AI cooperation method and cooperation system

Abstract

The embodiment of the invention discloses an AI cooperation method and system, the AI cooperation system comprises an interaction agent and a communication agent, the communication agent comprises a primary communication agent and a secondary communication agent, and the method comprises the following steps: if the interaction agent detects that a project initiator initiates a target project, starting the target project; if yes, inviting a first-level communication agent matched with the target project to join the target project; the first-level communication agent obtains total demand information of the target project and splits the target project according to the total demand information to obtain a plurality of sub-projects and a plurality of pieces of sub-demand information, and one sub-project is at least matched with one piece of sub-demand information; and the primary communication agent invites the secondary communication agent to join the target project according to the sub-demand information, so that the secondary communication agent completes the sub-project matched with the sub-demand information. The method can ensure that the final output result is consistent with the original demand information.

Description

Technical Field

[0001] The present invention relates to the field of AI technology, and in particular to an AI collaboration method and collaboration system. Background Art

[0002] AIGC (AI-Generated Content) refers to the automatic generation of text, images, audio, video and other content through artificial intelligence technology. At present, AIGC is usually achieved by using AI with deep thinking ability or targeted long text output ability in the automatic generation of long texts. This type of AI is generally used to output solutions to some complex problems, or some codes with strong internal logic but simple functional descriptions. When the target description is more complex and the text content is more scattered, this type of AI will lose some details due to too much information input during the interaction and insufficient TOKEN. At the same time, it is easy to cause the final output result to deviate greatly from the original demand. Summary of the invention

[0003] The embodiments of the present invention provide an AI collaboration method and a collaboration system, which aim to solve the problem that the final output results of the current AIGC in long text generation deviate greatly from the original requirements.

[0004] In a first aspect, an embodiment of the present invention provides an AI collaboration method, which is applied to an AI collaboration system, wherein the AI ​​collaboration system includes an interactive agent and a communication agent, wherein the communication agent includes a primary communication agent and a secondary communication agent, and the method includes: If the interactive agent detects that the project initiator initiates a target project, it invites the primary communication agent matching the target project to join the target project; The first-level communication agent obtains the total demand information of the target project, and disassembles the target project according to the total demand information to obtain multiple sub-projects and multiple sub-demand information, wherein one sub-project matches at least one sub-demand information; The primary communication agent invites the secondary communication agent to join the target project according to the sub-demand information so that the secondary communication agent completes the sub-project matching the sub-demand information.

[0005] In a second aspect, an embodiment of the present invention further provides an AI collaboration system, which is configured with any of the AI ​​collaboration methods described above, and the AI ​​collaboration system includes an interactive agent, a communication agent, and a program agent; the interactive agent is used to provide an interactive interface with the project initiator; the communication agent includes a primary communication agent and a secondary communication agent, and both the primary communication agent and the secondary communication agent are used to complete the target project; the program agent is used to complete predetermined functions related to the target project.

[0006] The embodiment of the present invention provides an AI collaboration method and collaboration system. The AI ​​collaboration system includes an interactive agent and a communication agent, wherein the communication agent includes a primary communication agent and a secondary communication agent, and the method includes: if the interactive agent detects that the project initiator initiates a target project, the primary communication agent matching the target project is invited to join the target project; the primary communication agent obtains the total demand information of the target project, and decomposes the target project according to the total demand information to obtain multiple sub-projects and multiple sub-demand information, wherein one sub-project matches at least one sub-demand information; the primary communication agent invites the secondary communication agent to join the target project according to the sub-demand information so that the secondary communication agent completes the sub-project matching the sub-demand information. According to an embodiment of the present invention, when a project initiator initiates a target project, the first-level communication agent that matches the target project can be invited to join the target project. The first-level communication agent can decompose the target project according to the total demand information to split the target project into multiple sub-projects, and split the total demand information into multiple sub-demand information, and one sub-project matches at least one sub-demand information. Then, the first-level communication agent invites secondary communication agents that match each sub-project to join the target project, so that multiple agents can collaborate to complete the target project at the same time to ensure that the final output result is consistent with the original demand information of the project initiator. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0008] Figure 1 is a flowchart of an AI collaboration method provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of a first sub-process of the AI ​​collaboration method provided by an embodiment of the present invention; Figure 3is a schematic diagram of a second sub-process of the AI ​​collaboration method provided in an embodiment of the present invention; Figure 4 3 is a schematic diagram of a third sub-process of the AI ​​collaboration method provided in an embodiment of the present invention; Figure 5 4 is a schematic diagram of a fourth sub-process of the AI ​​collaboration method provided in an embodiment of the present invention; Figure 6 is a schematic diagram of the fifth sub-process of the AI ​​collaboration method provided in an embodiment of the present invention; Figure 7 This is a sixth sub-process diagram of the AI ​​collaboration method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0009] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0010] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, operations, elements, components and / or groups thereof.

[0011] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms of "a", "an" and "the" are intended to include plural forms. It should also be further understood that the term "and / or" used in the specification of the present invention and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0012] See also Figure 1 , Figure 1 : is a flowchart of the AI ​​collaboration method provided by an embodiment of the present invention. The AI ​​collaboration method of the embodiment of the present invention applies an AI collaboration system, which includes an interactive agent, a communication agent and a program agent. The communication agent includes a primary communication agent and a secondary communication agent, which is used to control multiple different agents to collaborate to complete the target project. Figure 1 As shown, the method includes steps S110 to S130.

[0013] S110: If the interactive agent detects that the project initiator initiates a target project, it invites a primary communication agent matching the target project to join the target project.

[0014] In an embodiment of the present invention, the AI ​​collaboration system may include a variety of different types of agents, for example, it may include a communication agent, an interactive agent, and a program agent. The communication agent is mainly responsible for communicating with the AI ​​large model, and can use pre-made or configurable prompt words to process information from the system kernel and send the processing results to the program of the system kernel. The interactive agent is mainly used to provide an operation interface and a communication channel between the user and the system kernel. Through the interactive agent, the user and the communication agent can interact. The program agent is used to implement specific encoded functions, for example, to test the content encoded by the communication agent, which is usually expressed as traditional program coding. Among them, the system kernel is mainly used to manage and coordinate all agents. It also serves as a communication hub for all agents, responsible for receiving and forwarding messages, ensuring the flow of information between different agents, and the system kernel can also provide standardized structures and protocols, allowing agents to be deployed independently and interact with the kernel through two-way communication.

[0015] For a specific project, the communication agent may include a primary communication agent and a secondary communication agent, and the number of primary communication agents and secondary communication agents may be one or more. For example, a project may include a primary communication agent and multiple secondary communication agents, wherein, within the project, the primary communication agent is mainly used to complete the project and coordinate the work of the secondary communication agents, but outside the project, the various agents may have no connection.

[0016] The project initiator can be a user, and the user can create a project according to his or her own needs to complete the initiation of the target project. For example, a user can create a top-level project "Technical Book Writing" through an interactive agent, and the target project is "Technical Book Writing". After the target project is created, a first-level communication agent can be invited to join the target project, wherein the user can invite a designated first-level communication agent to join, or the interactive agent can invite a first-level communication agent that matches the target project to join. For example, for "Technical Book Writing", the first-level communication agent can be a "document generator", that is, the first-level communication agent corresponds to a large model that is good at generating documents. After inviting a first-level communication agent to join, the first-level communication agent is mainly responsible for the framework work of the target project. For example, the first-level communication agent can generate a directory based on the needs input by the user, and the generated directory includes multiple chapters.

[0017] S120, the first-level communication agent obtains the total demand information of the target project, and disassembles the target project according to the total demand information to obtain multiple sub-projects and multiple sub-demand information, wherein one of the sub-projects matches at least one of the sub-demand information.

[0018] In an embodiment of the present invention, the project initiator, that is, the user, can input original demand information through the interactive agent. The first-level communication agent understands the original demand information and obtains the total demand information of the target project, and decomposes the target project according to the total demand information to obtain multiple sub-projects and multiple sub-demand information.

[0019] The user can input relatively complex original demand information through the interactive agent, and the first-level communication agent can obtain and understand the original demand information, and then generate the total demand information after understanding, and ask the user to confirm whether the understanding is wrong through the interactive agent. When there is no problem, the first-level communication agent disassembles the target project according to the total demand information to obtain multiple sub-projects and multiple sub-demand information. For example, the target project can be "AI technical book writing", and the total demand information can include three parts: technical principles, application cases, and future trends. Then the first-level communication agent disassembles the target project according to the total demand information to obtain three sub-projects: AI technical principles, AI application cases, and AI future trends. The AI ​​technical principles correspond to the sub-demand information that explains the AI ​​principles, the AI ​​application cases correspond to the sub-demand information that explains the AI ​​applications, and the AI ​​future trends correspond to the sub-demand information that explains the AI ​​future trends. Each sub-project can correspond to multiple sub-demand information. For example, in the sub-project of AI technical principles, in addition to the sub-demand information that corresponds to the AI ​​principles, it can also contain sub-demand information that focuses on a certain point.

[0020] See also Figure 2 In some embodiments, such as the present embodiment, the step S120 may include steps S121-S122.

[0021] S121, the first-level communication agent disassembles the target project according to the total demand information to obtain a tree-like project set, wherein the tree-like project set includes a plurality of sub-projects; S122, the first-level communication agent allocates the sub-demand information to each sub-item according to the total demand information.

[0022] In an embodiment of the present invention, a user can create a target project through the system kernel. A target project can contain multiple sub-projects, and a tree-structured project collection can be realized by setting a "parent project". The creation of sub-projects can be created by the user, or by the first-level communication agent based on the total demand information and the target project. Complex projects can be disassembled through a tree-like project collection. For example, for the target project of "AI technical book writing", its sub-projects are AI technical principles, AI application cases, and AI future trends. One or more sub-projects can be created under each sub-demand based on the total demand information. In addition, for each project, it can also be divided into multiple stages, and each stage can be configured with a corresponding name. In each stage of a project, the goals within the stage need to be completed first, and after confirmation by the participants with audit responsibilities, the participants with process control responsibilities are responsible for opening a new stage. For example, each stage can be confirmed by the first-level communication agent and is responsible for opening a new stage.

[0023] By disassembling complex projects, a target project can be decomposed into multiple levels. Each level can include multiple sub-projects. The number of levels can be determined by the user. When the user thinks that the disassembly can be continued, the target project can be further disassembled until the user approves. When the project in the next level is completed, its results can be summarized as part of the project in the previous level, and traced back layer by layer until the goal of the original top-level project is completed.

[0024] S130, the primary communication agent invites the secondary communication agent to join the target project according to the sub-demand information so that the secondary communication agent completes the sub-project matching the sub-demand information.

[0025] In the embodiment of the present invention, the primary communication agent can invite the matching secondary communication agent to join the target project according to the sub-demand information, and the secondary communication agent is used to complete the sub-project corresponding to itself. By splitting the total demand information into multiple sub-demand information and completing them by the secondary communication agent associated with the sub-demand information, the information that a single agent needs to understand is reduced, ensuring that the final output information is consistent with the original demand information.

[0026] When there are multiple sub-projects, the user or the primary communication agent can coordinate the secondary communication agents to process the sub-projects in parallel. For example, the user can select multiple sub-projects at the same time and ask the corresponding communication agents to process the selected sub-projects in parallel at the same time, thereby improving the efficiency of processing the projects. The primary communication agent can also determine the projects that can be processed in parallel by itself, and then ask the user to confirm through the interactive agent. When the user approves, it will coordinate the corresponding communication agent to process the project in parallel. In addition, a sub-project can correspond to one or more secondary communication agents, and multiple secondary communication agents complete their respective tasks according to their respective sub-demand information.

[0027] For example, the target project is "AI Ethics and Social Research", and the total demand information includes table of contents, chapter content, case analysis and references. The first-level communication agent is a "book architect", which is responsible for structured content generation, such as generating table of contents, assigning chapter tasks and coordinating the work of secondary communication agents. Secondary communication agents can include "ethics experts" and "technical commentators". "Ethics experts" are used to write ethical theory analysis and review sensitive content, and "technical commentators" are used to explain the principles of AI technology. The program agent is a "literature crawler", which is used to automatically capture academic papers and organize them into reference format.

[0028] Users can create projects through interactive agents and upload original demand information, such as book outlines. Then users can invite "book architects" to join the target project. "Book architects" understand the original demand information and generate total demand information. After the user confirms that there are no problems, the target project is broken down into two sub-projects: introduction and technical foundation. At the same time, two sub-projects can be set under the introduction: the history of AI development and the origin of ethical issues. Under the technical foundation, two sub-projects are set: machine learning principles and model interpretability challenges. "Book architects" invite "ethics experts" and "technical commentators" to join the project, and assign the origin of ethical issues to "ethics experts" and the principle of machine learning to "technical commentators". That is, "ethics experts" are responsible for the sub-project of the origin of ethical issues, and "technical commentators" are responsible for the sub-project of the principle of machine learning. The rest of the sub-projects are in charge of "book architects". "Literature crawlers" are responsible for crawling relevant papers and organizing them into references. After writing each sub-project, the corresponding agent will be displayed to the user through the interactive agent, and the user will confirm whether it needs to be modified. If it needs to be modified, it will be adjusted according to the user's modification requirements until the user approves it. After all sub-projects are completed, the "book architect" will summarize all the content and generate a document, which will be saved in the resource warehouse.

[0029] In addition to writing books and papers, it is also possible to automatically collaborate on generating web front ends. For example, the target project is a "responsive website", the total demand information is to develop a responsive website, allowing users to provide real-time feedback and modify requirements through visual operations, the first-level communication agent is a "project manager", the secondary communication agents can have "front-end developers" and "UI designers", and the program agent can have a "tester". The "project manager" is responsible for structuring content and coordinating secondary communication agents, the "front-end developer" is responsible for generating HTML / CSS code, the "UI designer" is responsible for generating the initial interface diagram, and the "tester" deploys the code to the test environment, runs automated test cases for testing, and feeds back the test results.

[0030] Users can create projects and upload original demand information through interactive agents. The "project manager" outputs the total demand information based on the original demand information and disassembles it to obtain three sub-projects: home page, product page, and backend management. The "front-end developer" and "UI designer" are invited to participate in each sub-project, that is, the "front-end developer" and "UI designer" participate in the three sub-projects at the same time. The "front-end developer" generates HTML / CSS code and uploads it to the resource warehouse. Users can preview the page through the web debugging terminal, that is, the interactive agent, and click on the content that needs to be modified. For example, the user can click "the navigation bar color does not match the brand color", and the "front-end developer" can locate the corresponding position and adjust it according to the user's modification information, and then re-upload it. The "UI designer" generates the initial interface diagram. The user uses the interactive agent to mark "the button position needs to be moved 10px to the right", and the "UI designer" iterates and regenerates the initial interface diagram according to the user's modification information. When multiple sub-projects can be developed in parallel, the "project manager" can coordinate the corresponding secondary communication agents to develop in parallel. If it is necessary to develop step by step, the "project manager" can also coordinate the secondary communication agents of the next stage to participate in the development after completing one stage.

[0031] In the process of completing the project, the results generated by each agent need to be confirmed by the user. When the user confirms that it is passed, the work of the agent is considered completed. If the user confirms that it is not passed, the modification plan can be confirmed with the user in the form of a dialogue. For example, when the "ethics expert" completes the sub-project of the origin of ethical issues, the generated content will be submitted to the user for confirmation. If the user confirms that there is no problem, the relevant content can be stored in the preset resource warehouse. If the user confirms that the content needs to be modified, the place that needs to be modified and the modification opinions can be confirmed in the form of a dialogue, and then the relevant content can be regenerated.

[0032] See also Figure 3 In some embodiments, such as the present embodiment, the AI ​​collaboration method may further include steps S140-S144.

[0033] S140, the first-level communication agent obtains the original demand information of the project initiator, and generates a feedback text based on the original demand information; S141, the primary communication agent outputs the feedback text to the interactive agent so that the project initiator verifies the feedback text; S142, if the primary communication agent receives the first verification information fed back by the project initiator, the feedback text is used as the total demand information; S143, if the primary communication agent receives the second verification information fed back by the project initiator indicating that the verification has failed, modify the feedback text according to the second verification information to obtain a modified feedback text; S144, outputting the modified feedback text to the interactive agent so that the project initiator verifies the modified feedback text until the modified feedback text passes the verification.

[0034] In an embodiment of the present invention, the primary communication agent can communicate with the user through the interactive agent. After the user uploads the original demand information, the primary communication agent obtains the original demand information and understands it, and then outputs the feedback text to the interactive agent, so that the user can verify the feedback text. When the verification passes, the feedback text is used as the total demand information. When the verification fails, a dialogue can be conducted with the user to confirm where modifications need to be made and the modifications are made until the feedback text passes. Among them, the first verification information is used to indicate that the verification has passed, and the second verification information is used to indicate that the verification has failed.

[0035] It can be understood that in addition to the primary communication agent being able to confirm with the user whether there are any errors in understanding through feedback text, the secondary communication agent can also confirm whether there are any errors in its understanding of the sub-demand information through feedback text. When the understanding is correct, the secondary communication agent will start working. If there are any errors in understanding, it can communicate with the user through dialogue to make modifications until the user confirms that there are no errors.

[0036] See also Figure 4 In some embodiments, such as the present embodiment, the AI ​​collaboration method may further include steps S150-S152.

[0037] S150, the secondary communication agent confirms whether the invitation information is received; S151, if the secondary communication agent receives the invitation information, confirming whether to join the sub-project matching the invitation information according to the invitation information, wherein the invitation information includes the sub-demand information; S152: If the secondary communication agent confirms to join the sub-project, it associates itself with the sub-project to make itself a participant of the sub-project.

[0038] In an embodiment of the present invention, a primary communication agent or a user can invite a secondary communication agent to join a target project. When a secondary communication agent receives an invitation message, it can confirm whether to join the target project based on the sub-requirement information in the invitation message. When confirming, the secondary communication agent mainly confirms whether it matches the content in the sub-requirement information. For example, if the sub-requirement information is the principle of machine learning, and the invited secondary communication agent is a "technical commentator", then the two match, and the secondary communication agent can join the target project. If the invited secondary communication agent is an "ethics expert", then the two do not match, and the secondary communication agent can refuse to participate in the target project. In addition, the secondary communication agent can also actively participate in a target project. When the user or the primary communication agent agrees, it can participate in the project.

[0039] See also Figure 5 In some embodiments, such as the present embodiment, the AI ​​collaboration method may further include steps S153-S154.

[0040] S153, the participant includes a first state identifier and a second state identifier, and when the participant is configured as the first state identifier, the participant can obtain first-type conversation information of the target project, wherein the first-type conversation information is a public conversation of the target project; S154: When the participant is configured as the second state identifier, the participant may obtain second-type conversation information of the target project, wherein the second-type conversation information is a private conversation associated with the participant.

[0041] In the embodiment of the present invention, when an agent participates in a project, it becomes a participant of the project. The participant is associated with the project and is configured with a name and parameters. The name is used to distinguish different participants, and the parameters are used to distinguish the functions and settings of different participants. The participant information between different projects does not affect each other. In the same project, an agent can have multiple participant identities.

[0042] Participants and participation invitations are for a specific project. For a project, the agents or users participating in the project can be participants of the project. Participants can send participation invitations to invite other agents to join the project.

[0043] Participants can speak in the project and exchange information with other participants through speaking. Speech will be recorded in the database of the system kernel in the form of dialogue and pushed to relevant participants. For example, if agent A mentions agent B and agent C in his speech, the speech of agent A will be pushed to agent B and agent C at the same time. Speeches can include private speeches and public speeches, that is, private conversations and public conversations. In public speeches, a speech can mention 0 to multiple participants, that is, in public speeches, no participant can be mentioned, or multiple participants can be mentioned at the same time. In private speeches, a speech needs to mention at least one participant. Participants can configure whether each speech is a public speech or a private speech. In addition, in addition to text, a conversation can also be accompanied by an "additional data" to transmit formatted data related to this conversation. It is generally encoded in JSON format and sent with agreed attributes, so that participants can understand the relevant operation information of this speech simply and without objection, so as to perform corresponding actions.

[0044] Participants can have an "active" status. When the "active" status is set to enabled, that is, the participant is configured as the first status identifier, the participant can receive all public speeches in the project and all private speeches that mention themselves. When the "active" status is set to off, that is, the participant is configured as the second status identifier, the participant can only receive private speeches that mention themselves. Regardless of whether the participant is in the first status identifier or the second status identifier, the private speech that mentions the participant itself can be received by the participant, and if it is in the second status identifier, the participant cannot receive public speeches and can only receive private speeches that mention themselves.

[0045] In addition, all project participants can obtain all public speeches of the project through the system kernel, even those made before the participants joined the project. For example, if a secondary communication agent joins the project at time A, the secondary communication agent can obtain all public speeches of the project before time A through the system kernel, which helps the secondary communication agent understand the project situation.

[0046] See also Figure 6 In some embodiments, such as the present embodiment, the AI ​​collaboration method may further include steps S160-S162.

[0047] S160, the secondary communication agent generates a feedback text to the interactive agent according to the sub-demand so that the project initiator verifies the feedback text; S161, if the secondary communication agent receives the third verification information fed back by the project initiator indicating that the verification has passed, then encode the sub-project according to the feedback text; S162: If the secondary communication agent completes the encoding of the sub-project, the project file generated during the encoding process is stored in a preset resource warehouse.

[0048] In the embodiment of the present invention, the secondary communication agent can also generate feedback text according to the sub-demand information, which is convenient for users to verify. When the verification is passed, the secondary communication agent starts coding or writing documents. When the coding is completed, the project file is stored in the preset resource warehouse, where each project has a preset resource warehouse, and the participants of the project can access resources in the preset resource warehouse. Resources are stored in the form of files, and the file type, upload time, upload participants and other information are recorded for screening and viewing. The preset resource warehouse is used to save the original files, intermediate files, target files and related working files of the project.

[0049] See also Figure 7 In some embodiments, such as the present embodiment, the AI ​​collaboration method may further include steps S170-S171.

[0050] S170, if the primary communication agent receives the confirmation message sent by the project initiator confirming the completion of the project, then invite the program agent to join the target project; S171, the program agent verifies the target project to generate a verification result, and outputs the verification result to the interactive agent.

[0051] In the embodiment of the present invention, the program agent is an automated tool for implementing specific functions. It does not rely on AI models and only executes preset logic. For example, it adjusts the layout of documents to meet the requirements of publishers, automatically uploads or downloads files, connects to databases and API interfaces, synchronizes real-time data, detects code errors, etc. When the project is completed, a matching program agent can be invited to join, and the program agent verifies the target project, and then outputs the verification results to the interactive agent. If there are errors, the corresponding agent can make adjustments.

[0052] The present invention also provides an AI collaboration system, which is configured with the AI ​​collaboration method described in any one of the above embodiments, and the AI ​​collaboration system includes an interactive agent, a communication agent and a program agent; the interactive agent is used to provide an interactive interface with the project initiator; the communication agent includes a primary communication agent and a secondary communication agent, and the primary communication agent and the secondary communication agent are both used to complete the target project; the program agent is used to complete predetermined functions related to the target project.

[0053] Specifically, the AI ​​collaboration system may include a variety of different types of agents, for example, it may include communication agents, interactive agents and program agents. The communication agent is mainly responsible for communicating with the AI ​​large model, and can use pre-made or configurable prompt words to process information from the system kernel and send the processing results to the program of the system kernel. The interactive agent is mainly used to provide an operation interface and a communication channel between the user and the system kernel. Through the interactive agent, the user and the communication agent can interact. The program agent is used to implement specific coded functions, for example, to test the content encoded by the communication agent, which is usually expressed as traditional program coding. Among them, the system kernel is mainly used to manage and coordinate all agents. It also serves as a communication hub for all agents, responsible for receiving and forwarding messages, ensuring the flow of information between different agents, and the system kernel can also provide standardized structures and protocols, allowing agents to be deployed independently and interact with the kernel through two-way communication.

[0054] For a specific project, the communication agent may include a primary communication agent and a secondary communication agent, and the number of primary communication agents and secondary communication agents may be one or more. For example, a project may include a primary communication agent and multiple secondary communication agents, wherein, within the project, the primary communication agent is mainly used to complete the project and coordinate the work of the secondary communication agents, but outside the project, the various agents may have no connection.

[0055] The project initiator can be a user, and the user can create a project according to his or her own needs to complete the initiation of the target project. For example, a user can create a top-level project "Technical Book Writing" through an interactive agent, and the target project is "Technical Book Writing". After the target project is created, a first-level communication agent can be invited to join the target project, wherein the user can invite a designated first-level communication agent to join, or the interactive agent can invite a first-level communication agent that matches the target project to join. For example, for "Technical Book Writing", the first-level communication agent can be a "document generator", that is, the first-level communication agent corresponds to a large model that is good at generating documents. After inviting a first-level communication agent to join, the first-level communication agent is mainly responsible for the framework work of the target project. For example, the first-level communication agent can generate a directory based on the needs input by the user, and the generated directory includes multiple chapters.

[0056] The AI ​​collaboration method and collaboration system disclosed in the present invention can invite multiple secondary communication agents to join the target project, and the primary communication agent coordinates multiple secondary communication agents to complete the project, thereby simplifying the project and ensuring that the final output result is consistent with the original demand information.

[0057] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0058] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

[0059] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An AI collaboration method, characterized in that: Applied to an AI collaboration system, the AI ​​collaboration system includes an interactive agent and a communication agent, the communication agent includes a primary communication agent and a secondary communication agent, the method includes: If the interactive agent detects that the project initiator initiates a target project, it invites the primary communication agent matching the target project to join the target project; The first-level communication agent obtains the total demand information of the target project, and disassembles the target project according to the total demand information to obtain multiple sub-projects and multiple sub-demand information, wherein one sub-project matches at least one sub-demand information; The primary communication agent invites the secondary communication agent to join the target project according to the sub-demand information so that the secondary communication agent completes the sub-project matching the sub-demand information.

2. The method according to claim 1, characterized in that The step of decomposing the target project according to the total demand information to obtain a plurality of sub-projects and a plurality of sub-demand information includes: The first-level communication agent disassembles the target project according to the total demand information to obtain a tree-like project set, wherein the tree-like project set includes a plurality of sub-projects; The primary communication agent allocates the sub-demand information to each sub-item based on the total demand information.

3. The method according to claim 1, characterized in that The method further comprises: The first-level communication agent obtains the original demand information of the project initiator and generates a feedback text based on the original demand information; The primary communication agent outputs the feedback text to the interactive agent so that the project initiator can verify the feedback text; If the primary communication agent receives the first verification information fed back by the project initiator indicating that the verification has passed, the feedback text is used as the total demand information.

4. The method according to claim 3, characterized in that The method further comprises: If the primary communication agent receives the second verification information fed back by the project initiator indicating that the verification has failed, the feedback text is modified according to the second verification information to obtain a modified feedback text; The modified feedback text is output to the interactive agent so that the project initiator verifies the modified feedback text until the modified feedback text passes the verification.

5. The method according to claim 1, characterized in that The method further comprises: The secondary communication agent confirms whether the invitation information is received; If the secondary communication agent receives the invitation information, it confirms whether to join the sub-project matching the invitation information according to the invitation information, wherein the invitation information includes the sub-demand information; If the secondary communication agent confirms to join the sub-project, it will associate itself with the sub-project to make itself a participant of the sub-project.

6. The method according to claim 5, characterized in that The method further comprises: The participant includes a first state identifier and a second state identifier. When the participant is configured as the first state identifier, the participant can obtain first-type conversation information of the target project, wherein the first-type conversation information is a public conversation of the target project; When the participant is configured as the second state identifier, the participant can obtain the second type of conversation information of the target project, wherein the second type of conversation information is a private conversation associated with the participant.

7. The method according to claim 1, characterized in that The method further comprises: The secondary communication agent generates a feedback text to the interactive agent according to the sub-demand so that the project initiator verifies the feedback text; If the secondary communication agent receives the third verification information fed back by the project initiator indicating that the verification has passed, the sub-project is encoded according to the feedback text.

8. The method according to claim 7, characterized in that The method further comprises: If the secondary communication agent completes the encoding of the sub-project, the project file generated during the encoding process is stored in the preset resource warehouse.

9. The method according to claim 1, characterized in that The method further comprises: If the primary communication agent receives the confirmation message sent by the project initiator confirming the completion of the project, the program agent is invited to join the target project; The program agent verifies the target item to generate a verification result, and outputs the verification result to the interactive agent.

10. An AI collaboration system, characterized in that: The AI ​​collaboration system is configured with the AI ​​collaboration method according to any one of claims 1 to 9, and the AI ​​collaboration system comprises: Interactive agent, used to provide an interactive interface with the project initiator; A communication agent, including a primary communication agent and a secondary communication agent, wherein both the primary communication agent and the secondary communication agent are used to complete a target project; A program agent is used to complete predetermined functions related to the target project.

Images