Multi-agent configuration method and system based on agent search

Abstract

Disclosed are a multi-agent configuration method and system based on agent search. A multi-agent configuration method according to an embodiment may comprise the steps of: receiving a user query regarding a user's task; analyzing the user query to generate a search query for searching for an agent; and obtaining information on candidates of the agent by searching, using the search query, a vector database in which an embedding vector generated by embedding metadata of artificial intelligence model-based agents is stored.

Description

Method and system for configuring multi-agents based on agent search

[0001] The following description relates to a multi-agent configuration method and system based on agent discovery.

[0002] Large language models (LLMs) are a type of artificial intelligence (AI) trained on large corpora of text data to generate human-like responses to natural language input. These models consist of artificial neural networks with numerous parameters (typically billions of weights or more). These LLMs can be trained on large amounts of unlabeled text using self-supervised or semi-self-supervised learning.

[0003] In these LLM-based multi-agent systems, various agents are created and the performance of the AI ​​system is maximized through inter-agent dialogue. Conventional multi-agent technologies largely fall into two categories. The first is a fixed-form approach, where agents and their overall conversational flow are configured in a predefined format. The second approach dynamically generates agents based on a given problem (task) and then improves these dynamically generated agents through a critic agent.

[0004] First, the fixed-type approach suffers from severe limitations, as it cannot adequately handle diverse tasks, especially new ones, and requires significant manual intervention by human experts. Furthermore, the dynamic agent generation approach suffers from the problem of degrading overall performance due to the significant resource consumption required to create new agents and dynamically improve their performance.

[0005] A method and system for configuring a multi-agent based on agent search are provided.

[0006] A multi-agent configuration method of a multi-agent configuration system implemented by at least one computer device, comprising: receiving a user query regarding a user's task by at least one processor included in the at least one computer device; generating a search query for searching for an agent by analyzing the user query by the at least one processor; and obtaining information on candidate agents by searching a vector database storing an embedding vector generated by embedding metadata of agents based on an artificial intelligence model by the search query by the at least one processor.

[0007] According to one side, the vector database may be characterized by including a vector database based on Retrieval-Augmented Generation (RAG) for the artificial intelligence model.

[0008] According to another aspect, the metadata may be characterized by including at least one of a description of the agent, information about prompts to be used in the artificial intelligence model, information about external tools available to the agent, and information required to execute the agent.

[0009] According to another aspect, the multi-agent configuration method may further include a step of, by the at least one processor, accessing an external data source to obtain external data when the user query requires external data, and the step of generating the search query may be characterized in that the step of further analyzing the external data to generate the search query.

[0010] According to another aspect, the multi-agent configuration method may further include a step of selecting agents to perform the user's task by using the user query and information about the candidate agents, by the at least one processor; and a step of establishing a plan for the operation of the selected agents based on the user query and the selected agents, by the at least one processor.

[0011] According to another aspect, the plan may be characterized by including at least one of an initial input query to be delivered to each of the selected agents, an execution order of the selected agents, and a connection relationship between the selected agents.

[0012] According to another aspect, the multi-agent configuration method may further include a step of, by the at least one processor, accessing an external data source to obtain external data when the user query requires external data, wherein the step of selecting agents further utilizes the external data to select agents that will perform the user's task, and the step of establishing the plan may further comprise establishing the plan based on the external data.

[0013] According to another aspect, the multi-agent configuration method may further include: executing, by the at least one processor, multi-agents dynamically configured based on the list of selected agents and the established plan; compiling, by the at least one processor, processing results of each of the multi-agents to generate a final result and an answer for the user; and providing, by the at least one processor, at least one of the generated final result and answer to the user.

[0014] A computer program stored on a computer-readable recording medium is provided for executing the above method on a computer device in combination with a computer device.

[0015] A computer-readable recording medium having recorded thereon a program for executing the above method on a computer device is provided.

[0016] A multi-agent configuration system implemented by at least one computer device, comprising at least one processor implemented to execute instructions readable by the at least one computer device, wherein the multi-agent configuration system is characterized in that the at least one processor receives a user query regarding a user's task, analyzes the user query to generate a search query for searching for an agent, and searches a vector database storing an embedding vector generated by embedding metadata of agents based on an artificial intelligence model using the search query to obtain information on candidate agents.

[0017] An agent search method and system for configuring a multi-agent system can be provided.

[0018] FIG. 1 is a diagram illustrating an example of a network environment according to one embodiment of the present invention.

[0019] FIG. 2 is a block diagram illustrating an example of a computer device according to one embodiment of the present invention.

[0020] FIG. 3 is a diagram illustrating an example of an environment in which an agent search system operates according to one embodiment of the present invention.

[0021] FIG. 4 is a drawing showing an example of the overall configuration for a multi-agent configuration in one embodiment of the present invention.

[0022] FIG. 5 is a drawing showing another example of the overall configuration for a multi-agent configuration in one embodiment of the present invention.

[0023] FIG. 6 is a flowchart illustrating an example of a multi-agent configuration method according to one embodiment of the present invention.

[0024] Hereinafter, the detailed description will be given with reference to the attached drawings.

[0025] A multi-agent configuration system according to embodiments of the present invention may be implemented by at least one computer device. In this case, a computer program according to an embodiment of the present invention may be installed and run on the computer device, and the computer device may perform a multi-agent configuration method according to embodiments of the present invention under the control of the run computer program. The computer program described above may be stored on a computer-readable recording medium so as to be coupled to the computer device and cause the computer to execute the multi-agent configuration method.

[0026] FIG. 1 is a diagram illustrating an example of a network environment according to an embodiment of the present invention. The network environment of FIG. 1 represents an example including a plurality of electronic devices (110, 120, 130, 140), a plurality of servers (150, 160), and a network (170). FIG. 1 is merely an example for explaining the invention, and the number of electronic devices or servers is not limited to that of FIG. 1. In addition, the network environment of FIG. 1 merely illustrates one example of environments applicable to the present embodiments, and the environments applicable to the present embodiments are not limited to the network environment of FIG. 1.

[0027] The plurality of electronic devices (110, 120, 130, 140) may be fixed terminals or mobile terminals implemented as computer devices. Examples of the plurality of electronic devices (110, 120, 130, 140) include smart phones, mobile phones, navigation systems, computers, laptops, digital broadcasting terminals, PDAs (Personal Digital Assistants), PMPs (Portable Multimedia Players), tablet PCs, etc. For example, in FIG. 1, the shape of a smart phone is shown as an example of the electronic device (110), but in embodiments of the present invention, the electronic device (110) may actually mean one of various physical computer devices that can communicate with other electronic devices (120, 130, 140) and / or servers (150, 160) via a network (170) using a wireless or wired communication method.

[0028] The communication method is not limited, and may include not only a communication method that utilizes a communication network (e.g., a mobile communication network, a wired Internet, a wireless Internet, a broadcasting network) that the network (170) may include, but also short-range wireless communication between devices. For example, the network (170) may include any one or more of a personal area network (PAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN), a broadband network (BBN), the Internet, and the like. In addition, the network (170) may include any one or more of a network topology including, but not limited to, a bus network, a star network, a ring network, a mesh network, a star-bus network, a tree, or a hierarchical network.

[0029] Each server (150, 160) may be implemented as a computer device or multiple computer devices that communicate with multiple electronic devices (110, 120, 130, 140) via a network (170) to provide commands, codes, files, contents, services, etc. For example, the server (150) may be a system that provides services to multiple electronic devices (110, 120, 130, 140) connected via a network (170).

[0030] FIG. 2 is a block diagram illustrating an example of a computer device according to an embodiment of the present invention. Each of the multiple electronic devices (110, 120, 130, 140) or servers (150, 160) described above can be implemented by the computer device (200) illustrated in FIG. 2.

[0031] Such a computer device (200) may include a memory (210), a processor (220), a communication interface (230), and an input / output interface (240), as illustrated in FIG. 2. The memory (210) may be a computer-readable recording medium, and may include a random access memory (RAM), a read only memory (ROM), and a non-permanent mass storage device such as a disk drive. Here, the non-permanent mass storage device such as a ROM and a disk drive may be included in the computer device (200) as a separate permanent storage device distinct from the memory (210). In addition, an operating system and at least one program code may be stored in the memory (210). These software components may be loaded into the memory (210) from a computer-readable recording medium separate from the memory (210). Such a separate computer-readable recording medium may include a computer-readable recording medium such as a floppy drive, a disk, a tape, a DVD / CD-ROM drive, a memory card, etc. In another embodiment, the software components may be loaded into the memory (210) via a communication interface (230) other than a computer-readable recording medium. For example, the software components may be loaded into the memory (210) of the computer device (200) based on a computer program installed by files received over a network (170).

[0032] The processor (220) may be configured to process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. Instructions may be provided to the processor (220) via the memory (210) or the communication interface (230). For example, the processor (220) may be configured to execute instructions received according to program code stored in a storage device such as the memory (210).

[0033] The communication interface (230) may provide a function for the computer device (200) to communicate with other devices (e.g., the storage devices described above) via the network (170). For example, requests, commands, data, files, etc. generated by the processor (220) of the computer device (200) according to program codes stored in a recording device such as the memory (210) may be transmitted to other devices via the network (170) under the control of the communication interface (230). Conversely, signals, commands, data, files, etc. from other devices may be received by the computer device (200) via the communication interface (230) of the computer device (200) via the network (170). Signals, commands, data, etc. received via the communication interface (230) may be transmitted to the processor (220) or the memory (210), and files, etc. may be stored in a storage medium (the permanent storage device described above) that the computer device (200) may further include.

[0034] The input / output interface (240) may be a means for interfacing with an input / output device (250). For example, the input device may include a device such as a microphone, a keyboard, or a mouse, and the output device may include a device such as a display or a speaker. As another example, the input / output interface (240) may be a means for interfacing with a device that integrates input and output functions, such as a touchscreen. At least one of the input / output devices (250) may be configured as a single device with the computer device (200). For example, a touchscreen, a microphone, a speaker, etc. may be implemented in a form included in the computer device (200), such as a smartphone.

[0035] Additionally, in other embodiments, the computer device (200) may include fewer or more components than those illustrated in FIG. 2. However, it is not necessary to explicitly illustrate most conventional components. For example, the computer device (200) may be implemented to include at least some of the input / output devices (250) described above, or may further include other components such as a transceiver, a database, etc.

[0036] FIG. 3 is a diagram illustrating an example of an environment in which a multi-agent configuration system operates according to an embodiment of the present invention. The multi-agent configuration system (300) according to the present embodiment can be implemented using at least one computer device (for example, a computer device (200)), and can configure a multi-agent capable of processing a user's task based on an artificial intelligence model (for example, a large language model or large language model (LLM)) for a plurality of user terminals (310) accessing the multi-agent configuration system (300), and can process the user's task through the configured multi-agent to provide a response to the user. An agent can mean an independent entity that performs a task using an artificial intelligence model and / or an external tool by exchanging messages with other agents about a specific task based on an artificial intelligence model such as an LLM.

[0037] At this time, the multi-agent configuration system (300) can effectively manage a reusable agent pool, search for agents required to process a user's task, and dynamically configure a multi-agent to process the user's task, thereby efficiently increasing scalability. To this end, the multi-agent configuration system (300) can utilize a vector database (320) of Retrieval-Augmented Generation (RAG), which is commonly utilized in conjunction with an artificial intelligence model such as LLM. Here, the vector database (320) can be implemented externally to the multi-agent configuration system (300), as in the embodiment of FIG. 3, so that it can be accessed from the multi-agent configuration system (300) via a network. In another embodiment, the vector database (320) can also be implemented internally to the multi-agent configuration system (300).

[0038] The vector database (320) can store and manage embedding vectors generated by embedding metadata information of various agents. In some embodiments, the multi-agent configuration system (300) can also generate embedding vectors and store them in the vector database (320).

[0039] Agent information (metadata information) stored in the vector database (320) can be defined as shown in Table 1 below.

[0040] Description Description of the agent Prompt Prompt to be used in LLM Toolset Information about available external tools (e.g., code executor, debugger) Planning Information needed to actually run the agent

[0041] When a user creates a query for a desired task and inputs it into the multi-agent configuration system (300), the multi-agent configuration system (300) can analyze the query to generate a query (hereinafter, “search query”) to be used for agent search and necessary external knowledge information. At this time, the multi-agent configuration system (300) can search the vector database (320) through the search query to generate a list of agent candidates to be used. Thereafter, the multi-agent configuration system (300) can establish a plan for agents to be used and interactions between the agents based on the user’s query, external knowledge information, and the list of agent candidates. Thereafter, the multi-agent configuration system (300) can enable the multi-agents to operate according to the established plan, assist in message transmission between the multi-agents, and collect the processing results of the multi-agents to generate the final result and the corresponding answer and provide it to the user. FIG. 4 is a diagram showing an example of the entire configuration for the multi-agent configuration according to one embodiment of the present invention. FIG. 4 shows a task analyzer (410), a planner (420), a user (430), an external data source (440), a vector database (450), agent metadata (460), and a dynamic multi-agent execution module (470).

[0042] First, agent metadata (460) can be embedded and stored in a vector database (450). This embedding of agent metadata (460) can be performed by the multi-agent configuration system (300), but may also be performed by a separate entity depending on the embodiment. The vector database (450) can index metadata, such as agent descriptions, into embedding vectors and store them as agent information. When a search query is received, an embedding vector for the corresponding search query is generated, and based on this, associated agent information can be returned using an approximate nearest neighbor technique. For example, the vector database (450) can be implemented through open source solutions such as Weaviate, Milvus, and Qdrant.

[0043] When a user (430) inputs a query, the task analyzer (410) can analyze the query and generate a search query. At this time, the task analyzer (410) can search the vector database (450) using the generated search query to receive information on agent candidates. At this time, the task analyzer (410) can transmit the user's (430) query and information on the agent candidates to the planner (420). If the user's (430) query requires external data, the task analyzer (410) can access an external data source (440) to obtain the external data. In this case, the task analyzer (410) can further transmit the external data to the planner (420). Meanwhile, when external data is received, the search query can be generated further using the external data. This task analyzer (410) is a special agent based on LLM, which analyzes a user's (430) query, collects necessary external data from an external data source (440) such as the web or other external storage, and generates a search query to find an agent required to perform the task. This task analyzer (410) can be implemented using, for example, Langchain or LlamaIndex.

[0044] The planner (420) can select agents to be used using the user's (430) query, external data, and information about agent candidates, and can establish a plan for the agents' actions. At this time, the planner (420) can provide a list of selected agents (Agent List) and the established plan to the dynamic multi-agent execution module (470). This planner (420) is also a special agent based on LLM, and can establish a task execution plan by receiving the user's (430) query, collected external data, and agent candidates related to task execution. The plan can include an initial input query to be transmitted to the agents, an execution order, and a connection relationship between the agents. This planner (420) can be implemented through prompt engineering.

[0045] Additionally, the dynamic multi-agent execution module (470) can execute dynamically configured multi-agents based on a list of selected agents and an established plan. Furthermore, the dynamic multi-agent execution module (470) can generate answers to user (430) queries based on the results of executing the multi-agents, and then provide the generated answers to the user (430). This dynamic multi-agent execution module (470) can be implemented using a multi-agent platform such as Autogen.

[0046] FIG. 5 is a diagram illustrating another example of the overall configuration for a multi-agent configuration according to an embodiment of the present invention. Unlike the embodiment of FIG. 4, the embodiment of FIG. 5 illustrates an example in which an agent retriever (510) is configured separately. In the embodiment of FIG. 4, the task analyzer (410) directly searches for agent candidates in the vector database (450) through a search query, whereas in the embodiment of FIG. 5, the task analyzer (410) transmits the search query to the agent retriever (510), and the agent retriever (510) can search for agent candidates in the vector database (450) through the transmitted search query. Thereafter, the agent retriever (510) can transmit information about the searched agent candidates to the planner (420). In this case, the task analyzer (410) can transmit the user's (430) query and external data to the planner (420).

[0047] Below we describe an example of a detailed scenario where a multi-agent is configured to handle a specific task.

[0048] In this embodiment, the user's (430) query "<gitrepo url> Assume you have entered "Web Security Vulnerability Review".

[0049] In this case, the task analysis unit (410) can extract keywords related to "web security vulnerability review" and identify a Git repository URL. Thereafter, the task analysis unit (410) can clone the Git repository from the URL and store the code files in a local storage. In addition, the task analysis unit (410) can generate a search query to be used in the vector database (450) based on the code structure and the user query. Thereafter, the task analysis unit (410) can transmit the search query to the vector database (450) and receive information on agent candidates available for security vulnerability detection. In addition, the task analysis unit (410) can transmit the received information (the user's (430) query, code files as external data, and information on agent candidates) to the planner (420).

[0050] Considering the embodiment of FIG. 5 above, the search for agent candidates may be performed via the agent searcher (510). In this case, as previously described, information about the agent candidates may be transmitted to the planner (420) via the agent searcher (510).

[0051] The planner (420) can select agents. For example, the planner (420) can select the optimal agent among agent candidates through LLM. For example, for web security vulnerability review, an XSS (Cross Site Scripting) detection agent, a SQL injection (Structured Query Language injection) detection agent, etc. can be selected. In addition, the planner (420) can establish a work plan. For example, the planner (420) can plan how to cooperate with the selected agents. This plan can include definitions of the input, execution order, and interaction method of each agent. In addition, the planner (420) can transmit the list of selected agents and the final work plan to the dynamic multi-agent execution module (470).

[0052] The dynamic multi-agent execution module (470) can execute agents. For example, the dynamic multi-agent execution module (470) can execute agents on a list according to an established plan. At this time, the agents can perform code analysis of the transmitted code files according to their respective roles. In addition, the dynamic multi-agent execution module (470) can process message transmission between agents. For example, the dynamic multi-agent execution module (470) can process message transmission between agents to help the agents share analysis results and collaborate. In addition, the dynamic multi-agent execution module (470) can generate analysis results. For example, the dynamic multi-agent execution module (470) can synthesize the analysis results of all agents on the list to generate a final security vulnerability report. In addition, the dynamic multi-agent execution module (470) can provide the final results. For example, the dynamic multi-agent execution module (470) can provide the user with a generated final security vulnerability report, suggesting discovered security vulnerabilities and their remediation measures.

[0053] FIG. 6 is a flowchart illustrating an example of a multi-agent configuration method according to an embodiment of the present invention. The multi-agent configuration method according to the present embodiment can be performed by a multi-agent configuration system (300). The multi-agent configuration system (300) can be implemented by at least one computer device (200). At this time, the processor (220) of the computer device (200) can be implemented to execute control instructions according to the code of the operating system included in the memory (210) or the code of at least one computer program. Here, the processor (220) can operate according to the control instructions provided by the code stored in the computer device (200) to control the computer device (200) so that the multi-agent configuration system (300) implemented by the computer device (200) performs steps (610 to 680) included in the method of FIG. 6.

[0054] In step (610), the multi-agent configuration system (300) may receive a user query regarding the user's task. For example, the multi-agent configuration system (300) may provide a user interface through which a user accessing the system via a user terminal can input a user query, and may receive a user query entered through the user interface.

[0055] Depending on the embodiment, the user query may request external data. For example, "<gitrepo url> There may be a user query that requires work to be done through external data from an external data source, such as "review web security vulnerabilities." In this case, the multi-agent configuration system (300) can access the external data source and obtain the external data.

[0056] In step (620), the multi-agent configuration system (300) may analyze a user query to generate a search query for agent search. In some embodiments, if the user query requires external data, the multi-agent configuration system (300) may further analyze the external data to generate the search query.

[0057] In step (630), the multi-agent configuration system (300) can obtain information on candidate agents by searching a vector database that stores embedding vectors generated by embedding metadata of agents based on an artificial intelligence model using the search query. Here, the vector database may include a RAG-based vector database for an artificial intelligence model. In addition, the metadata may include at least one of a description of the agent, information on prompts to be used in the artificial intelligence model, information on external tools available to the agent, and information required to execute the agent.

[0058] Depending on the embodiment, the multi-agent configuration method may include steps (610) to (630). For example, the multi-agent configuration system (300) may obtain information about agent candidates and provide it to a separate system for operating multi-agents, and subsequent steps (640) to (680) may be performed by the separate system.

[0059] At step (640), the multi-agent configuration system (300) can select agents to perform the user's task using user queries and information about agent candidates. For example, the multi-agent configuration system (300) can select the optimal agent among agent candidates using an artificial intelligence model such as LLM.

[0060] At step (650), the multi-agent configuration system (300) may establish a plan for the actions of the selected agents based on a user query and the selected agents. Here, the plan may include at least one of an initial input query to be delivered to each of the selected agents, an execution order of the selected agents, and a connection relationship between the selected agents.

[0061] At step (660), the multi-agent configuration system (300) can dynamically execute multi-agents based on a list of selected agents and an established plan. For example, the multi-agent configuration system (300) can execute each agent according to the established plan, and each agent can perform tasks according to its role. In addition, the multi-agent configuration system (300) can support the agents to share and cooperate with each other by transmitting messages between agents.

[0062] At step (670), the multi-agent configuration system (300) can collate the processing results of each multi-agent and generate a final result and a response for the user. The multi-agent configuration system (300) can collate the sub-results processed by each multi-agent and synthesize them to generate a final result. Furthermore, the multi-agent configuration system (300) can generate a response to be delivered to the user based on the final result. For example, the response can be generated through an artificial intelligence model that receives a user query and the final result as input.

[0063] In step (680), the multi-agent configuration system (300) can provide the user with at least one of the generated final results and answers. Accordingly, the user can receive the results of processing the user's task through the multi-agents.

[0064] In this way, according to embodiments of the present invention, an agent search method and system for configuring a multi-agent system can be provided.

[0065] The systems or devices described above may be implemented as hardware components, or a combination of hardware components and software components. For example, the devices and components described in the embodiments may be implemented using one or more general-purpose computers or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing instructions and responding to them. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing device is sometimes described as being used alone; however, one of ordinary skill in the art will recognize that the processing device may include multiple processing elements and / or multiple types of processing elements. For example, a processing unit may include multiple processors, or a processor and a controller. Other processing configurations, such as parallel processors, are also possible.

[0066] Software may include a computer program, code, instructions, or a combination of one or more of these, which may configure a processing device to perform a desired operation or may independently or collectively command the processing device. The software and / or data may be embodied in any type of machine, component, physical device, virtual equipment, computer storage medium, or device for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on one or more computer-readable recording media.

[0067] The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program commands, data files, data structures, etc., singly or in combination. The medium may continuously store a computer-executable program, or may temporarily store it for execution or download. In addition, the medium may be various recording or storage means in the form of a single or multiple hardware combinations, and is not limited to a medium directly connected to a computer system, but may also be distributed over a network. Examples of the medium may include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and those configured to store program commands, including ROM, RAM, and flash memory. In addition, examples of other media may include recording or storage media managed by app stores that distribute applications, sites that supply or distribute various software, servers, etc. Examples of program instructions include machine language code, such as that produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

[0068] Although the embodiments described above have been described by way of limited examples and drawings, those skilled in the art will appreciate that various modifications and variations can be made based on the above teachings. For example, appropriate results can still be achieved even if the described techniques are performed in a different order than described, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different manner than described, or are replaced or substituted with other components or equivalents.

[0069] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

1. A multi-agent configuration method of a multi-agent configuration system implemented by at least one computer device, A step of receiving a user query regarding a user's task by at least one processor included in said at least one computer device; A step of generating a search query for searching an agent by analyzing the user query by at least one processor; and A step of obtaining information on candidate agents by searching a vector database storing embedding vectors generated by embedding metadata of agents based on an artificial intelligence model by at least one processor using the search query. A method for configuring a multi-agent including:

2. In paragraph 1, A multi-agent configuration method, characterized in that the above vector database includes a vector database based on RAG (Retrieval-Augmented Generation) for the above artificial intelligence model.

3. In paragraph 1, A multi-agent configuration method, wherein the metadata includes at least one of a description of the agent, information about a prompt to be used in the artificial intelligence model, information about external tools available to the agent, and information required to execute the agent.

4. In paragraph 1, A step of accessing an external data source and obtaining the external data by at least one processor when the user query requires external data. Including more, The steps for generating the above search query are: Generating the search query by further analyzing the above external data. A method for configuring a multi-agent comprising:

5. In paragraph 1, A step of selecting agents to perform the user's task by using the user query and information about the candidates of the agents, by the at least one processor; and A step of establishing a plan for the operation of the selected agents based on the user query and the selected agents by the at least one processor. A method for configuring a multi-agent comprising:

6. In paragraph 5, A multi-agent configuration method, wherein the plan includes at least one of an initial input query to be transmitted to each of the selected agents, an execution order of the selected agents, and a connection relationship between the selected agents.

7. In paragraph 5, A step of accessing an external data source and obtaining the external data by at least one processor when the user query requires external data. Including more, The steps for selecting the above agents are: Select agents that will perform the user's tasks by further utilizing the external data above, The steps to establish the above plan are: Establishing the above plan based further on the above external data A method for configuring a multi-agent comprising:

8. In paragraph 5, A step of executing multi-agents dynamically configured based on the list of selected agents and the established plan by at least one processor; A step of collating the processing results of each of the multi-agents by at least one processor to generate a final result and an answer for the user; and A step of providing at least one of the generated final results and answers to the user by at least one processor. A method for configuring a multi-agent comprising:

9. A computer program stored on a computer-readable recording medium for executing the method of any one of claims 1 to 8 on a computer device in combination with a computer device.

10. A computer-readable recording medium having recorded thereon a computer program for executing the method of any one of clauses 1 to 8 on a computer device.

11. In a multi-agent configuration system implemented by at least one computer device, At least one processor configured to execute instructions readable by at least one computer device, By at least one processor, Receive user queries about user tasks, Analyze the above user query to generate a search query for agent search, Obtaining information about candidate agents by searching a vector database that stores embedding vectors generated by embedding metadata of agents based on artificial intelligence models using the search query. A multi-agent configuration system featuring:

12. In paragraph 11, A multi-agent configuration system, characterized in that the above vector database includes a RAG (Retrieval-Augmented Generation) based vector database for the above artificial intelligence model.

13. In paragraph 11, A multi-agent configuration system, wherein the metadata includes at least one of a description of the agent, information about prompts to be used in the artificial intelligence model, information about external tools available to the agent, and information required to execute the agent.

14. In paragraph 11, By at least one processor, Select agents to perform the user's task using the information about the user's query and the candidates for the agent, Establishing a plan for the actions of the selected agents based on the user query and the selected agents. A multi-agent configuration system featuring:

15. In paragraph 14, By at least one processor, Execute dynamically configured multi-agents based on the list of selected agents and the established plan, Compile the processing results of each of the above multi-agents to generate a final result and an answer for the user, Providing at least one of the generated final results and answers to the user. A multi-agent configuration system featuring:

Images