Task result generation method and device based on multiple agents, equipment and medium

Abstract

The embodiment of the invention discloses a task result generation method and device based on multiple agents, equipment and a medium. A specific embodiment of the method comprises the steps of obtaining retail execution task information input by a target user; inputting the retail execution task information into a first agent for executing data processing to obtain a processed task data set; inputting the processed task data set and the retail execution task information into a second agent used for determining a task execution model to obtain task execution model information; inputting the task execution model information, the processed task data set and the retail execution task information into a third agent used for generating a retail task execution result to obtain the retail task execution result; and displaying the retail task execution result and the retail execution task information on an interactive display terminal corresponding to the multiple agents. According to the embodiment, the task execution result corresponding to the retail execution task information can be accurately and efficiently displayed through cooperative work among multiple agents.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and in particular to a method, apparatus, device and medium for generating task results based on multiple agents. Background Art

[0002] At present, with the continuous development of large language models and multi-agent technology, the application of multi-agent technology to various industries has become one of the main development trends. For the relevant execution of retail business, the method usually adopted is: first, the data set corresponding to the retail business is manually preprocessed to obtain a preprocessed data set. Then, the preprocessed data set is manually input into the business model related to the retail business to generate a business processing result. Among them, the business model is manually selected.

[0003] However, when the above method is adopted, the following technical problems often occur:

[0004] The generation efficiency of business processing results is low, resulting in lags in the interactive display terminal's display of retail task execution results, wasting a large amount of terminal display resources.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0006] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0007] Some embodiments of the present disclosure propose a multi-agent-based task result generation method, apparatus, device and medium to solve one or more of the technical problems mentioned in the above background technology section.

[0008] In a first aspect, some embodiments of the present disclosure provide a method for generating task results based on a multi-agent, comprising: obtaining retail execution task information input by a target user; inputting the retail execution task information into a first agent included in the multi-agent for performing data processing, so that the first agent can call a first large language model to perform data preprocessing of a target task data set to obtain a processed task data set, wherein the target task data set is a task data set corresponding to the retail execution task information; in response to determining to generate the processed task data set, inputting the processed task data set and the retail execution task information into a second agent included in the multi-agent for determining a task execution model. , so that the second agent calls the second largest language model to execute the execution model corresponding to the retail execution task information to obtain the task execution model information; in response to determining to generate the task execution model information, the task execution model information, the processed task data set and the retail execution task information are input into the third agent included in the multi-agent for generating the retail task execution result, so that the third agent inputs the processed task data set and the retail execution task information into the task execution model corresponding to the task execution model information to obtain the retail task execution result; the retail task execution result and the retail execution task information are displayed on the interactive display terminal corresponding to the multi-agent.

[0009] In a second aspect, some embodiments of the present disclosure provide a task result generation device based on a multi-agent, comprising: an acquisition unit, configured to acquire retail execution task information input by a target user; a first input unit, configured to input the above-mentioned retail execution task information into a first agent for performing data processing included in the above-mentioned multi-agent, so that the above-mentioned first agent calls a first large language model to perform data preprocessing of a target task data set to obtain a processed task data set, wherein the target task data set is a task data set corresponding to the above-mentioned retail execution task information; a second input unit, configured to, in response to determining to generate the above-mentioned processed task data set, input the above-mentioned processed task data set and the above-mentioned retail execution task information into a first agent for determining a task execution model included in the above-mentioned multi-agent. a second agent, for the second agent to call the second largest language model to execute the execution model corresponding to the retail execution task information to obtain the task execution model information; a third input unit, configured to, in response to determining to generate the task execution model information, input the task execution model information, the processed task data set and the retail execution task information to the third agent included in the multi-agent for generating the retail task execution result, so that the third agent can input the processed task data set and the retail execution task information to the task execution model corresponding to the task execution model information to obtain the retail task execution result; a display unit, configured to display the retail task execution result and the retail execution task information on the interactive display terminal corresponding to the multi-agent.

[0010] In a third aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner in the first aspect.

[0011] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any implementation manner in the first aspect is implemented.

[0012] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: through the multi-agent-based task result generation method of some embodiments of the present disclosure, through the collaborative work between multiple agents, the task execution results corresponding to the retail execution task information can be accurately and efficiently displayed. Specifically, the reason why the relevant task execution results are not accurate and efficient is that the generation efficiency of the business processing results is low, resulting in the freeze of the interactive display terminal displaying the retail task execution results, wasting a large number of terminal display resources. Based on this, the multi-agent-based task result generation method of some embodiments of the present disclosure first obtains the retail execution task information input by the target user to determine the retail task to be executed. Then, the above-mentioned retail execution task information is input into the first agent for performing data processing included in the above-mentioned multi-agent, so that the above-mentioned first agent calls the first large language model to perform data preprocessing of the target task data set, and obtains the processed task data set, wherein the target task data set is the task data set corresponding to the above-mentioned retail execution task information. Here, by the first agent calling the first large language model to automatically preprocess the data, the efficiency and speed of the data preprocessing can be greatly improved while ensuring the accuracy of the data preprocessing. Next, in response to determining to generate the above-mentioned processed task data set, the above-mentioned processed task data set and the above-mentioned retail execution task information are input into the second agent included in the above-mentioned multi-agent for determining the task execution model, so that the above-mentioned second agent calls the second largest language model to execute the execution model corresponding to the above-mentioned retail execution task information, and obtains the task execution model information. Here, by the second agent calling the second largest language model, the determination of the execution model can be accurately and efficiently realized. Further, in response to determining to generate the above-mentioned task execution model information, the above-mentioned task execution model information, the above-mentioned processed task data set and the above-mentioned retail execution task information are input into the third agent included in the multi-agent for generating the retail task execution result, so that the above-mentioned third agent inputs the above-mentioned processed task data set and the above-mentioned retail execution task information into the task execution model corresponding to the above-mentioned task execution model information, and obtains the retail task execution result. Here, on the basis of the third agent manipulating the third largest language model, the generation of the retail task execution result can be accurately and efficiently realized. Finally, the above-mentioned retail task execution result and the above-mentioned retail execution task information are displayed on the interactive display terminal corresponding to the above-mentioned multi-agent, so as to avoid the problem of display jamming in the interactive display terminal caused by the high generation time. In summary, by coordinating the execution among multiple agents, each task step corresponding to the retail execution task can be executed by each agent, which can greatly improve the execution efficiency and accuracy of the retail execution task. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0014] Figure 1 is a flowchart of some embodiments of the multi-agent-based task result generation method according to the present disclosure;

[0015] Figure 2 is a schematic diagram of the structure of some embodiments of the multi-agent-based task result generation device according to the present disclosure;

[0016] Figure 3 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0017] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0018] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0019] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0020] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0021] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0022] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0023] refer to Figure 1, shows a process 100 of some embodiments of the multi-agent based task result generation method according to the present disclosure. The multi-agent based task result generation method comprises the following steps:

[0024] Step 101: Obtain retail execution task information input by a target user.

[0025] In some embodiments, the execution subject of the above-mentioned multi-agent-based task result generation method can obtain the retail execution task information input by the target user through a wired connection or a wireless connection. Among them, the target user can be a user who requests to view the task execution results corresponding to the retail execution task. That is, the retail execution task information is determined and input by the target user. In practice, the retail execution task information can be the task information corresponding to the retail execution task. For example, the task information can be task description information or task identification. The retail execution task can be a task to be executed initiated by the target user related to the retail field. For example, the retail execution task can be a retail item quantity prediction task, or it can be a retail drone cargo replenishment task. Multiple agents can work together to complete the target task.

[0026] For another example, the corresponding retail execution task information can be customer analysis with high refund rates. That is, the corresponding multi-agent can use the observation and analysis capabilities of the big model to understand the user's problem, and then filter out the fields related to the refund amount, refund time, refunder, refund reason, etc., and analyze and process the missing values ​​and abnormal values ​​of the fields, identify and eliminate duplicate data, and identify and correct abnormal data.

[0027] Step 102, input the above-mentioned retail execution task information into the first agent for performing data processing included in the above-mentioned multi-agents, so that the above-mentioned first agent can call the first large language model to perform data preprocessing of the target task data set to obtain a processed task data set.

[0028] In some embodiments, the execution subject may input the retail execution task information into the first agent for performing data processing included in the multi-agent, so that the first agent may call the first large language model to perform data preprocessing of the target task data set to obtain the processed task data set. The first agent for performing data processing may be an agent for performing data preprocessing tasks. Data preprocessing may include but is not limited to at least one of the following: data normalization, data addition, and data missing supplementation. The target task data set is the task data set corresponding to the retail execution task information. That is, based on the task data set, the task execution result corresponding to the retail execution task is determined. The first large language model, the second large language model, and the third large language model may be the same model or may not be the same model. It may be specifically arranged according to the deployment model arrangement of the multi-agent. In practice, the first large language model may be an LLM (Large Language Model) model. For example, the first large language model may be a GPT-4 model.

[0029] In some optional implementations of some embodiments, the processed task data set is generated by the following steps:

[0030] The first step is to generate a query statement for field-related information and a task data query statement corresponding to the above retail execution task information. The query statement for field-related information can be a database statement for querying data fields related to the retail execution task. The field content corresponding to the data field affects the execution result of the task corresponding to the retail execution task. The task data query statement can be a database statement for querying task data sets related to the retail execution task.

[0031] The second step is to call the task database to query the first task data field set and the first field requirement information set corresponding to the above-mentioned field related information query statement, and the task data set corresponding to the above-mentioned task data query statement. Among them, the task database can be a database that stores task data sets related to various tasks. The task data set can be used to determine the task execution result of the corresponding task. There is a one-to-one correspondence between the first task data field in the first task data field set and the first field requirement information in the first field requirement information set. The first field requirement information can be a requirement for content preprocessing of the corresponding task data field content. For example, the first field requirement information can be that the field format of the task data field content is the target format. Here, each task data has at least one corresponding data field content. Each data field corresponds to a data feature of the task data.

[0032] The third step is to generate a field instruction representing the fields and field requirements corresponding to the retail execution task information. The field instruction may be an instruction for determining the applicable fields corresponding to the retail execution task and the field requirements corresponding to the applicable fields.

[0033] The fourth step is to input the field instructions into the first large language model to generate a second task data field set and a second field requirement information set.

[0034] The fifth step is to generate a target task data field set and a target field requirement information set based on the first task data field set, the second task data field set, the first field requirement information set and the second field requirement information set.

[0035] As an example, the execution subject may perform a task data field union of the first task data field set and the second task data field set to generate a target task data field set. The execution subject may perform a task data field union of the first field requirement information set and the second field requirement information set to generate a target field requirement information set.

[0036] As an example, the execution subject may intersect the first task data field set with the second task data field set to generate a target task data field set. The execution subject may intersect the first field requirement information set with the second field requirement information set to generate a target field requirement information set.

[0037] In the sixth step, the first language model is used to adjust the target task data fields in the task data set according to the target field requirement information set to generate a processed task data set.

[0038] Step 103, in response to determining to generate the above-mentioned processed task data set, input the above-mentioned processed task data set and the above-mentioned retail execution task information into the second agent included in the above-mentioned multi-agent for determining the task execution model, so that the above-mentioned second agent can call the second largest language model to execute the execution model corresponding to the above-mentioned retail execution task information to obtain the task execution model information.

[0039] In some embodiments, in response to determining to generate the above-mentioned processed task data set, the above-mentioned execution subject can input the above-mentioned processed task data set and the above-mentioned retail execution task information into the second agent included in the above-mentioned multi-agent for determining the task execution model, so that the above-mentioned second agent can call the second largest language model to execute the execution model corresponding to the above-mentioned retail execution task information, and obtain the task execution model information. The second agent is called to execute the subsequent task after the first agent completes the execution of the corresponding task. The task execution model can be a model that generates the corresponding task execution result according to the task data set. In practice, the task execution model can be a related task execution rule or a deep learning neural network model. The task execution model information can be the model information corresponding to the task execution model. Among them, the above-mentioned task execution model information can include but is not limited to at least one of the following: task execution model identification, task execution model parameters, and model description information of the task execution model. For example, for the retail execution task of predicting the refund rate of retail stores, the corresponding task execution model can be a Linear Regression linear regression method.

[0040] Here, the big language model is used to correspond to the ability to write programs. The big language model will select an algorithm that is more suitable for the task. For example, if a user asks a question about predicting the refund rate, the big language model will understand that this is a regression problem and will choose a regression algorithm. If a user asks to analyze whether a customer refund is benign or malignant, the big language model will understand that this is a classification problem and will choose a classification algorithm. If the user clearly states that the business requirement is to calculate only some rules, such as calculating the refund rate for this month, the big language model will also directly write a mathematical calculation program. This saves the work of manually trying to select algorithms, testing algorithms, and writing programs.

[0041] In some optional implementations of some embodiments, the task execution model information is generated by the following steps:

[0042] The first step is to generate a model information generation instruction representing the generation of a table-like model information for an execution model suitable for a retail execution task. The table-like model information may represent the display of various execution models suitable for a retail execution task in a table. The model information generation instruction may be an instruction to generate the table-like model information.

[0043] The second step is to input the model information generation instruction into the second language model to generate a first execution model information table, wherein each row in the first execution information table represents the model information of an execution model.

[0044] In a third step, the first execution model information table is displayed on the interactive display terminal, so that the target user can select corresponding execution model information on the interactive display terminal.

[0045] In a fourth step, in response to determining that the target user has selected the execution model information, the selected execution model information is determined as the task execution model information.

[0046] Step 5: In response to determining that the target user has not selected the execution model information, receive the execution model requirement information input by the target user in the model information customization area corresponding to the interactive display terminal. The model information customization area may be an area that supports the user to input requirements for various aspects of the model. For example, the execution model requirement information may include: model calculation amount requirements, model running time, model size, and model accuracy.

[0047] Step 6: Generate a second execution model information table using the second language model according to the execution model requirement information and the retail execution task information.

[0048] As an example, first, the execution subject may generate a model generation instruction for representing model information in a tabular form for generating an execution model applicable to a retail execution task and satisfying the corresponding requirements of the execution model requirement information. Then, the model generation instruction is input into the second language model to generate a second execution model information table.

[0049] In the seventh step, the second execution model information table is displayed on the interactive display terminal, so that the target user can select the corresponding execution model information on the interactive display terminal to obtain the task execution model information.

[0050] Step 104, in response to determining to generate the above-mentioned task execution model information, input the above-mentioned task execution model information, the above-mentioned processed task data set and the above-mentioned retail execution task information into the third agent included in the multi-agent for generating the retail task execution result, so that the above-mentioned third agent inputs the above-mentioned processed task data set and the above-mentioned retail execution task information into the task execution model corresponding to the above-mentioned task execution model information to obtain the retail task execution result.

[0051] In some embodiments, in response to determining to generate the above-mentioned task execution model information, the above-mentioned post-autonomous subject may input the above-mentioned task execution model information, the above-mentioned processed task data set and the above-mentioned retail execution task information into a third agent included in the multi-agent for generating a retail task execution result, so that the above-mentioned third agent inputs the above-mentioned processed task data set and the above-mentioned retail execution task information into the task execution model corresponding to the above-mentioned task execution model information to obtain the retail task execution result. Among them, the retail task execution result may be the task execution result corresponding to the retail task. In practice, for the retail execution task being a retail item unit quantity prediction task, the corresponding retail task execution result may be the predicted unit quantity of each retail item.

[0052] In some optional implementations of some embodiments, the above retail task execution result is generated by the following steps:

[0053] The first step is to obtain the task execution model parameter information and the task execution model structure corresponding to the above-mentioned task execution model information from the model database. Among them, the model database can be a database that stores model data related to each model. The task execution model parameter information can be each model parameter corresponding to the task execution model. For example, the task execution model is a deep learning model, and the corresponding task execution model parameter information can be each trained parameter information in the deep learning model. The task execution model structure can be the model structure of the task execution model. For example, the task execution model is a deep learning model, and the corresponding task execution model structure can be the model network structure of the deep learning model.

[0054] The second step is to build a task execution model based on the above task execution model parameter information and task execution model structure.

[0055] As an example, the task execution model parameter information and the task execution model structure may be correspondingly combined to generate a task execution model.

[0056] The third step is to input the task execution model information and the processed task data set into the task execution model to generate the retail task execution result.

[0057] Step 105, displaying the retail task execution result and the retail execution task information on the interactive display terminal corresponding to the multi-agent.

[0058] In some embodiments, the execution subject may display the retail task execution result and the retail task execution information on the interactive display terminal corresponding to the multi-agent. The interactive display terminal may be a terminal that supports related behavior interaction with the target user and displays the corresponding interaction result.

[0059] In some optional implementations of some embodiments, after step 105, the steps further include:

[0060] The first step is to determine the goods replenishment information corresponding to the above retail task execution result in response to determining that the above retail execution task information is the goods consumption prediction task information for the unmanned retail device. Among them, the unmanned retail device can be an unattended retail shelf that supports automatic pickup through payment. The goods consumption prediction task information can be task information that predicts the consumption of each placed goods in the unmanned retail device. The corresponding retail task execution result can be the estimated consumption of each placed goods on that day. The goods replenishment information can be determined based on the estimated consumption of each placed goods. That is, goods with high consumption are replenished more, and goods with low consumption are replenished less.

[0061] The second step is to control the target retail robot to perform the goods replenishment process corresponding to the above goods replenishment information. The target retail robot may be an automatic robot that supports automatic goods replenishment.

[0062] In some optional implementations of some embodiments, after step 105, the steps further include:

[0063] In the first step, in response to obtaining the actual retail task execution result for the above retail execution task information, a training sample between the above actual retail task execution result and the retail task execution result is generated. The actual retail task execution result may be the actual execution result corresponding to the retail execution task. For example, if the retail task is a goods consumption prediction task, the corresponding actual retail task execution result may be the actual consumption of each good after real statistics.

[0064] As an example, the execution entity may combine the actual retail task execution results, the business data set, the retail execution task information and the retail task execution results to generate combined information as a training sample.

[0065] The second step is to store the training samples in a multi-agent training sample library, where the multi-agent training sample library can be a training sample set used to store each large language model corresponding to the subsequent multi-agent for model training.

[0066] In the third step, in response to determining that the multi-agent training time has arrived, a training sample set under the retail execution task is obtained from the multi-agent training sample library as the retail training sample set. The multi-agent training time may be a time for determining the model training of each large language model corresponding to the multi-agent. The multi-agent training time may be determined periodically, that is, each large language model is periodically trained.

[0067] The fourth step is to input the above-mentioned retail training sample set into the fourth agent included in the above-mentioned multi-agent for determining the model training model, so that the above-mentioned fourth agent can call the fourth largest language model to perform model training of the first largest language model, the second largest language model and the third largest language model, and obtain the first largest language model after training, the second largest language model after training and the third largest language model after training.

[0068] As an example, first, a training instruction (prompt) is generated to represent the model training of the first language model, the second language model, and the third language model according to the retail training sample set. Then, the training instruction is input into the fourth language model to generate the first language model after training, the second language model after training, and the third language model after training.

[0069] The fifth step is to replace the first largest language model in the first agent with the first largest language model after training, replace the second largest language model in the second agent with the second largest language model after training, and replace the third largest language model in the third agent with the third largest language model after training.

[0070] Optionally, calling the fourth largest language model to perform model training of the first largest language model, the second largest language model, and the third largest language model to obtain the trained first largest language model, the trained second largest language model, and the trained third largest language model includes:

[0071] In the first step, for each retail training sample in the retail training sample set, the following first generation step is performed:

[0072] Sub-step 1: acquiring a target task data set corresponding to the retail training sample from a target database as a task data set to be displayed.

[0073] Sub-step 2: displaying the task data set to be displayed on the target data processing platform, so that the target data processing personnel can perform data processing on the task data set to be displayed on the target data processing platform to obtain a data set label.

[0074] Sub-step 3, generating data processing samples for retail execution task information, target task dataset and dataset labels.

[0075] The second step is to select first model training parameter information for the first language model on the target model training platform, wherein the first model training parameter information may include: model training method and model loss function.

[0076] The third step is to generate first training prompt information representing model training of the first large language model based on the first model training parameter information and the data processing sample set.

[0077] In the fourth step, the first training prompt information is input into the fourth largest language model to perform model training on the first largest language model to obtain the trained first largest language model.

[0078] Step 5: For each retail training sample in the retail training sample set, perform the following second generation step:

[0079] Sub-step 1: determining the task execution model information corresponding to the retail training sample.

[0080] Sub-step 2, determining the target number of task execution model information with the best execution model effect for the retail execution task information within the target historical time period as the task execution model information label set. The execution model effect can be determined based on the evaluation index output by each model. The model structures of each task execution model corresponding to the task execution model information label set are different.

[0081] Sub-step 3, generating a task model execution sample for the retail execution task information, the task execution model information, the task execution model information label set and the data processing sample set.

[0082] Step 6: Select the second model training parameter information for the second largest language model on the target model training platform. The second model training parameter information may include: model training method and model loss function.

[0083] The seventh step is to generate second training prompt information representing the model training of the second largest language model based on the second model training parameter information and the task model execution sample set.

[0084] In the eighth step, the second training prompt information is input into the fourth largest language model to perform model training on the second largest language model to obtain the trained second largest language model.

[0085] In the ninth step, the third model training parameter information for the third language model is selected on the target model training platform. The third model training parameter information may include: model training method and model loss function.

[0086] The tenth step is to generate third training prompt information representing model training of the third language model based on the retail execution task information, the task model execution sample set, the data processing sample set and the retail training sample set.

[0087] In the eleventh step, the third training prompt information is input into the fourth language model to perform model training on the third language model to obtain the trained third language model.

[0088] Optionally, as one of the invention points, it solves how to use a large language model to implement the adaptive training of each module in the model training process. Therefore, the present disclosure implements the accuracy and relevance model training of the first large language model, the second large language model and the third large language model by generating a task model execution sample set, a data processing sample set and a retail training sample set.

[0089] In some optional implementations of some embodiments, after step 105, the steps further include:

[0090] The first step is to sort the information corresponding to the data preprocessing of the target task data set by the first agent calling the first language model, the confirmation information of the execution model corresponding to the retail execution task information, the input information of the task execution model corresponding to the task execution model information input by the third agent into the processed task data set and the retail execution task information, the data set link corresponding to the processed task data set, the task execution model information and the retail task execution result to obtain an information sequence in response to receiving the target user selecting the retail task execution process display result at the interactive display terminal. The data set link can be a link to jump to the data table corresponding to the processed task data set. The information in the information sequence is sorted from early to late according to the information generation time.

[0091] The second step is to input the information sequence and the execution flowchart generation information into the fifth agent for generating the execution flowchart included in the multi-agent, so that the fifth agent can generate the execution flowchart corresponding to the information sequence. The execution flowchart can be a flowchart in the form of a directed acyclic graph, which represents the execution order of each information in the information sequence.

[0092] The third step is to display the above execution flow chart on the above interactive display terminal.

[0093] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: through the multi-agent-based task result generation method of some embodiments of the present disclosure, through the collaborative work between multiple agents, the task execution results corresponding to the retail execution task information can be accurately and efficiently displayed. Specifically, the reason why the relevant task execution results are not accurate and efficient is that the generation efficiency of the business processing results is low, resulting in the freeze of the interactive display terminal displaying the retail task execution results, wasting a large number of terminal display resources. Based on this, the multi-agent-based task result generation method of some embodiments of the present disclosure first obtains the retail execution task information input by the target user to determine the retail task to be executed. Then, the above-mentioned retail execution task information is input into the first agent for performing data processing included in the above-mentioned multi-agent, so that the above-mentioned first agent calls the first large language model to perform data preprocessing of the target task data set, and obtains the processed task data set, wherein the target task data set is the task data set corresponding to the above-mentioned retail execution task information. Here, by the first agent calling the first large language model to automatically preprocess the data, the efficiency and speed of the data preprocessing can be greatly improved while ensuring the accuracy of the data preprocessing. Next, in response to determining to generate the above-mentioned processed task data set, the above-mentioned processed task data set and the above-mentioned retail execution task information are input into the second agent included in the above-mentioned multi-agent for determining the task execution model, so that the above-mentioned second agent calls the second largest language model to execute the execution model corresponding to the above-mentioned retail execution task information, and obtains the task execution model information. Here, by the second agent calling the second largest language model, the determination of the execution model can be accurately and efficiently realized. Further, in response to determining to generate the above-mentioned task execution model information, the above-mentioned task execution model information, the above-mentioned processed task data set and the above-mentioned retail execution task information are input into the third agent included in the multi-agent for generating the retail task execution result, so that the above-mentioned third agent inputs the above-mentioned processed task data set and the above-mentioned retail execution task information into the task execution model corresponding to the above-mentioned task execution model information, and obtains the retail task execution result. Here, on the basis of the third agent manipulating the third largest language model, the generation of the retail task execution result can be accurately and efficiently realized. Finally, the above-mentioned retail task execution result and the above-mentioned retail execution task information are displayed on the interactive display terminal corresponding to the above-mentioned multi-agent, so as to avoid the problem of display jamming in the interactive display terminal caused by the high generation time. In summary, by coordinating the execution among multiple agents, each task step corresponding to the retail execution task can be executed by each agent, which can greatly improve the execution efficiency and accuracy of the retail execution task.

[0094] Further references Figure 2As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a multi-agent task result generation device, which are similar to Figure 1 Corresponding to the method embodiments shown, the multi-agent-based task result generation device can be specifically applied to various electronic devices.

[0095] like Figure 2 As shown, a task result generating device 200 based on a multi-agent includes: an acquisition unit 201, a first input unit 202, a second input unit 203, a third input unit 204 and a display unit 205. The acquisition unit 201 is configured to acquire the retail execution task information input by the target user; the first input unit 202 is configured to input the above-mentioned retail execution task information to the first agent for performing data processing included in the above-mentioned multi-agent, so that the above-mentioned first agent calls the first large language model to perform data preprocessing of the target task data set to obtain a processed task data set, wherein the target task data set is the task data set corresponding to the above-mentioned retail execution task information; the second input unit 203 is configured to, in response to determining to generate the above-mentioned processed task data set, input the above-mentioned processed task data set and the above-mentioned retail execution task information to the second agent for determining the task execution model included in the above-mentioned multi-agent, so that the above-mentioned second agent calls the first large language model to perform data preprocessing of the target task data set to obtain a processed task data set, wherein the target task data set is the task data set corresponding to the above-mentioned retail execution task information; the second input unit 203 is configured to, in response to determining to generate the above-mentioned processed task data set, input the above-mentioned processed task data set and the above-mentioned retail execution task information to the second agent for determining the task execution model included in the above-mentioned multi-agent, so that the second agent calls The execution model corresponding to the above-mentioned retail execution task information is executed with the second largest language model to obtain the task execution model information; the third input unit 204 is configured to, in response to determining to generate the above-mentioned task execution model information, input the above-mentioned task execution model information, the above-mentioned processed task data set and the above-mentioned retail execution task information into the third agent included in the multi-agent for generating the retail task execution result, so that the above-mentioned third agent inputs the above-mentioned processed task data set and the above-mentioned retail execution task information into the task execution model corresponding to the above-mentioned task execution model information to obtain the retail task execution result; the display unit 205 is configured to display the above-mentioned retail task execution result and the above-mentioned retail execution task information on the interactive display terminal corresponding to the above-mentioned multi-agent.

[0096] It can be understood that the units recorded in the multi-agent task result generation device 200 are similar to the reference Figure 1 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the multi-agent-based task result generation device 200 and the units contained therein, and will not be repeated here.

[0097] Reference below Figure 3 , which shows a structural schematic diagram of an electronic device (eg, an electronic device) 300 suitable for implementing some embodiments of the present disclosure. Figure 3The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0098] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0099] Typically, the following devices may be connected to the I / O interface 305: input devices 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 308 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 309. The communication devices 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0100] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 309, or installed from the storage device 308, or installed from the ROM 302. When the computer program is executed by the processing device 301, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0101] It should be noted that the computer-readable medium in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0102] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0103] The computer-readable medium may be included in the electronic device, or may exist independently without being installed in the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: obtains the retail execution task information input by the target user;

[0104] The retail execution task information is input into the first agent for performing data processing included in the multi-agent, so that the first agent calls the first large language model to perform data preprocessing of the target task data set to obtain a processed task data set, wherein the target task data set is the task data set corresponding to the retail execution task information; in response to determining to generate the processed task data set, the processed task data set and the retail execution task information are input into the second agent for determining the task execution model included in the multi-agent, so that the second agent calls the second large language model to execute the execution model corresponding to the retail execution task information to obtain the task execution model information; in response to determining to generate the task execution model information, the task execution model information, the processed task data set and the retail execution task information are input into the third agent for generating the retail task execution result included in the multi-agent, so that the third agent inputs the processed task data set and the retail execution task information into the task execution model corresponding to the task execution model information to obtain the retail task execution result; the retail task execution result and the retail execution task information are displayed on the interactive display terminal corresponding to the multi-agent.

[0105] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0106] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0107] The units described in some embodiments of the present disclosure may be implemented by software or hardware. The units described may also be provided in a processor, for example, may be described as: a processor includes an acquisition unit, a first input unit, a second input unit, a third input unit and a display unit. The names of these units do not, in some cases, constitute limitations on the units themselves, for example, the acquisition unit may also be described as a "unit for acquiring retail execution task information input by a target user".

[0108] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0109] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A method for generating task results based on multi-agent, comprising: Obtaining retail execution task information input by target users; Inputting the retail execution task information into a first agent for performing data processing included in the multi-agents, so that the first agent calls a first large language model to perform data preprocessing of a target task data set to obtain a processed task data set, wherein the target task data set is a task data set corresponding to the retail execution task information; In response to determining to generate the processed task data set, inputting the processed task data set and the retail execution task information into a second agent for determining a task execution model included in the multi-agent, so that the second agent calls the second language model to execute the execution model corresponding to the retail execution task information, and obtains the task execution model information; In response to determining to generate the task execution model information, inputting the task execution model information, the processed task data set and the retail execution task information to a third agent included in the multi-agent for generating a retail task execution result, so that the third agent inputs the processed task data set and the retail execution task information into a task execution model corresponding to the task execution model information to obtain a retail task execution result; The retail task execution result and the retail execution task information are displayed on the interactive display terminal corresponding to the multi-agent.

2. The method according to claim 1, wherein: The method further comprises: In response to determining that the retail execution task information is goods consumption prediction task information for an unmanned retail device, determining goods replenishment information corresponding to the retail task execution result; The target retail robot is controlled to execute the goods replenishment processing corresponding to the goods replenishment information.

3. The method according to claim 1, wherein: The method further comprises: In response to obtaining an actual retail task execution result for the retail execution task information, generating a training sample between the actual retail task execution result and the retail task execution result; Storing the training samples in a multi-agent training sample library; In response to determining that the multi-agent training time has arrived, obtaining a training sample set under the retail execution task from the multi-agent training sample library as a retail training sample set; Inputting the retail training sample set to a fourth agent included in the multi-agent for determining a model training model, so that the fourth agent calls a fourth language model to perform model training of the first language model, the second language model and the third language model, and obtains the trained first language model, the trained second language model and the trained third language model; The first largest language model in the first agent is replaced with the first largest language model after training, the second largest language model in the second agent is replaced with the second largest language model after training, and the third largest language model in the third agent is replaced with the third largest language model after training.

4. The method according to claim 1, wherein: The method further comprises: In response to receiving that the target user has selected a retail task execution process display result at the interactive display terminal, information is sorted to obtain an information sequence by performing processing information corresponding to data preprocessing of a target task data set by calling a first large language model, confirmation information corresponding to an execution model corresponding to the retail execution task information by calling a second large language model, input information corresponding to a task execution model inputted by a third agent into the processed task data set and the retail execution task information, a data set link corresponding to the processed task data set, task execution model information, and the retail task execution result; Inputting the information sequence and the execution flowchart generation information to a fifth agent for generating an execution flowchart included in the multi-agents, so that the fifth agent generates an execution flowchart corresponding to the information sequence; The execution flowchart is displayed on the interactive display terminal.

5. The method according to claim 1, wherein: The processed task dataset is generated by the following steps: Generate a field-related information query statement and a task data query statement corresponding to the retail execution task information; Calling a task database to query a first task data field set and a first field requirement information set corresponding to the field related information query statement, and a task data set corresponding to the task data query statement; Generate a field instruction representing corresponding fields and field requirements for determining retail execution task information; Inputting the field instructions into the first large language model to generate a second task data field set and a second field requirement information set; Generate a target task data field set and a target field requirement information set according to the first task data field set, the second task data field set, the first field requirement information set and the second field requirement information set; Using the first large language model, field adjustment is performed on each target task data field in the task data set according to the target field requirement information set to generate a processed task data set.

6. The method according to claim 1, wherein: The task execution model information is generated by the following steps: Generate a model information generation instruction representing model information in a tabular form for generating an execution model suitable for a retail execution task; Inputting the model information generation instruction into the second language model to generate a first execution model information table; Displaying the first execution model information table on the interactive display terminal so that the target user can select corresponding execution model information on the interactive display terminal; In response to determining that the target user has selected execution model information, determining the selected execution model information as task execution model information; In response to determining that the target user has not selected the execution model information, receiving the execution model requirement information input by the target user in the model information customization area corresponding to the interactive display terminal; Generate a second execution model information table using the second language model according to the execution model requirement information and the retail execution task information; The second execution model information table is displayed on the interactive display terminal so that the target user can select corresponding execution model information on the interactive display terminal to obtain the task execution model information.

7. The method according to claim 1, wherein: The retail task execution result is generated by the following steps: Acquire task execution model parameter information and task execution model structure corresponding to the task execution model information from a model database; Constructing a task execution model according to the task execution model parameter information and the task execution model structure; The task execution model information and the processed task data set are input into the task execution model to generate the retail task execution result.

8. A multi-agent-based task result generation device, comprising: An acquisition unit, configured to acquire retail execution task information input by a target user; A first input unit is configured to input the retail execution task information into a first agent for performing data processing included in the multi-agents, so that the first agent calls a first large language model to perform data preprocessing of a target task data set to obtain a processed task data set, wherein the target task data set is a task data set corresponding to the retail execution task information; A second input unit is configured to, in response to determining to generate the processed task data set, input the processed task data set and the retail execution task information to a second agent included in the multi-agent for determining a task execution model, so that the second agent calls the second language model to execute the execution model corresponding to the retail execution task information, and obtains the task execution model information; a third input unit, configured to, in response to determining to generate the task execution model information, input the task execution model information, the processed task data set and the retail execution task information to a third agent included in the multi-agent for generating a retail task execution result, so that the third agent inputs the processed task data set and the retail execution task information into the task execution model corresponding to the task execution model information to obtain the retail task execution result; The display unit is configured to display the retail task execution result and the retail execution task information on the interactive display terminal corresponding to the multi-agent.

9. An electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer readable medium having a computer program stored thereon, wherein: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images