call decision result of the task request, where the call decision result is used to indicate whether it is necessary to process the task request by calling a tool; the task request is processed based on the initial processing result and the call decision result to obtain the target processing result, achieving the improvement of task processing efficiency; it is easy to notice that through the decision-making reasoning of the intelligent decision-making model and directly obtaining the initial processing result of the task and the judgment of whether a tool needs to be called, this process significantly reduces the time of manual judgment and decision-making. The intelligent decision-making model can quickly understand the context and requirements of the task request. Based on its powerful language understanding and reasoning capabilities, it directly gives processing strategies and tool call suggestions, shortening the preliminary preparation time of task processing. It can combine the initial processing result and the call decision result to automatically call tools, further accelerating the execution stage of task processing, reducing manual intervention, and improving the degree of automation of the process, thereby effectively improving task processing efficiency, and further solving the technical problem of low task processing efficiency in the related art.

[0071] In the above embodiments of the present application, the task request is processed based on the initial processing result and the call decision result to obtain the target processing result, including: in response to the call decision result indicating that the task request needs to be processed by calling a tool, determining a preset task process and a tool call plan based on the initial processing result, and processing the task request based on the preset task process and the tool call plan to obtain the target processing result; in response to the call decision result indicating that the task request does not need to be processed by calling a tool, determining the initial processing result as the target processing result.

[0072] The above-mentioned preset task process is a series of steps and procedures preset according to the task nature and the tool call plan when the intelligent decision-making model determines that a tool needs to be called to process the task request. These processes include a series of automated operations or service executions before and after the tool call, aiming to solve complex tasks in a structured and automated manner. The preset task process provides a specific operation guide for processing task requests with complex or special requirements, ensures the smooth and efficient tool call, and helps improve the accuracy and speed of the processing result.

[0073] The above-mentioned tool call plan refers to the strategy and method for calling external tools or services generated by the intelligent decision-making model according to the specific requirements of the task request. This includes details such as tool selection, call method, call parameters, etc. The tool call plan provides guidance for the use of tools in the preset task process, ensures the pertinence and effectiveness of the tool call, helps collect key information, execute specific operations or verify data, thereby improving the accuracy and efficiency of task processing.

[0074] The above-mentioned target processing result is the solution or processing effect finally generated by the system after the task request has been preliminarily processed and possibly called a tool, which meets the solution requirements of the task request and provides a clear and reasonable conclusion or action plan for the user. The target processing result is the final output of the entire processing process, combining the initial processing result and the actual effect of the tool call, providing a comprehensive solution based on intelligent decision-making and automated services, which is directly related to whether the user's problem is effectively solved, as well as the user's satisfaction and experience quality.

[0075] The system adopts a flexible strategy to process the task request according to the decision of the intelligent decision-making model, aiming to solve the problem in an efficient manner and improve the user experience. If the task request is relatively simple and no additional tool intervention is required, the system will directly use the initial processing result of the intelligent decision-making model as the final target processing result, quickly respond to the user's needs, and reduce unnecessary processing links. However, for complex situations, the intelligent decision-making model will trigger a tool call, and the system will call the corresponding tool for in-depth processing according to the preset process and plan.

[0076] In the scenario of after-sales service disputes, the reasoning ability of large language models can be used to resolve disputes in after-sales service, reduce the burden on human customer service, and improve processing efficiency. When no additional tools are required for dispute resolution, the judgment of the intelligent decision-making model can be directly converted into the target processing result and quickly feedback to the user, achieving efficient processing of simple scenarios. When in-depth investigation or evidence review is needed for disputes, the system will automatically call relevant tools, such as evidence review tools, communication and coordination tools, etc., according to the preset task process and tool call plan, to ensure the proper resolution of complex disputes, improving both the accuracy of processing and the user experience.

[0077] Exemplarily, the target client (consumer) initiates a return application. The system constructs a prompt by analyzing the task request and then inputs the prompt into the intelligent decision-making model. The model determines that the return application meets the conditions for direct processing and no additional tools need to be called. At this time, the processing process of the system will directly adopt the initial processing result given by the model, such as approving the return and initiating the refund process, without the intervention of the preset task process and tool call plan. Finally, the consumer receives a notification that the return process has been automatically initiated by the system, and the refund is directly credited to the account without the intervention of human customer service, achieving a fast and automated service experience.

[0078] On the other hand, if the intelligent decision-making model discovers a controversial point when analyzing the return application, such as the description of the damaged condition of the product not matching the evidence, the model will output a call decision result to call the evidence review tool. In response to this result, the system automatically calls the evidence review tool according to the preset task process for photo comparison and damage degree analysis. Once the evidence review is completed, the system will combine the review result and the initial processing result to generate the final target processing result, such as requiring the consumer to provide more detailed proof of product damage, or directly rejecting the return application that does not meet the conditions, ensuring the accuracy and fairness of the processing.

[0079] Through the above steps, the automation level and processing efficiency of after-sales service can be improved, while ensuring the accuracy and fairness of complex dispute resolution. For simple and clear disputes, the system can respond quickly and directly provide solutions, significantly shortening the processing time and enhancing user satisfaction. For complex disputes, through the judgment of the intelligent decision-making model and the execution of the preset task process, the system can automatically call necessary tools for in-depth processing, improving the accuracy of dispute resolution and the user experience. Through this intelligent and automated processing method, the platform can not only effectively handle the large-scale dispute resolution requirements, but also significantly reduce the operation cost, achieving a balance between service quality and cost control.

[0080] Through the above steps and mechanisms, it is possible to intelligently determine whether to call external tools or services according to the complexity of the task request, ensuring the flexibility and pertinence of the processing flow. Whether it is a quick response to simple tasks or a deep solution to complex tasks, it can be user-centered, providing an efficient and accurate service experience, and achieving the goals of service intelligence and automation.

[0081] In the above embodiments of the present application, the task request is processed based on the preset task process and the tool call scheme to obtain the target processing result, including: calling at least one tool associated with the preset task process based on the tool call scheme; generating a preset to-be-processed task of the preset task process based on the at least one tool; executing the preset to-be-processed task, and obtaining the target processing result based on the execution result of the preset to-be-processed task and the initial processing result.

[0082] The above-mentioned preset to-be-processed task is a series of specific operations or services generated based on the tool call scheme and to be executed by the called tool. These tasks are to supplement information, verify data, or perform specific operations to assist the intelligent decision-making model in more accurately processing the task request.

[0083] The preset to-be-processed task is a specific manifestation of the execution of the tool call scheme, ensuring that the decision-making basis of the intelligent decision-making model is supplemented or verified, and contributing to the accuracy and reliability of the final target processing result.

[0084] After receiving the tool call scheme of the intelligent decision-making model, the system calls at least one tool related to the task process according to the preset task process. These tools generate preset to-be-processed tasks according to their respective functions, such as evidence review, data query, communication and coordination, etc. Then, the system executes these preset to-be-processed tasks, collects or generates the required information, performs specific operations, and finally obtains the execution result of the tool.

[0085] The execution result of the preset to-be-processed task will be fed back to the system, combined with the initial processing result of the intelligent decision-making model, and comprehensively analyzed and decision-making to generate the target processing result. This process ensures the integrity of the processing flow and the accuracy of the processing result. Even when facing complex and ambiguous task requests, it is possible to call external tools, collect necessary information or perform specific operations to assist the model in outputting a decision closer to the actual needs.

[0086] In an optional embodiment, based on the judgment of the intelligent decision-making model, external tools can be flexibly called to generate preset tasks to be processed, execute tasks and integrate execution results, and finally form target processing results to meet the needs of users or service requesters. According to the nature and complexity of the task request, the selection of tools and the execution of tasks can be intelligently determined to ensure the pertinence and efficiency of the processing process. For simple tasks, the process may directly apply the initial processing results; for complex tasks, the process calls tools to generate and execute preset tasks to be processed, collect key information to assist in the generation of final decisions, and realize the intelligence and automation of the processing process.

[0087] In the after-sales dispute scenario, when the intelligent decision-making model determines that a dispute needs additional information or operations, the system will call the corresponding evidence review tools, communication and coordination tools, etc., to generate preset tasks to be processed, such as reviewing the damaged product photos provided by consumers, communicating with merchants about returns, etc. After executing these tasks, the system will combine the collected information and execution results with the initial processing results of the model to generate the final target processing results, such as approving returns, rejecting returns and explaining the reasons, etc.

[0088] The processing flow in the embodiments of the present application significantly improves the efficiency and accuracy of after-sales service dispute handling, reduces reliance on manual customer service, and also improves user experience. Through the automated tool call and the execution of preset pending tasks, the system can quickly collect the information needed to handle complex disputes, reduce the ambiguity and subjectivity of decision-making, and improve the fairness and efficiency of dispute resolution. In addition, the implementation of this process can also significantly reduce the frequency of manual intervention and the cost of manual service, while maintaining high service standards and user satisfaction, achieving a dual improvement in technology application and user experience.

[0089] For example, suppose a consumer files a return application with the platform after receiving the goods, citing quality issues with the goods. After receiving this task request, the system analyzes the intelligent decision model and concludes that the quality of the goods needs to be verified. Based on this call decision result, the system calls the evidence review tool and generates a preset pending task to review the damaged goods photos uploaded by the consumer to confirm the actual status of the goods. The evidence review tool confirms through image recognition and analysis that the goods do have serious quality issues and need to be returned.

[0090] Subsequently, the system executed the preset task to be processed, collected the execution results of the evidence review tool, combined them with the preliminary processing results, and finally generated the target processing result: approving the consumer's return, notifying the merchant to proceed with the return process, and at the same time informing the consumer of the return process and the estimated time. This process does not require the intervention of a human customer service and is completely automated by the system, which not only improves the processing speed but also ensures the objectivity and accuracy of the processing results, meets the service needs of consumers, and enhances the overall after-sales service experience.

[0091] In the above embodiments of the present application, the prompt word is input into the intelligent decision-making model, and the intelligent decision-making model is used to perform decision-making reasoning on the prompt word to obtain the initial processing result and the call decision result of the task request, including: inputting the prompt word into the intelligent decision-making model, using the intelligent decision-making model to perform decision-making reasoning on the prompt word to obtain the output result; in response to the intelligent decision-making model meeting the preset iteration condition, determining the initial processing result and the call decision result based on the output result; in response to the intelligent decision-making model not meeting the preset iteration condition, adjusting the prompt word based on the output result to obtain the adjusted prompt word, and inputting the adjusted prompt word into the intelligent decision-making model, using the intelligent decision-making model to perform decision-making reasoning on the adjusted prompt word until the output result meets the preset iteration condition.

[0092] The above-mentioned preset iteration condition is a condition used by the intelligent decision-making model to judge whether it is necessary to repeat the decision-making reasoning process until a specific standard is met when processing a task request. In this technical solution, the preset iteration condition may be based on the number of iterations, or the call decision result in the output result does not require further tool calls for processing. The preset iteration condition ensures that the decision-making process of the intelligent decision-making model can effectively converge and can also process task requests in complex situations, and achieves the accuracy and integrity of the decision through loop iteration.

[0093] The system first inputs the constructed prompt word into the intelligent decision-making model, and the model performs decision-making reasoning and outputs a result. If the output result meets the preset iteration condition, that is, the decision of the model is accurate enough and no further information collection or tool call is required, the system will directly determine the initial processing result and the call decision result based on this output result. This means that for simple or directly judgmental task requests, the intelligent decision-making model can give accurate processing suggestions at one time without repeated iteration, improving the processing efficiency.

[0094] However, if the output result of the intelligent decision-making model does not meet the preset iteration conditions, that is, the model believes that there is still uncertainty in the current decision, or the processing of the task request requires more in-depth information or operations, the system will adjust the prompt words based on the output result, generate new prompt words, and re-enter them into the intelligent decision-making model for decision-making reasoning. This process will be repeated until the output result of the model meets the preset iteration conditions, that is, the accuracy and integrity of the decision are achieved, or it is confirmed that no further tool calls are required. Through this loop iteration mechanism, the system can handle more complex or ambiguous task requests and ensure the accuracy and reliability of the processing results.

[0095] Through decision-making reasoning, the model initially processes the task request and outputs information including the initial processing result and the call decision result. As a control mechanism for the decision-making process, the preset iteration conditions ensure that the model's decision can be iterated according to the nature and complexity of the task request to achieve a better processing effect. This mechanism not only improves the processing efficiency but also ensures the accuracy and reliability of the processing results. Even in the face of complex tasks, through loop iteration, information can be gradually supplemented or specific operations can be performed, and finally a solution that meets the requirements can be obtained.

[0096] In the scenario of after-sales service disputes, in order to reduce the burden on artificial customer service and improve the processing efficiency. The combined use of the intelligent decision-making model and the preset iteration conditions can flexibly handle task requests of different complexities and ensure efficient and accurate dispute resolution. For simple and straightforward tasks, the model can make a decision once and quickly give the processing result; while for complex or information-incomplete tasks, the model will gradually improve the information through the iteration process until the processing standard is reached, thus realizing the intelligence and flexibility of the automated processing process.

[0097] By combining the intelligent decision-making model with the preset iteration conditions, the efficiency and accuracy of after-sales service dispute processing have been significantly improved. For simple tasks, the model can directly give accurate processing results, avoiding unnecessary iterations and improving the processing speed; while for complex tasks, the model gradually improves the processing information through loop iteration until the preset iteration conditions are met, achieving in-depth processing and accurate decision-making, and enhancing the fairness and efficiency of dispute resolution. This mechanism effectively balances the automation and accuracy of processing, reduces the service cost of artificial customer service, and improves the user experience at the same time.

[0098] Exemplarily, assume that a consumer initiates a return application to the platform after receiving the goods, stating that the goods have quality problems. After the system receives this task request, it constructs a prompt and inputs it into the intelligent decision-making model. In the first decision-making inference, if the model determines based on the current information that there is no need to further call tools for processing, that is, the decision result of the call is not to call, or the decision-making process has reached the preset number of iterations, the processing result output by the model will be directly used as the initial processing result. For example, directly approve the return and initiate the refund process, or reject the return and give specific reasons, without further information collection or tool call operations. However, if the intelligent decision-making model does not meet the preset iteration conditions after the first decision-making inference, for example, the model is still unable to determine the quality problem of the goods or needs more information for decision-making, the system will adjust the prompt based on the output result of the model. For example, add an instruction requiring the consumer to provide a quality certificate for the goods. The system then generates the adjusted prompt and re-inputs it into the intelligent decision-making model for decision-making inference.

[0099] This process will be executed cyclically until the output result of the intelligent decision-making model meets the preset iteration conditions. For example, if the model confirms the quality problem of the goods in the second decision-making inference and the decision result of the call indicates that there is no need to further call tools, the system will determine this decision result and generate a target processing result based on this, such as approving the return and informing the consumer of the return process and required materials, and at the same time notifying the merchant to prepare to receive the return.

[0100] Through the above steps and mechanisms, not only the processing efficiency of simple tasks is improved, but also the accuracy and depth of complex task processing are ensured, achieving a double improvement in technology application and user experience.

[0101] In the above example, the preset iteration conditions ensure that the intelligent decision-making model can accurately judge the return application. Whether directly processing simple requests or gradually processing complex requests, it can ensure the accuracy and fairness of the processing results, while reducing unnecessary tool calls and avoiding waste of resources. The ultimate goal is to provide an efficient and satisfactory after-sales service experience and improve the overall service quality and user satisfaction of the platform. Through the iterative decision-making of the intelligent model, complex return cases are deeply processed, while simple cases are avoided from being over-processed.

[0102] In the above embodiments of the present application, in response to receiving a task request sent by a target client, constructing a prompt based on the task request includes: in response to receiving the task request, determining whether the task request meets the conditions for automated processing; in response to the task request meeting the conditions for automated processing, constructing a prompt based on the task request.

[0103] The above-mentioned target client refers to the entity that initiates a task request, which can be a consumer, a merchant, or any service requester who needs to resolve a dispute. The target client is the object of the entire technical solution service. The task request sent by it is the starting point of the process, triggering the start of the subsequent automated processing process.

[0104] The above-mentioned automated processing conditions are a set of rules or conditions used to determine whether a task request is suitable for processing through an intelligent decision-making model and automated tools. As a control mechanism for the process, the automated processing conditions ensure that the system can identify and screen out task requests suitable for automated processing, avoiding ineffective operations on requests that are not suitable for automated processing, and improving the pertinence and efficiency of processing.

[0105] When the system receives a task request sent by the target client, it will first judge the automated processing conditions for the task request. This judgment process aims to identify the nature and requirements of the request to determine whether it can be processed through an intelligent decision-making model and automated tools without the intervention of a human customer service.

[0106] If the task request meets the automated processing conditions, it means that the content, nature, and requirements of the request can be solved through the preset intelligent decision-making model and automated tools. The system will enter the next step and construct a prompt word based on the task request. The key information of the task request can be converted into a format that the intelligent decision-making model can understand and process, providing input for the model's decision-making reasoning, ensuring that the model can accurately understand the requirements of the task and give corresponding processing results.

[0107] Through the judgment of the automated processing conditions, the system can effectively screen out tasks suitable for automated processing, avoiding improper automated processing of complex, special, or sensitive tasks, and ensuring the fairness and accuracy of dispute resolution. At the same time, the construction of the prompt word provides clear and accurate input for the intelligent decision-making model, enabling the model to make decision-making reasoning based on the specific content and requirements of the task request, generating an initial processing result and a tool invocation plan, providing a direction and strategy for subsequent dispute resolution.

[0108] In the scenario of after-sales disputes, the judgment of the automated processing conditions and the process of constructing a prompt word based on the task request can ensure that the intelligent decision-making model can accurately identify and process dispute requests suitable for automated processing, while leaving complex or tasks that require in-depth human judgment to professional customer service personnel, realizing an efficient processing mode of human-machine collaboration.

[0109] By means of the mechanism for judging automated processing conditions and constructing prompt words based on task requests, the efficiency and accuracy of dispute handling have been significantly improved. Meanwhile, the dependence on human customer service has been reduced, and service costs have been decreased. This mechanism ensures the rapid and automatic resolution of simple disputes and the in-depth manual handling of complex disputes. Through the collaboration between intelligence and human resources, the efficiency and comprehensiveness of dispute handling have been achieved, enhancing the user experience.

[0110] Exemplarily, assume that the target client (consumer) initiates a dispute request on the grounds that the goods do not match the description. After receiving this request, the system first judges the conditions for automated processing. If the request contains clear evidence of the goods not matching the description, such as pictures or specific descriptions, and does not belong to special or complex situations, the system determines that the task request meets the conditions for automated processing.

[0111] Subsequently, the system constructs a prompt word based on the task request. The prompt word contains key information such as product information, specific content and relevant evidence of the consumer's feedback, and processing requirements of the request. This prompt word is then input into the intelligent decision-making model, and the model conducts decision-making reasoning based on the information in the prompt word to generate a preliminary processing result and a tool invocation plan, such as directly approving the return, rejecting the return and requiring the merchant to provide a detailed description of the goods, etc.

[0112] Through the above steps and mechanisms, the present application realizes the automated preliminary handling of after-sales service disputes, can quickly respond to simple and clear requests, reduces the burden on human customer service, and decreases service costs. At the same time, the possibility of human intervention is reserved for complex or controversial requests, ensuring the fairness and accuracy of dispute handling and enhancing the overall user experience and satisfaction.

[0113] The judgment of automated processing conditions ensures that the intelligent model can handle simple disputes suitable for automated operations, while complex disputes are guaranteed to be handled deeply and fairly through the intervention of human customer service. This human-machine collaborative processing mode not only improves the efficiency of dispute handling but also maintains the high quality of handling.

[0114] In the above embodiments of the present application, constructing a prompt word based on a task request includes: extracting the feature information of the task request and querying the associated information of the task request; determining the tool information involved in the task request according to the task type of the task request, where the tool information is used to represent the information of at least one tool associated with the task request; constructing a prompt word based on the feature information, associated information, and tool information.

[0115] The above-mentioned feature information refers to the key details extracted from the task request, including but not limited to the background of the request, problem description, identity information of the requester, time node of the request, specific features of goods or services, etc., which are used for the intelligent decision-making model to understand the essence and requirements of the request. The extraction of feature information provides a basis for the intelligent decision-making model to process the task request, ensuring that the model can make decision reasoning based on the specific content of the task request and providing an accurate foundation for subsequent automated processing.

[0116] The above-mentioned associated information refers to any additional information associated with the task request, including but not limited to historical work order information, transaction records of users and merchants, detailed descriptions of goods, industry standards, etc. These information can help the intelligent decision-making model more comprehensively understand the background and environment of the task request and make more accurate decisions. The query and integration of associated information provide richer context and background knowledge for the intelligent decision-making model, ensuring the comprehensiveness and accuracy of decisions and avoiding decision-making mistakes based on incomplete information.

[0117] It should be noted that the above-mentioned associated information can also be data generated during the asynchronous processing of the task request. For example, in the scenario of after-sales disputes, it can be the relevant content negotiated between the buyer and the seller. This associated information, as a reference content for constructing the prompt word, can improve the accuracy of the prompt word.

[0118] The above-mentioned tool information refers to the detailed information of at least one tool related to the task request processing determined according to the task request type, including the function of the tool, operation method, call parameters, applicable scenarios, etc., which are used for the intelligent decision-making model to judge whether a specific tool needs to be called and how to call these tools for task processing. The determination of tool information provides a basis for generating the preset task process and tool call plan, ensuring that the intelligent decision-making model can intelligently decide whether to call a tool and which tools to call according to the nature and requirements of the task request, realizing the automation and efficiency improvement of task processing.

[0119] When the system recognizes that a task request meets the conditions for automated processing, it will enter the step of constructing a prompt word based on the task request. First, the system will deeply analyze the task request and extract the key feature information therein, such as request type, problem description, relevant evidence, etc., to provide a decision basis for the intelligent decision-making model. Subsequently, the system will automatically query the associated information related to the task request, such as user transaction records, processing results of similar disputes, applicable industry standards, etc., to enrich the decision context of the model and improve the accuracy of decisions.

[0120] Next, the system will determine the tool information that may need to be called according to the task type of the task request, such as evidence review tools, communication and coordination tools, data query tools, etc., to support the generation of subsequent preset task processes and tool call plans. Finally, based on the extracted feature information, queried associated information, and determined tool information, the system constructs a prompt containing task requirements, context information, and tool call strategies for the decision-making and reasoning of the intelligent decision-making model, ensuring that the model can accurately understand the task request, make decisions based on sufficient information, and provide a direction and basis for subsequent dispute resolution.

[0121] In the scenario of after-sales service disputes, the process of constructing a prompt based on the task request ensures that the intelligent decision-making model can accurately understand the essence and requirements of each dispute request, providing an accurate basis for subsequent automated processing. By extracting feature information, querying associated information, and determining tool information, the system can provide a comprehensive and accurate task description for the intelligent decision-making model, enabling the model to make fair and accurate processing results based on sufficient information during decision-making, reducing the dependence on human customer service, and improving the efficiency and accuracy of dispute resolution.

[0122] By the process of constructing a prompt based on the task request, the accuracy and comprehensiveness of dispute resolution are significantly improved, the dependence on human customer service is reduced, and the service cost is lowered. This mechanism ensures that the intelligent decision-making model can make decisions based on sufficient and accurate information, avoiding decision-making mistakes based on incomplete information and improving the quality of dispute resolution. At the same time, through the automated processing process, the efficiency of dispute resolution is enhanced, the burden on human customer service is reduced, the user experience is improved, demonstrating the significant value and potential of the technical solution in the field of service automation.

[0123] Exemplarily, the target client (consumer) initiates a return application on the grounds that the product has quality problems. After receiving this request, the system first extracts the feature information of the task request, including the description of the product, the consumer's return reason, the time node of the return application, etc.; subsequently, the system queries the associated information related to the task request, such as the transaction record between the consumer and the merchant, the quality assurance information of the merchant's products, the processing results of similar return cases, etc.; next, according to the task type of the return application, the system determines the tool information that may need to be called, such as product quality review tools, communication and negotiation tools, etc.

[0124] Based on the above-extracted feature information, queried association information, and determined tool information, the system constructs a prompt that includes the background of the return application, problem description, relevant evidence, and a possible tool invocation strategy. Subsequently, the prompt is input into the intelligent decision-making model, and the model makes decision inferences based on the information in the prompt to generate a preliminary processing result and a tool invocation plan. This process ensures that the intelligent decision-making model can comprehensively and accurately understand the specific situation of the return application, make decisions based on sufficient information, improve the efficiency and accuracy of return processing, reduce the burden on human customer service, and enhance the user experience.

[0125] In the above example, the process of constructing the prompt based on the task request not only improves the efficiency of the intelligent model in processing return requests but also ensures the depth and accuracy of the processing. This mechanism realizes the automated preliminary processing of return disputes through the combination of the intelligent model and tool information, retains the intervention of human customer service for complex requests, and improves the overall efficiency and quality of return processing through human-machine collaboration.

[0126] In the above embodiments of the present application, the method further includes: in response to the task request not meeting the conditions for automated processing, adding the task request to a target queue, where the task requests in the target queue are processed by a preset client.

[0127] The above-mentioned target queue is a queue for storing and managing task requests that do not meet the conditions for automated processing. These tasks may require processing by human customer service due to their complexity, sensitivity, or uniqueness. The target queue serves as a buffer between the automated processing flow and the manual processing flow, ensuring that task requests that are not processed automatically can be effectively managed and orderly assigned to the preset client, achieving the continuity and efficiency of task processing.

[0128] The above-mentioned preset client refers to the artificial customer service system or personnel preset in the system for processing task requests that are complex, sensitive, or not applicable to the automated process. As a supplement to the automated process, the preset client provides the ability to perform in-depth manual processing for task requests that cannot be processed by automated means, ensuring that each task request can be appropriately processed and meeting the needs of various dispute resolutions.

[0129] When a task request is identified as not meeting the conditions for automated processing, the system automatically adds these task requests to the target queue. The target queue, as the entry point for the manual processing flow, manages and sorts these task requests to ensure that they can be efficiently assigned to the preset client for processing.

[0130] The preset client, usually composed of professionally trained customer service staff or advanced processing systems, will extract task requests from the target queue for in-depth analysis and manual processing. These customer service staff or systems can handle more complex and sensitive disputes and provide more personalized and meticulous services to meet the needs of resolving special or complex disputes.

[0131] In the after-sales dispute scenario, for dispute requests that do not meet the conditions for automated processing, the system ensures that complex disputes are resolved through in-depth manual handling by adding them to the target queue and having the preset client handle them, avoiding possible errors or improper handling that may be caused by automated processing, while maintaining high service standards and user satisfaction. Through the combination of the target queue and the preset client, the comprehensiveness and efficiency of dispute resolution have been significantly improved. For simple and straightforward requests, the automated process can provide fast and accurate processing, reducing service costs; while for complex and sensitive requests, the cooperation of the target queue and the preset client ensures that the requests can be analyzed in depth and processed in a personalized manner, improving service quality and user satisfaction.

[0132] Exemplarily, assume that the target client (consumer) initiates a return application on the grounds that the product has major quality problems. After receiving this request, the system, through the judgment of the automated processing conditions, finds that the request content contains a claim for rights and interests, which is a complex situation and not suitable for automated processing. Therefore, the system adds this return application to the target queue. The target queue manages and sorts this task request, and then the preset client, that is, the professionally trained customer service staff, extracts this task request from the queue for in-depth analysis and manual processing. The customer service staff will communicate with the consumer and the merchant to understand the specific situation of the dispute, review relevant evidence, and ensure the fairness and legality of the processing result.

[0133] Through the above steps and mechanisms, a solution for in-depth manual processing is provided for those complex return cases that cannot be processed through the automated process. This mechanism ensures the in-depth analysis and fair resolution of complex disputes, avoids problems that may be caused by automated processing, and at the same time maintains high service quality and user satisfaction. Through the hierarchical combination of automated and manual processing, the efficiency and comprehensiveness of dispute resolution are achieved.

[0134] In the above example, the combined use of the target queue and the preset client ensures that complex return cases can be processed manually in a deep and fair manner, while the automated processing mechanism focuses on simple and straightforward requests. This hierarchical processing method effectively improves the overall efficiency of dispute resolution, reduces the need for professional customer service staff, and at the same time ensures service quality.

[0135] In the above embodiments of the present application, the method further includes: in response to the task request not meeting the conditions for automated processing, or the intelligent decision-making model not generating an initial processing result, sending the task request to the target terminal; receiving the target processing result corresponding to the task request returned by the target terminal.

[0136] The above-mentioned target terminal refers to the terminal in the system used to receive and process those task requests that are not suitable for automated processing or for which the automated processing process fails to produce a processing result, and can be composed of professional human customer service personnel or advanced processing systems.

[0137] As a supplement to the automated process, the target terminal ensures that all task requests not processed by the automated process can be properly handled. Especially for those complex, sensitive or information-incomplete requests, through the intervention of human customer service, the depth and fairness of the processing are guaranteed.

[0138] When the system determines that a task request does not meet the conditions for automated processing, or the intelligent decision-making model fails to generate a preliminary processing result when processing the request, the system will automatically send the task request to the target terminal. The receiving personnel or advanced processing system of the target terminal can deeply analyze the specific content of the task request, including but not limited to the background of the request, problem description, relevant evidence, etc., and perform manual processing and in-depth analysis to ensure the fairness and depth of dispute resolution.

[0139] After receiving the task request, the target terminal will perform manual processing on it, including but not limited to communicating with the requester and relevant parties, collecting more information, invoking professional tools for verification, etc., and finally generating a target processing result for the task request. The processing result will include dispute resolution suggestions, operation instructions, notification information, etc., to ensure that the request is properly resolved.

[0140] Through the combination of the automated process and the manual processing of the target terminal, the comprehensiveness and efficiency of task request processing are achieved. For those task requests that are not suitable for automated processing or for which automated processing fails to produce a result, the manual processing mechanism of the target terminal can provide in-depth analysis and precise processing, ensuring the fairness and depth of dispute resolution, while avoiding possible errors or improper processing caused by automated processing, and maintaining high service standards and user satisfaction.

[0141] In the scenario of after-sales disputes, the reasoning ability of the large language model (LLM) can be used to resolve after-sales service disputes and reduce service costs. With this goal, for dispute requests that do not meet the conditions for automated processing, or for cases where the intelligent decision-making model fails to generate a preliminary processing result, the system ensures the in-depth resolution of complex disputes by sending the task request to the target terminal for manual processing, while maintaining the advantages of automated processing, achieving a balance between the efficiency and quality of the technical solution in dispute resolution.

[0142] By combining an automated process with manual handling at the target terminal, the efficiency and quality of dispute resolution are improved. For simple and straightforward disputes, the automated process can provide fast and accurate handling, reducing service costs; while for complex, sensitive or information-incomplete disputes, the intervention of the target terminal ensures that requests can be deeply analyzed and personalized, improving service quality and user satisfaction.

[0143] Exemplarily, assume that the target terminal (consumer) initiates a return application on the grounds of an indescribable quality problem with the product, which involves complex issues of consumer rights. After receiving this request, the system, through the judgment of automated processing conditions, finds that the request content contains descriptions of legal issues and complex situations, which is a scenario not suitable for automated processing. Therefore, the system forwards this return application to the target terminal, that is, it is processed by customer service staff with dispute resolution experience. The customer service staff deeply analyzes the background information of the request, communicates with the consumer in detail to understand the specific problems of the product and the specific demands of the consumer, and also communicates with the merchant to collect the merchant's feedback and evidence. The customer service staff also queries relevant laws and regulations and conducts in-depth analysis to ensure the fairness of the processing result. Finally, the customer service staff generates a target processing result for this return application, which may include suggestions for return, refund or repair. This processing result is fed back to the system through the target terminal, and then the system notifies the consumer and the merchant to execute the processing result and resolve the dispute.

[0144] Through the above steps and mechanisms, a solution for in-depth manual processing is provided for those complex return cases that are not suitable for automated processing. This mechanism ensures the in-depth analysis and fair resolution of complex disputes, avoids problems that may be brought about by automated processing, and at the same time maintains high service quality and user satisfaction, reflecting the comprehensiveness and flexibility of the technical solution in the field of after-sales service automation.

[0145] In the above example, the intervention and processing of the target terminal ensure that complex return cases can be deeply and fairly processed manually, while the automated processing mechanism focuses on simple and straightforward requests. This hierarchical processing method effectively improves the overall efficiency of dispute resolution and maintains service quality.

[0146] Figure 3 is a schematic diagram of a model processing flow according to an embodiment of the present application, as Figure 3 shown, the consumer on the consumer side is the starting point of the process, referring to an individual or entity that submits a service application or encounters a dispute. The application platform intervenes to enable the consumer to submit their needs or disputes through a specific application platform, and the platform starts the process processing after receiving it.

[0147] Automated decision-making in the offline dispute handling side enables the platform to first attempt to preliminarily process the application through preset automated rules or algorithms to quickly resolve some simple or standardized issues. Routing assignment, based on the nature of the application and the results of the preliminary processing, assigns the application to the corresponding processing queues, such as automated processing, manual processing, or large language model processing. The work order assignment time marks the time when the application is received and assigned, which is used for subsequent tracking and management. The manual queue and the agent queue mean that if the application cannot be resolved automatically, it will be assigned to the manual queue or the queue assisted by the agent for processing. Manual processing and manual negotiation mean that for situations requiring human intervention, the processing personnel will communicate directly with the consumer to collect more information or negotiate a solution. The judgment notice means that after the problem is solved or the solution is reached, the consumer will receive the final processing result or judgment.

[0148] Event perception on the large language model processing side includes agent intervention events, project association events, and asynchronous task events. These events trigger the large language model processing process and may involve specific situations or tasks during the processing. Event perception means that the large language model can identify and understand the specific situations of various processing events. The agent executor is responsible for executing the decisions or action plans generated by the large language model. Multi-round decision-making means that the large language model may need to conduct multi-round analysis and decision-making to ensure the accuracy and comprehensiveness of the processing solution. Tools and prompt words are the auxiliary tools and guiding information provided to the large language model to help it better understand and generate solutions. The use of tools can be achieved through process tools. The large language model can refer to the model used to process complex tasks and provide intelligent suggestions, that is, the above-mentioned intelligent decision-making model. The model output event is the event generated after the large language model processing, which may require further manual review or the cooperation of other tools to complete the entire processing process.

[0149] After the results generated after large language model or manual processing are output, further analysis and understanding of the large language model output or manual processing results can be carried out to ensure that they meet the requirements. It can be judged whether there is a solution in the output results, that is, to check whether there are ready-made solutions or processing strategies that can be applied. If the preliminary analysis shows no solution, the process will turn to more detailed or manual processing.

[0150] Furthermore, R & D personnel can manage prompt words, model versions, process orchestration, and tools, and update the configuration on the processing side of the large language model according to the adjustments. Operation personnel can adjust long - process tools according to task processes, task bindings, and information collection and release configurations. The long - process tools include various tool types such as task decision - making, process initiation, process handling, and information feedback. Process handling includes notification confirmation, negotiation, evidence presentation, and collaboration, etc. Among them, notification confirmation includes outbound notifications, information confirmation, etc.; negotiation includes outbound negotiation, etc.; evidence presentation includes evidence distribution, evidence review, etc.; collaboration includes collaboration work orders and information recovery, etc. The above - mentioned different types of tools can process and feedback data. It should be noted that the buyer and seller can also negotiate with each other, and the result information of the negotiation can be fed back to the asynchronous task event of event perception, so that the content of the negotiation can be referred to when constructing prompt words, thereby improving the construction accuracy of prompt words.

[0151] Figure 4 is a judgment link architecture diagram according to an embodiment of the present application, as Figure 4 shown, including a gateway layer, core capabilities, underlying capabilities, and underlying dependencies. The gateway layer includes a communication protocol (MQ protocol), a call framework (asynchronous HSF), and model result output. The core capabilities include a proxy executor, prompt word capabilities, memory information, and tool duration. The underlying capabilities include work order processing capabilities, handover to human, dispute judgment, cost model capabilities, and large language models. The underlying dependencies include service capabilities and algorithm capabilities.

[0152] The MQ protocol in the gateway layer is at the level of system communication with the outside world, using the Message Queue protocol (MQ) to process dispute requests from consumers or merchants. This is an asynchronous communication mechanism, which helps to improve the system's response speed and processing capacity. Asynchronous HSF is to execute service calls by asynchronously invoking the High Speed Service Framework (HSF), especially when interacting with other internal services, asynchronous HSF ensures non - blocking and high - efficiency task processing. MQ model result output is that after the large language model processes the dispute request, its result is output through the message queue protocol, ensuring that the result can be received and processed by other components in the system in a timely and reliable manner.

[0153] The proxy executor in the core capabilities includes intervention management, task management, and decision management. Among them, intervention management includes rule admission and work order intervention, task management includes task creation and task follow-up, and decision management includes single-round decision-making and multi-round decision-making; rule admission is used to determine which dispute work orders are suitable for automated processing and which require manual intervention, work order intervention is used to automatically start the processing flow for work orders suitable for automation, task creation is used to create processing tasks according to the specific situation of the dispute, task follow-up is used to monitor the progress of the task to ensure that the task is correctly executed, single-round decision-making is used to judge whether a dispute handling result can be directly generated, and multi-round decision-making is used to support multi-round interactions for situations that require more in-depth analysis to achieve a more accurate judgment.

[0154] Feature extraction in the prompt word capabilities can extract key information from dispute work orders and provide input for the model. Retrieval-Augmented Generation (abbreviated as RAG) retrieves relevant information from the knowledge base to assist the model in making more practical judgments. The dynamic template dynamically generates templates called by the model according to the dispute type and features, improving the pertinence of the judgment.

[0155] Short-term memory in the memory information is used to store the context of the current work order and the recent tool call information to help the agent model understand the current state. Long-term memory includes the knowledge base system and the project rule set, providing long-term historical data and rules for the model to make decisions.

[0156] The tool market provides a series of tools for the proxy processor to call, including asynchronous task classes, audits, outbound negotiations, collaborative work orders, etc. These tools cover various requirements for dispute handling.

[0157] The work order processing ability in the underlying capabilities is used to suspend the work order when the dispute work order is not suitable for automated processing to avoid its misprocessing or causing incorrect judgments. Handing over to the human is when the work order requires in-depth manual intervention, the system can transfer the work order to the human customer service for processing to ensure that the dispute is properly resolved. Dispute judgment is the built-in dispute judgment logic of the system, serving as a supplementary or confirmation mechanism for the model judgment to ensure the fairness and accuracy of the judgment result. The cost model ability is used to evaluate the costs in the dispute handling process to help improve resource allocation and processing strategies.

[0158] The underlying dependencies include service capabilities and algorithm capabilities, including multiple internal systems or modules. These modules are the underlying support for the system operation, responsible for providing data, tools, or services to support the realization of intelligent judgment.

[0159] Such as Figure 4As shown, by leveraging the inference ability of the model and a series of auxiliary tools, the full process automation from the reception, analysis, decision-making to execution of dispute requests is achieved. For simple scenarios, the system can quickly generate judgment results, while for complex scenarios, it can suspend the work order and transfer it to manual processing in a timely manner to ensure the quality and efficiency of dispute resolution.

[0160] Figure 5 It is a flowchart of a task request processing method according to an embodiment of the present application. As Figure 5 shown, the method may include the following steps:

[0161] Step S502, in response to an input instruction on the operation interface, display a task request sent by a target client on the operation interface, and construct a prompt word based on the task request;

[0162] When an input instruction can be issued through the operation interface, the system responds to this instruction and first displays the task request to be processed sent by the target client on the operation interface. This display process provides complete details of the task request, including the nature of the request, problem description, relevant evidence, etc., ensuring that the processing personnel can comprehensively understand the situation of the task request.

[0163] Subsequently, the system constructs a prompt word based on the displayed task request. The process of constructing the prompt word involves extracting key feature information of the task request, querying associated information related to the request, and determining tool information that may need to be called according to the request type. The construction of the prompt word ensures that the intelligent decision-making model can accurately understand the essence of the task request and make decision reasoning based on sufficient information, providing a direction for subsequent processing.

[0164] Step S504, in response to a processing instruction on the operation interface, display the target processing result of the task request on the operation interface.

[0165] Among them, the target processing result is obtained by processing the task request based on the initial processing result and the call decision result. The initial processing result and the call decision result are obtained by inputting the prompt word into the intelligent decision-making model and using the intelligent decision-making model to make decision reasoning on the prompt word. The call decision result is used to indicate whether it is necessary to process the task request by calling a tool.

[0166] After completing the processing of the task request, when the processor issues a processing instruction through the operation interface, the system responds to this instruction and displays the processing result of the task request on the operation interface.

[0167] Through the above steps, in response to an input instruction on the operation interface, a task request sent by a target client is displayed on the operation interface, and a prompt word is constructed based on the task request; in response to a processing instruction on the operation interface, a target processing result of the task request is displayed on the operation interface, where the target processing result is obtained by processing the task request based on an initial processing result and a call decision result, and the initial processing result and the call decision result are obtained by inputting the prompt word into an intelligent decision-making model and using the intelligent decision-making model to perform decision-making reasoning on the prompt word. The call decision result is used to indicate whether it is necessary to process the task request by calling a tool, achieving the improvement of task processing efficiency; it is easy to notice that by performing decision-making reasoning through the intelligent decision-making model and directly obtaining the initial processing result of the task and the judgment of whether a tool needs to be called, this process significantly reduces the time for manual judgment and decision-making. The intelligent decision-making model can quickly understand the context and requirements of the task request, and based on its powerful language understanding and reasoning capabilities, directly give a processing strategy and tool call suggestion, shortening the preliminary preparation time of task processing. It can combine the initial processing result and the call decision result to automatically call the tool, further accelerating the execution stage of task processing, reducing manual intervention, and improving the degree of automation of the process, thereby effectively improving task processing efficiency, and further solving the technical problem of low task processing efficiency in the related art.

[0168] Figure 6 is a flowchart of a method for processing a task request according to an embodiment of the present application, as Figure 6 shown, the method may include the following steps:

[0169] Step S602, obtain a task request by calling a first interface, and construct a prompt word based on the task request;

[0170] Wherein, the first interface includes a first parameter, and the parameter value of the first parameter includes the task request.

[0171] The above-mentioned first interface may be an interface for data interaction between a cloud server and a client, and the task request can be passed into an interface function as the first parameter of the interface function to achieve the purpose of uploading the task request to the cloud server.

[0172] Step S604, input the prompt word into an intelligent decision-making model, and use the intelligent decision-making model to perform decision-making reasoning on the prompt word to obtain an initial processing result and a call decision result of the task request;

[0173] Wherein, the call decision result is used to indicate whether it is necessary to process the task request by calling a tool.

[0174] Step S606, process the task request based on the initial processing result and the call decision result to obtain a target processing result;

[0175] Step S608, output the target processing result by calling the second interface.

[0176] Among them, the second interface includes a second parameter, and the parameter value of the second parameter includes the target processing result.

[0177] The above-mentioned second interface can be an interface for data interaction between the cloud server and the client. The cloud server can pass the target processing result into the interface function as the second parameter of the interface function to achieve the purpose of sending the target processing result to the client.

[0178] Through the above steps, obtain the task request by calling the first interface, and construct a prompt based on the task request. Among them, the first interface includes a first parameter, and the parameter value of the first parameter includes the task request; input the prompt into the intelligent decision-making model, use the intelligent decision-making model to make decision reasoning on the prompt, and obtain the initial processing result of the task request and the call decision result. The call decision result is used to indicate whether it is necessary to process the task request by calling a tool; process the task request based on the initial processing result and the call decision result to obtain the target processing result; output the target processing result by calling the second interface. Among them, the second interface includes a second parameter, and the parameter value of the second parameter includes the target processing result, which improves the task processing efficiency. It is easy to notice that through the decision reasoning of the intelligent decision-making model, the initial processing result of the task and the judgment of whether a tool needs to be called are directly obtained. This process significantly reduces the time for manual judgment and decision-making. The intelligent decision-making model can quickly understand the context and requirements of the task request. Based on its powerful language understanding and reasoning capabilities, it directly gives processing strategies and tool call suggestions, shortening the preliminary preparation time of the task processing. It can combine the initial processing result and the call decision result to automatically call tools, further accelerating the execution stage of the task processing, reducing manual intervention, and improving the automation degree of the process, thereby effectively improving the task processing efficiency, and further solving the technical problem of low task processing efficiency in the related technology.

[0179] 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 for analysis, stored data, displayed data, 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 relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions, and corresponding operation entrances are provided for users to choose to authorize or refuse.

[0180] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0181] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of various embodiments of this application.

[0182] According to an embodiment of this application, there is also provided a task request processing device for implementing the above task request processing method. Figure 7 It is a schematic diagram of a task request processing device according to an embodiment of this application, as Figure 7 shown. The device 700 includes: a construction module 702, an inference module 704, and a processing module 706.

[0183] Among them, the construction module is used to construct a prompt word based on the task request in response to receiving a task request sent by a target client; the inference module is used to input the prompt word into an intelligent decision-making model, and use the intelligent decision-making model to perform decision inference on the prompt word to obtain an initial processing result of the task request and a call decision result, where the call decision result is used to indicate whether it is necessary to process the task request by calling a tool; the processing module is used to process the task request based on the initial processing result and the call decision result to obtain a target processing result.

[0184] It should be noted here that the above construction module 702, inference module 704, and processing module 706 correspond to steps S202 to S206 in the above embodiments. The instances and application scenarios realized by the three modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should be noted that the above modules or units can be hardware components or software components stored in a memory and processed by one or more processors, and the above modules can also be part of a device and can run in the server 10 provided in the above embodiments.

[0185] In the above embodiments of the present application, the processing module is further configured to, in response to a call decision result indicating that it is necessary to process the task request by invoking a tool, determine a preset task process and a tool invocation plan based on the initial processing result, and process the task request based on the preset task process and the tool invocation plan to obtain a target processing result; in response to a call decision result indicating that it is not necessary to process the task request by invoking a tool, determine the initial processing result as the target processing result.

[0186] In the above embodiments of the present application, the processing module is further configured to invoke at least one tool associated with the preset task process based on the tool invocation plan; generate a preset task to be processed for the preset task process based on the at least one tool; execute the preset task to be processed, and obtain a target processing result based on the execution result of the preset task to be processed and the initial processing result.

[0187] In the above embodiments of the present application, the inference module is further configured to input a prompt word into the intelligent decision-making model, use the intelligent decision-making model to perform decision-making inference on the prompt word to obtain an output result; in response to the intelligent decision-making model satisfying a preset iteration condition, determine the initial processing result and the call decision result based on the output result; in response to the intelligent decision-making model not satisfying the preset iteration condition, adjust the prompt word based on the output result to obtain an adjusted prompt word, and input the adjusted prompt word into the intelligent decision-making model, and use the intelligent decision-making model to perform decision-making inference on the adjusted prompt word until the output result satisfies the preset iteration condition.

[0188] In the above embodiments of the present application, the construction module is further configured to, in response to receiving a task request, determine whether the task request meets the conditions for automated processing; in response to the task request meeting the conditions for automated processing, construct a prompt word based on the task request.

[0189] In the above embodiments of the present application, the construction module is further configured to extract the feature information of the task request and query the associated information of the task request; determine the tool information involved in the task request according to the task type of the task request, where the tool information is used to represent the information of at least one tool associated with the task request; construct a prompt word based on the feature information, the associated information, and the tool information.

[0190] In the above embodiments of the present application, the device is further configured to, in response to the task request not meeting the conditions for automated processing, add the task request to a target queue, where the task requests in the target queue are processed by a preset client.

[0191] In the above embodiments of the present application, the device is further configured to, in response to the task request not meeting the conditions for automated processing, or the intelligent decision-making model not generating an initial processing result, send the task request to a target terminal; receive the target processing result corresponding to the task request returned by the target terminal.

[0192] According to an embodiment of the present application, there is also provided a task request processing device for implementing the above-mentioned task request processing method. Figure 8 It is a schematic diagram of a task request processing device according to an embodiment of the present application, as Figure 8 shown. The device 800 includes: a first display module 802 and a second display module 804.

[0193] Among them, the first display module is used to respond to an input instruction acting on the operation interface, display a task request sent by a target client on the operation interface, and construct a prompt word based on the task request; the second display module is used to respond to a processing instruction acting on the operation interface, and display a target processing result of the task request on the operation interface, where the target processing result is obtained by processing the task request based on an initial processing result and a call decision result, and the initial processing result and the call decision result are obtained by inputting the prompt word into an intelligent decision-making model and using the intelligent decision-making model to perform decision-making reasoning on the prompt word, and the call decision result is used to indicate whether it is necessary to process the task request by calling a tool.

[0194] It should be noted here that the above-mentioned first display module 802 and second display module 804 correspond to steps S502 to S504 in the above embodiment. The instances and application scenarios implemented by the two modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment. It should be noted that the above modules or units can be hardware components or software components stored in a memory and processed by one or more processors, and the above modules can also be part of a device and can run in the server 10 provided in the above embodiment.

[0195] According to an embodiment of the present application, there is also provided a task request processing device for implementing the above-mentioned task request processing method. Figure 9 It is a schematic diagram of a task request processing device according to an embodiment of the present application, as Figure 9 shown. The device 900 includes: an acquisition module 902, an inference module 904, a processing module 906, and an output module 908.

[0196] Among them, the acquisition module is used to obtain a task request by calling a first interface and construct a prompt word based on the task request. The first interface includes a first parameter, and the parameter value of the first parameter includes the task request. The inference module is used to input the prompt word into an intelligent decision-making model, and use the intelligent decision-making model to make a decision inference on the prompt word to obtain an initial processing result of the task request and a call decision result, where the call decision result is used to indicate whether it is necessary to process the task request by calling a tool. The processing module is used to process the task request based on the initial processing result and the call decision result to obtain a target processing result. The output module is used to output the target processing result by calling a second interface. The second interface includes a second parameter, and the parameter value of the second parameter includes the target processing result.

[0197] It should be noted here that the above-mentioned acquisition module 902, inference module 904, processing module 906, and output module 908 correspond to steps S602 to S608 in the above embodiment. The examples and application scenarios implemented by the four modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment. It should be noted that the above modules or units can be hardware components or software components stored in a memory and processed by one or more processors. The above modules can also be part of a device and can run in the server 10 provided in the above embodiment.

[0198] It should be noted that the preferred implementation schemes involved in the above embodiments of the present application are the same as the schemes, application scenarios, and implementation processes provided in the above embodiments, but are not limited to the schemes provided in the above embodiments.

[0199] An embodiment of the present application can provide a computing device. Figure 10 It is a structural block diagram of a computing device according to an embodiment of the present application. As Figure 10 shown, the computing device 100 may include: one or more ( Figure 10 only one is shown in the figure) processors 102, a memory 104, a storage controller, and a peripheral interface.

[0200] The above-mentioned computing device can be understood as an integrated intelligent terminal, including but not limited to a server, a desktop computer, a PC (Personal Computer), a model all-in-one machine, etc. And, the above-mentioned model in the above embodiments of the present application may be pre-set in the computing device.

[0201] Specifically, the computing device can pre-set various types of models, including but not limited to models in the fields of natural language processing, visual processing, speech processing, code processing, multi-modal task processing, etc., so as to provide diverse model selection. In different product forms, the computing device can support one or more model usage methods, including but not limited to model training, model invocation, model fine-tuning, model deployment, model inference and application, etc. In some product forms, the computing device also supports model management, including but not limited to multi-type model management (supporting the management of various types of models such as discriminative and generative models), model version control (supporting the control of different model versions), model evaluation (evaluating the performance and effect of the model based on model evaluation tools), etc. In other product forms, the computing device can also create applications based on the model, provide API invocation capabilities, and can call the model into the created application through the API interface. At the same time, an application management tool is provided to realize the management and monitoring of the application.

[0202] Furthermore, the computing device can also include data management (supporting the creation and management of model tuning data sets), a training center (providing rich training resources to help users learn and master AI technologies), and basic control capabilities (providing enterprise-level basic control capabilities to ensure the security and efficient operation of the system). Through the above functions, a comprehensive and integrated AI development, training, deployment, and application device is provided.

[0203] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, to implement the methods in the above embodiments. The memory can include high-speed random access memory, and can also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory can further include memories remotely set relative to the processor, and these remote memories can be connected to the terminal A through a network. Examples of the above network include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and their combinations.

[0204] The processor can call the executable program stored in the memory through the transmission device to execute the method of any one of the above embodiments.

[0205] An embodiment of the present application can provide an electronic device. Figure 11 It is a structural block diagram of an electronic device according to an embodiment of the present application. As Figure 11 shown, the electronic device can include: an input / output device 112; a memory 114 and a processor 116, wherein the processor 116 is connected to the input / output device 112 and the memory 114 through a bus 118.

[0206] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, to implement the methods in the above embodiments. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory may further include a memory remotely set relative to the processor, and these remote memories can be connected to terminal A through a network. Examples of the above network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and their combinations.

[0207] The processor can call the executable program stored in the memory through the transmission device to execute the method of any one of the above embodiments.

[0208] Those of ordinary skill in the art can understand that Figure 11 the structure shown is only schematic, and the computing device can also be a terminal device such as a smart phone, a tablet computer, a personal digital assistant, and a mobile Internet device (MID), a PAD, etc. The Figure 11 it does not limit the structure of the above computing device. For example, the computing device 100 may further include more or fewer components (such as a network interface, a display device, etc.) than those shown in the figure, or have a configuration different from that Figure 11 shown.

[0209] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware of the terminal device through a program, and the program can be stored in a computer-readable storage medium. The storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0210] The embodiments of the present application also provide a computer-readable storage medium. Optionally, in this embodiment, the above computer-readable storage medium can be used to save the program code executed by the method provided in the above embodiments.

[0211] Optionally, in this embodiment, the above storage medium may be located in the computing device.

[0212] Optionally, in this embodiment, the computer-readable storage medium is set to store an executable program, and when the executable program runs, it controls the device where the computer-readable storage medium is located to execute the method of any one of the above embodiments.

[0213] An embodiment of the present application also provides a computer program product. Optionally, in this embodiment, the above computer program product may include a computer program, and when the computer program is executed by a processor, the method provided in the above embodiment is implemented.

[0214] An embodiment of the present application also provides a computer program product. Optionally, the above computer program product may include a non-volatile computer-readable storage medium, and the non-volatile computer-readable storage medium may be used to store a computer program, and when the computer program is executed by a processor, the method provided in the above embodiment is implemented.

[0215] An embodiment of the present application also provides a computer program. Optionally, in this embodiment, when the computer program is executed by a processor, the method provided in the above embodiment is implemented.

[0216] In the above embodiments of the present application, the descriptions of the various embodiments have their own emphases. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0217] In the several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces, and the indirect coupling or communication connection of units or modules can be in an electrical or other form.

[0218] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0219] In addition, the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0220] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media such as USB flash drives, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0221] The above are only the preferred embodiments of this application. It should be noted that for those of ordinary skill in the art, without departing from the principle of this application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of this application.

Claims

1. A task request processing method, characterized in that: include: In response to receiving a task request sent by a target client, constructing a prompt word based on the task request; Input the prompt word into the intelligent decision model, use the intelligent decision model to perform decision reasoning on the prompt word, and obtain the initial processing result and the calling decision result of the task request, wherein the calling decision result is used to indicate whether the task request needs to be processed by calling a tool; The task request is processed based on the initial processing result and the calling decision result to obtain a target processing result.

2. The task request processing method according to claim 1, characterized in that: Processing the task request based on the initial processing result and the call decision result to obtain a target processing result includes: In response to the calling decision result that the task request needs to be processed by calling a tool, a preset task flow and a tool calling scheme are determined based on the initial processing result, and the task request is processed based on the preset task flow and the tool calling scheme to obtain the target processing result; In response to the calling decision result being that the task request does not need to be processed by a calling tool, the initial processing result is determined to be the target processing result.

3. The task request processing method according to claim 2, characterized in that: Processing the task request based on the preset task flow and the tool calling scheme to obtain the target processing result includes: Calling at least one tool associated with the preset task process based on the tool calling scheme; Generate a preset task to be processed of the preset task flow based on the at least one tool; The preset task to be processed is executed, and the target processing result is obtained based on the execution result of the preset task to be processed and the initial processing result.

4. The task request processing method according to claim 3, characterized in that: Inputting the prompt word into the intelligent decision model, using the intelligent decision model to perform decision reasoning on the prompt word, and obtaining the initial processing result and the calling decision result of the task request, including: Inputting the prompt word into the intelligent decision model, and using the intelligent decision model to perform decision reasoning on the prompt word to obtain an output result; In response to the intelligent decision model satisfying a preset iteration condition, determining the initial processing result and the calling decision result based on the output result; In response to the intelligent decision-making model not satisfying the preset iteration condition, the prompt word is adjusted based on the output result to obtain an adjusted prompt word, and the adjusted prompt word is input into the intelligent decision-making model, and the intelligent decision-making model is used to perform decision reasoning on the adjusted prompt word until the output result satisfies the preset iteration condition.

5. The task request processing method according to any one of claims 1 to 4, characterized in that: In response to receiving a task request sent by a target client, constructing a prompt word based on the task request, including: In response to receiving the task request, determining whether the task request satisfies a condition for automated processing; In response to the task request satisfying the condition of the automated processing, the prompt word is constructed based on the task request.

6. The task request processing method according to claim 5, characterized in that: Constructing prompt words based on the task request, including: Extracting characteristic information of the task request, and querying associated information of the task request; Determining tool information involved in the task request according to a task type of the task request, wherein the tool information is used to represent information of at least one tool associated with the task request; The prompt word is constructed based on the feature information, the association information and the tool information.

7. The task request processing method according to claim 5, characterized in that: The method further comprises: In response to the task request not satisfying the automated processing condition, the task request is added to a target queue, wherein the task requests in the target queue are processed by a preset client.

8. The task request processing method according to claim 5, characterized in that: The method further comprises: In response to the task request not satisfying the automated processing condition, or the intelligent decision model not generating the initial processing result, sending the task request to a target terminal; The target processing result corresponding to the task request returned by the target terminal is received.

9. A task request processing method, characterized in that: include: In response to an input instruction acting on the operation interface, the task request sent by the target client is displayed on the operation interface, and a prompt word is constructed based on the task request; In response to the processing instruction acting on the operation interface, the target processing result of the task request is displayed on the operation interface, wherein the target processing result is obtained by processing the task request based on the initial processing result and the calling decision result, the initial processing result and the calling decision result are obtained by inputting the prompt word into the intelligent decision model and performing decision reasoning on the prompt word using the intelligent decision model, and the calling decision result is used to indicate whether the task request needs to be processed by calling the tool.

10. A task request processing method, characterized in that: include: Acquire a task request by calling a first interface, and construct a prompt word based on the task request, wherein the first interface includes a first parameter, and a parameter value of the first parameter includes the task request; Inputting the prompt word into an intelligent decision model, using the intelligent decision model to perform decision reasoning on the prompt word, and obtaining an initial processing result and a call decision result of the task request, wherein the call decision result is used to indicate whether the task request needs to be processed by a call tool; Processing the task request based on the initial processing result and the calling decision result to obtain a target processing result; The target processing result is output by calling a second interface, wherein the second interface includes a second parameter, and a parameter value of the second parameter includes the target processing result.

11. A computing device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 10 when running.

12. An electronic device, characterized in that: include: A memory storing an executable program; A processor, connected to the memory via a bus, and configured to run the program, wherein the program executes the method described in any one of claims 1 to 10 when running.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 10.

14. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 10.

Images