A multi-agent group chat collaboration method, device, equipment, medium and product

Abstract

The application discloses a multi-agent group chat cooperation method and device, equipment, medium and product. The multi-agent group chat cooperation method comprises the following steps: in response to receiving a user's inquiry text, performing semantic analysis on the inquiry text to obtain an inquiry label, performing vectorization processing on the inquiry label to obtain an inquiry feature vector; performing similarity matching on the inquiry feature vector and a plurality of pre-stored function description vectors of agents in an agent database to obtain a plurality of candidate agents; based on the inquiry text and the function description information of the candidate agents in the agent database, filtering the candidate agents through a large language model to generate an agent list participating in the group chat; based on the agent list, establishing an agent group chat, and using the group chat arrangement agent to control the speaking order of each agent in the agent group chat to cooperatively respond to the inquiry text. The technical scheme of the embodiment of the application can realize efficient recall of agents and dynamic construction of a multi-agent group chat.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, medium and product for multi-agent group chat collaboration. Background Technology

[0002] With the rapid development of artificial intelligence technology, especially the breakthrough in natural language processing capabilities centered on Large Language Models (LLMs), intelligent question-answering systems, dialogue assistants, and task-oriented dialogue systems based on LLMs have been widely applied in multiple fields such as finance, education, healthcare, and government affairs.

[0003] Most mainstream dialogue systems are based on a single agent architecture, relying on a single general language model to handle all tasks. When faced with multi-task concurrency, cross-domain knowledge fusion, or deep reasoning requirements, they often fail to meet user needs.

[0004] To address the aforementioned issues, existing technologies propose multi-agent collaborative reasoning and task execution. However, most existing multi-agent systems employ a pre-defined set of agents and a static role binding mechanism, meaning that the identity and division of labor of each agent are determined during initialization, lacking the ability to be invoked and dynamically organized as needed during actual dialogue. Summary of the Invention

[0005] This invention provides a multi-agent group chat collaboration method, apparatus, device, medium, and product to solve the problem that the preset intelligent agent set and static role binding mechanism lack on-demand calling and dynamic organization capabilities. It realizes the ability to call some intelligent agents from a massive heterogeneous intelligent agent set on demand according to the current user's query needs, and dynamically organize the called intelligent agents to respond to queries.

[0006] According to one aspect of the present invention, a multi-agent group chat collaboration method is provided, comprising: In response to receiving a user's query text, semantic analysis is performed on the query text to obtain query tags, and the query tags are vectorized to obtain query feature vectors; The query feature vector is matched with the functional description vectors of multiple agents pre-stored in the agent database to obtain multiple candidate agents; Based on the query text and the functional description information of the candidate agents in the agent database, a list of agents participating in the group chat is generated by filtering the candidate agents through a large language model. Based on the list of agents, an agent group chat is established, and a group chat orchestration agent is used to control the speaking order of each agent in the agent group chat in order to collaboratively respond to the query text.

[0007] According to another aspect of the present invention, a multi-agent group chat collaboration device is provided, comprising: The feature vector construction module is used to respond to the received user's query text, perform semantic analysis on the query text to obtain query tags, and vectorize the query tags to obtain query feature vectors; The candidate agent determination module is used to perform similarity matching between the query feature vector and the functional description vectors of multiple agents pre-stored in the agent database to obtain multiple candidate agents; The agent list construction module is used to filter the candidate agents based on the query text and the functional description information of the candidate agents in the agent database, and generate a list of agents participating in the group chat by using a large language model. A multi-agent group chat collaboration module is used to establish an agent group chat based on the agent list, and to use group chat orchestration agents to control the speaking order of each agent in the agent group chat in order to collaboratively respond to the query text.

[0008] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the multi-agent group chat collaboration method according to any embodiment of the present invention.

[0009] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the multi-agent group chat collaboration method according to any embodiment of the present invention.

[0010] According to another aspect of the present invention, a computer program product is provided, including a computer program that, when executed by a processor, implements the multi-agent group chat collaboration method of any embodiment of the present disclosure.

[0011] The technical solution of this invention, in response to receiving a user's query text, performs semantic analysis on the query text to obtain query tags, and vectorizes the query tags to obtain query feature vectors. The query feature vectors are then matched with the functional description vectors of multiple agents pre-stored in an agent database to obtain multiple candidate agents. Based on the query text and the functional description information of the candidate agents in the agent database, a large language model is used to filter the candidate agents, generating a list of agents participating in a group chat. Based on the agent list, an agent group chat is established, and a group chat orchestration agent is used to control the speaking order of each agent in the group chat to collaboratively respond to the query text. By combining agent database retrieval and a large language model, efficient agent recall is achieved, multi-agent group chats are dynamically constructed, and the accuracy and efficiency of agent collaboration are improved.

[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart of a multi-agent group chat collaboration method provided in Embodiment 1 of the present invention; Figure 2a This is a flowchart of a multi-agent group chat collaboration method provided in Embodiment 2 of the present invention; Figure 2b This is a flowchart of a group chat arrangement method provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the structure of a multi-agent group chat collaboration device according to Embodiment 3 of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device that implements the multi-agent group chat collaboration method of the present invention. Detailed Implementation

[0015] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0016] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0017] Example 1 Figure 1 This is a flowchart of a multi-agent group chat collaboration method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where agent retrieval and dynamic group chat construction are performed by combining knowledge base retrieval and a large language model. This method can be executed by a multi-agent group chat collaboration device, which can be implemented in hardware and / or software and can be configured in various general-purpose computing devices. Figure 1 As shown, the method includes: S110. In response to receiving the user's query text, perform semantic analysis on the query text to obtain query tags, and then vectorize the query tags to obtain query feature vectors.

[0018] Query text is used to represent user intent or questions. Query text is natural language text input by the user into a large language model. For example, the query text could be "I'm bad at math, can you teach me math?"

[0019] Inquiry tags are structured semantic descriptions generated by semantic analysis of inquiry text, which can represent the user's core intent and key needs. For example, inquiry tags can be a collection of one or more of the following: keywords, phrases, or short sentences.

[0020] In this embodiment of the invention, after receiving the query text input by the user, a large language model can first be invoked to perform deep semantic understanding on the query text, obtaining functional tags or descriptive phrases describing the user's core intent as query tags. Further, the query tags are input into an embedding model for vectorization processing to obtain query feature vectors.

[0021] In a specific example, if the user inputs a query text that reads, "Please plan a seven-day trip to City A that combines historical and cultural experiences with modern art, while keeping the budget on track," the system first invokes a large language model to perform semantic analysis on the query text. Specifically, the large language model analyzes the query's intent, domain, specific tasks involved, and constraints, and outputs a set of query tags, such as "travel planning, City A, seven days, historical and cultural attractions, modern art venues, budget control," etc. Then, by vectorizing these query tags, a query feature vector is obtained, which represents the core functional requirements of the query text.

[0022] S120. Perform similarity matching between the query feature vector and the functional description vectors of multiple agents pre-stored in the agent database to obtain candidate agents.

[0023] The agent database is a pre-built database containing agent metadata. Each record in the agent database is associated with at least the agent's functional description information, functional description vector, and calling interface. Among them, the functional description information is a description of the agent's capabilities, expertise, and applicable scenarios in natural language text form.

[0024] In this embodiment of the invention, a functional description vector of an agent is extracted sequentially from the agent database. The similarity between the functional description vector and the query feature vector is calculated, for example, by calculating cosine similarity, to obtain the similarity between the functional description vector and the query feature vector of each agent. Based on a preset similarity threshold or a preset number of agents N, the N agents with the highest similarity, or agents with similarity higher than the similarity threshold, are selected as candidate agents. Nearest neighbor retrieval based on semantic vector space can quickly narrow down the matching range from a large number of agents.

[0025] S130. Based on the query text and the functional description information of candidate agents in the agent database, a list of agents participating in the group chat is generated by filtering the candidate agents through a large language model.

[0026] In this embodiment of the invention, based on the identified candidate agents, functional description information corresponding to the candidate agents is extracted from the agent database. The query text and the functional description information of the candidate agents are then input into a large language model. The large language model further filters the candidate agents to select the agents ultimately participating in the group chat, forming an agent list. Specifically, the large language model can determine whether there are agents with highly overlapping functions among the candidate agents based on their functional description information, and select representative agents (e.g., the most functionally complete) from among multiple overlapping agents. In addition, the large language model can analyze the matching degree between the functional description information of the candidate agents and the query text, and filter agents based on the matching degree. The filtered agents constitute the agent list.

[0027] S140. Based on the list of agents, establish an agent group chat, and use a preset arrangement method to control the speaking order of each agent in the agent group chat to collaboratively respond to the query text.

[0028] Intelligent agent group chat is a group chat dialogue in a collaborative dialogue environment, in which users, multiple intelligent agents from the intelligent agent list, and the group chat orchestration intelligent agent responsible for coordination participate.

[0029] In this embodiment of the invention, a group chat list of participating agents is initialized based on the agents in the agent list, and an agent group chat is established based on the agent group chat list. The agents included in the initial query-response agent set are the same as those in the agent list.

[0030] Furthermore, the group chat orchestration agent takes over the process control and executes an iterative orchestration loop. Specifically, in each orchestration round, the current response agent is determined based on the group chat orchestration agent, and the current response agent is invoked to generate response content for the query text.

[0031] The response content is then evaluated for quality, and the subsequent orchestration loop is determined based on the evaluation results: If the current response agent cannot be invoked normally, for example, by throwing an error during invocation, or if the evaluation result of the response content indicates that the response content does not match the function of the current response agent (e.g., the response content is English knowledge, but the function description is a math teaching agent), then the current response agent is removed from the current group chat. This prevents the agent from participating in subsequent user-initiated queries. Simultaneously, the current response agent is also removed from the query response agent set, preventing it from generating response content for the current round of queries. The process then returns to executing the group chat orchestration agent. The operation involves selecting the current responding agent from the set of query-response agents. If the evaluation result indicates that the response content does not meet the completeness standard, for example, if the query text is "Please plan a seven-day trip to City A that combines historical and cultural experiences with modern art, while keeping the budget under control," but the travel plan in the response content only includes historical and cultural attractions, then the current responding agent is removed from the set of query-response agents, and the operation of selecting the current responding agent from the set of query-response agents via the group chat orchestration agent is returned. If the evaluation result indicates that the response content meets the completeness standard, then the current group chat round is terminated, the response content is returned to the user, and the speaking right is transferred.

[0032] The technical solution of this invention, in response to receiving a user's query text, performs semantic analysis on the query text to obtain query tags, and vectorizes the query tags to obtain query feature vectors. The query feature vectors are then matched with the functional description vectors of multiple agents pre-stored in an agent database to obtain multiple candidate agents. Based on the query text and the functional description information of the candidate agents in the agent database, a large language model is used to filter the candidate agents and generate a list of agents participating in a group chat. Based on the agent list, an agent group chat is established, and a group chat orchestration agent is used to control the speaking order of each agent in the group chat to collaboratively respond to the query text. By combining agent database retrieval and a large language model, the accuracy of agent matching with user intent is improved, achieving efficient agent recall, dynamically constructing multi-agent group chats, and improving the accuracy and efficiency of agent collaboration.

[0033] Example 2 Figure 2a This is a flowchart of a multi-agent group chat collaboration method provided in Embodiment 2 of the present invention. This embodiment further refines the above embodiment, providing specific steps for establishing an agent group chat based on the agent list, and using a group chat orchestration agent to control the speaking order of each agent in the group chat to collaboratively respond to the query text. For example... Figure 2a As shown, the method includes: S210. In response to receiving the user's query text, perform semantic analysis on the query text to obtain query tags, and vectorize the query tags to obtain query feature vectors.

[0034] Optionally, before performing semantic analysis on the query text in response to receiving the user's query text, the method further includes: Obtain basic information about the agent, including the agent's name and original functional description text; Based on the aforementioned basic information, semantic analysis is performed using a large language model to generate extended knowledge information for the intelligent agent. Based on the basic information and the extended knowledge information, functional description information is generated by integrating them through the large language model. The functional description information is vectorized to generate a corresponding functional description vector, and the functional description vector, functional description information, and corresponding intelligent agent interaction interface information are associated and stored in the intelligent agent database.

[0035] In this optional embodiment, a specific method is provided for constructing an agent database before performing semantic analysis on the query text received from a user: First, the agent's agent name and original functional description text are obtained as the agent's basic information. Then, after performing deep semantic analysis and reasoning on the basic information through a large language model, supplementary extended information on the original functional description is generated as the agent's extended knowledge information.

[0036] Furthermore, the agent name, the original functional description text, and extended knowledge information are input into the large language model, which integrates this information to generate an enhanced functional description text that is semantically rich, comprehensive in content, and standardized in expression, which serves as the functional description information.

[0037] Finally, the functional description information is vectorized to generate corresponding functional description vectors. These vectors, along with the functional description information and the corresponding agent's interaction interface information, are then stored in the agent database. The extended knowledge information generated by the large language model compensates for the potential limitations, ambiguity, or insufficient information in the original description information. The enhanced functional description information more completely expresses the agent's true capabilities, making the position of its functional description vector in the semantic space more precise. This, in turn, improves the recall and accuracy of subsequent vector similarity matching.

[0038] Optionally, based on the basic information and the extended knowledge information, functional description information is generated by integrating them through the large language model, including: Based on the functional description template, the basic information and the extended knowledge information are integrated through the large language model to generate the functional description information.

[0039] The functional description template is a pre-defined structured text framework used to standardize the content composition and organization format of the final generated functional description information. For example, the functional description template is: "[Agent Name] is an agent focused on [Core Domain], whose main capabilities include: [List of Capabilities], and is good at handling [Specific Task Types]".

[0040] In this optional embodiment, a specific method is provided for integrating basic information and the extended knowledge information through a large language model to generate functional description information: The large language model, based on the constraints of a functional description template, uses discrete, potentially redundant, or non-standard basic information and extended knowledge information as source data, recombines them to fill the template, and generates coherent, complete, and standardized descriptive text as functional description information. Information integration through a functional description template forces all agents' final functional description information to adhere to a unified content framework and format, eliminating problems such as unclear information dimensions or ambiguous expressions caused by free descriptions, and improving the reliability of subsequent retrieval.

[0041] Optionally, embodiments of the present invention further include: Periodically retrieve the list of online intelligent agents; The agents in the online agent list are compared with the agents stored in the agent database to identify differences. Based on the differences, perform agent update operations on the agent database.

[0042] In this optional embodiment, a method for updating the agent database is provided: a scheduled task is set to trigger a synchronization process at a preset time interval, for example, every 10 minutes. The interface of the agent management platform is called to obtain a list of online agents containing the entire database at the current time. The obtained list of online agents is compared with the records already stored in the agent database to determine discrepancies. These discrepancies may include agents that are not in the online agent list but exist in the agent database, agents that are not in the agent database but exist in the online agent list, and agents whose online agent list has been updated but whose agent database has not.

[0043] Finally, based on the discrepancies, update operations are performed on the agent database, including agent deletion, agent addition, and agent updates. Regular agent synchronization ensures the accuracy of the agent database and reduces manual maintenance costs.

[0044] S220. Perform similarity matching between the query feature vector and the functional description vectors of multiple agents pre-stored in the agent database to obtain multiple candidate agents.

[0045] S230. Based on the query text and the functional description information of candidate agents in the agent database, a large language model is used to filter the candidate agents and generate a list of agents participating in the group chat.

[0046] S240. Based on the agents in the agent list, initialize the query response agent set.

[0047] In this embodiment of the invention, after the list of agents is selected, a group chat is established, such as... Figure 2b As shown, the query response agent set is first initialized based on the agents in the agent list. The query response agent set is a list dynamically maintained during the group chat based on the response status of each agent participating in the group chat. The agents in the initialized query response agent set are the same as those in the agent list.

[0048] S250. Based on the query text and the functional description information of each agent in the query response agent set, select the current response agent from the query response agent set through group chat orchestration agents, and call the current response agent to generate response content for the query text.

[0049] In this embodiment of the invention, the query text and the functional description information of each agent in the query response agent set are jointly input into the group chat orchestration agent. The group chat orchestration agent performs semantic analysis to determine the suitability of each agent's sports query text participating in the group chat, and selects the agent with the highest suitability in the current round as the current response agent. Then, the current response agent is invoked to generate response content for the query text.

[0050] Optionally, based on the query text and the functional description information of each agent in the query-response agent set, the current responding agent is selected from the query-response agent set through group chat orchestration, including: By orchestrating intelligent agents in group chat, historical dialogue information associated with the query text is obtained, and semantic analysis is performed on the query text, historical dialogue information, and functional description information of each intelligent agent in the query response intelligent agent set. Based on the semantic analysis results, the current response agent is determined from the set of query response agents.

[0051] In this optional embodiment, a specific method is provided for selecting the current responding agent from the set of query-response agents by a group chat orchestration agent, based on the query text and the functional description information of each agent in the query-response agent set: The group chat orchestration agent obtains historical dialogue information associated with the query text, that is, the query text and the response content of the responding agents within the group from the start of the current group chat to the current moment. Then, the orchestration agent performs joint semantic analysis on the query text, historical dialogue information, and the functional description information of each agent in the query-response agent set to determine the fit between each agent and the query text. Finally, the agent with the highest fit is selected as the current responding agent.

[0052] Unlike existing speaking mechanisms based on fixed roles or preset order, the intelligent scheduling based on dynamic dialogue context in this invention re-evaluates the current task state and agent capability matching degree each time the agent is selected, ensuring that the agent that speaks in each round is the most suitable agent at the moment.

[0053] S260. Evaluate the quality of the response content, update the query response agent set based on the evaluation results, and return to execute the operation of selecting the current response agent from the query response agent set through the group chat orchestration agent until the end condition of the current group chat round is met.

[0054] In this embodiment of the invention, after obtaining the response content of the current responding agent, the group chat orchestration agent evaluates the quality of the response content. Based on the evaluation result, it updates the query-response agent set and returns to execute the operation of selecting the current responding agent from the query-response agent set through the group chat orchestration agent, until the end condition of the current group chat round is met. The evaluation result may include whether the response content matches the overall function of the current response and whether the response content meets the completeness standard. The end condition of the current group chat round includes detecting that the query-response agent set is empty when returning to execute the selection of the current responding agent from the query-response agent set, or that the response content meets the completeness standard. By using a group speaking strategy, it is possible to integrate multi-source knowledge and cross-skill perspectives to answer complex questions, effectively solving the problem of blind spots in the knowledge coverage of a single agent.

[0055] Optionally, the response content is evaluated for quality, and the query response agent set is updated based on the evaluation results, including: If the evaluation result indicates that the current response agent cannot be invoked normally or the response content does not match the functional description information corresponding to the current response agent, then the current response agent will be removed from the current group chat and the set of query response agents. If the evaluation result indicates that the response content does not meet the completeness standard, the response content will be returned to the user, and the current response agent will be removed from the query response agent set. If the evaluation result indicates that the response content meets the completeness standard, the current group chat round will be terminated, and the response content will be returned to the user.

[0056] In this optional embodiment, a specific method is provided for evaluating the quality of the response content and updating the query response agent set based on the evaluation result: If the evaluation result indicates that the current response agent cannot be invoked normally, for example, by directly reporting an error or returning an empty response, or if the response content does not match the functional description information corresponding to the current response agent, for example, if a "travel planning agent" returns medical advice, it can be determined that the current response agent is not suitable to continue participating in this round of group chat. In this case, the current response agent is permanently deleted from the current group chat, and the operation of initializing the query response agent set is returned.

[0057] If the evaluation result indicates that the response content does not meet the completeness standard, the response content is returned to the user, the current responding agent is removed from the query response agent set, and the process returns to select the next agent to supplement the response. If the evaluation result indicates that the response content meets the completeness standard, the current group chat round is terminated, and the response content is returned to the user. By introducing an instant quality evaluation process, abnormal behavior can be detected and handled immediately after an agent speaks, removing agents with incompatible functions or those that cannot be invoked from the group chat, thus preventing their low-quality output from interfering with user perception or consuming subsequent scheduling resources.

[0058] The technical solution of this invention initializes a query response agent set based on agents in an agent list. Then, based on the query text and the functional description information of each agent in the query response agent set, a group chat orchestration agent selects the current response agent from the query response agent set and calls the current response agent to generate response content for the query text. The response content is then evaluated for quality. Based on the evaluation results, the query response agent set is updated, and the operation of selecting the current response agent from the query response agent set via the group chat orchestration agent is returned to continue until the end condition of the current group chat round is met. By re-evaluating the matching degree between the current query text and agent capabilities with each selection by the orchestration agent, the most suitable agent is determined to reply, which can improve the reliability of the response content.

[0059] Example 3 Figure 3 This is a schematic diagram of a multi-agent group chat collaboration device provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes: The feature vector construction module 310 is used to respond to the received user's query text, perform semantic analysis on the query text to obtain query tags, and vectorize the query tags to obtain query feature vectors. The candidate agent determination module 320 is used to perform similarity matching between the query feature vector and the functional description vectors of multiple agents pre-stored in the agent database to obtain multiple candidate agents; The agent list construction module 330 is used to filter the candidate agents based on the query text and the functional description information of the candidate agents in the agent database, and generate a list of agents participating in the group chat through a large language model. The multi-agent group chat collaboration module 340 is used to establish an agent group chat based on the agent list, and to use group chat orchestration agents to control the speaking order of each agent in the agent group chat in order to collaboratively respond to the query text.

[0060] The technical solution of this invention, in response to receiving a user's query text, performs semantic analysis on the query text to obtain query tags, and vectorizes the query tags to obtain query feature vectors. The query feature vectors are then matched with the functional description vectors of multiple agents pre-stored in an agent database to obtain multiple candidate agents. Based on the query text and the functional description information of the candidate agents in the agent database, a large language model is used to filter the candidate agents, generating a list of agents participating in a group chat. Based on the agent list, an agent group chat is established, and a group chat orchestration agent is used to control the speaking order of each agent in the group chat to collaboratively respond to the query text. By combining agent database retrieval and a large language model, efficient agent recall is achieved, multi-agent group chats are dynamically constructed, and the accuracy and efficiency of agent collaboration are improved.

[0061] Optional, a multi-agent group chat collaboration module 340, including: The list initialization unit is used to initialize the query response agent set based on the agents in the agent list; The response agent determination unit is used to select the current response agent from the set of query response agents based on the query text and the functional description information of each agent in the query response agent set, through group chat orchestration agents, and call the current response agent to generate response content for the query text; The group chat collaboration unit is used to evaluate the quality of the response content, update the query response agent set based on the evaluation results, and return to execute the operation of selecting the current response agent from the query response agent set through the group chat orchestration agent until the end condition of the current group chat round is met.

[0062] Optional, group chat collaboration unit, specifically used for: If the evaluation result indicates that the current response agent cannot be invoked normally or the response content does not match the functional description information corresponding to the current response agent, then the current response agent will be removed from the current group chat and the set of query response agents. If the evaluation result indicates that the response content does not meet the completeness standard, the response content will be returned to the user, and the current response agent will be removed from the query response agent set. If the evaluation result indicates that the response content meets the completeness standard, the current group chat round is terminated, and the response content is returned to the user.

[0063] Optional multi-agent group chat collaboration devices also include: The basic information acquisition module is used to acquire basic information of the agent before performing semantic analysis on the query text received from the user. The basic information includes the agent name and the original functional description text. The extended knowledge determination module is used to generate extended knowledge information of the agent by performing semantic analysis through a large language model based on the basic information. The function description information determination module is used to generate function description information by integrating the basic information and the extended knowledge information through the large language model. The agent database construction module is used to vectorize the functional description information to generate corresponding functional description vectors, and associate and store the functional description vectors, functional description information and corresponding agent interaction interface information in the agent database.

[0064] Optional multi-agent group chat collaboration devices also include: The online agent acquisition module is used to periodically acquire a list of online agents. The difference recognition module is used to compare the agents in the online agent list with the agents stored in the agent database and identify the differences. The database update module is used to perform agent update operations on the agent database based on the differences.

[0065] Optionally, the response agent determining unit is specifically used for: By orchestrating intelligent agents in group chat, historical dialogue information associated with the query text is obtained, and semantic analysis is performed on the query text, historical dialogue information, and functional description information of each intelligent agent in the query response intelligent agent set. Based on semantic analysis, the current response agent is determined from the set of query response agents.

[0066] Optional, the function description information determination module is specifically used for: Based on the functional description template, the basic information and the extended knowledge information are integrated through the large language model to generate the functional description information.

[0067] The multi-agent group chat collaboration device provided in the embodiments of the present invention can execute the multi-agent group chat collaboration method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0068] In the technical solution of this invention, the information collected is information and data authorized by the user or fully authorized by all parties. The collection, storage, use, processing, transmission, provision, disclosure and application of related data all comply with the relevant laws, regulations and standards of relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0069] Example 4 According to embodiments of the present invention, the present invention also provides an electronic device, a readable storage medium, and a computer program product.

[0070] Figure 4 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, application processors, blade application processors, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0071] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0072] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0073] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as multi-agent group chat collaboration methods.

[0074] In some embodiments, the multi-agent group chat collaboration method can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the multi-agent group chat collaboration method described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the multi-agent group chat collaboration method by any other suitable means (e.g., by means of firmware).

[0075] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0076] Computer programs used to implement the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs can be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or application.

[0077] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0078] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0079] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data application processors), or computing systems that include middleware components (e.g., application application processors), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0080] A computing system can include clients and applications. Clients and applications are generally geographically separated and typically interact via a communication network. The client-application relationship is established by computer programs running on the respective computers and having a client-application relationship with each other. An application can be a cloud application, also known as a cloud computing application or cloud server, which is a hosting product within the cloud computing application ecosystem. It addresses the shortcomings of traditional physical servers and VPS applications, such as high management difficulty and weak business scalability.

[0081] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0082] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A multi-agent group chat collaboration method, characterized in that, include: In response to receiving a user's query text, semantic analysis is performed on the query text to obtain query tags, and the query tags are vectorized to obtain query feature vectors; The query feature vector is matched with the functional description vectors of multiple agents pre-stored in the agent database to obtain multiple candidate agents; Based on the query text and the functional description information of the candidate agents in the agent database, a list of agents participating in the group chat is generated by filtering the candidate agents through a large language model. Based on the list of agents, an agent group chat is established, and a group chat orchestration agent is used to control the speaking order of each agent in the agent group chat in order to collaboratively respond to the query text.

2. The method according to claim 1, characterized in that, Based on the list of agents, an agent group chat is established, and a group chat orchestration agent is used to control the speaking order of each agent in the agent group chat to collaboratively respond to the query text, including: Initialize the query response agent set based on the agents in the agent list; Based on the query text and the functional description information of each agent in the query response agent set, the group chat orchestration agent selects the current response agent from the query response agent set and calls the current response agent to generate response content for the query text; The response content is evaluated for quality. Based on the evaluation results, the query response agent set is updated, and the operation of selecting the current response agent from the query response agent set through the group chat orchestration agent is returned to be executed until the end condition of the current group chat round is met.

3. The method according to claim 2, characterized in that, The quality of the response content is evaluated, and the query response agent set is updated based on the evaluation results, including: If the evaluation result indicates that the current response agent cannot be invoked normally or the response content does not match the functional description information corresponding to the current response agent, then the current response agent will be removed from the current group chat and the set of query response agents. If the evaluation result indicates that the response content does not meet the completeness standard, the response content will be returned to the user, and the current response agent will be removed from the query response agent set. If the evaluation result indicates that the response content meets the completeness standard, the current group chat round is terminated, and the response content is returned to the user.

4. The method according to claim 1, characterized in that, Before performing semantic analysis on the query text received from the user, the method further includes: Obtain basic information about the agent, including the agent's name and original functional description text; Based on the aforementioned basic information, semantic analysis is performed using a large language model to generate extended knowledge information for the intelligent agent. Based on the basic information and the extended knowledge information, functional description information is generated by integrating them through the large language model. The functional description information is vectorized to generate a corresponding functional description vector, and the functional description vector, functional description information, and corresponding intelligent agent interaction interface information are associated and stored in the intelligent agent database.

5. The method according to claim 1, characterized in that, Also includes: Periodically retrieve the list of online intelligent agents; The agents in the online agent list are compared with the agents stored in the agent database to identify differences. Based on the differences, perform agent update operations on the agent database.

6. The method according to claim 2, characterized in that, Based on the query text and the functional description information of each agent in the query response agent set, the current response agent is selected from the query response agent set through group chat orchestration agents, including: By orchestrating intelligent agents in group chat, historical dialogue information associated with the query text is obtained, and semantic analysis is performed on the query text, historical dialogue information, and functional description information of each intelligent agent in the query response intelligent agent set. Based on the semantic analysis results, the current response agent is determined from the set of query response agents.

7. The method according to claim 4, characterized in that, Based on the aforementioned basic information and extended knowledge information, functional description information is generated through integration using the large language model, including: Based on the functional description template, the basic information and the extended knowledge information are integrated through the large language model to generate the functional description information.

8. A multi-agent group chat collaboration device, characterized in that, include: The feature vector construction module is used to respond to the received user's query text, perform semantic analysis on the query text to obtain query tags, and vectorize the query tags to obtain query feature vectors; The candidate agent determination module is used to perform similarity matching between the query feature vector and the functional description vectors of multiple agents pre-stored in the agent database to obtain multiple candidate agents; The agent list construction module is used to filter the candidate agents based on the query text and the functional description information of the candidate agents in the agent database, and generate a list of agents participating in the group chat by using a large language model. A multi-agent group chat collaboration module is used to establish an agent group chat based on the agent list, and to use group chat orchestration agents to control the speaking order of each agent in the agent group chat in order to collaboratively respond to the query text.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the multi-agent group chat collaboration method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the multi-agent group chat collaboration method according to any one of claims 1-7.

11. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the multi-agent group chat collaboration method according to any one of claims 1-7.

Images