Multi-user large model interaction method, device and platform based on tree structure

Abstract

The invention relates to a multi-user large model interaction method, device and platform based on a tree structure, and the method comprises the steps: responding to a session sharing request for interaction between a first user and a generative large model, forming nodes according to the interaction content of each round, and constructing a father-son association relationship between the nodes according to the sequence of the interaction of each round, forming a session tree; when a second user performs additional interaction with the generative large model at any target node of the session tree, branching at the target node according to the session information to form session branches; and updating the session tree based on the session branches. According to the method, substantial improvement is realized in the aspects of improving the dynamism, the flexibility, the collaboration, the manageability and the like of the session content, and efficient exploration and collaboration in a multi-user large model interaction process can be effectively met.

Description

Technical Field

[0001] The present invention relates to the technical fields of natural language processing and large model interaction, and particularly relates to a multi-user large model interaction method, device and platform based on a tree structure. Background Art

[0002] At present, large model service providers usually provide large model interaction services to users through web pages or mobile applications, and the interaction mode is based on the basic mode of one question and one answer.

[0003] In this mode, each user input and model output constitute the smallest information unit in the interaction process; several smallest units form a complete conversation in chronological order. Each user has an independent set of conversations, and the conversation contents between different users are isolated from each other and cannot be accessed or interfered with each other.

[0004] On this basis, some service providers further provide a historical message modification function, that is, users are allowed to modify the sent messages, and use the modified content and the previous historical messages as the context to regenerate new model outputs. The scope of modification permissions varies on different platforms: some platforms only allow the modification of the most recent user message, while some platforms support the modification of any historical message.

[0005] In addition, some service providers also provide a conversation sharing function, which allows users to share the content of a certain conversation with others in the form of generating sharing links, pictures, documents, and applets. The user who receives the share can copy the shared content to their own account by scanning the code or other means, and continue to ask questions based on this copy, but the connection with the original conversation is disconnected.

[0006] In summary, first, the current sharing function essentially generates a copy of the conversation at a specific moment. Once the sharing is completed, even if new conversation content is generated in the original conversation later, the shared copy will not be updated synchronously, resulting in information isolation and inability to evolve continuously.

[0007] Second, the content shared by most platforms is only for reading, or even if continued questioning is allowed, it is achieved by copying the copy, which is equivalent to disconnecting the connection with the original conversation. The subsequent conversation development is completely isolated and cannot be integrated into the original context tree, lacking continuity and collaboration.

[0008] Third, the existing implementations are usually based on linear conversation history, and users can only proceed step by step along the fixed timeline. If they want to modify the content of a certain node in the history or continue to ask questions from it, they often need to copy a new conversation, resulting in a large number of repeated operations and a fragmented user experience.

[0009] Fourth, the sharing mechanism is designed for one-way dissemination (from the sharer to the recipient), lacking the ability for multiple people to collaborate and explore based on the same conversation context. It is impossible for multiple people to pose questions around the same topic and share knowledge accumulation and exploration paths based on the same context tree.

[0010] Fifth, since it is a static copy, once generated, the content format and structure are basically fixed, making it difficult to support subsequent advanced functions such as flexible appending, sorting, and archiving. Summary of the Invention

[0011] The present invention provides a multi-user large model interaction method, device, and platform based on a tree structure, aiming to solve problems such as content staticization, inability to dynamically update, limited branch exploration, and insufficient multi-person collaboration ability in the existing large model interaction system during session sharing and collaboration.

[0012] In a first aspect, the present application provides a multi-user large model interaction method based on a tree structure, the method comprising: Responding to a session sharing request for a first user to interact with a generative large model, forming nodes based on the interaction content of each round, and constructing a parent-child association relationship between the nodes according to the order between the interactions of each round to form a session tree; Whenever a second user conducts an additional interaction with the generative large model at any target node of the session tree, forking at the target node according to the session information to form a session branch; and updating the session tree based on the session branch.

[0013] In a second aspect, the present application further provides a multi-user large model interaction device based on a tree structure, the device comprising: A session tree initial construction unit, configured to respond to a session sharing request for a first user to interact with a generative large model, form nodes based on the interaction content of each round, and construct a parent-child association relationship between the nodes according to the order between the interactions of each round to form a session tree; A session tree expansion unit, configured to whenever a second user conducts an additional interaction with the generative large model at any target node of the session tree, fork at the target node according to the session information to form a session branch; and update the session tree based on the session branch.

[0014] In a third aspect, the present application further provides a session sharing platform, which deploys the above-mentioned multi-user large model interaction device based on a tree structure; The multi-user large model interaction device based on a tree structure accesses a generative large model through a large model interaction interface to call the generative large model through the large model interaction interface to provide large model interaction services for users; The multi-user large model interaction device based on a tree structure is also communicatively connected to a database, and the database is used to store and read session data.

[0015] In a fourth aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned multi-user large model interaction method based on a tree structure are implemented.

[0016] In a fifth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the above-mentioned multi-user large model interaction method based on a tree structure are implemented.

[0017] The above-mentioned at least one technical solution adopted in the embodiments of the present application can achieve the following beneficial effects: First, the present application supports the dynamic synchronization and real-time update of session content. The shared session is no longer a static copy, but is synchronized and updated as the original session content increases or is modified, so that the shared content remains up-to-date, improving the real-time performance and consistency of information sharing; Second, the present application supports the modification of any node and multi-branch exploration. Users can arbitrarily select a node in the session history to modify or append questions, and the system automatically generates a new branch from this node to form a tree-like evolution structure, supporting multi-path exploration and divergent thinking, breaking through the limitations of traditional linear conversations; Third, the present application improves the efficiency of multi-person collaboration and co-creation. Multiple users can independently carry out explorations based on the same shared session, each forming a branch while retaining logical associations, greatly improving the collaboration efficiency and innovation space based on the same topic; Fourth, the present application enhances the scalability and manageability of session content. Based on the tree-like structure organization method, it can support subsequent diverse expansion functions, such as session summary, branch merging, and structured display, etc., significantly improving the management efficiency and utilization value of large-scale session data; In summary, the present application has achieved substantial improvements over the prior art in terms of improving the dynamics, flexibility, collaboration, and manageability of session content, and can effectively meet the needs of efficient exploration and collaboration in the process of multi-user large model interaction. Description of the Drawings

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0019] Figure 1 Shows a schematic structural diagram of a session sharing platform according to an embodiment of the present application; Figure 2 Shows a schematic flowchart of a multi-user large model interaction method based on a tree structure according to an embodiment of the present application; Figure 3 Shows a schematic structural diagram of an initial session tree according to an embodiment of the present application; Figure 4 Shows a schematic structural diagram of an updated session tree according to another embodiment of the present application; Figure 5 Shows a schematic structural diagram of a session tree updated again according to still another embodiment of the present application; Figure 6 Shows a schematic flowchart of a multi-user large model interaction method based on a tree structure according to another embodiment of the present application; Figure 7 Shows a schematic structural diagram of a multi-user large model interaction device based on a tree structure according to an embodiment of the present application; Figure 8 Shows a schematic structural diagram of a computer device according to an embodiment of the present application. Detailed implementation manners

[0020] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0021] The concept of the present invention lies in: by introducing a session data organization method based on tree structure management, the dynamic synchronization of session content, the flexible modification of historical nodes and branch generation are realized, significantly improving the controllability, collaboration and exploration efficiency of multiple users in the process of large model interaction.

[0022] Specifically, for the dynamic synchronization problem, the existing session sharing method is based on static copies and cannot reflect the subsequent evolution of the original session. The present application proposes a sharing mechanism that can be dynamically updated and synchronized in real time with the original session to ensure that the shared content is always kept updated.

[0023] For the problem of flexible modification and branch generation of historical nodes, in the prior art, users can only interact linearly and cannot freely modify and branch historical nodes. The present application uses a tree structure to manage the conversation history, enabling users to continue interacting from any node and generate independent branches, supporting multi-path exploration.

[0024] Regarding the problem of multi - person collaboration, the content interaction after sharing in existing systems is restricted and isolated, lacking the ability for multiple people to jointly explore based on the same conversation context. This application supports multiple people to freely explore based on a shared conversation tree, forming their own independent branches and maintaining a logical association with the original conversation, enhancing the collaboration experience.

[0025] Regarding the problem of version control and backtracking, existing technologies lack a fine - grained session version control and backtracking mechanism. This application supports users to backtrack arbitrarily and compare the evolution processes of different branches by explicitly managing each interaction node, facilitating knowledge organization and reuse.

[0026] Regarding the problems of scalability and manageability, and the poor scalability of static copies, this application is based on a tree - shaped conversation structure, supports subsequent diverse expansion of conversation content, and greatly improves the maintainability and utilization value of session data.

[0027] This application is mainly applied to scenarios where multiple users interact with a generative large - model. Among them, the generative large - model can be understood as an interactive generative large - model constructed based on natural language processing methods. The generative large - model can be any one of the existing technologies, and this application does not make any limitations in this regard.

[0028] Figure 1 shows a schematic structural diagram of a session sharing platform according to an embodiment of the present application. From Figure 1 it can be seen that the session sharing platform 100 deploys a tree - structure - based multi - user large - model interaction device of the present application. This device is used to implement the business logic of the tree - structure - based multi - user large - model interaction method of the present application (see Figure 7 ).

[0029] The tree - structure - based multi - user large - model interaction device can be deployed in the core service layer of the session sharing platform 100. The core service layer also has an API gateway. The role of the API gateway is to route user requests. In this application, for the session sharing requests of users, when they reach the API gateway of the core service layer of the session sharing platform 100, the API gateway routes them to the tree - structure - based multi - user large - model interaction device. The tree - structure - based multi - user large - model interaction device provides session tree orchestration services and has the business logic for implementing the tree - structure - based multi - user large - model interaction method provided by the present application.

[0030] And from Figure 1It can be seen that the multi-user large model interaction device based on a tree structure accesses a generative large model through a large model interaction interface, and calls the generative large model through the large model interaction interface to provide large model interaction services for users. Specifically, in some embodiments, the multi-user large model interaction device based on a tree structure can implement the interaction between users and the generative large model on the session sharing platform 100 through technical means such as "preparing context requests" and "returning model inference results".

[0031] Moreover, the multi-user large model interaction device based on a tree structure is also communicatively connected to a database, and the database provides data storage services. Specifically, the multi-user large model interaction device based on a tree structure performs storage and reading operations on data such as session record data, session trees, session branches, and nodes through the database.

[0032] Please refer to Figure 1 again. The method of the present application can be applied to the session sharing platform 100, specifically executed by a multi-user large model interaction device based on a tree structure. Multiple users are network communicatively connected to the session sharing platform 100, and each user can communicate about topics on the session sharing platform 100. The session sharing platform can be presented in various forms such as community forums and official accounts. The present application does not limit this. And the session sharing platform "accesses" a generative large model, and the access method can be any one of the following ways: through the API interface method ( Figure 1 as shown), or using an SDK developed by the official or a third party to encapsulate the API call process, and customized fine-tuning, etc. The present application does not limit this.

[0033] A user interacts with the generative large model, and then can share and publish his / her session with the generative large model in the form of a session tree to the session sharing platform. Other users, even the original user who initiated the session, can "continue to ask questions" at any node of the session tree, thus realizing the common participation of multiple users in the session, and having strong scalability.

[0034] It should be noted that Figure 1 This is only an exemplary illustration of the application of the present application. The present application is not limited to the above scenarios and system architectures. Any system or platform or architecture that can implement the business logic of the multi-user large model interaction method based on a tree structure of the present application is acceptable.

[0035] Figure 2 shows a schematic flow chart of a multi-user large model interaction method based on a tree structure according to an embodiment of the present application. From Figure 2 it can be seen that this embodiment includes steps S210 to S220: Step S210, in response to a session sharing operation in which a first user interacts with a generative large model, form nodes based on the interaction content of each round, and construct a parent-child association relationship between the nodes according to the order between each round of interactions, so as to form a session tree.

[0036] A user, denoted here as the first user A, interacts with a generative large model on a topic. During the interaction, the first user A usually has at least one round of interaction with the generative large model. "Interaction" generally refers to various data carriers input by the user into the generative large model, including but not limited to text, images, videos, and other forms of files and structured or unstructured data. The generative large model responds according to the user's input, and the specific form of the interaction is not limited in this application. For the convenience of description below, the text interaction method of the user asking questions and the generative large model answering is used for narration. For example, one round of interaction usually means that the first user A asks a question Q1, and the generative large model makes a response to the question Q1 and gives an answer A1. The question Q1 and the answer A1 form a Q&A pair. Here, the interaction content between the first user A and the generative large model is denoted as the first session information.

[0037] After the first user A finishes interacting with the generative large model on a topic, the session can be shared. The sharing method can adopt any one of the existing technologies, such as by generating a sharing link, picture, document, small program, etc. The session can be shared to a session sharing platform. For example, the previous session sharing platform can be a community forum or a public account, etc.

[0038] Specifically, a "Share" virtual button can be provided on the front-end interface of the interaction between the first user A and the generative large model. When the user clicks the virtual button, a "session sharing request" is generated and sent to the session sharing platform at the same time, so as to realize the session sharing operation.

[0039] After receiving the session sharing request, the session sharing platform constructs a conversation tree based on the content to be shared by the first user A and shares the session in the form of a tree structure.

[0040] The session tree is constructed by the following method: forming nodes of the session tree according to the interaction content of each round, and constructing a parent-child association relationship between the nodes according to the order between each round of interactions.

[0041] The definitions of nodes (Checkpointer) and node identities (Checkpointer ID) are as follows: A node refers to a specific session record, which serves as a fixed checkpoint in the session history. Every time a user asks a question and the generative model generates a reply, a new node (Checkpointer) is formed, and a unique node identity (Checkpointer ID) is assigned to the node for subsequent location and reference.

[0042] The sharing of the session tree in this step can be understood as initial session sharing, and the session tree can be specifically constructed according to the following method: obtaining the session record data of each round of interaction of the first user, and structuring the session record data into session information; taking the session information of a round of interaction as a node, and assigning a node identity to the node; for any node, marking the previous node adjacent to the node as its parent node, and marking the subsequent node adjacent to the node as a child node, so as to construct a parent-child association relationship between nodes; assigning an identity to the formed session tree, and attributing it to the first user.

[0043] Please refer to Figure 3 , Figure 3 A schematic diagram of the structure of an initial session tree according to an embodiment of the present application is shown. Figure 3 It can be seen that the first user A initiated three rounds of interaction, namely question Q1, question Q2, and question Q3, and the order of questions was question Q1, question Q2, and question Q3; the generative large model responded respectively, namely answer A1, answer A2, and answer A3. Question Q1 and answer A1 are a question-answer pair, question Q2 and answer A2 are a question-answer pair, and question Q3 and answer A3 are a question-answer pair.

[0044] After obtaining the record data of question Q1, question Q2, question Q3, answer A1, answer A2 and answer A3, this information can be structured to form session information. Then, the session information of each round of interaction is used as a node Checkpointer, and a node identity is assigned to it. For example, here, the node identity assigned to question Q1 and answer A1 is recorded as Checkpointer ID1, the node identity assigned to question Q2 and answer A2 is recorded as Checkpointer ID2, and the node identity assigned to question Q3 and answer A3 is recorded as Checkpointer ID3.

[0045] The session information may include but is not limited to: user messages, model reply messages, and context information; wherein the context information includes: at least one of system messages input to the large model, historical messages, session rounds, and user information.

[0046] A user message can be understood as a question posed by the user to the generative large model, and the model's response message can be understood as a reply from the generative large model to the user's question. The conversation information can include not only the conversation content between the user and the generative large model, but also some other information, such as context information, which includes, but is not limited to, at least one of the system message input to the large model, historical messages, conversation turns, and user information.

[0047] Then, establish the association relationship between Checkpointer ID1, Checkpointer ID2, and Checkpointer ID3. The relationship between Checkpointer ID1, Checkpointer ID2, and Checkpointer ID3 is that of parent and child nodes. Specifically, Checkpointer ID1 is the parent node of Checkpointer ID2, and Checkpointer ID2 is the parent node of Checkpointer ID3. In other words, Checkpointer ID2 is the child node of Checkpointer ID1, and Checkpointer ID3 is the child node of Checkpointer ID2. A parent node can have multiple child nodes, while a child node has only one parent node. The connection of this parent-child association relationship represents the process of evolving from one conversation state to another.

[0048] Finally, assign a session tree identity identifier to the formed session tree. Currently, each node in the formed session tree can be understood as the root node of the entire session tree and is attributed to the first user A.

[0049] The session sharing platform publishes this session tree for other users to browse.

[0050] Furthermore, some embodiments of the present application support personalized display definitions. For example, there is a virtual "Hide" button beside each node of the session tree. If the user does not want to display a certain question, they can click this virtual button, and the session sharing platform will ignore this node. As before, if the user does not want to display Checkpointer ID2, they can choose to hide Checkpointer ID2. The session sharing platform will re-establish the association relationship between Checkpointer ID3 and Checkpointer ID1, mark Checkpointer ID1 as the parent node of Checkpointer ID3, and hide Checkpointer ID2 when displaying.

[0051] Step S220: Whenever a second user has an additional interaction with the generative large model at any target node of the session tree, fork at the target node according to the session information to form a session branch; and update the session tree based on the session branch.

[0052] All users can view the publicly available session tree on the session sharing platform. Here, the users are denoted as second users. It should be noted that the first user and the second user can be the same or different. That is, the second user can be any user other than the first user, or the second user can be the first user itself. In other words, any user can ask further questions, and this application does not make any restrictions in this regard.

[0053] This application provides a "follow-up question" mechanism. Next to each node of the publicly available session tree, there is a virtual button labeled "Continue to Ask". The second user can click on this virtual button to continue the interaction with the question at that node. The second user can continue to ask questions at that node and interact with the generative large model. For example, user B Figure 3 asks a follow-up question at the node Checkpointer ID2 of the shown session tree. The second user first asks question Q4, and the generative large model gives answer A4; the second user then continues to ask question Q5, and the generative large model gives answer A5. Here, the content of the interaction between the second user B and the generative large model is denoted as the second session information. After the second user finishes the conversation, they can click on the virtual button labeled "Publish" to share their session on the session sharing platform.

[0054] The session sharing platform performs "Fork" according to the session content of the second user. Fork means that the user initiates a new interaction based on a certain historical node Checkpointer, and the platform automatically generates a new session branch (Thread) based on this node, forming a tree-like topological structure. After each Fork, the user continues the conversation in their own Thread without affecting the original historical content.

[0055] After the second user clicks "Publish", a publish request will be generated. The session sharing platform responds to the publish request of the second user and forks at the target node of the follow-up question according to the second session information generated by the interaction between the second user B and the generative large model. Here, it is assumed that the target node of the follow-up question is Checkpointer ID2, to form a session branch Thread; and update the original session tree based on the formed session branch and share the updated session tree.

[0056] The specific operation of forking is as follows: Obtain the session record data of each round of interaction of the second user, and structure the session record data into session information; construct a session branch with the target node as the starting node. In the session branch, take the session information of one round of interaction as a node. For any node, mark the adjacent previous node as its parent node and the adjacent subsequent node as its child node to construct the parent-child association relationship between nodes; assign a session branch identity identifier to the formed session branch and attribute it to the second user.

[0057] After obtaining the record data of questions Q4, Q5, answers A4, and A5, these information can be structured to form session information. Then, take the session information of each round of interaction as a node Checkpointer and assign a node identity identifier to it. For example, assign the node identity identifier Checkpointer ID4 to questions Q4 and answer A5, and assign the node identity identifier Checkpointer ID5 to questions Q5 and answer A5. The session information may include, but is not limited to, user messages, model reply messages, and context information; where the context information includes at least one of the system message input to the large model, historical messages, session rounds, and user information.

[0058] Then, construct the association relationship between each node. Specifically, take the node Checkpointer ID2 as the starting node. In the order of asking questions, mark Checkpointer ID2 as the parent node of Checkpointer ID4 (i.e., Checkpointer ID4 is the child node of Checkpointer ID2), and mark Checkpointer ID4 as the parent node of Checkpointer ID5 (i.e., Checkpointer ID5 is the child node of Checkpointer ID4). Checkpointer ID2, Checkpointer ID4, and Checkpointer ID5 form a session branch (Thread). It can be seen that the session branch refers to an independent follow-up question clue continued by a certain user based on a certain node. Each Thread represents a session branch evolving according to the user interaction trajectory starting from a certain node, with an independent life cycle and context. That is, each session branch Thread in the tree structure retains its starting node and its subsequent evolution path.

[0059] Similarly, a unique session branch identity identifier (Thread ID) will be assigned to the formed session branch. Here, the session branch formed by Checkpointer ID2, Checkpointer ID4, and Checkpointer ID5 is assigned the session branch identity identifier Thread ID1, and it is attributed to the second user B. After the second user B finishes publishing, a session tree as shown in Figure 4 is formed. As can be seen from Figure 4 , based on Figure 3 , a session branch ThreadID1 is added to the session tree, and it belongs to the second user B.

[0060] Moreover, in some embodiments of the present application, when there are multiple second users, different second users can parallelly ask questions based on the same target node or different target nodes to form different session branches.

[0061] Similarly, after the second user B finishes publishing the session tree, another second user C also asks a question at Checkpointer ID2 and poses question Q6. The generative large model gives an answer A6. After the second user C clicks to publish, after the session sharing platform obtains the record data of question Q6 and answer A6, these information can be structured to form session information. Then, the session information of each round of interaction is used as a node Checkpointer, and a node identity identifier is assigned to it. For example, here the node identity identifier assigned to question Q6 and answer A6 is denoted as Checkpointer ID6. Then, the association relationships between each node are constructed. Specifically, the node is used as the starting node of Checkpointer ID2, and Checkpointer ID2 is marked as the parent node of Checkpointer ID6 (that is, Checkpointer ID6 is the child node of Checkpointer ID2). Checkpointer ID2 and Checkpointer ID6 form a session branch (Thread). Here, the session branch formed by Checkpointer ID2 and Checkpointer ID6 is assigned the session branch identity identifier ThreadID2, and it is attributed to the second user C.

[0062] After the second user C finishes publishing, a session tree as shown in Figure 5 is formed. As can be seen from Figure 5 , based on Figure 4 , a session branch Thread ID2 is added to the session tree, and it belongs to the second user C.

[0063] Similarly, the second user can also make a follow-up question at Checkpointer ID3, thus achieving a fork at Checkpointer ID3 and forking a session branch (not shown in the figure).

[0064] And so on. Whenever a new user asks a question, this step S220 is executed. Each complete session is presented in a tree structure. The newly generated session record belongs to the personal session branch of the initiating user and is also retained in the overall tree structure.

[0065] Finally, update the session tree according to the formed session branches. On the one hand, add the new session branch to the tree structure of the entire session tree, and for the nodes of the newly added session branch, support the function of continued follow-up questions.

[0066] As mentioned above, in the Thread ID1 of the second user B, it contains the node Checkpointer ID4. If there is a new user, the second user D, he can also continue to ask questions based on Checkpointer ID4.

[0067] It can be seen that when a user hopes to perform a "continued follow-up" operation on any historical node, the system allows the user to fork based on this node and create a new Thread. This Thread represents a new session path starting from this historical node. The platform will logically use this original node as the starting point of the new session branch to form a tree-like branch structure and maintain the consistency of the original context.

[0068] In addition, in some embodiments of the present application, a user not only has the right to continue to ask questions about the nodes in the session branch initiated by himself, but also has the right to edit; other users not only have the right to continue to ask questions about the session branch initiated by this user, but also have the right to edit. The following will separately describe the editing scenarios of a user for a node.

[0069] First, the editing of a node in the session branch belonging to a user himself In some embodiments of the present application, the method further includes: in response to a modification operation of any first user or second user on any historical node belonging to himself, freezing and hiding the historical node and its subsequent nodes, and marking the historical node as the edited version and retaining the original record for subsequent version tracing; generating a replacement node according to the interaction content between the user and the generative large model based on the modification operation, and using the replacement node to update the historical node in a covering manner.

[0070] If a user wishes to modify the user question content of a historical node belonging to themselves, the platform can adopt the following strategies: The platform first freezes the original historical node content, marks it as the edited version, and retains the original record for subsequent version tracing. Moreover, both the original historical node and the content of the subsequent nodes of this historical node are hidden.

[0071] The platform regards the modified message as the new version of this node, regenerates the content of this node, deletes or hides the large model reply content on the original subsequent path, and makes the new node and its subsequent nodes appear as a new Thread branch.

[0072] The modification operation only affects the Thread owned by the current user and does not affect the branches generated by other users based on the same node, ensuring data isolation and consistency in a multi-user collaboration environment. Furthermore, this application also supports the tracing back and reply of historical versions. Specifically, in some embodiments, the method further includes: in response to any user's operation of viewing the versions of any historical node belonging to themselves, displaying multiple versions of the historical node; in response to the user's operation of tracing back to the edited version of the historical node, switching the historical node to the edited version.

[0073] The platform also supports users to view the historical version records of any node. For the edited node Checkpointer, users can switch to view the original content and the modified content, and can choose to restore to any historical version, thereby providing a complete version rollback ability. At the same time, the platform also records the meta-information of all edit operations, including the operation time, the initiating user, the original content, etc., improving the auditability and traceability of the system.

[0074] Second, the editing of a node in the session branch belonging to a user himself In some embodiments of this application, the method further includes: in response to any user's modification operation on any historical node belonging to others, generating a new node and making a fork to form a new session branch, with the new session branch starting from the historical node; attributing the new session branch to the user who performs the modification operation.

[0075] If a user modifies the historical node of another user, the original node and session branch will be retained. On this basis, a new node is generated according to the user's modification content and the reply content generated by the generative large model, and a fork is made at this historical node. Specifically, the fork starts from the new node, and the generated new session branch is attributed to the user who performs the modification operation.

[0076] It can be seen that the embodiments of the present application provide perfect version control and traceability capabilities. The conversation history is explicitly managed in the form of nodes and branches. Users can trace back to any historical node at any time, view the evolution path, and compare the development of different branches, which helps with knowledge organization, result comparison, and reuse. In this way, all the forked conversation branches Threads under a certain topic are displayed in the form of a unified tree structure aggregation for all users to browse. Each Thread evolves independently, but from an overall perspective, they jointly form a complete topic conversation tree.

[0077] Moreover, in some embodiments of the present application, multiple conversation trees can be displayed simultaneously, and different conversation trees are isolated from each other. That is, it supports parallel management of several conversation trees. Each tree represents an independent topic, and the conversation trees of each topic are isolated from each other, and are managed, browsed, and interacted with separately without affecting each other.

[0078] From the above content, it can be seen that: First, dynamic conversation update and tree-like evolution. The conversation content dynamically expands with the user's follow-up questions, presenting a clear tree structure and intuitively showing the exploration process.

[0079] Second, realizing multi-person collaboration with data isolation. Each user has an independent follow-up Thread without interference from each other, and at the same time can jointly build a rich topic conversation tree.

[0080] Third, any node can be forked for flexible exploration. It supports continuing to ask questions at any historical node, expanding multiple exploration paths, and promoting in-depth thinking from multiple angles.

[0081] Fourth, unified platform visual management. The platform centrally displays all topics and their evolution context, supports browsing, searching, and archiving by topic, and improves management and usage efficiency.

[0082] Figure 6 The flowchart of the multi-user large model interaction method based on a tree structure according to another embodiment of the present application is shown. As can be seen from Figure 6 For a user A, as a creator, his workflow is as shown in the left figure of Figure 6 He has a conversation with the generative large model, thus generating a series of conversation records. These conversation records serve as the basis for the subsequent nodes of the generated conversation tree. Virtual buttons can be provided between user A and the generative large model. When it is retrieved that user A decides to share, a conversation sharing request is generated, and the conversation sharing request is sent to the conversation sharing platform together with the conversation records.

[0083] The processing flow of the platform is as shown in Figure 6As shown in the middle figure, after the platform receives the shared content, it constructs an initial session tree. The initial session tree contains several nodes of User A and an initial session branch Thread A, and the initial session tree is publicly visible on the session sharing platform.

[0084] A user B, who is a user of the session sharing platform and is denoted as a platform user here. He can browse the session tree and choose to initiate an additional interaction at any node. When User B makes a post, the session sharing platform forks at the selected node where the follow-up question is initiated for User B to construct a new session branch Thread B. User B can have a new conversation with the generative large model within his own session branch.

[0085] The session sharing platform updates the initial session tree according to the newly generated session branch of User B, so that all users can see a dynamically growing and multi-branched shared session knowledge base.

[0086] From Figure 2 the method shown in, first, this application supports the dynamic synchronization and real-time update of session content. The shared session is no longer a static copy, but is synchronized and updated as the original session content increases or is modified, so that the shared content remains up-to-date, improving the real-time performance and consistency of information sharing. Second, this application supports the modification of any node and multi-branch exploration. Users can arbitrarily select a node in the session history to modify or append questions. The system automatically generates a new branch from this node to form a tree-like evolution structure, supporting multi-path exploration and divergent thinking, breaking through the limitations of traditional linear conversations. Third, this application improves the efficiency of multi-person collaboration and co-creation. Multiple users can independently conduct explorations based on the same shared session, each forming a branch while retaining logical associations, greatly improving the collaboration efficiency and innovation space based on the same topic. Fourth, this application enhances the scalability and manageability of session content. Based on the tree-like structure organization method, it can support subsequent diverse expansion functions, such as session summary, branch merging, and structured display, etc., significantly improving the management efficiency and utilization value of large-scale session data. In summary, this application has achieved substantial improvements compared with the prior art in terms of improving the dynamics, flexibility, collaboration, and manageability of session content, and can effectively meet the needs of users for efficient exploration and collaboration during the interaction with large models.

[0087] Figure 7 shows a schematic structural diagram of a multi-user large model interaction device based on a tree structure according to an embodiment of the present application. From Figure 7 it can be seen that the multi-user large model interaction device 700 based on a tree structure includes: The session tree initial construction unit 710 is configured to, in response to a session sharing request for a first user to interact with a generative large model, form nodes based on the interaction content of each round, and construct a parent-child association relationship between the nodes according to the order between the interactions of each round, so as to form a session tree; The session tree extension unit 720 is configured to, whenever a second user performs an additional interaction with the generative large model at any target node of the session tree, fork at the target node according to the session information to form a session branch; and update the session tree based on the session branch.

[0088] In some embodiments of the present application, the above device further includes: a backtracking unit, configured to, in response to a modification operation of any user on any historical node belonging to itself, freeze and hide the historical node and its subsequent nodes, mark the historical node as an edited version, and retain the original record for subsequent version tracing; generate a replacement node according to the interaction content between the user and the generative large model based on the modification operation, and update the historical node in a covering manner using the replacement node.

[0089] In some embodiments of the present application, in the above device, the backtracking unit is further configured to, in response to a version viewing operation of any user on any historical node belonging to itself, display multiple versions of the historical node; and in response to a backtracking operation of the user on the edited version of the historical node, switch the historical node to the edited version.

[0090] In some embodiments of the present application, in the above device, the backtracking unit is further configured to, in response to a modification operation of any user on any historical node belonging to others, generate a new node and fork to form a new session branch, with the new session branch starting from the historical node; and attribute the new session branch to the user who performs the modification operation.

[0091] In some embodiments of the present application, in the above device, the session tree initial construction unit 710 is configured to obtain the session record data of each round of interaction of the first user, and structure the session record data into session information; use the session information of one round of interaction as a node, and assign a node identity identifier to the node; for any node, mark the adjacent previous node as its parent node, and mark the adjacent subsequent node as its child node, so as to construct a parent-child association relationship between the nodes; assign a session tree identity identifier to the formed session tree and attribute it to the first user.

[0092] In some embodiments of the present application, in the above-mentioned device, the conversation tree expansion unit 720 is configured to obtain the conversation record data of each round of interaction of the second user, and structure the conversation record data into conversation information; construct a conversation branch starting from the target node, in the conversation branch, take the conversation information of one round of interaction as a node, and for any node, mark the adjacent previous node as its parent node and the adjacent subsequent node as its child node to construct the parent-child association relationship between the nodes; assign an identity identifier to the formed conversation branch and attribute it to the second user.

[0093] In some embodiments of the present application, in the above-mentioned device, the conversation information includes: user messages, model reply messages, and context information; wherein, the context information includes at least one of: system messages input to the large model, historical messages, conversation rounds, and user information.

[0094] In some embodiments of the present application, in the above-mentioned device, the conversation tree initial construction unit 710 is further configured to simultaneously display multiple conversation trees, and different conversation trees are isolated from each other.

[0095] In some embodiments of the present application, in the above-mentioned device, the conversation tree expansion unit 720 is configured to when there are multiple second users, different second users can parallelly ask questions based on the same target node or different target nodes to form different conversation branches.

[0096] In some embodiments of the present application, in the above-mentioned device, the first user and the second user can be the same or different.

[0097] For the specific limitations of the multi-user large model interaction device based on the tree structure, reference can be made to the limitations of the multi-user large model interaction method based on the tree structure in the above text, which will not be elaborated here. Each module in the above-mentioned multi-user large model interaction device based on the tree structure can be implemented in whole or in part by software, hardware, and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

[0098] In one embodiment, a computer device is provided, and the internal structure diagram of the computer device can be as Figure 8As shown in the figure. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, it realizes the functions or steps of a multi-user large model interaction method based on a tree structure.

[0099] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, it realizes the steps of the multi-user large model interaction method based on a tree structure in the present application.

[0100] It should be noted that for the functions or steps that the above computer-readable storage medium or computer device can achieve, reference can be made to the relevant descriptions in the foregoing method embodiments. To avoid repetition, they will not be described in detail here.

[0101] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or an external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0102] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0103] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A multi-user large model interaction method based on a tree structure, characterized in that, The method includes: In response to a session sharing request for a first user to interact with a generative large model, forming nodes based on the interaction content of each round, and constructing a parent-child association relationship between the nodes according to the order between each round of interactions to form a session tree; Whenever a second user has an additional interaction with the generative large model at any target node of the session tree, bifurcating at the target node according to the session information to form a session branch; and updating the session tree based on the session branch.

2. The method according to claim 1, wherein The method further includes: In response to a modification operation by any user on any historical node belonging to themselves, freezing and hiding the historical node and its subsequent nodes, and marking the historical node as an edited version and retaining the original record for subsequent version tracing; Generating a replacement node according to the interaction content between the user and the generative large model based on the modification operation, and using the replacement node to update the historical node in a covering manner.

3. The method according to claim 2, characterized in that The method further includes: In response to a version viewing operation by any user on any historical node belonging to themselves, displaying multiple versions of the historical node; In response to a backtracking operation by the user on the edited version of the historical node, switching the historical node to the edited version.

4. The method according to any one of claims 1 to 3, characterized in that, The session information includes: user messages, large model reply messages, and context information; Wherein, the context information includes at least one of: system messages input to the large model, historical messages, session rounds, and user information.

5. The method according to claim 1, characterized in that, The method further includes: In response to a modification operation by any user on any historical node belonging to others, generating a new node and bifurcating to form a new session branch, with the historical node as the starting node of the new session branch; Assigning the new session branch to the user who performed the modification operation.

6. The method according to claim 1, characterized in that, The forming of nodes based on the interaction content of each round, and the constructing of a parent-child association relationship between the nodes according to the order between each round of interactions to form a session tree specifically includes: Obtaining the session record data of each round of interaction of the first user, and structuring the session record data into session information; Regarding the session information of one round of interaction as a node, and assigning a node identity identifier to the node; For any node, marking the adjacent previous node as its parent node, and marking the adjacent subsequent node as its child node to construct a parent-child association relationship between the nodes; Assigning a session tree identity identifier to the formed session tree and attributing it to the first user.

7. The method according to claim 1, wherein The bifurcating at the target node according to the session information to form a session branch specifically includes: Obtaining the session record data of each round of interaction of the second user, and structuring the session record data into session information; Constructing a session branch with the target node as the starting node. In the session branch, regarding the session information of one round of interaction as a node, and for any node, marking the adjacent previous node as its parent node, and marking the adjacent subsequent node as its child node to construct a parent-child association relationship between the nodes; Assign a session branch identity to the formed session branch and attribute it to the second user.

8. The method according to claim 1, characterized in that, The method further includes: simultaneously displaying multiple session trees, and isolating different session trees from each other.

9. The method according to claim 1, wherein When there are multiple second users, different second users can perform additional interactions in parallel based on the same target node or different target nodes to form different session branches.

10. The method according to claim 1, characterized in that, The first user and the second user can be the same or different.

11. A multi-user large model interaction device based on a tree structure, characterized in that, The apparatus includes: A session tree initial construction unit, configured to respond to a session sharing request for a first user to interact with a generative large model, form nodes according to the interaction content of each round, and construct a parent-child association relationship between the nodes according to the order between the interactions of each round to form a session tree; A session tree extension unit, configured to whenever a second user performs an additional interaction with the generative large model at any target node of the session tree, fork at the target node according to the session information to form a session branch; and update the session tree based on the session branch.

12. A session sharing platform, characterized in that, The session sharing platform deploys the tree structure-based multi-user large model interaction apparatus according to claim 11; The tree structure-based multi-user large model interaction apparatus accesses a generative large model through a large model interaction interface to call the generative large model through the large model interaction interface to provide large model interaction services for users; The tree structure-based multi-user large model interaction apparatus is also communicatively connected to a database to perform storage and reading operations of session data through the database.

13. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, the steps of the tree structure-based multi-user large model interaction method according to any one of claims 1 to 10 are implemented.

14. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, the steps of the tree structure-based multi-user large model interaction method according to any one of claims 1 to 10 are implemented.

Images