Virtual team service processing method and related equipment

Abstract

The invention provides a virtual team service processing method, which comprises the following steps of: inputting a to-be-processed service into a virtual team, and performing service processing iteration on the to-be-processed service through a large language model according to the level sequence of the virtual action roles based on action data of each virtual action role, service processing message bodies of all the virtual action roles are obtained, and each virtual action role corresponds to one service processing message body; and performing service file write-in processing on the service processing message body of each virtual action role to obtain a service file of each virtual person, and when the virtual action role of the lowest level completes the service file write-in processing, obtaining a final processing result of the to-be-processed service. By setting the virtual team to execute the business processing, the business processing of different levels is realized by using the action data of the virtual action role and the semantic understanding ability of the large language model, and the business processing efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular, to a method for processing virtual team services and related devices. Background Art

[0002] In the existing service processing mode, for a service, it usually requires personnel from multiple departments to cooperate to complete. The overall service is split into service modules at different levels, and corresponding employees are scheduled by departments at different levels to process. In such a service processing process, only when the personnel at the upper level are proficient in the service and understand the capabilities of the employees can the service be processed efficiently. Otherwise, the service processing efficiency will decrease instead. At the same time, a large-scale scheduling of employees is required, resulting in high labor and time costs. Therefore, the service processing efficiency of the existing service processing template is not high. Summary of the Invention

[0003] An embodiment of the present invention provides a method for processing virtual team services, aiming to solve the problem of low service processing efficiency of the existing service processing template. By setting up a virtual team to perform service processing, using the action data of virtual action roles and the semantic understanding ability of large language models, service processing at different levels is realized. Service files are generated using the service processing message bodies of different virtual action roles to obtain the final processing result of the service to be processed. By replacing the existing manpower with a virtual team, the work content of each position can be efficiently completed, reducing the labor cost of service processing and improving the service processing efficiency.

[0004] In a first aspect, an embodiment of the present invention provides a method for processing virtual team services. The virtual team includes virtual action roles at different levels, and each virtual action role has action data corresponding to its own level. The virtual team performs service processing in order of level from high to low. The method includes:

[0005] Input the service to be processed into the virtual team. Based on the action data of each virtual action role, the service to be processed is iteratively processed by a large language model in order of the level of the virtual action role from high to low to obtain the service processing message bodies of each virtual action role. Each virtual action role corresponds to a service processing message body;

[0006] Perform service file writing processing on the service processing message bodies of each virtual action role to obtain the service files of each virtual person. When the virtual action role with the lowest level completes the service file writing processing, the final processing result of the service to be processed is obtained.

[0007] Optionally, based on the action data of each of the virtual action roles, the business processing of the business to be processed is iterated through a large language model in the order of the ranks of the virtual action roles, and the business processing message bodies of each of the virtual action roles are obtained, including:

[0008] In the current iteration round, detect whether there is a business processing message body of the current virtual action role;

[0009] If there is a business processing message body of the current virtual action role, then based on the action data of the next virtual action role and the historical business processing message body, the business processing of the business to be processed is performed through a large language model to obtain the business processing message body of the next virtual action role;

[0010] If there is no business processing message body of the current virtual action role, then based on the action data of the current virtual action role and the historical business processing message body, the business processing of the business to be processed is performed through a large language model to obtain the business processing message body of the current virtual action role.

[0011] Optionally, the business processing of the business to be processed is performed through a large language model based on the action data of the current virtual action role and the historical business processing message body to obtain the business processing message body of the current virtual action role, including:

[0012] Determine whether the number of the action data of the current virtual action role is multiple;

[0013] If the number of the action data of the current virtual action role is single, then determine the single action data of the current virtual action role as the target action data of the current virtual action role;

[0014] If the number of the action data of the current virtual action role is multiple, then based on the multiple action data and the historical business processing message body, generate a first prompt text, and the first prompt text includes selection prompts for the multiple action data;

[0015] Input the first prompt text into the large language model to obtain the target action data of the current virtual action role;

[0016] Based on the target action data and the historical business processing message body, the business processing of the business to be processed is performed through a large language model to obtain the business processing message body of the current virtual action role.

[0017] Optionally, the business processing of the business to be processed is performed through a large language model based on the target action data and the historical business processing message body to obtain the business processing message body of the current virtual action role, including:

[0018] If the target action data contains a preset action keyword, generate a second prompt text based on the action keyword and the historical business processing message body;

[0019] If the target action data does not contain a preset action keyword, determine the action category of the target action data;

[0020] Determine the class prompt template of the target action data according to the action category of the target action data, and different action categories correspond to different class prompt templates;

[0021] Generate a second prompt text based on the class prompt template of the target action data and the historical business processing message body;

[0022] Input the second prompt text into the large language model to perform business processing on the business to be processed, and obtain the business processing message body of the current virtual action role.

[0023] Optionally, performing business file writing processing on the business processing message bodies of each virtual action role to obtain the writing files of each virtual person. When the virtual action role with the lowest rank completes the business file writing processing, obtaining the final processing result of the business to be processed, including:

[0024] In the current iteration round, determine whether the rank of the current virtual action role is the lowest rank;

[0025] If the rank of the current virtual action role is not the lowest rank, format the business processing message body of the current virtual action role and write it into a business file;

[0026] If the rank of the current virtual action role is the lowest rank, write the business processing message body of the current virtual action role into a business file to obtain the final processing result of the business to be processed.

[0027] Optionally, the virtual team further includes a virtual verification role, and the method further includes:

[0028] Based on the action data and the business processing message body of the virtual action role, perform role positioning judgment and requirement positioning judgment on the virtual action role through the virtual verification role to obtain the role positioning judgment result and the requirement positioning judgment result of the virtual action role;

[0029] Determine whether the virtual action role passes the inspection according to the role positioning judgment result and the requirement positioning judgment result of the virtual action role.

[0030] Second aspect, embodiments of the present invention further provide a virtual team business processing device. The virtual team includes virtual action roles of different ranks, and each virtual action role has action data corresponding to its own rank. The virtual team processes business in order of rank from high to low. The virtual team business processing device includes:

[0031] A first processing module, configured to input a business to be processed into the virtual team, and based on the action data of each virtual action role, perform business processing iteration on the business to be processed through a large language model in order of rank from high to low of the virtual action roles, to obtain business processing message bodies of each virtual action role, and each virtual action role corresponds to a business processing message body;

[0032] A second processing module, configured to perform business file writing processing on the business processing message bodies of each virtual action role, to obtain business files of each virtual person. When the virtual action role with the lowest rank completes the business file writing processing, the final processing result of the business to be processed is obtained.

[0033] Third aspect, embodiments of the present invention provide a virtual team business processing system, including a request terminal, a large language model, and the virtual team business processing device as described in the embodiments of the present invention. The request terminal generates a business to be processed and inputs it into the virtual team business processing device for business processing. After the virtual business processing device processes the business to be processed through the large language model, the obtained final processing result is returned to the request terminal.

[0034] Fourth aspect, embodiments of the present invention provide an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps in the virtual team business processing method provided by the embodiments of the present invention are implemented.

[0035] Fifth aspect, embodiments of the present invention provide a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the virtual team business processing method provided by the embodiments of the invention are implemented.

[0036] In an embodiment of the present invention, the business to be processed is input into the virtual team. Based on the action data of each virtual action role, the business to be processed is iteratively processed by a large language model in the order of the rank of the virtual action role from high to low, and a business processing message body for each virtual action role is obtained. Each virtual action role corresponds to a business processing message body. The business processing message bodies of each virtual action role are processed for writing business files, and business files for each virtual person are obtained. When the virtual action role with the lowest rank completes the processing of writing the business file, the final processing result of the business to be processed is obtained. By setting up a virtual team to perform business processing, using the action data of virtual action roles and the semantic understanding ability of the large language model, business processing at different ranks is realized. Using the business processing message bodies of different virtual action roles to generate business files, the final processing result of the business to be processed is obtained. By using a virtual team to replace the existing manpower, the work content of each position can be efficiently completed, the labor cost of business processing can be reduced, and the efficiency of business processing can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0038] Figure 1 is an architecture diagram of a virtual team business processing system provided by an embodiment of the present invention;

[0039] Figure 2 is an interaction schematic diagram of a virtual team business processing system provided by an embodiment of the present invention;

[0040] Figure 3 is a structural schematic diagram of an environmental memory provided by an embodiment of the present invention;

[0041] Figure 4 is a schematic diagram of a virtual role performing business processing provided by an embodiment of the present invention;

[0042] Figure 5 is a flowchart of a virtual team business processing method provided by an embodiment of the present invention;

[0043] Figure 6 is a structural schematic diagram of a virtual team business processing device provided by an embodiment of the present invention;

[0044] Figure 7 is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] As Figure 1 shown Figure 1 is an architecture diagram of a virtual team service processing system provided by an embodiment of the present invention. The virtual team service processing system includes: a request terminal, a large language model, and a virtual team service processing device. The request terminal generates a service to be processed and inputs it into the virtual team service processing device for service processing. After the virtual service processing device processes the service to be processed through the large language model, it returns the obtained final processing result to the request terminal.

[0047] Specifically, the above virtual team can be configured according to the team requirements of existing services, and one virtual team can be configured for one service scenario. The virtual team may include virtual action roles and virtual verification roles. Among them, the virtual action roles are configured according to job levels. The virtual action roles can be used to perform service processing corresponding to their job levels. The virtual verification roles are used to verify the accuracy of the virtual action roles and, when the accuracy does not reach the expected accuracy, control the corresponding virtual action roles to re-perform the corresponding service processing.

[0048] The virtual action roles are configured with action data corresponding to their job levels, so that the virtual action roles can perform functions corresponding to their job levels according to the corresponding action data. Thus, by completing the functions of each virtual action role, each virtual role can call the large language model interface to complete the corresponding service processing. One or more prompt templates can be created for the functions of each virtual role. These prompt templates will be used to generate the prompt text that needs to be input into the large language model, so that the large language model can output results related to the prompt text according to the prompt text. After the prompt of each virtual role is input into the large language model, the results returned by the large language model will be saved in the environmental memory. And the next virtual role will write the results of the previous virtual role into its own prompt text to ensure that the large language model can read the context to generate corresponding results.

[0049] The above large language model (LLM) can be large language models that can be commercially used or open-sourced in the market, such as GPT, ERNIE Bot, and Pangu Model. The virtual character can call the large language model API to request the corresponding large language model to output. It can be understood that in the specific implementation of this application, when it comes to large language model-related technologies, when the embodiments in this application are applied to specific products or technologies, the training, use, and processing of large language model-related technologies need to comply with relevant laws, regulations, and standards of relevant countries and regions.

[0050] The requesting terminal can include an interaction interface, and the user can submit the corresponding business to be processed to the virtual team business processing device through the interaction interface. The virtual team processes the business in order of rank from high to low to obtain the final processing result of the business to be processed. Each virtual character calls the large language model interface at least once to process the business.

[0051] Taking software development as an example for illustration, the system needs to first define virtual characters for each rank in the software development team. The virtual characters that can be included are: Team Boss (Boss), Product Manager (ProductManager), Architect (Architect), Project Manager (ProjectManager), Development Engineer (Engineer), Process Verifier (Verifier), and Tester (Testers). The virtual characters in the virtual team can be defined according to the actual team requirements. In the virtual team, the Process Verifier is used to verify the accuracy of the other virtual characters.

[0052] After constructing the virtual characters required for the virtual team, functions can be assigned to each virtual character, emphasizing what each virtual character should do. For example:

[0053] BOSS: Propose a requirement (receive external business to be processed).

[0054] ProductManager: Customize a product according to the BOSS's requirement, list the product description and requirements.

[0055] Architect: Determine the system framework according to the ProductManager's requirements, be responsible for designing the API, designing code files, and determining the programming language for development.

[0056] ProjectManager: Allocate development tasks, improve team efficiency, and deliver with quality and quantity guaranteed.

[0057] Engineer: Write elegant, readable, extensible, and efficient code.

[0058] Testers: Test the code written by the Engineer.

[0059] Please continue to refer to Figure 2 , Figure 2 is an interaction schematic diagram of a virtual team service processing system provided by an embodiment of the present invention. In Figure 2Among them, the virtual action roles include: BOSS, ProductManager, Architect, ProjectManager, Engineer. After receiving the external business to be processed, the BOSS generates a BOSS requirement based on the external business to be processed. This BOSS requirement can be the description text of the business to be processed itself, or the requirement text processed by the large language model. If the BOSS requirement is the description text of the business to be processed itself, the BOSS does not need to request the large language model for processing. If the BOSS requirement is the requirement text processed by the large language model, the BOSS needs to request the large language model to generate the corresponding requirement text according to the business to be processed. Specifically, the BOSS can preset a prompt template, fill in the business to be processed into the BOSS's prompt template, and assemble to obtain the BOSS's prompt text prompt. Then, the BOSS's prompt text is input into the large language model through the large language model API, and the requirement text (business processing message body) of the business to be processed is output through the large language model. After obtaining the BOSS requirement, the BOSS requirement is written into the environmental memory, and the business processing process of the next virtual role, ProductManager, is entered. The ProductManager accesses the environmental memory, obtains the historical business processing message body (BOSS requirement) from the environmental memory, writes the BOSS requirement into the ProductManager's prompt template, and assembles to obtain the ProductManager's prompt text prompt. Then, the ProductManager's prompt text is input into the large language model through the large language model API, and the business processing message body of the ProductManager is output through the large language model. After obtaining the business processing message body of the ProductManager, the business processing message body of the ProductManager is written into the environmental memory, and the business processing message body is formatted and written into the business file. The above formatting process refers to processing the business processing message body according to the business processing rules of the ProductManager to generate the corresponding business file. The business processing process of the next virtual role, Architect, is entered. The Architect accesses the environmental memory, obtains the historical business processing message body (BOSS requirement and the business processing message body of the ProductManager) from the environmental memory, writes the historical business processing message body into the Architect's prompt template, and assembles to obtain the Architect's prompt text prompt. Then, the Architect's prompt text is input into the large language model through the large language model API, and the business processing message body of the Architect is output through the large language model.After obtaining the business processing message body of Architect, write the business processing message body of Architect into the environmental memory, and format the business processing message body and write it into the business file. The above formatting process refers to processing the business processing message body according to the business processing rules of Architect to generate the corresponding business file. Enter the business processing process of the next virtual role, ProjectManager. ProjectManager accesses the environmental memory, obtains the historical business processing message bodies (BOSS requirements, the business processing message body of ProductManager, and the business processing message body of Architect) from the environmental memory, writes the historical business processing message bodies into the prompt template of Architect, assembles to obtain the prompt text prompt of ProjectManager, inputs the prompt text of ProjectManager into the large language model through the large language model API, and outputs the business processing message body of ProjectManager through the large language model. After obtaining the business processing message body of ProjectManager, write the business processing message body of Architect into the environmental memory, and format the business processing message body and write it into the business file. The above formatting process refers to processing the business processing message body according to the business processing rules of ProjectManager to generate the corresponding business file. Enter the business processing process of the next virtual role, Engineer. Engineer accesses the environmental memory, obtains the historical business processing message bodies (BOSS requirements, the business processing message body of ProductManager, the business processing message body of Architect, and the business processing message body of ProjectManager) from the environmental memory, writes the historical business processing message bodies into the prompt template of Engineer to obtain the prompt text prompt of Engineer, inputs the prompt text of Engineer into the large language model through the large language model API, and outputs the business processing message body of Engineer through the large language model. After obtaining the business processing message body of Engineer, write the business processing message body of Engineer into the environmental memory, and format the business processing message body and write it into the business file. The above formatting process refers to processing the business processing message body according to the business processing rules of Engineer to generate the corresponding business file. Thus, the overall processing process for a business to be processed is completed. After Engineer generates the corresponding business file, the final processing result of the business to be processed is obtained.

[0060] It should be noted that the above environmental memory can be a message queue or a message list. In the environmental memory, virtual action roles of different ranks are arranged in the order of rank, or in the order of writing of business processing message bodies. Please refer to Figure 3 , Figure 3 is a schematic structural diagram of an environmental memory provided by an embodiment of the present invention. In Figure 3 , the business processing message bodies are arranged according to the rank.

[0061] Furthermore, please refer to Figure 4 , Figure 4 is a schematic diagram of a virtual role performing business processing provided by an embodiment of the present invention. Taking the Engineer in Figure 2 as an example, as shown in Figure 4 , the specific process includes the following steps:

[0062] S1. Start.

[0063] S2. The BOSS provides the requirement content.

[0064] S3. Initialize the entity classes of virtual roles ProductManager, Architect, ProjectManager, and Engineer and assign Actions.

[0065] S4. Create a company entity.

[0066] S5. Enter the virtual role.

[0067] S6. Insert the business processing message bodies of each virtual role into the environment.

[0068] S7. Fetch messages from the environmental memory.

[0069] S8. Determine whether the last business processing message body in the environmental memory is the business processing message body of the Engineer; if so, go to step S9, if not, go to step S10.

[0070] S9. End.

[0071] S10. Determine whether there is a business processing message body of the current virtual role in the environmental memory. If there is, go to step S7 to perform the business processing process of the next virtual role; if not, go to step S11.

[0072] S11. Obtain the action data of the current virtual role.

[0073] S12. Determine whether there are multiple Action data for the current virtual role; if not, go to step S13; if so, go to step S14.

[0074] S13. Set the action data Action to be executed as the first one and proceed to step S17.

[0075] S14. Combine multiple action data Actions according to the prompt template to obtain the first prompt text prompt, where the first prompt text includes a prompt word for instructing the large language model to select the action data Action.

[0076] S15. Invoke the large language model and input the first prompt text prompt into the large language model.

[0077] S16. Determine the action data Action to be executed according to the value returned by the large language model and proceed to step S17. The corresponding relationship between the pre-set action data Action to be executed and the value returned by the large language model is set in advance.

[0078] S17. Execute the virtual character action data Action method.

[0079] S18. Determine whether there is an action keyword in the current virtual character's action data Action. If not, proceed to step S22; if so, proceed to step S19.

[0080] S19. Search and summarize the prompts according to the template to obtain the keyword prompt text.

[0081] S20. Invoke the large language model API to process the keyword prompt text.

[0082] S21. Return the appropriate prompt word prompt and proceed to step S22.

[0083] S22. Synthesize the prompt word according to the template in the action data Action class to obtain the second prompt text prompt.

[0084] S23. Invoke the large language model API to process the second prompt text prompt to obtain the return content of the large language model.

[0085] S24. Determine whether the current virtual character is an Engineer. If not, proceed to step S25; if so, proceed to step S26.

[0086] S25. Format the business processing message body returned by the large language model and proceed to step S26.

[0087] S26. Write the returned business processing message body into the corresponding business file and proceed to step S6.

[0088] In the embodiments of the present invention, a virtual team is set up to perform business processing. By using the action data of virtual action roles and the semantic understanding ability of large language models, business processing at different levels is achieved. Business files are generated using the business processing message bodies of different virtual action roles, and the final processing result of the business to be processed is obtained. By replacing the existing human resources with a virtual team, the work content of each position can be efficiently completed, the labor cost of business processing can be reduced, and the efficiency of business processing can be improved.

[0089] Please refer to Figure 5 , Figure 5 which is a flowchart of a virtual team business processing method provided by an embodiment of the present invention. As Figure 5 shown, the virtual team business processing method includes the following steps:

[0090] 501. Input the business to be processed into the virtual team. Based on the action data of each virtual action role, perform business processing iteration on the business to be processed through a large language model in the order of the levels of the virtual action roles, and obtain the business processing message bodies of each virtual action role.

[0091] In the embodiments of the present invention, the above virtual team business processing method can be applied to the above virtual team business processing system. The above virtual team may include virtual action roles. Among them, the virtual action roles are configured according to levels, and the virtual action roles can be used to perform business processing corresponding to their levels. The above virtual action roles have corresponding levels, and there may be at least one virtual action role for each level. Or rather, each virtual action role at each level corresponds to at least one action data, and each action data can implement at least one function of that level.

[0092] The above business to be processed can be input into the virtual team by the user through a request terminal. After receiving the external business to be processed, the virtual team can perform business processing iteration in the order of the levels of the virtual action roles. The above business processing iteration can be understood as that the input data used in the current iteration round includes the output data in the previous iteration round. In this way, when each virtual action role performs business processing, the business processing results of the upstream levels will be considered. The business processing result of each virtual action role can also be called a business processing message body, and each virtual action role corresponds to a business processing message body.

[0093] The virtual action role is configured with action data corresponding to the rank, so that the virtual action role can execute the functions of the corresponding rank according to the corresponding action data. Thus, by completing the functions of each virtual action role, each virtual action role can call the large language model interface to complete the corresponding business processing. One or more prompt templates can be created for the functions of each virtual action role, and these prompt templates will be used to generate the prompt text that needs to be input into the large language model, so that the large language model can output results related to the prompt text according to the prompt text. After the prompt of each virtual action role is input into the large language model, the results returned by the large language model will be saved in the environmental memory. And the next virtual action role will write the results of the previous virtual action role into its own prompt text to ensure that the large language model can read the context to generate the corresponding results.

[0094] The above large language model can be a commercially available or open-source large language model such as GPT, Wenxin Yiyan, Pangu Model, etc. The virtual role can call the large language model API to request the output of the corresponding large language model, and then obtain the business processing message body of the virtual action role. Since the business processing message body is processed by the large language model according to the action data of the corresponding virtual action role for the business to be processed, the business message body output by the large language model contains data related to the action data of the virtual action role. For example, if the function of the virtual action role is to generate code according to the project manager's requirements, the large language model will generate a business processing message body containing the code corresponding to the project manager's requirements.

[0095] 502. Perform business file writing processing on the business processing message bodies of each virtual action role to obtain the business files of each virtual person. When the virtual action role with the lowest rank completes the business file writing processing, the final processing result of the business to be processed is obtained.

[0096] In the embodiment of the present invention, since the business processing message body is processed by the large language model according to the action data of the corresponding virtual action role for the business to be processed, the business message body output by the large language model contains data related to the action data of the virtual action role. Therefore, the data related to the action data of the virtual action role in the business message body can be extracted and written into the business file of the virtual person. For example, if the function of the virtual action role is to generate code according to the project manager's requirements, the large language model will generate a business processing message body containing the code corresponding to the project manager's requirements, and the code content is extracted from the business processing message body and written into the corresponding business file to obtain a business file containing the code.

[0097] Since the business process of the virtual team is executed in the order of rank, the business process of a business to be processed starts from the virtual operator with the highest rank and ends with the virtual operator with the lowest rank. After the virtual action role with the lowest rank completes the processing of writing the business document, the final processing result of the business to be processed can be obtained.

[0098] Furthermore, after each virtual action role completes its own business processing, the relevant content in the corresponding business message body can be extracted and written into the corresponding business document. After the virtual action role with the lowest rank completes the processing of writing the business document, the final processing result of the business to be processed can be obtained. It is also possible to write the business message bodies of all virtual action roles into the business document in sequence after all virtual action roles complete their business processing, and obtain the final business document as the final processing result of the business to be processed.

[0099] In the embodiment of the present invention, the business to be processed is input into the virtual team. Based on the action data of each virtual action role, the business to be processed is iteratively processed by the large language model in the order of the rank of the virtual action role, and the business processing message bodies of each virtual action role are obtained. Each virtual action role corresponds to a business processing message body; the business processing message bodies of each virtual action role are processed for writing into the business document, and the business documents of each virtual person are obtained. When the virtual action role with the lowest rank completes the processing of writing the business document, the final processing result of the business to be processed is obtained. By setting up a virtual team to execute business processing, using the action data of virtual action roles and the semantic understanding ability of the large language model, business processing at different ranks is realized. The business document is generated using the business processing message bodies of different virtual action roles, and the final processing result of the business to be processed is obtained. By using the virtual team to replace the existing manpower, the work content of each position can be efficiently completed, the labor cost of business processing can be reduced, and the efficiency of business processing can be improved.

[0100] Optionally, in the step of iteratively processing the business to be processed by the large language model in the order of the rank of the virtual action role based on the action data of each virtual action role to obtain the business processing message bodies of each virtual action role, it is possible to detect whether there is a business processing message body of the current virtual action role in the current iteration round; if there is a business processing message body of the current virtual action role, then based on the action data of the next virtual action role and the historical business processing message body, the business to be processed is processed by the large language model to obtain the business processing message body of the next virtual action role; if there is no business processing message body of the current virtual action role, then based on the action data of the current virtual action role and the historical business processing message body, the business to be processed is processed by the large language model to obtain the business processing message body of the current virtual action role.

[0101] In an embodiment of the present invention, the business processing of each virtual action character corresponds to an iteration round, and the current iteration round corresponds to the current virtual action character. The above-mentioned current iteration round may not include the first-round iteration, or may include the first-round iteration. In the case where the current iteration round does not include the first-round iteration, the virtual action character with the highest rank is only responsible for receiving the business to be processed. In the case where the current iteration round includes the first-round iteration, the virtual action character with the highest rank may receive the business to be processed, and based on the action data of the virtual action character with the highest rank, the business to be processed is subjected to the first-round business processing through a large language model to obtain the first-round business processing message body.

[0102] The virtual action character is configured with the action data corresponding to the rank, so that the virtual action character can perform the functions corresponding to the rank according to the corresponding action data. Thus, by completing the functions of each virtual action character, each virtual action character can call the large language model interface to complete the corresponding business processing. One or more prompt templates can be created for the functions of each virtual action character. These prompt templates will be used to generate the prompt text that needs to be input into the large language model, so that the large language model can output the results related to the prompt text according to the prompt text. After the prompt of each virtual action character is input into the large language model, the results returned by the large language model will be saved in the environmental memory. And the next virtual action character will write the results of the previous virtual action character into its own prompt text to ensure that the large language model can read the context to generate the corresponding results.

[0103] When entering the business processing flow of the current virtual action character, it can be queried in the environmental memory whether there is a business processing message body of the current virtual action character. If there is no business processing message body of the current virtual action character in the environmental memory, it means that the current virtual action character has not completed the corresponding business processing flow.

[0104] Specifically, in the case where there is no business processing message body of the current virtual action role in the environmental memory, the historical business processing message body can be obtained from the environmental memory. The above historical business processing message body is the business processing message body of the virtual action role at the upstream rank of the current virtual action role. Of course, if the current virtual action role is the virtual action role at the highest rank, there is no historical business processing message body. At this time, only the action data and the business to be processed of the virtual action role at the highest rank need to be concatenated into the corresponding prompt text, and the prompt text is input into the large language model. The large language model processes the business to be processed and outputs the business processing message body of the virtual action role at the highest rank. If the current virtual action role is not the virtual action role at the highest rank, after obtaining the historical business processing message body, the historical business processing message body and the action data of the current virtual action role can be assembled into the corresponding prompt text, and the prompt text is input into the large language model. The large language model processes the business to be processed and outputs the business processing message body of the current virtual action role.

[0105] If there is a business processing message body of the current virtual action role in the environmental memory, it means that the current virtual role has completed the corresponding business processing. Therefore, the business processing process of the next virtual action role can be entered. The business processing processes of each virtual action role are the same, and the difference lies in the different action data. Currently, if the current virtual action role is the virtual action role at the lowest rank, the business processing of the virtual team will end.

[0106] By detecting whether there is a business processing message body of the current virtual action role to determine whether the current virtual action role has completed the processing, it is possible to avoid a virtual action role repeating the same business processing process, improve the process simplicity of the virtual team, and thus improve the business processing efficiency of the virtual team.

[0107] Optionally, in the step of performing business processing on the business to be processed through a large language model based on the action data of the current virtual action role and the historical business processing message body to obtain the business processing message body of the current virtual action role, it is possible to determine whether the number of action data of the current virtual action role is multiple; if the number of action data of the current virtual action role is single, then determine the single action data of the current virtual action role as the target action data of the current virtual action role; if the number of action data of the current virtual action role is multiple, then generate a first prompt text based on the multiple action data and the historical business processing message body, where the first prompt text includes selection prompts for the multiple action data; input the first prompt text into the large language model to obtain the target action data of the current virtual action role; and perform business processing on the business to be processed through the large language model based on the target action data and the historical business processing message body to obtain the business processing message body of the current virtual action role.

[0108] In the embodiments of the present invention, there may be at least one virtual action role corresponding to each rank, or in other words, there is at least one action data corresponding to each rank of virtual action role, and each action data can implement at least one function of that rank.

[0109] When the current virtual action role has multiple action data, the large language model can be used to combine the context semantics of the historical business processing message body to select the most appropriate one or more action data from the multiple action data as the target action data. Specifically, a prompt template can be used to prompt the large language model to select the target action data from the multiple action data. The above prompt template can be as follows:

[0110] "You are a XXXX (rank of the current virtual action role), and your action data are:

[0111] 1. XXX.

[0112] 2. XXXXX.

[0113] ……

[0114] N. XXXXXXX.

[0115] Your upstream rank has made the following requirements:

[0116] (Historical business processing message body).

[0117] Please select the most suitable function from your action data according to the requirements made by the upstream rank and only output the serial number of the function. Do not output any other data except the serial number of the function."

[0118] The respective action data of the current virtual action role and the historical business processing message body can be filled into a prompt template to obtain a prompt text. The prompt text is input into a large language model, and the corresponding serial number is output by the large language model. The corresponding action data is obtained according to the serial number as the target action data.

[0119] In a possible embodiment, when there are multiple action data Actions for the current virtual action role, a policy can be set to intervene in the selection result of the large language model. For example, it is specified that the large language model only needs to answer a number between 0 - N (the number of Actions). The most appropriate stage is selected according to the understanding of the historical business processing message body. The answer can be a number without adding any other text, and each number corresponds to an action data Action. If there is no conversation record, 0 can be selected, and the large language model does not need to answer any other questions or add any other information in the answer. A virtual verification role Verifier can be used to verify the returned number. If the returned number is not a number or the number is not within the range of 0 - N, the historical business processing message body is retrieved again to request the large language model. The large language model can be made to iterate internally N times according to the above content, and the result after iteration is used as the target action data.

[0120] After obtaining the target action data, the target action data can be assembled with the historical business processing message body to obtain the corresponding prompt text. The prompt text is input into the large language model so that the large language model processes the business to be processed according to the prompt text, and the business processing message body of the current virtual action role is obtained.

[0121] Specifically, the target action data and the historical business processing message body can be assembled through a prompt template to obtain the corresponding prompt text. The same or different prompt templates can be set for different virtual action roles. The above prompt template can be:

[0122] "You are a XXXX (rank of the current virtual action role), and your target action data is: XXX.

[0123] Your upstream rank has the following requirements:

[0124] (Historical business processing message body).

[0125] Please generate your business processing result and output it according to the requirements of the upstream rank, combining your target action data. The business processing result should meet the requirements of the upstream rank and your target action rules. Other requirements are...".

[0126] Fill the target action data and the historical business processing message body into the prompt model to obtain a prompt text, and input the prompt text into a large language model so that the large language model processes the business to be processed according to the prompt text, and obtain the business processing message body of the current virtual action role.

[0127] Optionally, in the step of processing the business to be processed by the large language model based on the target action data and the historical business processing message body to obtain the business processing message body of the current virtual action role, if the target action data contains a preset action keyword, a second prompt text can be generated based on the action keyword and the historical business processing message body; if the target action data does not contain a preset action keyword, determine the action category of the target action data; determine the class prompt template of the target action data according to the action category of the target action data, and different action categories correspond to different class prompt templates; generate a second prompt text based on the class prompt template of the target action data and the historical business processing message body; input the second prompt text into the large language model to process the business to be processed, and obtain the business processing message body of the current virtual action role.

[0128] In the embodiment of the present invention, after obtaining the target action data, it can be further determined whether the target action data contains a preset action keyword. The above action keyword can be a keyword or prompt word set by a person, and can be specifically set through a serpapi_api_key file. Specifically, it can be detected whether there is a serpapi_api_key file in the target action data. If there is a serpapi_api_key file in the target action data, it means that the target action data contains a preset action keyword. If there is no serpapi_api_key file in the target action data, it means that the target action data does not contain a preset action keyword.

[0129] Further, if there is a serpapi_api_key file in the target action data, keyword prompt text can be obtained according to template search and summary prompts, call the large language model API to process the keyword prompt text, and return a suitable prompt word prompt through the large language model. Generate a second prompt text based on the prompt word prompt and the historical business processing message body.

[0130] If there is no serpapi_api_key file in the target action data, the action category of the target action data can be determined; the class prompt template of the target action data can be determined according to the action category of the target action data, and different action categories correspond to different class prompt templates; synthesize prompt words according to the class prompt template, and generate a second prompt text based on the prompt words synthesized according to the class prompt template and the historical business processing message body.

[0131] After obtaining the second prompt text, input the second prompt text into the large language model so that the large language model processes the business to be processed according to the second prompt text, and obtains the business processing message body of the current virtual action role.

[0132] In the embodiments of the present invention, the second prompt text can be flexibly generated according to preset action keywords.

[0133] Optionally, when performing business file writing processing on the business processing message bodies of each virtual action role to obtain the written files of each virtual person, and when the virtual action role with the lowest rank completes the business file writing processing to obtain the final processing result of the business to be processed, it can be judged in the current iteration round whether the rank of the current virtual action role is the lowest rank; if the rank of the current virtual action role is not the lowest rank, format the business processing message body of the current virtual action role and write it into the business file; if the rank of the current virtual action role is the lowest rank, write the business processing message body of the current virtual action role into the business file to obtain the final processing result of the business to be processed.

[0134] In the embodiments of the present invention, the business process of the virtual team is executed in the order of rank. Therefore, the business process of a business to be processed starts from the virtual action personnel with the highest rank and ends with the virtual action personnel with the lowest rank. After the virtual action role with the lowest rank completes the business file writing processing, the final processing result of the business to be processed can be obtained.

[0135] When entering the business processing process of the current virtual action role, it can be judged whether the current virtual action role is the virtual action role with the lowest rank. If the rank of the current virtual action role is the lowest rank, there is no need to format the business processing message body of the current virtual action role, but directly write the relevant data in the business processing message body into the business file according to the function rules of the current virtual action role, and end the business processing process. If the rank of the current virtual action role is not the lowest rank, continue the iteration of business processing until the rank of the current virtual action role is the lowest rank.

[0136] The above formatting refers to processing the business processing message body according to the preset format of the virtual action role, so that the obtained business processing message body becomes data in the preset format. For example, the formatting of BOSS will process the corresponding business processing message body into the format corresponding to the BOSS requirements. The ProductManager will process the corresponding business message body into the format of customizing a product and listing the product description and requirements. The Architect will process the corresponding business message body into the format of determining the system framework, being responsible for designing the API, designing the code files, and determining the programming language for development, etc. Different virtual action roles correspond to different formatting processes.

[0137] Optionally, the virtual team further includes a virtual verification role. In the virtual team business processing method, based on the action data of the virtual action role and the business processing message body, the virtual verification role can also perform role positioning judgment and requirement positioning judgment on the virtual action role to obtain the role positioning judgment result and requirement positioning judgment result of the virtual action role; according to the role positioning judgment result and requirement positioning judgment result of the virtual action role, it is determined whether the virtual action role passes the verification.

[0138] In the embodiment of the present invention, in the virtual team, there is also a virtual verification role, which can also be called the process verifier Verifier. The virtual verification role is used to verify the accuracy of the virtual action role. The virtual verification role can be used to verify each virtual action role. Specifically, the Verifier role needs to re-verify it. According to the set verification rules, the action data of the virtual action role, the business processing message body, and the verification rules are assembled into a verification prompt text promt, and the verification prompt text promt is provided to the large language model. The result returned by the large language model is compared with the result of the virtual action role. The comparison rules are as follows: if the output is the same business processing message body as the virtual action role to be tested, the role positioning is correct; if the output execution of the position is similar to the execution process of the virtual action role to be tested, the requirement function is correct.

[0139] If the role positioning is correct and the requirement positioning is correct, it is determined that the virtual action role passes the verification. If the role positioning is incorrect and / or the requirement positioning is incorrect, it is determined that the virtual action role fails the verification, and the virtual action role is controlled to re-perform the business processing process to obtain a more accurate business processing message body, thereby making the final business processing result more accurate.

[0140] As Figure 6 shown, the embodiment of the present invention provides a virtual team business processing device, and the virtual team business processing device includes:

[0141] The first processing module 601 is configured to input the business to be processed into the virtual team, and based on the action data of each virtual action role, perform iterative business processing on the business to be processed through a large language model in the order of the ranks of the virtual action roles, so as to obtain the business processing message bodies of each virtual action role, and each virtual action role corresponds to a business processing message body;

[0142] The second processing module 602 is configured to perform business file writing processing on the business processing message bodies of each virtual action role to obtain the business files of each virtual person. When the virtual action role with the lowest rank completes the business file writing processing, the final processing result of the business to be processed is obtained.

[0143] Optionally, the first processing module 501 is further configured to detect whether there is a business processing message body of the current virtual action role in the current iteration round; if there is a business processing message body of the current virtual action role, then based on the action data of the next virtual action role and the historical business processing message body, perform business processing on the business to be processed through a large language model to obtain the business processing message body of the next virtual action role; if there is no business processing message body of the current virtual action role, then based on the action data of the current virtual action role and the historical business processing message body, perform business processing on the business to be processed through a large language model to obtain the business processing message body of the current virtual action role.

[0144] Optionally, the first processing module 501 is further configured to determine whether the number of the action data of the current virtual action role is multiple; if the number of the action data of the current virtual action role is single, then determine the single action data of the current virtual action role as the target action data of the current virtual action role; if the number of the action data of the current virtual action role is multiple, then based on the multiple action data and the historical business processing message body, generate a first prompt text, where the first prompt text includes selection prompts for the multiple action data; input the first prompt text into the large language model to obtain the target action data of the current virtual action role; based on the target action data and the historical business processing message body, perform business processing on the business to be processed through a large language model to obtain the business processing message body of the current virtual action role.

[0145] Optionally, the first processing module 501 is further configured to: if the target action data contains a preset action keyword, generate a second prompt text based on the action keyword and the historical business processing message body; if the target action data does not contain a preset action keyword, determine the action category of the target action data; determine the class prompt template of the target action data according to the action category of the target action data, where different action categories correspond to different class prompt templates; generate a second prompt text based on the class prompt template of the target action data and the historical business processing message body; input the second prompt text into the large language model to perform business processing on the to-be-processed business, and obtain the business processing message body of the current virtual action role.

[0146] Optionally, the second processing module 502 is further configured to: in the current iteration round, determine whether the rank of the current virtual action role is the lowest rank; if the rank of the current virtual action role is not the lowest rank, format the business processing message body of the current virtual action role and write it into a business file; if the rank of the current virtual action role is the lowest rank, write the business processing message body of the current virtual action role into a business file to obtain the final processing result of the to-be-processed business.

[0147] Optionally, the virtual team further includes a virtual verification role, and the apparatus further includes:

[0148] A first verification module, configured to perform role positioning judgment and requirement positioning judgment on the virtual action role through the virtual verification role based on the action data and the business processing message body of the virtual action role, and obtain the role positioning judgment result and the requirement positioning judgment result of the virtual action role;

[0149] A second verification module, configured to determine whether the virtual action role passes the verification according to the role positioning judgment result and the requirement positioning judgment result of the virtual action role.

[0150] The virtual team business processing apparatus provided in the embodiments of the present invention can implement each process implemented by the virtual team business processing method in the above method embodiments, and can achieve the same beneficial effects. To avoid repetition, it will not be elaborated here.

[0151] See Figure 7 , Figure 7 is a schematic structural diagram of an electronic device provided in an embodiment of the present invention. As Figure 7 shown, it includes: a memory 702, a processor 701, and a computer program of the virtual team business processing method stored on the memory 702 and executable on the processor 701, where:

[0152] The processor 701 is used to call a computer program stored in the memory 702 and execute the following steps:

[0153] Input the business to be processed into the virtual team, and based on the action data of each virtual action role, perform business processing iteration on the business to be processed through a large language model in the order of the ranks of the virtual action roles, to obtain business processing message bodies of each virtual action role, and each virtual action role corresponds to a business processing message body;

[0154] Perform business file writing processing on the business processing message bodies of each virtual action role to obtain business files of each virtual person. When the virtual action role with the lowest rank completes the business file writing processing, obtain the final processing result of the business to be processed.

[0155] Optionally, the processor 701 executes the step of performing business processing iteration on the business to be processed through a large language model in the order of the ranks of the virtual action roles based on the action data of each virtual action role, to obtain business processing message bodies of each virtual action role, including:

[0156] In the current iteration round, detect whether there is a business processing message body of the current virtual action role;

[0157] If there is a business processing message body of the current virtual action role, then perform business processing on the business to be processed through a large language model based on the action data of the next virtual action role and the historical business processing message body, to obtain the business processing message body of the next virtual action role;

[0158] If there is no business processing message body of the current virtual action role, then perform business processing on the business to be processed through a large language model based on the action data of the current virtual action role and the historical business processing message body, to obtain the business processing message body of the current virtual action role.

[0159] Optionally, the processor 701 executes the step of performing business processing on the business to be processed through a large language model based on the action data of the current virtual action role and the historical business processing message body, to obtain the business processing message body of the current virtual action role, including:

[0160] Determine whether the number of the action data of the current virtual action role is multiple;

[0161] If the number of the action data of the current virtual action role is single, then determine the single action data of the current virtual action role as the target action data of the current virtual action role;

[0162] If the number of the action data of the current virtual action character is multiple, generate a first prompt text based on the multiple action data and the historical service processing message body, where the first prompt text includes selection prompts for the multiple action data;

[0163] Input the first prompt text into the large language model to obtain the target action data of the current virtual action character;

[0164] Based on the target action data and the historical service processing message body, perform service processing on the service to be processed through the large language model to obtain the service processing message body of the current virtual action character.

[0165] Optionally, the processor 701 executes the step of performing service processing on the service to be processed through the large language model based on the target action data and the historical service processing message body to obtain the service processing message body of the current virtual action character, including:

[0166] If the target action data contains a preset action keyword, generate a second prompt text based on the action keyword and the historical service processing message body;

[0167] If the target action data does not contain a preset action keyword, determine the action category of the target action data;

[0168] Determine the class prompt template of the target action data according to the action category of the target action data, and different action categories correspond to different class prompt templates;

[0169] Generate a second prompt text based on the class prompt template of the target action data and the historical service processing message body;

[0170] Input the second prompt text into the large language model to perform service processing on the service to be processed to obtain the service processing message body of the current virtual action character.

[0171] Optionally, the processor 701 executes the step of performing service file writing processing on the service processing message bodies of the virtual action characters to obtain the writing files of the virtual personnel. When the virtual action character with the lowest rank completes the service file writing processing, obtain the final processing result of the service to be processed, including:

[0172] In the current iteration round, determine whether the rank of the current virtual action character is the lowest rank;

[0173] If the rank of the current virtual action character is not the lowest rank, format the service processing message body of the current virtual action character and write it into the service file;

[0174] If the rank of the current virtual action role is the lowest rank, write the service processing message body of the current virtual action role into the service file to obtain the final processing result of the service to be processed.

[0175] Optionally, the virtual team further includes a virtual verification role, and the method executed by the processor 701 further includes:

[0176] Based on the action data and the service processing message body of the virtual action role, perform role positioning judgment and requirement positioning judgment on the virtual action role through the virtual verification role to obtain the role positioning judgment result and requirement positioning judgment result of the virtual action role.

[0177] Determine whether the virtual action role passes the verification according to the role positioning judgment result and requirement positioning judgment result of the virtual action role.

[0178] The electronic device provided by the embodiment of the present invention can implement each process implemented by the virtual team service processing method in the above method embodiment, and can achieve the same beneficial effects. To avoid repetition, it will not be elaborated here.

[0179] The embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements each process of the virtual team service processing method provided by the embodiment of the present invention, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0180] Those of ordinary skill in the art can understand that all or part of the processes of implementing the method in the above embodiment can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. Among them, the computer-readable storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0181] The above-disclosed are only the preferred embodiments of the present invention. Of course, the scope of the rights of the present invention cannot be limited thereby. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope covered by the present invention.

Claims

1. A method for processing virtual team services, characterized in that, The virtual team includes virtual action roles at different levels, each virtual action role having action data corresponding to its own level. The virtual team processes business in order of the level from high to low. The virtual team business processing method includes: Input the business to be processed into the virtual team. Based on the action data of each virtual action role, perform iterative business processing on the business to be processed through a large language model in the order of the level of the virtual action role from high to low, and obtain the business processing message bodies of each virtual action role. Each virtual action role corresponds to a business processing message body; Perform business file writing processing on the business processing message bodies of each virtual action role to obtain the business files of each virtual person. When the virtual action role with the lowest level completes the business file writing processing, obtain the final processing result of the business to be processed.

2. The method according to claim 1, characterized in that, The performing iterative business processing on the business to be processed through a large language model in the order of the level of the virtual action role from high to low based on the action data of each virtual action role, and obtaining the business processing message bodies of each virtual action role includes: In the current iteration round, detect whether there is a business processing message body of the current virtual action role; If there is a business processing message body of the current virtual action role, perform business processing on the business to be processed through a large language model based on the action data of the next virtual action role and the historical business processing message body, and obtain the business processing message body of the next virtual action role; If there is no business processing message body of the current virtual action role, perform business processing on the business to be processed through a large language model based on the action data of the current virtual action role and the historical business processing message body, and obtain the business processing message body of the current virtual action role.

3. The method according to claim 1, characterized in that, The performing business processing on the business to be processed through a large language model based on the action data of the current virtual action role and the historical business processing message body, and obtaining the business processing message body of the current virtual action role includes: Determine whether the number of the action data of the current virtual action role is multiple; If the number of the action data of the current virtual action role is single, determine the single action data of the current virtual action role as the target action data of the current virtual action role; If the number of the action data of the current virtual action role is multiple, generate a first prompt text based on the multiple action data and the historical business processing message body. The first prompt text includes selection prompts for the multiple action data; Input the first prompt text into the large language model to obtain the target action data of the current virtual action role; Perform business processing on the business to be processed through a large language model based on the target action data and the historical business processing message body, and obtain the business processing message body of the current virtual action role.

4. The method according to claim 3, characterized in that, Based on the target action data and historical business processing message bodies, the business of the to-be-processed business is processed through a large language model to obtain the business processing message body of the current virtual action role, including: If the target action data contains a preset action keyword, a second prompt text is generated based on the action keyword and the historical business processing message body; If the target action data does not contain a preset action keyword, the action category of the target action data is determined; According to the action category of the target action data, a class prompt template of the target action data is determined, and different action categories correspond to different class prompt templates; A second prompt text is generated based on the class prompt template of the target action data and the historical business processing message body; The second prompt text is input into the large language model to process the to-be-processed business to obtain the business processing message body of the current virtual action role.

5. The method according to any one of claims 1 to 4, characterized in that, Performing business file writing processing on the business processing message bodies of each virtual action role to obtain the written files of each virtual person. When the virtual action role with the lowest rank completes the business file writing processing, the final processing result of the to-be-processed business is obtained, including: In the current iteration round, it is judged whether the rank of the current virtual action role is the lowest rank; If the rank of the current virtual action role is not the lowest rank, the business processing message body of the current virtual action role is formatted and written into a business file; If the rank of the current virtual action role is the lowest rank, the business processing message body of the current virtual action role is written into a business file to obtain the final processing result of the to-be-processed business.

6. The method according to any one of claims 1 to 5, characterized in that, The virtual team further includes a virtual verification role, and the method further includes: Based on the action data and business processing message body of the virtual action role, the virtual verification role performs role positioning judgment and requirement positioning judgment on the virtual action role to obtain the role positioning judgment result and requirement positioning judgment result of the virtual action role; According to the role positioning judgment result and requirement positioning judgment result of the virtual action role, it is determined whether the virtual action role passes the verification.

7. A virtual team service processing device, characterized in that, The virtual team includes virtual action roles with different ranks, and each virtual action role has action data corresponding to its own rank. The virtual team performs business processing in sequence according to the rank from high to low. The device includes: A first processing module, configured to input a to-be-processed business into the virtual team, and based on the action data of each virtual action role, perform business processing iteration on the to-be-processed business through a large language model in the order of the ranks of the virtual action roles from high to low, to obtain the business processing message bodies of each virtual action role, and each virtual action role corresponds to a business processing message body; A second processing module, configured to perform business file writing processing on the business processing message bodies of each of the virtual action characters to obtain business files of each of the virtual personnel, and when the virtual action character with the lowest rank completes the business file writing processing, obtain the final processing result of the to-be-processed business.

8. A virtual team service processing system, characterized in that, Comprising a request terminal, a large language model, and the virtual team business processing device as claimed in claim 7, wherein the request terminal generates a to-be-processed business and inputs it into the virtual team business processing device for business processing, and after the virtual business processing device processes the to-be-processed business through the large language model, returns the obtained final processing result to the request terminal.

9. An electronic device, characterized in that, Comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the steps in the virtual team business processing method as claimed in any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium, characterized in that A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps in the virtual team business processing method as claimed in any one of claims 1 to 6 are implemented.

Images