Intelligent agent participated group chat session method and device, electronic equipment and storage medium

Abstract

The invention provides an agent-participated group chat session method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence such as large models, agents, human-computer interaction and intelligent agents. The method is applied to a main agent and comprises the steps of determining a target dialogue needing to be processed in a group chat form in dialogue content initiated by a first user; determining an additional group chat participation object which at least comprises one sub-agent and selectively comprises a second user according to the dialogue demand of the target dialogue; creating a group chat session entry containing the first user, the group chat participating object and the main agent in the session where the target session is located; and in response to the first user entering the target group chat session through the group chat session entry, bringing the preceding text information related to the target session into the target group chat session, and controlling the sub-agent to reply the first user in the target group chat session based on the target session and the preceding text information. According to the method, the group chat session can be automatically created based on the perceived group chat demand, and the cooperation efficiency is improved.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, specifically to the fields of artificial intelligence technology such as large models, intelligent agents, human-computer interaction, and intelligent agents, and particularly to a method, device, electronic device, computer-readable storage medium, and computer program product for group chat sessions involving intelligent agents. Background Technology

[0002] With the continuous development of artificial intelligence, agent technology has received increasing attention and has gradually become an important research topic in the field of artificial intelligence.

[0003] Most current intelligent agents provide conversational interaction with users in point-to-point chat windows in applications similar to instant messaging (IM), which makes it difficult to meet diverse collaboration needs. Summary of the Invention

[0004] This disclosure provides a method, apparatus, electronic device, computer-readable storage medium, and computer program product for group chat sessions involving intelligent agents.

[0005] In a first aspect, embodiments of this disclosure propose a group chat session method involving intelligent agents, applied to a main intelligent agent. The method includes: determining a target dialogue that needs to be processed in the form of a group chat from the dialogue content initiated by a first user; determining additional group chat participants based on the dialogue requirements of the target dialogue; wherein the group chat participants include at least one sub-intelligent agent distinct from the main intelligent agent and selectively include a second user distinct from the first user; creating a group chat session entry point in the session containing the first user, the group chat participants, and the main intelligent agent; in response to the first user entering the target group chat session through the group chat session entry point, bringing the context information related to the target dialogue into the target group chat session, and controlling the sub-intelligent agent to reply to the first user in the target group chat session based on the target session and the context information.

[0006] Secondly, embodiments of this disclosure propose a group chat session device involving intelligent agents, applied to a main intelligent agent. The device includes: a target dialogue determination unit, configured to determine a target dialogue that needs to be processed in the form of a group chat from the dialogue content initiated by a first user; a group chat participant determination unit, configured to determine additional group chat participants based on the dialogue requirements of the target dialogue; wherein the group chat participants include at least one sub-intelligent agent distinct from the main intelligent agent and selectively include a second user distinct from the first user; a group chat session entry creation unit, configured to create a group chat session entry containing the first user, group chat participants, and the main intelligent agent in the session where the target dialogue is located; and a context import and control response unit, configured to, in response to the first user entering the target group chat session through the group chat session entry, import context information related to the target dialogue into the target group chat session and control the sub-intelligent agent to respond to the first user in the target group chat session based on the target session and the context information.

[0007] Thirdly, embodiments of this disclosure provide an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to implement a group chat session method involving intelligent agents as described in the first aspect.

[0008] Fourthly, embodiments of this disclosure provide a non-transitory computer-readable storage medium storing computer instructions that enable a computer, when executed, to implement a group chat session method involving intelligent agents as described in the first aspect.

[0009] Fifthly, embodiments of this disclosure provide a computer program product including a computer program that, when executed by a processor, can implement the steps of the group chat session method involving intelligent agents as described in the first aspect.

[0010] The group chat solution involving intelligent agents disclosed herein proactively identifies target dialogues requiring group chat processing within a one-on-one conversation with the first user. Based on the dialogue requirements, it dynamically determines the group chat participants, including specialized sub-agents (providing targeted capabilities) and optional second users (connected to social relationships). It then directly creates a native group chat entry point containing all participants within the one-on-one conversation and automatically synchronizes the one-on-one chat context information after the user enters the group chat. This achieves a seamless migration from one-on-one to group chat, eliminating the need for users to manually create groups, add members, and forward history. Simultaneously, sub-agents directly reply to users in the group chat based on the target dialogue content and synchronized context information, maintaining the continuity and naturalness of the dialogue. This solution satisfies both professional problem-solving needs and integrates social collaboration scenarios, significantly improving the efficiency and user experience of group chat collaboration.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0012] Other features, objects, and advantages of this disclosure will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is an exemplary system architecture to which this disclosure can be applied; Figure 2 A flowchart illustrating a group chat method involving intelligent agents, provided as an embodiment of this disclosure; Figure 3 A flowchart illustrating a method for determining a first target dialogue and determining group chat participants, provided in this embodiment of the disclosure; Figure 4 A branch diagram illustrating four different scenarios of group chat session methods provided in the embodiments of this disclosure; Figure 5 A flowchart of a method for determining a second target dialogue, determining group chat participants, and controlling the replies of group chat participants provided in this embodiment of the disclosure; Figure 6 A flowchart illustrating a method for determining and presenting cause information provided in an embodiment of this disclosure; Figure 7 A flowchart illustrating a cross-platform group chat method between a first user and a second user, provided as an embodiment of this disclosure; Figures 8-1 to 8-4 These are all schematic diagrams of different interfaces in the group chat session scheme provided in the embodiments of this disclosure; Figure 9 A structural block diagram of a group chat device involving intelligent agents provided in an embodiment of this disclosure; Figure 10 This is a schematic diagram of the structure of an electronic device suitable for performing a group chat session method involving intelligent agents, as provided in an embodiment of this disclosure. Detailed Implementation

[0013] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding; these should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description. It should be noted that, unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0014] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0015] Figure 1 An exemplary system architecture 100 is shown, in which embodiments of the intelligent agent participation methods, apparatuses, electronic devices, and computer-readable storage media of this disclosure can be applied.

[0016] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0017] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various applications for enabling information communication between the terminal devices 101, 102, and 103 and server 105 can be installed. These applications include interactive dialog applications, large model tool platform applications, and instant messaging applications.

[0018] Terminal devices 101, 102, and 103 and server 105 can be either hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with displays, including but not limited to smartphones, tablets, laptops, and desktop computers. When terminal devices 101, 102, and 103 are software, they can be installed in the aforementioned electronic devices, and can be implemented as multiple software programs or software modules, or as a single software program or software module; no specific limitation is made here. When server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When server 105 is software, it can be implemented as multiple software programs or software modules, or as a single software program or software module; no specific limitation is made here.

[0019] Terminal devices 101, 102, 103, and server 105 can provide various services through built-in applications. Taking a dialogue interaction application (with a main agent running on it) that can provide collaborative group chat services based on a large model as an example, terminal devices 101, 102, 103, and server 105 can achieve the following effects when running this dialogue interaction application: First, in the dialogue content initiated by the first user, the target dialogue that needs to be processed in the form of a group chat is determined; then, according to the dialogue requirements of the target dialogue, additional group chat participants are determined, including at least one sub-agent different from the main agent and, selectively, a second user different from the first user; next, a group chat session entry containing the first user, group chat participants, and the main agent is created in the session where the target dialogue is located; next, in response to the first user entering the target group chat session through the group chat session entry, the context information related to the target dialogue is brought into the target group chat session, and the sub-agent is controlled to reply to the first user in the target group chat session based on the target session and the context information.

[0020] Since analyzing and determining the target dialogue and invoking and controlling the intelligent agent to respond requires significant computing resources and capabilities, the group chat methods involving intelligent agents provided in the subsequent embodiments of this disclosure are generally executed by a server 105 with strong computing power and abundant computing resources (i.e., the underlying capabilities provided by the main intelligent agent are provided by the base model deployed on the server 105). Correspondingly, the group chat device involving intelligent agents is also generally located in the server 105. However, it should also be noted that when the terminal devices 101, 102, and 103 also possess sufficient computing power and resources, the terminal devices 101, 102, and 103 can also complete the aforementioned calculations performed by the server 105 through the dialogue interaction applications installed on them, thereby outputting the same results as the server 105. Especially when multiple terminal devices with different computing capabilities exist simultaneously, but the dialogue-based application determines that the terminal device it is using has strong computing power and sufficient remaining computing resources, the terminal device can perform the aforementioned calculations, thereby appropriately reducing the computing pressure on server 105. Correspondingly, the group chat device involving the intelligent agent can also be set up in terminal devices 101, 102, and 103. In this case, the exemplary system architecture 100 may also exclude server 105 and network 104.

[0021] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0022] Please refer to Figure 2 , Figure 2A flowchart of a group chat method involving intelligent agents is provided for embodiments of this disclosure, wherein process 200 includes the following steps: Step 201: In the dialogue initiated by the first user, identify the target dialogue that needs to be processed in the form of a group chat; This step aims to address the execution of group chat methods involving intelligent agents (e.g., the main body of the process). Figure 1 The dialogue interaction application installed on the terminal devices 101, 102, 103 or server 105 shown, and the main intelligent agent running on the dialogue interaction application is based on a large model, actively identifies the target dialogue that needs to be processed in the form of a group chat in the one-on-one chat dialogue of the first user.

[0023] To achieve this goal, the aforementioned executing agent (i.e., the main intelligent agent) can, on the one hand, analyze the dialogue content through a natural language processing model to identify whether it contains the intent of "requiring multi-user participation." Such intents usually imply semantics of "collaboration," "invitation," or "joint problem-solving," such as "Let Xiaohong take a look at this solution," "Ask Xiaoming to help solve the child's problem," or even implicit expressions like "This problem needs to be discussed with others." On the other hand, for dialogues that do not explicitly mention multi-user participation, the main intelligent agent can assess their complexity through task decomposition and domain cross-analysis. For example, it can break down "planning a weekend family trip" into multiple sub-tasks such as "booking flights, booking hotels, and planning itineraries" (if the number of sub-tasks exceeds a preset threshold, such as 2), or determine whether the problem involves cross-domain knowledge such as "transportation + accommodation + attractions." If any of these conditions are met, it is determined to be a high-complexity target dialogue that requires group chat processing. Regardless of the aspect, the entire evaluation process can combine the context (such as the roles "Xiaohong" and "Xiaoming" mentioned in the previous dialogue) and the user's historical behavior (such as the user choosing group chat when dealing with similar problems before) to avoid misjudgment. For example, if the first user has invited the teacher to join the group chat multiple times in the "child tutoring" question, when the user mentions "the child can't do the math problem" again, the main agent can quickly identify it as the target dialogue.

[0024] In practical applications, the main agent can achieve target dialogue recognition through a combination of a "rule engine + machine learning model". For "multi-user participation intent", the rule engine matches "collaboration keywords" (such as "together", "find XX", "invite") in the dialogue, or uses a fine-tuned intent recognition model to determine whether the dialogue belongs to the category of "requiring the participation of others". For "high-complexity needs", the main agent can extract task steps through dependency parsing (such as "booking flights → booking hotels → planning itineraries" as three sub-tasks), or use a domain tag library (such as "transportation", "accommodation", "attractions") to determine whether cross-domain involvement is involved. If either condition is met, it is determined to be the target dialogue. For example, when the first user sends "Help me find Xiaoming and the teacher to tutor my child's math problems", the main agent can identify the "multi-user participation intent" through the keywords "together" and "find Xiaoming and the teacher"; when the first user sends "I want to plan a weekend trip that includes Disneyland, hotels, and transportation", the main agent identifies the "high-complexity need" through task decomposition (three sub-tasks) and domain intersection ("attractions", "accommodation", "transportation"), both of which are determined to be target dialogues.

[0025] To improve the accuracy and flexibility of recognition, the lead agent can also combine fuzzy intent processing and user feedback mechanisms. For example, for dialogues that do not explicitly express "multi-user participation" or "high complexity" (such as "Help me look at this problem"), the lead agent can confirm the user's needs by asking follow-up questions (such as "Do you need to discuss this with others?"), or infer whether a group chat is needed based on the user's historical behavior (such as using group chat for similar questions in the past). At the same time, the lead agent can also record the user's feedback on "target dialogue recognition" (such as the user refusing to join a group chat), and optimize the intent recognition and complexity evaluation models through reinforcement learning to improve long-term accuracy. In addition, the lead agent should avoid over-triggering group chats. That is, for simple requests (such as "What's the weather like tomorrow?"), even if they contain keywords related to "multi-user participation" (such as "Tell Xiaohong about tomorrow's weather"), there is no need to switch to a group chat. The agent can reply directly in a one-on-one chat and synchronize the information with Xiaohong to ensure the rational use of resources.

[0026] Step 202: Determine additional group chat participants based on the dialogue requirements of the target conversation; Based on step 201, this step aims to have the aforementioned executing entity determine additional group chat participants according to the dialogue requirements of the target dialogue. These group chat participants include at least one sub-agent that is distinct from the main agent and optionally a second user that is distinct from the first user. That is, the group chat participants do not necessarily include a second user, but must include at least one sub-agent.

[0027] From a technical perspective, the determination of the sub-agent and the second user follows the logic of "intent-ability matching" and "need-social relevance," respectively. For sub-agents, which act as carriers of professional capabilities, the main agent can first use a natural language processing model to analyze the core intent of the target dialogue (e.g., the core of "tutoring a child in math" is "solving elementary school math olympiad problems"), and then retrieve a pre-built "agent capability library" (containing information such as domain labels, capability descriptions, and response styles of each agent) to match the sub-agent most relevant to the intent. This process can comprehensively consider the agent's capability coverage (e.g., whether it can handle specific problem types like "chicken and rabbit in the same cage"), user preferences (e.g., the user previously selected an agent with a "friendly style"), and response efficiency (e.g., prioritizing online agents). For the second user, who acts as a social contact, the main agent can analyze the user's social graph (e.g., contacts in the address book and chat history) and the context of the target dialogue (e.g., the user previously mentioned "Xiaoming is the child's classmate") to discover users relevant to the need (e.g., "Xiaoming" corresponding to "tutoring a child in math"), while also considering social closeness (e.g., the frequency of Xiaoming's chat with the user), need relevance (e.g., whether Xiaoming has participated in similar tutoring), and willingness to participate (e.g., whether Xiaoming frequently responds to such requests).

[0028] In practice, the main agent can implement this logic through a process of "intent classification + resource retrieval." Taking the target dialogue "Help me find Xiaoming and his teacher to tutor my child's math problems" as an example: the main agent first uses an intent classification model to identify the core intent as "math tutoring." Then, it searches the "agent capability library" and matches sub-agents tagged with "elementary math," "friendly style," and whose capability description includes "solving Olympiad math problems" (such as the "Little Math Pro" agent). Simultaneously, it analyzes the context to find that "Xiaoming is the child's classmate, and they often do homework together," and searches the address book to confirm that Xiaoming is a frequently used contact, thus identifying him as the second user. At this point, the sub-agent provides professional math tutoring capabilities, while the second user, Xiaoming, is connected to the child's social relationships. The combination of these two elements allows the group chat to possess both problem-solving capabilities and a natural social atmosphere.

[0029] To improve accuracy and user experience, the main agent can incorporate dynamic adjustments and user preference factors: For sub-agents, if multiple options match the user's intent (e.g., the "Little Math Pro" agent and the "Strict Math Teacher" agent), the main agent can sort them based on the user's historical preferences (e.g., previously selected "Friendly Style") or the current scenario (e.g., 8 PM is more suitable for a relaxed explanation) and select the agent that best meets the user's needs. For the second user, if multiple potential targets exist (e.g., "Xiaoming" and "Xiaohong"), recommendations can be made based on social frequency (e.g., Xiaoming chats more frequently) or willingness to participate (e.g., Xiaohong has previously actively participated in tutoring). If the recommended target does not respond, other relevant users can be dynamically adjusted for recommendations (e.g., "Xiaogang"). Furthermore, the main agent should avoid over-recommendation: If the target dialogue does not require social interaction (e.g., "Check the weather"), only sub-agents (e.g., the "Weather Query Agent") can be recommended; if social interaction is required but the user has not mentioned a specific target (e.g., "Find friends to play ball with"), "fuzzy recommendations" (e.g., "Recommend Xiaozhang, whom you have recently contacted") can be used to guide the user's choice, maintaining flexibility.

[0030] In other words, the step of identifying participants in a group chat provided in this section is a crucial step in achieving a seamless transition from one-on-one chat to group chat. Its core is the combination of "professional capability carriers (sub-agents) + social associates (second users)" to fully satisfy users' dual needs for "efficient problem-solving" and "natural social collaboration." The design logic of this step stems from an understanding of the essence of group chat: it should not only be a tool-based dialogue window, but also a collaborative space with professional support and emotional connection.

[0031] It should be noted that both the main intelligent agent and the sub-intelligent agent are intelligent agents, which refer to a system or entity that can perceive the environment and make autonomous decisions and execute tasks. In some scenarios, they can also be referred to as "intelligent agents." An "intelligent agent" can be understood not only as an entity, but also as a service that intelligently uses a single intelligent agent or multiple intelligent agents in conjunction to provide services that meet the user's needs, based on the user's expressed intentions.

[0032] Step 203: Create a group chat session entry point in the session containing the target dialogue, including the first user, group chat participants, and the main agent; Building upon step 202, this step aims to have the aforementioned executing entity create a group chat session entry point within the target dialogue session, containing the first user, group chat participants, and the main intelligent agent. This step serves as the core component for achieving a seamless migration from "one-on-one chat" to "group chat." Essentially, it integrates the "starting threshold" of group chat collaboration into the user's current dialogue scenario, preventing users from interrupting their train of thought due to "switching interfaces or manually creating groups." In other words, when a user makes a request in a one-on-one chat, they should be able to directly enter a group chat containing all necessary roles within the same session window without any additional operations.

[0033] From a technical perspective, the solution provided in this step aims to address two core issues: the "nativeness" of the entry point and the "completeness" of the participants. "Nativeness" means that the presentation of the group chat entry point must conform to the inherent style of conversational interactive applications (such as the "group chat invitation card" provided in instant messaging application A, or the "conversation link" provided in instant messaging application B), avoiding user resistance due to "unfamiliar interactive elements." "Completeness" means that the group chat pointed to by the entry point must pre-include the first user (the request initiator), group chat participants (sub-agents + second users, providing professional skills and social connections), and the main agent (the coordinator, ensuring the orderly operation of the group chat). This way, users do not need to manually add members after entering the group chat and can directly begin collaboration. This process relies on the application's "group chat pre-creation" capability; that is, the main agent creates the group chat containing all participants in the background before generating the entry point, and the entry point only exists as an "access channel."

[0034] In practical terms, the main agent can generate group chat entry points through a process of "scenario adaptation + resource preloading." Taking the target dialogue "Help me find Xiaoming and a teacher to tutor my child's math problems" as an example: When a user initiates this dialogue in a one-on-one chat with a friend, the main agent recognizes it as the target dialogue. First, it creates a group chat in the background, named "Xiaoming's Math Tutoring Group," and adds the first user (the user), the group chat participants ("Little Math Master Agent" + "Xiaoming"), and the main agent ("Group Chat Assistant") to the group chat. Then, in the user's one-on-one chat with their friend, a native-style group chat entry card is generated: the top of the card displays the group name "Xiaoming's Math Tutoring Group," the middle displays the avatars of the participants (the user, Xiaoming, "Little Math Master Agent," and "Group Chat Assistant"), and the bottom has a prominent "Enter Group Chat" button. This card design perfectly matches the group chat invitation style of IM applications; users can immediately understand its function without additional learning.

[0035] To enhance the usability and flexibility of entry points, the main agent can also personalize and optimize them for specific scenarios. For example, group names can be dynamically generated based on the content of the target conversation (e.g., "Children's English Essay Editing Group" or "Weekend Family Trip Planning Group"), allowing users to immediately understand the topic of the group chat; the entry point's placement can be aligned with the conversation context (e.g., immediately following the target conversation message to avoid users having to search); the entry point's validity period can be set to "valid within the demand period" (e.g., the "Tutoring Children in Math" entry point is valid for 24 hours and automatically hides after expiration to avoid redundancy). Furthermore, if a user declines to join the group chat (e.g., by clicking the "Cancel" button), the main agent can mark the entry point as "declined" and retain it for a period (e.g., 7 days), allowing the user to re-enter if they change their mind later, thus improving flexibility.

[0036] Step 204: In response to the first user entering the target group chat session through the group chat session entry, bring the context information related to the target dialogue into the target group chat session, and control the sub-agent to reply to the first user in the target group chat session based on the target session and the context information.

[0037] Building upon step 203, this step assumes that the first user enters the target group chat through the group chat entry point. The aim is for the aforementioned execution entity to bring relevant contextual information into the target group chat and control the sub-agent to reply to the first user within the target group chat based on the target conversation and the contextual information. Specifically, when a user clicks the group chat entry point (e.g., the "Enter Group Chat" button), the dialogue interaction application can detect the user's entry action through a conversation state monitoring mechanism (e.g., front-end event listeners or back-end interface callbacks) and trigger the main agent to execute subsequent operations. The core objective of this step is to eliminate the "repetitive description cost" for users in group chats. That is, after entering the group chat, users do not need to retell the background of the problem; the sub-agent can directly generate targeted responses based on historical information from individual chats and current needs, maintaining the natural continuation of the dialogue.

[0038] First, the lead agent needs to bring contextual information related to the target dialogue into the group chat. This "contextual information" doesn't necessarily mean all historical messages from the one-on-one chat; it can simply be content directly related to the target dialogue, filtered through semantic association analysis. For example, if the target dialogue is "Help me find Xiaoming and a teacher to tutor my child's math problems," the lead agent will filter out key background information from the one-on-one chat, such as "the child's math problem is a chicken-and-rabbit problem" and "couldn't solve it last night." If the target dialogue is "planning a family trip this weekend," it will filter out core needs such as "the child wants to go to Disneyland" and "budget of 3000 yuan." This information will be presented in a way that matches the application's style (e.g., "[Context Synchronization] It was mentioned before that the child's math problem is a chicken-and-rabbit problem, which couldn't be solved last night"), without interfering with the current flow of the group chat, while allowing all participants to quickly access the context.

[0039] Next, the main agent controls the sub-agents to generate responses based on the target conversation (the user's initial request in the one-on-one chat, such as "tutoring my child's math problems") and the preceding information. The sub-agents' responses need to combine the "core request" with "background details." For example, regarding the "chicken and rabbit problem," the sub-agent might reply: "Hello! You mentioned earlier that your child had a chicken and rabbit problem in a cage, which they couldn't solve yesterday. I can help explain it. These types of problems usually use the assumption method, such as assuming all the animals are chickens, calculating the difference in the number of feet, and then dividing by the difference in the number of feet for each animal to get the number of rabbits. Do you need me to explain it step by step?" This response addresses the target conversation's request while also incorporating the preceding information, making it direct and targeted.

[0040] At the technical implementation level, "introducing contextual information" requires addressing the issue of "relevance filtering." The main agent can filter out content directly related to the target dialogue through keyword matching (such as "chicken and rabbit in the same cage" or "math problem") or semantic similarity calculation (such as cosine similarity), avoiding irrelevant information from interfering with the group chat. For example, in a one-on-one chat, a message like "homework is due tomorrow" is background information, but it has low relevance to "problem-solving methods" and may not be synchronized (or placed as secondary information in "more context"). As for "controlling sub-agent responses," it is necessary to ensure that the response timing is appropriate (e.g., sent within 1-2 seconds of the user entering), the content is accurate (verified by a rules engine to ensure the response contains target keywords and fits the contextual background), and that irrelevant content is avoided (e.g., recommending math tutoring books unless the user specifically requested them).

[0041] The solution provided in this step eliminates the cost of repetitive descriptions for users by "introducing the preceding information" and enables the immediate output of professional skills by "controlling the sub-agent's response." The combination of these two approaches ensures good interactive continuity and targeted problem-solving in group chat collaboration from the moment a user enters the chat. For example, if a user mentions in a one-on-one chat, "My child's math problem is a chicken-and-rabbit problem, and I couldn't solve it yesterday," upon entering the group chat, the preceding information is automatically synchronized, and the sub-agent immediately replies with the solution. The user no longer needs to repeat the problem and can directly begin tutoring. This experience is both efficient and natural, meeting the core needs of dialogue-based interactive applications.

[0042] The group chat method involving intelligent agents provided in this disclosure actively identifies target dialogues that need to be processed in a group chat format during a one-on-one chat with a first user. It dynamically determines the group chat participants, including specialized sub-intelligent agents (providing targeted capabilities) and optional second users (associating with social relationships), based on the dialogue requirements. Then, it directly creates a native group chat entry point containing all participants within the one-on-one chat session. After the user enters the group chat, the one-on-one chat context information is automatically synchronized, achieving a seamless migration from one-on-one chat to group chat without requiring the user to perform tedious manual operations such as creating a group, adding members, and forwarding history. Simultaneously, the sub-intelligent agents directly reply to the user in the group chat based on the target dialogue content and synchronized context information, maintaining the continuity and naturalness of the dialogue. This method satisfies the needs of professional problem-solving while integrating social collaboration scenarios, significantly improving the efficiency and user experience of group chat collaboration.

[0043] Considering that target conversations requiring group chat processing may have different scenarios, please refer to [link / reference needed] for a deeper understanding of this section. Figure 3 , Figure 3 A flowchart of a method for determining a first target dialogue and determining group chat participants provided in this disclosure embodiment, wherein process 300 includes the following steps: Step 301: Identify the conversations that express the intention to require multi-user participation as the primary target conversations; The first target dialogue described in this embodiment focuses on the user's expressed desire for others to participate in solving the problem, and its recognition logic is based on natural language intent analysis. The main agent can determine whether the dialogue contains a need for multi-user participation through collaborative keywords such as "together," "find XX," "invite," and "jointly," or a fine-tuned intent recognition model. For example, when a user sends "Let Xiaohong and Xiaoming help me look at this project plan," keywords such as "together" and "help me look" directly point to the intent of multi-user participation, and the main agent can quickly identify it as the first target dialogue; if a user sends "This problem needs to be discussed with professionals and friends," "discussing professionals and friends" implies a need for multi-role participation, and the intent classification model can determine that it belongs to "collaborative intent," thus also identifying it as the first target dialogue.

[0044] In practice, the main intelligent agent needs to optimize recognition accuracy by combining contextual information. If a user has previously invited friends to participate multiple times in a "project solution discussion" scenario, when they mention "let's take a look at the solution together" again, the main intelligent agent can strengthen its judgment of "multi-user participation intent" by analyzing the user's historical behavior profile, thus reducing misjudgments. In addition, for implicit intents such as "I can't handle this problem by myself," the main intelligent agent can supplement the recognition through semantic association (e.g., associating "can't handle it by myself" with "needs help from others") to ensure coverage of more scenarios.

[0045] Step 302: Based on the dialogue requirements of the first target dialogue, identify other users who need to participate in the conversation as second users and identify the intelligent agents used to resolve the dialogue requirements as sub-intelligent agents.

[0046] The participants in the group chat include: the second user and the sub-agent.

[0047] Building upon step 301, when the main agent identifies the first target dialogue (i.e., a dialogue where the user explicitly expresses a need for multi-user participation, such as "Let Xiaohong and the project team help me analyze the market risks of the new solution"), the core task is to transform the user's "willingness to collaborate" into specific collaborative roles, namely, the second user (other real users who need to participate) and the sub-agent (AI agents that need to join), thereby constructing a group chat participation system of "real users + AI" collaboration. The design goal of this step is to enable effective collaboration to begin immediately after the group chat is started, avoiding the tedious process of users manually adding members.

[0048] The core of the primary goal dialogue is "multi-user participation," therefore, the determination of the second user and sub-agents needs to revolve around "role completion of dialogue needs." The second user is a real user related to the dialogue needs (distinct from the first user, i.e., the initiator of the need). Their role is to provide social connections or human support in specific areas; for example, "Xiao Hong" has experience collaborating with users, and "people in the project team" are familiar with the project background. At the technical level, the primary agent can analyze dialogue intent (extracting role requirements such as "collaboration partners" and "project members" from "analyzing the solution together") and combine this with the user's social graph (the user's contact network, including contact tags, historical interaction records, etc.) to match real users related to the needs. For example, if a user mentions "people in the project team," the primary agent can retrieve members marked as "project team" from their contacts (such as "Xiao Li" and "Xiao Zhang"); if they mention "Xiao Hong," they can directly extract "Xiao Hong" from the social graph as the second user. Sub-agents are AI agents related to the dialogue needs (distinct from the primary agent). Their role is to provide professional capabilities; for example, a "market risk assessment agent" can provide data statistics and risk assessment capabilities. Technically, the main agent extracts the core intent of the dialogue request (extracting the "market risk assessment" request from "analyzing the market risks of the solution"), searches the agent capability library (a database storing the skills and capability descriptions of each AI agent), and matches it with sub-agents possessing the corresponding professional capabilities. For example, if the request is "analyzing market risks," the main agent can select a "market risk assessment agent" (tagged with "solution analysis" and "risk assessment," with capability descriptions including "assessing risks using data models" and "generating risk reports"); if the user previously preferred a "detailed style," then agents with a "detailed" response style will be prioritized to match user habits.

[0049] Taking a user's message, "Please have Xiaohong and the project team help me analyze the market risks of the new solution," as an example, the execution flow of the main intelligent agent is as follows: First, the dialogue is parsed using a natural language processing model to extract the core requirement as "analyzing the market risks of the new solution," requiring the roles of "collaboration partners (Xiaohong, the project team)" and "professional analytical capabilities (market risk assessment)." Next, a second user is determined, extracting "Xiaohong" (with shared collaborative experience) from the social graph, and searching for "Xiao Li" and "Xiao Zhang" (familiar with the project background) marked as "project team" in the contacts. Then, a sub-agent is determined, selecting the "market risk assessment agent" from the capability library (meeting the "market risk assessment" requirement, with a "detailed" response style, matching user preferences). Finally, the group chat participants include the first user (the user themselves), the second user (Xiaohong, Xiao Li, Xiao Zhang), the sub-agent (market risk assessment agent), and the main intelligent agent (coordinating the group chat), ensuring that collaboration can begin immediately after the group chat starts.

[0050] To improve flexibility and accuracy, the main agent can also try to introduce a dynamic adjustment mechanism: if the second user (such as "Xiao Hong") does not respond in time, a substitute user with high relevance to Xiao Hong and meeting the needs (such as "Xiao Ming," also a project team member) can be recommended from the social graph; if a sub-agent (such as the "market risk assessment agent") is offline, it can be switched to the backup "business analysis agent" (with similar assessment capabilities) to avoid group chat interruption. Furthermore, a priority ranking mechanism can be adopted: when looking for a second user, ranking is based on "social closeness (such as chat frequency) + need relevance (such as solution analysis experience)," prioritizing the most likely user to participate; when looking for a sub-agent, ranking is based on "response efficiency (such as online status) + user satisfaction (such as previous feedback ratings)," selecting the optimal agent. At the same time, it is still necessary to consider mechanisms to avoid over-recommendation: if the need only requires a small number of users (such as "Let Xiao Hong help me look at the solution"), too many second users will not be recommended (only Xiao Hong will be invited); if a sub-agent can solve the problem (such as "market risk assessment"), multiple similar agents will not be recommended (to avoid group chat bloat).

[0051] exist Figure 3 Based on the illustrated embodiment, since the second user has also joined the target group chat session, the dialogue within the group chat becomes slightly more complex and chaotic. To avoid confusion when the question-response roles, including the main agent and sub-agents, respond to dialogue requests from the first and second users, please refer to [further details needed]. Figure 4 , Figure 4 Branching diagrams for four different group chat session methods provided in embodiments of this disclosure: First, based on the content of the group chat dialogue initiated by the first user and / or the second user in the target group chat session, determine the dialogues that need to be replied to by the main agent or the sub-agent.

[0052] In essence, the main AI agent needs to identify content from group chat messages that requires a response from either the main AI agent or a sub-AI agent. From a technical perspective, this step can rely on natural language intent recognition. For example, machine learning models can analyze the semantics of group chat messages to determine if they fall into categories such as "questioning," "requesting help," or "needing professional advice" (e.g., "What are the market risks of this solution?" is a question; "Help me optimize the process" is a request for help). In practice, the main AI agent can improve recognition accuracy by combining the group chat context (e.g., whether previous conversations involved AI support) and user history (e.g., users' habit of asking AI professional questions). For instance, if the group chat previously discussed "solution analysis," and a user sends "Does the market risk part need to be checked again?", the main AI agent can quickly determine that this is a dialogue requiring an AI response.

[0053] Based on "whether to instruct the recipient" and "number of sub-agents", the main agent will handle the pending dialogue in four different ways: Scenario 1: In response to a dialogue to be replied to that does not indicate a reply recipient and there is only one sub-agent, the sub-agent is invoked to reply to the first user and / or the second user who initiated the dialogue to be replied to in the target group chat session based on the dialogue information related to the dialogue to be replied to in the target group chat session.

[0054] When a user doesn't specify "who should reply" (e.g., not mentioning any agent), but there's only one sub-agent in the group chat (e.g., the "market risk assessment agent"), the main agent doesn't need to make additional judgments and directly calls that sub-agent to reply. Technically, this is a "unique solution" logic, meaning there's no room for choice; it directly matches the only professional role. In practice, for example, if the group chat only has the "market risk assessment agent," and the user sends "What are the market risks of this solution?", the main agent immediately instructs that agent to reply: "Based on data model analysis, the market risk score for this solution is 3.2 / 5 (medium risk). The main risk lies in the recent price war strategy of competitors; we recommend supplementing the competitor's response plan." This approach is suitable for scenarios with a "single professional need," avoiding user confusion due to choices.

[0055] Scenario 2: In response to a dialogue to be replied to that does not indicate a reply target and there are multiple different sub-agents, determine the target sub-agent or combination of target sub-agents that matches the dialogue to be replied to among the multiple sub-agents; wherein, the combination of target sub-agents includes at least two different target sub-agents, and the different target sub-agents are combined in the order of reply.

[0056] When a matching target sub-agent is identified, the target sub-agent can be invoked to reply to the first user and / or the second user who initiated the dialogue in the target group chat session based on the dialogue information related to the dialogue to be replied to in the target group chat session. When a matching combination of target sub-agents is identified, different target sub-agents can be invoked sequentially according to the reply order to reply to the first user and / or the second user who initiated the reply in the target group chat session based on the dialogue information related to the dialogue to be replied to in the target group chat session.

[0057] In other words, when a user does not explicitly request a response but there are multiple sub-agents in a group chat (such as a "market risk assessment agent," a "process optimization agent," or a "cost estimation agent"), the main agent needs to match the sub-agents according to the needs of the dialogue to be responded to, or combine multiple sub-agents to respond sequentially. This objective can be achieved through a demand-capability matching model at the technical level: extracting core demands from the dialogue to be responded to (e.g., "What are the market risks and process optimization suggestions for this solution?" extracting the demands of "market risk assessment" and "process optimization"), then searching the sub-agents' capability tag library (each sub-agent is tagged with skills such as "market risk," "process optimization," and "cost estimation"), and matching the corresponding sub-agent.

[0058] If the requirement is of a single type (such as "process optimization suggestion"), only the corresponding sub-agent (such as "process optimization agent") can be invoked; if the requirement is of multiple types (such as needing both "market risk assessment" and "process optimization"), multiple sub-agents can be combined and responded in order of "requirement priority" (such as discussing risk first, then optimization).

[0059] In practical terms, for example, if a user sends a message like "Help me analyze the market risks and process design of this solution," the main AI agent can first call the "Market Risk Assessment AI Agent" to reply with the risk part, and then call the "Process Optimization AI Agent" to reply with the process part, ensuring that the reply covers all needs.

[0060] Scenario 3: In response to the pending dialogue indicating that the target sub-agent is the reply object, the target sub-agent is invoked to reply to the first user and / or the second user who initiated the pending dialogue in the target group chat session based on the dialogue information related to the pending dialogue in the target group chat session; wherein, the target agent is a unique sub-agent or a designated sub-agent among multiple different sub-agents.

[0061] When a user explicitly mentions a specific sub-agent (e.g., "@Market Risk Assessment Agent, please analyze the market risks of this solution"), the main agent directly invokes that sub-agent to reply. Technically, this is based on "explicit intent" logic, meaning the user has already specified a professional role, eliminating the need for additional matching. In practice, for example, if a user mentions the "Process Optimization Agent" saying, "This process has too many approval steps; can you optimize it?", the main agent can immediately have the "Process Optimization Agent" reply: "Based on process design principles, the approval steps can be simplified from 5 to 3 (merging departmental review and general manager approval), which is expected to shorten the approval time by 2 days. Do you need me to provide specific optimization solutions?" This approach is suitable for scenarios where the user "knows what they need," improving the accuracy of the response.

[0062] Scenario 4: In response to the pending dialogue, the main agent is instructed to act as the responder. An assignment notification dialogue is generated in the target group chat to assign a sub-agent to respond. The sub-agent is invoked to respond to the first user and / or the second user who initiated the pending dialogue in the target group chat based on the dialogue information related to the pending dialogue in the target group chat.

[0063] When a user mentions the main AI agent (e.g., "@GroupChatAssistant, please analyze the market risks of this solution"), the main AI agent's primary responsibility is to assign tasks to specialized sub-AI agents, not to provide direct replies. In technical terms, the main AI agent acts as a "coordinator," not a "professional solver." It needs to assign specific tasks to sub-AI agents with the appropriate capabilities and inform the user, making it clear who is handling the problem. In practice, for example, if a user mentions the main AI agent saying, "@GroupChatAssistant, please check the market risks of this solution," the main AI agent first replies: "A market risk assessment AI agent has been assigned to analyze it for you, please wait." (Assignment notification dialogue), and then invokes the "market risk assessment AI agent" to provide the specific details. This design avoids unprofessional responses caused by the main AI agent trying to "manage everything," while giving the user a sense of security that "a more professional agent is better suited to handle my problem."

[0064] The core logic of the scenarios described in this embodiment is the precise matching of "user intent + AI role." When the user's needs are clear, the corresponding AI agent is directly invoked; when the user's needs are unclear, the most suitable AI agent is matched through intelligent analysis; and when the user seeks the main AI agent, a mechanism is used to involve specialized sub-AI agents. This design allows sub-AI agents in group chats to respond in a way that better meets user expectations, preserving the sense of "user-led" collaboration while leveraging the advantages of "specialized sub-AI agents." This effectively solves the problems of "inaccurate responses from the main AI agent" and "users not knowing who to contact" in group chats. For example, when a user sends a message in a group chat asking, "What are the market risks and process optimization suggestions for this solution?", the main AI agent can automatically match the "market risk assessment AI agent" and the "process optimization AI agent" to respond in sequence, allowing the user to obtain all the necessary information at once without repeated inquiries, thus demonstrating the value and effectiveness of collaboration.

[0065] Different from Figure 3 Please also refer to the shown cutting angle. Figure 5 , Figure 5 A flowchart of a method for determining a second target dialogue, determining group chat participants, and controlling the responses of group chat participants provided in this embodiment of the disclosure is included in the following steps: Step 501: Identify conversations in which the dialogue content expresses a need to solve a problem with an actual complexity exceeding a preset complexity threshold as the second target conversation; The second target dialogue described in this embodiment addresses the "high-complexity problem-solving needs that cannot be handled by one-way chat," and its judgment logic is based on task decomposition and domain cross-analysis. At the technical principle level, the main intelligent agent can use dependency parsing or task extraction models such as Seq2Seq to decompose the dialogue needs into sub-tasks (e.g., "planning a weekend family trip" is decomposed into "booking flights → booking hotels → selecting attractions → arranging itineraries"), and count indicators such as the number of sub-tasks, the number of domains involved (e.g., "transportation," "accommodation," and "tourism" belong to different domains), and the amount of information required (e.g., requiring users to provide multiple information points such as budget, time, and preferences). If these indicators exceed a preset complexity threshold (e.g., number of sub-tasks ≥ 3, number of domains involved ≥ 2, amount of information required ≥ 4), it is determined to be a second target dialogue. For example, a user sends a message asking, "Help me plan a weekend family trip, including booking flights, hotels, and attraction tickets, and also considering the children's interests and the elderly's rest time." This request is broken down into four sub-tasks, involving four areas: "transportation," "accommodation," "tourism," and "family needs." The amount of information required includes "travel time," "budget," "children's interests," and "elderly's rest habits," all of which exceed the preset threshold (e.g., the number of sub-tasks ≥ 3). Therefore, it is determined to be the second target dialogue.

[0066] In practical terms, this preset threshold can be dynamically adjusted based on user attributes. For novice users (those new to the application), the threshold is set to "number of subtasks ≥ 2" (indicating weak ability to handle complex problems); for expert users (those who frequently plan their schedules), the threshold is increased to "number of subtasks ≥ 4" (to avoid excessive triggering of group chats). Simultaneously, the main agent can optimize the threshold setting based on user feedback. For example, if a user considers a request "not complex" but it is still judged as a secondary target dialogue, the main agent lowers its threshold to improve judgment accuracy.

[0067] Step 502: Divide the complete task corresponding to the problem-solving requirements into multiple sub-tasks, and determine the agents that execute each sub-task as sub-agents; The participants in the group chat include: each sub-agent used to perform the sub-tasks.

[0068] Step 503: Control each sub-agent to reply to the user in the target group chat in the order of execution of each sub-task, based on the target session, the preceding information, and the execution results of the previous sub-tasks.

[0069] After matching the sub-agents, the master agent controls them to respond sequentially according to the logical order of the sub-tasks, ensuring that the result of each sub-task can be passed to the next sub-agent. For example, in the travel planning process, the master agent first calls the flight booking agent, which, based on the user's weekend travel needs and family preferences (such as needing adjacent seats), replies with flight information and completes the booking, storing the flight time and other results as the "execution result of the preceding sub-task"; next, it calls the hotel recommendation agent, inputting the flight time and family preferences (such as needing a family room or being near the airport), recommends suitable hotels and completes the booking, storing the hotel information; then, it calls the attraction selection agent, inputting the hotel location and family preferences (such as attractions suitable for children), recommending attractions and purchasing tickets; finally, it calls the itinerary planning agent, inputting the results of the previous three, and generates a detailed itinerary. The response of each sub-agent depends on the result of the previous sub-task, forming a progressive solution process.

[0070] To improve flexibility and accuracy, the main agent can also introduce some optimization mechanisms: for example, if the user has special requirements for the order of a certain subtask (such as "select attractions first and then book hotels," because the location of attractions affects hotel selection), the main agent can adjust the order of subtasks by learning user preferences; if a sub-agent cannot perform a task due to a malfunction (such as the flight booking agent going offline), the main agent can replace the backup agent to ensure that the process is not interrupted; in addition, the main agent can automatically pass the results of the previous sub-agent to the next, avoiding the user from re-entering information (such as the flight booking time being automatically passed to the hotel recommendation agent, so that the user does not need to say "the flight time is Saturday morning at 8:00").

[0071] The core value of the approach provided in this embodiment for the second-target dialogue scenario lies in breaking down complex needs into sub-tasks that can be solved step by step. Through the division of labor and sequential collaboration of specialized sub-agents, large model tools can efficiently solve highly complex problems. In other words, users do not need to plan step by step themselves; they only need to state their needs to complete all tasks progressively, making "complex travel planning" "simple and efficient."

[0072] In group chat scenarios, it's common to encounter situations where the first user (the request initiator) and the second user (other participating users) send similar messages within a short span of a few tens of seconds, while the sub-agent responsible for replying (such as the agent handling flight ticket inquiries) hasn't yet processed the first initiator's message. In other words, the sub-agent responsible for replying needs to handle multiple similar requests, and the conventional approach of replying independently to each request often leads to content duplication, thus disrupting the conciseness of the group chat information. To address this issue, this embodiment also provides a specific implementation method: In response to the group chat participants including at least one second user, and the first user and the second user respectively initiating similar group chat dialogues with an actual similarity exceeding a preset similarity threshold within a preset time window, a sub-agent matching the conversation requirements of the similar group chat dialogue is invoked to simultaneously reply to the similar group chat dialogues initiated by the first user and the second user in the target group chat session by generating a single reply content based on the conversation information related to the similar group chat dialogue in the target group chat session; wherein, the preset time window is when the similar group chat dialogue initiated later is received before the sub-agent matching the previously initiated similar group chat dialogue has completed generating the reply content.

[0073] To handle similar dialogues, the first step is to determine whether the "semantic similarity" of the two pieces of content exceeds a pre-defined standard (e.g., cosine similarity ≥ 0.8). Technically, this can be achieved by converting the dialogue into semantic vectors to calculate similarity, or by keyword matching (e.g., both dialogues contain core words such as "weekend," "Beijing," and "airfare"). For example, in a group chat, if the first user says, "How much is a plane ticket to Beijing this weekend?", and the second user says 15 seconds later, "How much is a plane ticket to Beijing this weekend?", the keywords and semantics of the two dialogues highly overlap, and thus they are considered similar dialogues.

[0074] The "preset time window" refers to the time period from when the first similar dialogue is identified to when the sub-agent begins to reply (usually set based on the average response time of the sub-agent, such as 30 seconds). If the interval between two dialogues is too long (e.g., more than 5 minutes), even if the content is similar, they may be independent requests and do not need to be merged; however, if they are within the time window (e.g., within 30 seconds), it indicates that they are repetitive requests within a short period of time and need to be integrated. For example, if the sub-agent takes an average of 20 seconds to generate a reply, and the time window is set to 30 seconds, then any dialogue initiated later will be included in the merging scope as long as it is sent within 30 seconds.

[0075] When a similar dialogue is identified and falls within a time window, the main agent will invoke the corresponding sub-agent to generate a response that balances relevance and integration. For example, regarding the aforementioned flight ticket question, the sub-agent might reply: "@UserA @UserB, the price of a flight to Beijing this weekend is between 800-1200 yuan, depending on the airline and departure time. Would you like me to check specific flights for you?" (That is, a single reply is presented in the target group chat by simultaneously referencing both the previously initiated and subsequently initiated similar group chat dialogues.) This reply both @ two users (making each person feel that the agent is responding to them), merges similar requests (avoiding duplicate replies), and retains space for further interaction (asking if they need specific flight information).

[0076] To make the mechanism more flexible, some optimizations can be tried: For example, when more users (e.g., three) ask similar questions in a short period of time, the main agent can merge their questions and reply to all users in one message (e.g., "@User A @User B @User C, the price of air tickets to Beijing this weekend is between 800-1200 yuan. Do you need me to do a batch search for you?"). If users have personalized preferences (e.g., User A prefers early morning flights, and User B prefers late night flights), the reply can incorporate these preferences (e.g., "@User A (early morning flight preference) @User B (late night flight preference), the price of early morning flights to Beijing this weekend is about 900 yuan, and the price of late night flights is about 800 yuan. Do you need me to book for you?"). In addition, the time window can be dynamically adjusted according to the load of the sub-agents. When the sub-agents are handling many requests and responding slowly, the time window can be appropriately enlarged (e.g., extended from 30 seconds to 45 seconds) to avoid missing similar conversations in a short period of time; when the load is low, the window can be reduced to ensure timely processing.

[0077] The core of the solution provided in this embodiment lies in efficient collaboration and ensuring information conciseness: when multiple people in a group chat ask the same question within a short period of time, the intelligent agent will not reply to each person individually, but will respond simultaneously with a single message. For example, users do not need to wait for multiple repetitive replies, the group chat will not become cluttered due to an overabundance of messages, and the sub-agents will also save resources from repetitive processing.

[0078] Specifically, if the preference configuration information of the first or second user contains a configuration item that disallows merging replies to similar conversations, a sub-agent matching the content of the similar group chat can be invoked to reply to the similar group chat initiated by the first and second users respectively in the target group chat by generating different reply content based on the conversation information related to the content of the similar group chat in the target group chat.

[0079] In a group chat, when the first and second users send similar messages, the AI ​​agent would normally merge the replies to save time. However, if the user's preferences explicitly state "no merging," the AI ​​needs to reply separately. This is to respect the user's individual preferences and ensure the AI's response matches the user's expectations. User preferences are personalized options they set within the application, such as "whether to allow merging replies to similar questions," and are stored in their user profiles. If a user selects "no merging," for example, by checking "each question must be replied to separately," it means they want their questions to receive independent responses and don't want them grouped with others' similar questions. Before processing similar conversations, the AI ​​agent checks the preferences of relevant users (including the two people who sent the messages). If even one person has set "no merging," the logic for replying separately is triggered. This step requires real-time checks of the user's latest settings, such as reading the "merge reply permission" field from the database, to ensure the decision is based on the user's current intentions.

[0080] For example, in a group chat, user A asks, "How much are the high-speed rail tickets to Shanghai this weekend?", and user B replies 10 seconds later, "Approximately how much do high-speed rail tickets to Shanghai cost this weekend?". The main agent uses a semantic similarity algorithm (e.g., the cosine similarity of vectors generated by BERT exceeds 0.8) to determine that these two conversations are similar. Then, it checks user preferences and finds that user A previously set "Do not merge" (they dislike merging their own questions with others'). At this point, even if the two conversations are very similar and within the time window, the main agent will still call the corresponding sub-agent (e.g., the high-speed rail ticket query agent) to reply to each user separately. The reply to user A might be, "@UserA, High-speed rail tickets to Shanghai this weekend, for example, G123, second class seat 300 yuan, first class seat 500 yuan," while the reply to user B might be, "@UserB, High-speed rail tickets to Shanghai this weekend are between 300 and 500 yuan. Tickets for popular trains at 8 am are in high demand; it is recommended to book in advance." This satisfies both user A's requirement for a separate reply and user B's need, and the replies have different focuses and avoid repetition.

[0081] Furthermore, if two users have conflicting preferences—for example, user A allows merging while user B does not—then a stricter rule is usually applied: if even one user disallows it, the merging is not performed. For instance, if a user receives a merge response and says, "From now on, my questions will be answered separately," the main agent can automatically update their preferences and respond separately to similar conversations from that user in the future. Also, when responding separately, the sub-agent can adjust the content based on the user's characteristics. For example, if user A is a frequent high-speed rail ticket buyer, the response could mention, "Your usual second-class seat costs 300 yuan." If user B is a new user, the response could add, "I recommend choosing second-class seats; they offer better value." This way, even with separate responses, there won't be much repetition, and the responses will be more targeted.

[0082] Based on any of the above embodiments, please also refer to Figure 6 , Figure 6 A flowchart of a method for determining and presenting cause information provided in this disclosure embodiment is included in process 600, which includes the following steps: Step 601: In response to any second user being added to the target group chat session by the first user, determine the reason information for the second user being added to the target group chat session by the first user in the target group chat session; The reason information is determined based on the historical dialogue in the target group chat session and the action information of the first user to add a group chat participant.

[0083] Among them, the target group chat is the currently ongoing group chat (such as "Weekend Travel Planning Group"); the second user is a new member invited to join the group chat by the first user; the reason information is the reason why the first user invited the second user to join the group (such as "You mentioned before that you like weekend travel"); the history of the conversation is the previous message record in the target group chat (such as the group chat discussing "climbing Mount Tai next week"); the action information is the operation or accompanying information when the first user invited the user (such as the "add reason" option selected when inviting the user, i.e., "travel partner", or the note when inviting the user "I would like to invite you to plan the trip together").

[0084] From a technical perspective, the process of generating causal information by the main agent can be divided into two steps: information collection and integration. First, the main agent can analyze the historical dialogue of the target group chat using natural language processing technology to extract keywords or topics related to the second user (such as the second user previously saying in the group chat, "I want to go out for a walk this weekend," or the group chat is discussing "weekend travel plans"). At the same time, it collects the action information of the first user when inviting others (such as the "reason for inviting" tag selected in the application when inviting others, such as "travel partner," or the custom note entered when inviting others, such as "You previously said you like mountain climbing, let's go to Mount Tai together"). Then, the main agent integrates these two parts of information to generate causal information that is relevant to the scenario and personalized (such as "You previously mentioned in the group that you wanted to go out for a walk this weekend, so you were invited to join this travel planning group," or "You like mountain climbing, and this group is planning a trip to Mount Tai next week, so you were invited").

[0085] In a specific practical scenario, for example: User 1 invites User 2 to a "Weekend Travel Planning Group." User 2 had previously messaged in the group chat saying, "I'd like to find a place to go hiking this weekend." When User 1 invited User 2, they selected "travel buddy" as the reason in the app's "Add Group Member" interface. The main AI agent will then extract these two pieces of information: "Weekend hiking" from the historical conversation and "travel buddy" from the action information, and generate the reason information: "You previously mentioned wanting to find a place to go hiking this weekend. This group is planning a trip to Mount Tai next week, so you were invited to join."

[0086] Step 602: Present the reason information to the second user who enters the target group chat session.

[0087] The main agent can send a message to the second user entering the target group chat: "@Second User, you have been invited to join the 'Weekend Travel Planning Group'. Reason: You previously mentioned wanting to go hiking this weekend, and this group is planning a trip to Mount Tai next week." Through application permission settings, this message can be made visible only to the second user (other members cannot see it). This way, the second user, upon entering the group chat, can immediately understand why they were added and quickly integrate into the discussion. Alternatively, this reason information can be set to be visible only to the first user who performed the invitation, allowing them to correct any errors in the reason information they find. In this case, the reason information should first be visible to the first user and their confirmation of its correctness before being sent to the second user.

[0088] Furthermore, the main AI agent can adjust the tone or content based on the second user's preferences. If the second user prefers brevity, it generates a message like, "You mentioned you like hiking, so you were invited to join a travel group." If they prefer more detail, it adds, "The group is planning a trip to Mount Tai next week and would like to invite you to join." The main AI agent can also provide the first user with more options for "reasons for inviting others" (such as "hobbies," "work requirements," "friend's recommendation," etc.), or allow the first user to input a custom reason. The main AI agent then uses this information to generate the reason information, which better reflects the first user's true intentions.

[0089] The core of the solution provided in this embodiment is to enable new members to quickly understand their connection to the group chat and enhance their sense of belonging by providing personalized reasons for joining. Simultaneously, by setting the visibility to only the second user, the privacy of the first user's intention to invite others is protected (e.g., not wanting other members to know the specific reasons for inviting). This design aligns with the "user-centric" interaction principle, making the group chat joining process more transparent and user-friendly.

[0090] Furthermore, the second user joins the target group chat session on the first platform by receiving a group chat invitation message sent to the second platform via a cross-platform message format from the first user on the first platform. In other words, when the first user and the second user are on different platforms, they can send group chat invitation messages using a cross-platform message format.

[0091] Based on any of the above embodiments, considering that as the dialogue between the various group chat participants in the group chat session gradually unfolds and the communication deepens, there may be a situation where the new dialogue request initiated by any user in the target group chat session does not match any of the sub-intelligent agents currently included in the target group chat session. In this case, the aforementioned execution entity can continue to determine the new sub-intelligent agent corresponding to the new dialogue request, and pull the new sub-intelligent agent into the target group chat session as a new group chat participant. The new sub-intelligent agent is then invoked to reply to the user who initiated the new group chat session in the target group chat session based on the session information related to the new group chat dialogue in the target group chat session. Here, any user includes the first user and / or the second user, and the new sub-intelligent agent includes the subordinate sub-intelligent agent of any of the currently included sub-intelligent agents.

[0092] In group chat scenarios within conversational applications, when a user (whether the initiator or a participant) sends a new message, and none of the existing agents in the chat (such as the previously added "travel planning agent") can meet the message's requirements, the primary agent can initiate a "dynamic agent addition" process. This involves finding a specialized agent capable of fulfilling the new requirement to join the chat and respond to the user. The core of this design is to allow the collaborative capabilities of group chat agents to "expand on demand," ensuring that no matter what new needs the user raises, a corresponding specialized agent can respond.

[0093] For example, suppose there's a "Weekend Paris Travel Planning Group" with a main AI agent, User A (the one who started the group chat), User B (a participant), and a previously joined "Travel Planning AI agent." This agent can help plan itineraries and recommend hotels, but it can't check the weather. User B sends a message: "Can you check the weather in Paris tomorrow?" The main AI agent first needs to understand that the core requirement of this message is "check the weather in Paris tomorrow." Then it checks if the current "Travel Planning AI agent" can handle this request, but obviously it can't, because its capability tag doesn't include "weather check." At this point, the main AI agent needs to search the "global AI agent library" for an AI that can check the weather.

[0094] At this point, the main agent discovers that the "Travel Planning Agent" has a "sub-agent," essentially a specialized version called the "International Travel Weather Agent." This agent is specifically designed for international travel weather inquiries. It not only checks the weather but also provides suggestions based on the travel scenario, such as "It will rain lightly in Paris tomorrow; we recommend bringing an umbrella. You could adjust your itinerary to include indoor activities." This is more relevant to the current group chat's "travel" theme than the ordinary "General Weather Inquiry Agent." Therefore, the main agent prioritizes this "sub-agent," adding it to the "Weekend Paris Travel Planning Group" and sending a notification: "In response to the request to 'check the weather in Paris tomorrow,' the 'International Travel Weather Agent' has been invited to join the group chat." Then, it invokes this agent to reply to user B: "@User B, the weather in Paris tomorrow (July 15th) is light rain, with temperatures ranging from 18-22°C and a wind force of level 2. We recommend bringing a lightweight umbrella, and you could change your itinerary to visiting the Louvre to avoid getting wet."

[0095] From a technical perspective, when the main AI agent handles this type of situation, it can be done in three steps: First, it uses natural language processing technology to parse the user's new message and extract the core needs (such as "Paris tomorrow" and "weather"). Then, it checks the "capability library" of the sub-agents in the current group chat (each AI has its own skill tags, such as "itinerary arrangement and hotel recommendation" for the "travel planning AI agent") to see if there is one that can handle "weather query". If not, it searches for a corresponding new AI agent in the global AI agent library. If there is a "sub-sub-agent" (i.e., a more specialized version) of the current sub-agent, it prioritizes that one because it is more suitable for the group chat scenario. For example, the "international travel weather AI agent" of the "travel planning AI agent" understands the weather needs during travel better than the general weather AI agent.

[0096] Furthermore, if the current sub-agent does not have a corresponding subordinate sub-agent, the master agent can generate one in real time based on the user's new needs. For example, the "travel planning agent" originally did not have an "international travel weather agent," but if the user asks about "Paris weather," the master agent can automatically combine the capabilities of "travel planning" and "weather query" to generate this subordinate sub-agent. Even further, the master agent can also select a new agent based on the user's historical preferences. If the user previously preferred "detailed and suggestive" responses, the "international travel weather agent" will be selected over the "general weather agent" because it can provide travel suggestions.

[0097] The core of the solution provided in this embodiment is to enable the capabilities of sub-agents in a group chat to be "expanded on demand." That is, no matter what new needs a user raises, as long as there is a corresponding specialized sub-agent (including a sub-version of the current sub-agent), the main agent can quickly add it to the group chat, allowing the specialized sub-agent to do its job. For example, if a user asks about the weather in a travel group, the main agent won't let the sub-agent responsible for planning the itinerary give a half-hearted answer; instead, it will find a more specialized weather sub-agent. This ensures the correct answer and makes group chat collaboration more efficient.

[0098] In cross-platform scenarios of chat applications, the following situation often occurs: A first user (e.g., "Li Si" on instant messaging platform A) wants to invite a second user (e.g., "Zhang San" on instant messaging platform B) to join a group chat on the first platform (A, e.g., a "Weekend Travel Planning Group"). However, the second user doesn't have the first platform installed (A, but has a private chat relationship with the first user on the second platform (B, B)). That is, the first and second users have a need for a cross-platform group chat session, but are limited by the barriers between the platforms. To attempt to solve this problem, this embodiment also provides a specific implementation method; please refer to [link to implementation details]. Figure 7 , Figure 7 A flowchart of a cross-platform group chat method between a first user and a second user provided in this embodiment of the disclosure, wherein process 700 includes the following steps: Step 701: In response to the fact that the first platform is not installed or deployed on the second user's terminal device, the second platform is installed or deployed on the first user's terminal device, and the first user and the second user have established a point-to-point communication relationship on the second platform, the group chat joining intention expressed by the second user in the target session interface between the second user and the first user on the second platform is mapped to the operation of adding the corresponding virtual second user to the target group chat session on the first platform. Step 702: Translate the group chat conversations associated with the virtual second user that appear in the target group chat conversation into the target conversation interface in a cross-platform message format, and translate the second user's reply to the translated group chat conversation in the target conversation interface into the virtual second user's conversation content in the target group chat conversation in a cross-platform message format.

[0099] First, it's important to clarify that the first platform and the second platform are two different instant messaging platforms (such as Platform A and Platform B mentioned above), with different message formats and user systems. Cross-platform message formatting involves converting a message from one platform (such as a group chat invitation on WeChat) into a format that the other platform can recognize (such as "text + link" on Platform B, where the link contains the group chat topic and member list), allowing messages to be transmitted between the two platforms. The virtual second user is a "stand-in" for the second user on the first platform (such as "B-Zhang San" on Platform A), with the same nickname, avatar, and the real user on the second platform, used to speak in the group chat on the first platform. The target conversation interface is the private chat window between the second user and the first user on the second platform (such as the chat box between "Zhang San" and "Li Si" on Platform B), where the second user receives and replies to group chat messages.

[0100] From a technical perspective, firstly, the first user sends a group chat invitation on the first platform (Platform A). The main agent converts the invitation into a cross-platform message (e.g., "Li Si invites you to join the 'Weekend Travel Planning Group' on Platform A, click to see details") and sends it to the second user's private chat window on the second platform (Platform B). Since the second user doesn't have the first platform (Platform A) installed, the main agent creates a virtual second user on the first platform (e.g., "B-Zhang San") and adds them to the target group chat (the "Weekend Travel Planning Group" on Platform A). Next, messages in the group chat (e.g., "Wang Wu" in Platform A says, "Going to Mount Tai next week, anyone want to go?") are converted by the main agent into the format of the second platform (Platform B) (e.g., "Weekend Travel Planning Group > Wang Wu: Going to Mount Tai next week, anyone want to go?") and sent to the second user's private chat window on Platform B. Once the second user replies in a private chat on platform B (e.g., "I want to go, what time is the meeting?"), the main AI agent then converts the reply into a message on the first platform (platform A) and sends it back to the group chat on platform A using the identity of the virtual second user ("DingTalk-Zhang San"). At this point, the people in the WeChat group (e.g., Li Si, Wang Wu) will see the reply from "B-Zhang San," just as if Zhang San were actually on platform A.

[0101] For a concrete example: Li Si is a user of platform A. He wants to invite Zhang San, a user of platform B, to join a "Weekend Travel Planning Group." However, Zhang San doesn't have platform A installed, but the two have private chats on platform B. Li Si clicks "Invite Member" on platform A, selects Zhang San's platform B account, and the main AI agent converts the invitation into a platform B message, sending it to Zhang San's private chat window on platform B. The message reads, "Li Si invites you to join the 'Weekend Travel Planning Group' on platform A. They're discussing a trip to Mount Tai next week." Zhang San replies in the platform B private chat, "Okay, I want to join." The main AI agent then quickly creates a virtual user named "B-Zhang San" on platform A, with the same avatar as the Zhang San on platform B and a nickname containing the "B" identifier, and then adds this virtual user to the "Weekend Travel Planning Group." A while later, Wang Wu sent a message in the group chat on platform A: "Going to Mount Tai next week, meeting at the train station at 8 am?" Upon seeing this, the main agent immediately converted the message into text on platform B and sent it to Zhang San's private chat window on platform B, writing: "Weekend Travel Planning Group > Wang Wu: Going to Mount Tai next week, meeting at the train station at 8 am?" Zhang San saw this and replied in the private chat on platform B: "No problem, I'll bring hiking poles." The main agent received the reply and immediately converted it into a message on platform A, sending it back to the group chat on platform A under the identity "B-Zhang San". At this point, Li Si and Wang Wu in group A saw the reply from "B-Zhang San" and started chatting with Zhang San. For example, Wang Wu said, "Then we'll wait for you at the train station entrance at 7:30 am." The main agent then converted this message back into text on platform B and sent it to Zhang San, who replied again, and the cycle continued.

[0102] To make this mechanism more user-friendly, the virtual second user's nickname and profile picture should be exactly the same as the real user on the second platform (for example, if Zhang San's profile picture on platform B is a mountain climbing photo, then the virtual user on platform A should also use this picture). This way, people in the group will see the virtual user just like they are seeing the real Zhang San, and won't feel unfamiliar. Furthermore, it should support cross-platform transmission of non-text messages, such as emoticons, images, and voice messages from platform A, which can be converted to the corresponding format on platform B (for example, a "smiley face" emoticon from platform A can be converted to a "smiley face" emoticon on platform B, and an image from platform A can be converted to an image link on platform B, which can be viewed with a single click). This makes the messages more vivid and interactive. Also, permissions should be synchronized. For example, if Zhang San sets "do not receive group chat @ messages" in DingTalk, then the virtual user "B-Zhang San" on platform A will not receive group chat @ messages. This ensures consistent permission logic and avoids inconvenience for Zhang San. Additionally, you can add a reason for the invitation in cross-platform messages, such as "You mentioned before on platform B that you like hiking, so we're inviting you to join this travel group." This way, when Zhang San sees it, he immediately understands why he was invited and can more easily integrate into the group chat.

[0103] The core of the solution provided in this embodiment is to break down the barriers between platforms, allowing users to participate in group chats without switching or installing new communication platforms. For example, Zhang San may not have platform A installed, but can still participate in a group chat on platform A through a private chat window on platform B. His replies will be synchronized to the group chat on platform A, as if he were actually on platform A. This not only makes it convenient for users but also maintains the continuity of group chat interaction, preventing the chat rhythm from being interrupted due to platform issues. For users who frequently use multiple different communication platforms, this will save a significant amount of time switching between platforms.

[0104] To further enhance the understanding of the solutions provided in the above embodiments of this application, the following description and illustration are provided in conjunction with some specific examples: Suppose that Ms. Li (mother) makes a request in the 1-on-1 main dialogue provided by XX Dialogue Assistant: "I need to tutor my child on the chicken and rabbit problem, but I don't know how to teach it" (see [link to dialogue assistant]). Figure 8-1 (User input). At this point, the XX dialogue assistant (main intelligent agent) judges that: "Chicken and Rabbit Tutoring" requires multi-person collaboration (child's friends, teachers) and multi-role capabilities (problem-solving guidance, error analysis), which meets the triggering conditions for "automatic group creation / group creation".

[0105] The XX chat assistant directly generates a native group invitation card in the main conversation (corresponding to...). Figure 8-1 The "Invite You to Join Group Chat" module in the chat window displays the message: "'XX Dialogue Assistant' invites you to join the group chat 'Xiaoming's Math Tutoring.' Enter to view details." The card format is identical to the "Join Group Invitation" in one-on-one chats on other or current instant messaging platforms. Ms. Li does not need to be redirected or manually create a group; she can simply click to join.

[0106] After joining the group, the XX dialogue assistant automatically brings the context of the 1v1 main conversation ("tutoring the chicken and rabbit problem" and "the child's wrong questions") into the group chat to ensure task continuity. The initial members of the group include: Ms. Li, the child's friend (Xiao Zhang), and the tutor (a vertical intelligent agent "mathematical problem-solving assistant" invited by the XX dialogue assistant).

[0107] After the group chat is started, the group chat assistant (i.e., the main intelligent agent) acts as the autonomous coordinator, continuously understanding the group chat message flow. When a new professional need arises in the group, the process of "bringing vertical agents (i.e., the sub-intelligent agents described in the above embodiments) into the group" is automatically triggered: Scenario 1: Medical consultation needs (see also) Figure 8-2 A member in the group, Xiao Zhang (a friend of the child), sent a message: "I sprained my ankle playing basketball recently, so I can't go out to play this weekend... It seems to be getting more and more swollen." Figure 8-2 (User input in the group). The group chat assistant (i.e., the main AI agent) determines that "swelling from a sprained ankle" requires medical expertise, and there is currently no relevant AI agent in the group. Therefore, the group chat assistant automatically executes: Select the matching vertical category Agent: "XX Health Butler" (Medical Specialist Intelligent Agent); Sending a group invitation: The group will display "Group chat assistant invites Wenxin Health Manager to join the group chat" (corresponding to...) Figure 8-2 (System prompts in the system). Routing context: Synchronize Xiao Zhang's "swelling ankle" message to XX Health Manager; The health manager posted in the group: "If swelling continues to worsen after a sprained ankle, stop all activity immediately and seek medical attention as soon as possible... Temporary treatments include rest, ice packs, pressure bandages, and elevation of the affected limb." Figure 8-2 (The agent's response in the text).

[0108] In group chats, users can trigger agent collaboration through explicit @ (precise invocation) or natural dialogue (implicit autonomy). The group chat assistant (main agent) is responsible for scheduling the speaking order and pace to ensure the readability of the group chat. Scenario 1: Explicit @ trigger (corresponding to Figure 8-3 ): Group admin Kevin (Ms. Li's friend) sent a message: "@GroupChatAssistant, please search for recent AI-related news for me, and then ask @Kevin's personal assistant to help analyze and provide insights." Figure 8-3 (Image input by the user).

[0109] After receiving the @ command, the group chat assistant (master agent) first executes the "search news" task and outputs: "Recent AI News: Baichuan Intelligence releases large-scale medical model... CES exhibition shifts towards practical application" (corresponding to...) Figure 8-3 (Reply in the group chat assistant) Next, the group chat assistant automatically @Kevin's personal assistant (the vertical AI agent "AI Industry Analysis Assistant") and routes the news context to it; Kevin's personal assistant's analysis: "The commercialization of AI is accelerating, with healthcare and consumer scenarios emerging as new growth areas... This is a long-term positive for the healthy development of the industry." Figure 8-3 (Personal assistant reply).

[0110] Scenario 2: Implicit Autonomous Triggering (corresponding to) Figure 8-4 ): The group discussed brainstorming slogans for the coffee brand "Duka". Figure 8-4 (User input from the website), member Xiao Zhang said, "It's best to combine Chinese and English," and Xuexue said, "And it should be simple, easy to understand, and catchy." (corresponding to...) Figure 8-4 (User replies).

[0111] The group chat assistant (master agent) automatically understands the "slogan brainstorming" needs without needing to be @mentioned, and combines the brand name "DuKa" (the connotation of "Du" being: moderate, daily) to output an autonomous reply: "@Kevin @Xuexue @XiaozhangDuKa: Daily Sip, daily moderate, taking into account both the English memorability and the connotation of the brand name 'Du,' concise, catchy, and easy to spread" (corresponding to...) Figure 8-4 (Reply from the group chat assistant).

[0112] As can be seen from the above examples, the technical solution provided in this embodiment has the following technical effects compared with existing similar technologies: 1) The process of joining the group chat from the main conversation is completely consistent with WeChat. Ms. Li does not need to learn any new operations and can join the group by simply clicking the invitation card; 2) The group chat assistant automatically invites the medical manager to the group and dispatches AI to analyze news. Users do not need to manually @ or select plugins; 3) The combination of explicit @ (precise invocation) and implicit autonomy (natural dialogue) satisfies both precise needs and maintains the naturalness of the group chat; 4) Specialized intelligent agents (health manager, AI analysis assistant) are responsible for their professional fields, avoiding the generalization error of a single robot "doing everything"; 5) The intelligent agents participate as "group members" and complete the collaboration in the same message stream. Ms. Li can clearly perceive "who is solving what problem" (e.g., "the health manager is answering the ankle sprain question" and "the personal assistant is analyzing AI news").

[0113] Further reference Figure 9 As an implementation of the methods shown in the above figures, this disclosure provides an embodiment of a group chat device involving intelligent agents, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0114] like Figure 9As shown, the group chat device 500 involving intelligent agents in this embodiment may include: a target dialogue determination unit 901, a group chat participant determination unit 902, a group chat session entry creation unit 903, and a context import and control reply unit 904. The target dialogue determination unit 901 is configured to determine the target dialogue that needs to be processed in the form of a group chat from the dialogue content initiated by the first user; the group chat participant determination unit 902 is configured to determine additional group chat participants according to the dialogue requirements of the target dialogue; wherein the group chat participants include at least one sub-agent different from the main agent and selectively include a second user different from the first user; the group chat session entry creation unit 903 is configured to create a group chat session entry containing the first user, group chat participants, and the main agent in the session where the target dialogue is located; the context import and control response unit 904 is configured to, in response to the first user entering the target group chat session through the group chat session entry, import context information related to the target dialogue into the target group chat session, and control the sub-agent to reply to the first user in the target group chat session based on the target session and the context information.

[0115] In this embodiment, the specific processing and technical effects of the target dialogue determination unit 901, the group chat participant determination unit 902, the group chat session entry creation unit 903, and the context import and control response unit 904 in the group chat conversation device 900 involving intelligent agents can be found in the following references. Figure 2 The relevant descriptions of steps 201-204 in the corresponding embodiments will not be repeated here.

[0116] In some other optional implementations of this embodiment, the target dialogue determination unit 901 may include: The first target dialogue determination subunit is configured to identify conversations whose dialogue content expresses an intention to require multi-user participation as the first target dialogue; or The second target dialogue determination subunit is configured to identify conversations in which the dialogue content expresses a need to solve a problem with actual complexity exceeding a preset complexity threshold as the second target dialogue. The target dialogue includes either the first target dialogue or the second target dialogue.

[0117] In some other optional implementations of this embodiment, the group chat participant determination unit 902 may include: The second user is identified as a participating subunit and configured to respond to the target dialogue as the first target dialogue. Based on the dialogue requirements of the first target dialogue, other users who need to participate in the conversation are identified as the second user, and the intelligent agents used to resolve the dialogue requirements are identified as sub-intelligent agents. Among them, the participants in the group chat include: the second user and the sub-intelligent agents.

[0118] In some other optional implementations of this embodiment, the group chat device 900 in which the intelligent agent participates may further include: The pending dialogue determination unit is configured to determine the pending dialogues that require a response from the main agent or a sub-agent based on the content of the group chat dialogue initiated by the first user and / or the second user in the target group chat session. The unique sub-agent reply unit is configured to respond to a dialogue that does not indicate a reply target and only one sub-agent exists, by invoking the sub-agent to reply to the first user and / or the second user who initiated the dialogue in the target group chat session based on the dialogue information related to the dialogue to be replied to in the target group chat session.

[0119] In some other optional implementations of this embodiment, the group chat device 900 in which the intelligent agent participates may further include: The target sub-agent and combination determination unit is configured to, in response to a dialogue to be replied to not indicating a reply object and the existence of multiple different sub-agents, determine a target sub-agent or a combination of target sub-agents that matches the dialogue to be replied to from among the multiple sub-agents; wherein, the combination of target sub-agents includes at least two different target sub-agents, and the different target sub-agents are combined in the order of reply; The target sub-agent response unit is configured to invoke the target sub-agent to reply to the first user and / or the second user who initiated the dialogue in the target group chat session, based on dialogue information related to the dialogue to be replied to in the target group chat session; or The target sub-agents are combined into sequential reply units, which are configured to call different target sub-agents in the reply order to reply to the first user and / or the second user who initiated the reply in the target group chat session based on the dialogue information related to the dialogue to be replied to in the target group chat session.

[0120] In some other optional implementations of this embodiment, the group chat device 900 in which the intelligent agent participates may further include: The pending dialogue determination unit is configured to determine the pending dialogues that require a response from the main agent or a sub-agent based on the content of the group chat dialogue initiated by the first user and / or the second user in the target group chat session. The designated sub-agent reply unit is configured to, in response to a dialogue indicating that the target sub-agent is the reply object, invoke the target sub-agent to reply to the first user and / or the second user who initiated the dialogue in the target group chat session based on the dialogue information related to the dialogue to be replied to in the target group chat session; wherein, the target agent is a unique sub-agent or a designated sub-agent among multiple different sub-agents.

[0121] In some other optional implementations of this embodiment, the group chat device 900 in which the intelligent agent participates may further include: The pending dialogue determination unit is configured to determine the pending dialogues that require a response from the main agent or a sub-agent based on the content of the group chat dialogue initiated by the first user and / or the second user in the target group chat session. The allocation notification dialogue response unit is configured to generate an allocation sub-agent to respond to the main agent of the dialogue to be replied to in response to the main agent as the reply object in the target group chat session. The sub-agent reply unit is configured to invoke the sub-agent to reply to the first user and / or the second user who initiated the dialogue to be replied to in the target group chat session based on the dialogue information related to the dialogue to be replied to in the target group chat session.

[0122] In some other optional implementations of this embodiment, the group chat participant determination unit 902 may include: The multi-sub-agents determine the participating sub-units, which are configured to respond to the target dialogue as the second target dialogue. The complete task corresponding to the problem-solving requirements is broken down into multiple sub-tasks, and the agents that execute each sub-task are determined as each sub-agent. Among them, the group chat participants include: each sub-agent used to execute each sub-task. Correspondingly, the above-mentioned input and control response unit 903 includes a control response subunit configured to control the sub-agent to reply to the user in the target group chat session based on the target session and the above-mentioned information. The control response subunit can be further configured to: The control agents respond to the user in the target group chat in the order of execution of each sub-task, based on the target conversation, the preceding information, and the execution results of the previous sub-tasks.

[0123] In some other optional implementations of this embodiment, the group chat device 900 in which the intelligent agent participates may further include: The similar group chat dialogue merging and reply unit is configured to respond to situations where the group chat participants also include at least one second user, and the first user and the second user each initiate a similar group chat dialogue with an actual similarity exceeding a preset similarity threshold within a preset time window. This unit calls a sub-agent matching the session requirements of the similar group chat dialogue to simultaneously reply to the similar group chat dialogue initiated by the first user and the second user in the target group chat session by generating a single reply based on the dialogue information related to the similar group chat dialogue in the target group chat session. The preset time window is defined as receiving the later-initiated similar group chat dialogue before the sub-agent matching the earlier-initiated similar group chat dialogue has completed generating its reply.

[0124] In some other optional implementations of this embodiment, a single reply is presented in the target group chat session by simultaneously referencing both the earlier and later similar group chat conversations.

[0125] In some other optional implementations of this embodiment, the group chat device 900 in which the intelligent agent participates may further include: The independent reply unit for similar group chat dialogues is configured to respond to a configuration item in the preference configuration information of the first user or the second user that disallows merging replies to similar dialogues. It calls a sub-agent that matches the content of the similar group chat dialogue and, based on the dialogue information related to the content of the similar group chat dialogue in the target group chat dialogue, replies to the similar group chat dialogue initiated by the first user and the second user respectively by generating different reply content in the target group chat dialogue.

[0126] In some other optional implementations of this embodiment, the group chat device 900 in which the intelligent agent participates may further include: The new sub-agent addition and response unit is configured to respond to a new group chat dialogue request initiated by any user in the target group chat session that does not match any sub-agent currently included in the target group chat session. It then determines a new sub-agent corresponding to the new dialogue request, adds the new sub-agent as a new group chat participant to the target group chat session, and invokes the new sub-agent to respond to the user who initiated the new group chat dialogue in the target group chat session based on the session information related to the new group chat dialogue. Here, "any user" includes the first user and / or the second user, and "new sub-agent" includes the subordinate sub-agent of any currently included sub-agent.

[0127] In some other optional implementations of this embodiment, the group chat device 900 in which the intelligent agent participates may further include: The reason information determination unit is configured to determine the reason information for the second user being added to the target group chat session by the first user in response to any second user being added to the target group chat session by the first user; wherein the reason information is determined based on the historical dialogue in the target group chat session and the action information of the first user performing to add a group chat participant. The cause information presentation unit is configured to present cause information in the target group chat session to the second user who enters the target group chat session; wherein the cause information is set to be visible only to the second user.

[0128] In some other optional implementations of this embodiment, the second user is added to the target group chat session on the first platform by the group chat invitation information sent to the second platform by the first user on the first platform through a cross-platform message format.

[0129] In some other optional implementations of this embodiment, the group chat device 900 in which the intelligent agent participates may further include: The cross-platform mapping unit is configured to respond to situations where the first platform is not installed or deployed on the second user's terminal device, the second platform is installed or deployed on the first user's terminal device, and the first user and the second user establish a point-to-point communication relationship on the second platform, by mapping the group chat joining intention expressed by the second user in the target session interface between the second user and the first user on the second platform to the operation of adding the corresponding virtual second user to the target group chat session on the first platform. The cross-platform paraphrasing unit is configured to paraphrase the group chat dialogue associated with the virtual second user in the target group chat session in a cross-platform message format on the target session interface, and to paraphrase the second user's reply to the paraphrased group chat dialogue in the target session interface in a cross-platform message format as the virtual second user's dialogue content in the target group chat session.

[0130] This embodiment exists as a device embodiment corresponding to the above method embodiment. The group chat conversation device with intelligent agents provided in this embodiment actively identifies the target dialogue that needs to be processed in the form of a group chat in the one-on-one chat with the first user. Based on the dialogue requirements, it dynamically determines the group chat participants, including professional sub-intelligent agents (providing targeted capabilities) and optional second users (associating social relationships). Then, it directly creates a native group chat entry containing all participants in the one-on-one chat and automatically synchronizes the one-on-one chat context information after the user enters the group chat, realizing a seamless migration from one-on-one chat to group chat without requiring the user to perform tedious operations such as manually creating a group, adding members, and forwarding history. At the same time, the sub-intelligent agents directly reply to the user in the group chat based on the target dialogue content and the synchronized context information, maintaining the continuity and naturalness of the dialogue. It not only meets the needs of professional problem solving but also integrates social collaboration scenarios, significantly improving the efficiency of group chat collaboration and user experience.

[0131] According to embodiments of this disclosure, this disclosure also provides an electronic device, the electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to implement the group chat session method involving intelligent agents described in any of the above embodiments when executed.

[0132] According to embodiments of this disclosure, this disclosure also provides a readable storage medium storing computer instructions that, when executed by a computer, enable the implementation of the group chat session method involving intelligent agents as described in any of the above embodiments.

[0133] According to embodiments of this disclosure, this disclosure also provides a computer program product that, when executed by a processor, can implement the group chat session method involving intelligent agents as described in any of the above embodiments.

[0134] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, 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 present disclosure described and / or claimed herein.

[0135] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 602 or a computer program loaded into random access memory (RAM) 603 from storage unit 608. RAM 603 may also store various programs and data required for the operation of device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0136] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0137] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 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 computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the agent-involved group chat session method. For example, in some embodiments, the agent-involved group chat session method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the agent-involved group chat session method described above can be performed. Alternatively, in other embodiments, computing unit 601 may be configured by any other suitable means (e.g., by means of firmware) to perform a group chat session method involving agents.

[0138] Various embodiments 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), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments 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.

[0139] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may 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 server.

[0140] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. 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 fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0141] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. 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).

[0142] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user 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., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0143] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and Virtual Private Server (VPS) services, such as high management difficulty and weak business scalability.

[0144] According to the technical solution of this disclosure, the main intelligent agent actively identifies the target dialogue that needs to be processed in the form of a group chat in the one-on-one chat with the first user. Based on the dialogue requirements, it dynamically determines the group chat participants, including professional sub-intelligent agents (providing targeted capabilities) and optional second users (associating social relationships). Then, it directly creates a native group chat entry point containing all participants in the one-on-one chat session, and automatically synchronizes the one-on-one chat context information after the user enters the group chat, realizing a seamless migration from one-on-one chat to group chat without requiring the user to perform tedious operations such as manually creating a group, adding members, and forwarding history. At the same time, the sub-intelligent agents directly reply to the user in the group chat based on the target dialogue content and the synchronized context information, maintaining the continuity and naturalness of the dialogue. This not only meets the needs of professional problem solving, but also integrates social collaboration scenarios, significantly improving the efficiency of group chat collaboration and user experience.

[0145] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0146] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. 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 disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for group chat sessions involving intelligent agents, wherein, Applied to the main intelligent agent, including: In the conversation initiated by the first user, identify the target conversation that needs to be handled in the form of a group chat. Additional group chat participants are determined based on the dialogue requirements of the target dialogue; wherein, the group chat participants include at least one sub-agent that is different from the main agent and, selectively, a second user that is different from the first user; Create a group chat session entry point in the session containing the first user, the group chat participants, and the main agent; In response to the first user entering the target group chat session through the group chat session entry point, the preceding information related to the target conversation is brought into the target group chat session, and the sub-agent is controlled to reply to the first user in the target group chat session based on the target conversation and the preceding information.

2. The method according to claim 1, wherein, The process of identifying the target dialogue that needs to be processed in a group chat format from the dialogue content initiated by the first user includes: Conversations in the dialogue content that express the intention to require multi-user participation are identified as the first target dialogue; or Conversations in which the dialogue content expresses a need to solve a problem with an actual complexity exceeding a preset complexity threshold are identified as the second target dialogue. The target dialogue includes either the first target dialogue or the second target dialogue.

3. The method according to claim 2, wherein, The step of determining additional group chat participants based on the dialogue requirements of the target dialogue includes: In response to the target dialogue being the first target dialogue, based on the dialogue requirements of the first target dialogue, other users who need to participate in the conversation are identified as the second user, and the intelligent agent used to resolve the dialogue requirements is identified as the sub-intelligent agent; wherein, the group chat participants include: the second user and the sub-intelligent agent.

4. The method according to claim 3, further comprising: Based on the content of the group chat dialogue initiated by the first user and / or the second user in the target group chat session, determine the dialogues that need to be replied to by the main agent or the sub-agent; In response to the fact that the dialogue to be replied to does not indicate a reply target and there is only one sub-agent, the sub-agent is invoked to reply to the first user and / or the second user who initiated the dialogue to be replied to in the target group chat session based on the dialogue information related to the dialogue to be replied to in the target group chat session.

5. The method according to claim 4, further comprising: In response to the fact that the dialogue to be replied to does not indicate a reply object and there are multiple different sub-agents, a target sub-agent or a combination of target sub-agents that matches the dialogue to be replied to is determined from among the multiple sub-agents; wherein, the combination of target sub-agents includes at least two different target sub-agents, and the different target sub-agents are combined in the order of reply; The target sub-agent is invoked to reply to the first user and / or the second user who initiated the pending reply dialogue in the target group chat session based on the dialogue information related to the pending reply dialogue in the target group chat session; or According to the reply order, different target sub-agents are sequentially invoked to reply to the first user and / or the second user who initiated the dialogue to be replied to in the target group chat session based on the dialogue information related to the dialogue to be replied to in the target group chat session.

6. The method according to claim 3, further comprising: Based on the content of the group chat dialogue initiated by the first user and / or the second user in the target group chat session, determine the dialogues that need to be replied to by the main agent or the sub-agent; In response to the pending dialogue indicating that the target sub-agent is the reply object, the target sub-agent is invoked to reply to the first user and / or the second user who initiated the pending dialogue in the target group chat session based on the dialogue information related to the pending dialogue in the target group chat session; wherein, the target sub-agent is a unique sub-agent or a designated sub-agent among multiple different sub-agents.

7. The method according to claim 3, further comprising: Based on the content of the group chat dialogue initiated by the first user and / or the second user in the target group chat session, determine the dialogues that need to be replied to by the main agent or the sub-agent; In response to the pending dialogue indicating that the main agent is the respondent, an allocation notification dialogue is generated in the target group chat session to assign the sub-agent to reply. The sub-agent is invoked to reply to the first user and / or the second user who initiated the dialogue in the target group chat session based on the dialogue information related to the dialogue to be replied to in the target group chat session.

8. The method according to claim 2, wherein, The step of determining additional group chat participants based on the dialogue requirements of the target dialogue includes: In response to the target dialogue being the second target dialogue, the complete task corresponding to the problem-solving requirement is divided into multiple sub-tasks, and the agents that execute each sub-task are identified as each sub-agent; wherein, the group chat participants include: each sub-agent used to execute each sub-task; Correspondingly, the control of the sub-agent to reply to the user in the target group chat session based on the target session and the aforementioned information includes: The control system ensures that each of the sub-agents responds to the user in the target group chat session according to the execution order of each sub-task, based on the target session, the preceding information, and the execution results of the previous sub-tasks.

9. The method according to claim 1, further comprising: In response to the group chat participants including at least one second user, and the first user and the second user respectively initiating similar group chat dialogues with an actual similarity exceeding a preset similarity threshold within a preset time window, a sub-agent matching the session requirements of the similar group chat dialogue is invoked to simultaneously reply to the similar group chat dialogues initiated by the first user and the second user in the target group chat session based on the dialogue information related to the similar group chat dialogue in the target group chat session, by generating a single reply content; wherein, the preset time window is when the similar group chat dialogue initiated later is received before the sub-agent matching the previously initiated similar group chat dialogue has completed generating the reply content.

10. The method according to claim 9, wherein, The content of a single reply is presented in the target group chat session by simultaneously referencing both the earlier and later similar group chat conversations.

11. The method of claim 9, further comprising: In response to the fact that the preference configuration information of the first user or the second user contains a configuration item that does not allow merging replies to similar conversations, a sub-agent matching the content of the similar group chat is invoked to reply to the similar group chat initiated by the first user and the second user respectively in the target group chat by generating different reply content based on the conversation information related to the content of the similar group chat in the target group chat.

12. The method according to claim 1, further comprising: In response to a new group chat request initiated by any user in the target group chat session that does not match any of the sub-agents currently included in the target group chat session, a new sub-agent corresponding to the new dialogue request is determined, and the new sub-agent is added to the target group chat session as a new group chat participant. Furthermore, the new sub-agent is invoked to reply to the user who initiated the new group chat session in the target group chat session based on the session information related to the new group chat request. Here, the "any user" includes the first user and / or the second user, and the new sub-agent includes the subordinate sub-agent of any of the currently included sub-agents.

13. The method according to any one of claims 1-12, further comprising: In response to any second user being added to the target group chat session by the first user, reason information for the second user being added to the target group chat session by the first user is determined in the target group chat session; wherein, the reason information is determined based on the historical dialogue in the target group chat session and the action information of the first user performing to add a group chat participant; The reason information is presented to the second user who enters the target group chat session; wherein the reason information is set to be visible only to the second user.

14. The method according to claim 13, wherein, The second user joins the target group chat session on the first platform by sending a group chat invitation message to the second platform via a cross-platform message format from the first user on the first platform.

15. The method of claim 14, further comprising: In response to the fact that the first platform is not installed or deployed on the second user's terminal device, the second platform is installed or deployed on the first user's terminal device, and the first user and the second user have established a point-to-point communication relationship on the second platform, the group chat intention expressed by the second user in the target session interface between the second user and the first user on the second platform is mapped on the first platform to add the corresponding virtual second user to the target group chat session. The group chat dialogue associated with the virtual second user that appears in the target group chat session is relayed to the target session interface in a cross-platform message format, and the reply content of the second user in the target session interface to the relayed group chat dialogue is relayed to the dialogue content of the virtual second user in the target group chat session in a cross-platform message format.

16. A group chat device involving intelligent agents, wherein, Applied to the main intelligent agent, including: The target dialogue determination unit is configured to determine the target dialogue that needs to be processed in the form of a group chat from the dialogue content initiated by the first user. The group chat participant determination unit is configured to determine additional group chat participants based on the dialogue requirements of the target dialogue; wherein, the group chat participants include at least one sub-agent that is different from the main agent and selectively include a second user that is different from the first user; The group chat session entry creation unit is configured to create a group chat session entry containing the first user, the group chat participants, and the main agent in the session where the target dialogue is located. The above-mentioned input and control response unit is configured to, in response to the first user entering the target group chat session through the group chat session entry, input the above-mentioned information related to the target dialogue into the target group chat session, and control the sub-agent to reply to the first user in the target group chat session based on the target session and the above-mentioned information.

17. An electronic device comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the group chat session method involving intelligent agents as described in any one of claims 1-15.

18. A non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform a group chat session method involving an agent as described in any one of claims 1-15.

19. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the group chat session method involving an agent according to any one of claims 1-15.

Images