Interaction method and interaction device based on large model and vehicle

Abstract

The invention provides an interaction method and an interaction device based on a large model and a vehicle, the method relates to the field of vehicles, the method is applied to a vehicle enterprise platform, the vehicle enterprise platform carries the large model, the method comprises the following steps: obtaining input content of a user, and analyzing the input content to obtain N tasks, N being an integer greater than 0; determining a target task chain from a plurality of task chains associated with the large model according to the N tasks; and processing the N tasks based on the target task chain to generate target output content corresponding to the input content. According to the method, the intelligent degree of large model interaction in a car business scene can be improved.

Description

Technical Field

[0001] The present application relates to the field of vehicles, and more specifically, to a large model-based interaction method, interaction device, and vehicle in the field of vehicles. Background Art

[0002] By applying big models to automotive business scenarios, the intelligence level and business innovation capabilities of automotive businesses can be improved.

[0003] However, existing large-model interaction methods fail to meet the complex business logic requirements of the automotive sector and are unable to link underlying model capabilities with business processes, resulting in a low level of intelligence for large-model interactions in automotive business scenarios. Therefore, improving the intelligence of large-model interactions in automotive business scenarios is a technical challenge that needs to be addressed. Summary of the Invention

[0004] The present application provides an interaction method, interaction device and vehicle based on a large model, which can improve the intelligence level of large model interaction in automobile business scenarios.

[0005] In a first aspect, a large model-based interaction method is provided. The method is applied to an automotive enterprise platform equipped with a large model. The method includes:

[0006] Get the user's input content, parse the input content to obtain N tasks, where N is an integer greater than 0;

[0007] Based on N tasks, determine the target task chain from multiple task chains associated with the large model;

[0008] Process N tasks based on the target task chain to generate target output content corresponding to the input content.

[0009] In an embodiment of the present application, the user input content is parsed, and the target task chain associated with the large model is matched based on the N tasks obtained from the parsing; and the N tasks are processed according to the target task chain to generate the target output content. Compared with the prior art, when processing complex tasks (tasks with a large number of processing flows composed of multiple tasks), the user needs to input prompt words multiple times so that the model can process them in sequence according to the processing flow of the complex tasks. This solution can determine the target task chain from multiple pre-configured task chains based on the user's input content, and automatically execute the N tasks corresponding to the input content according to the target task chain, so that the large model can cope with scenarios that require multiple rounds of interaction and multi-task collaboration, ensuring that the large model can handle complex business in the vehicle field, thereby improving the intelligence of model interaction.

[0010] In conjunction with the first aspect, in certain implementations of the first aspect, the large model is associated with M agents, each of the M agents has a different processing capability, and the target task chain is associated with Q target agents among the M agents, where M ≥ Q > 1, and M and Q are integers;

[0011] Determining a target task chain from a plurality of task chains associated with the large model according to the N tasks, including: determining Q target intelligent agents from M intelligent agents associated with the large model according to the N tasks; and determining the task chains corresponding to the Q target intelligent agents as the target task chain;

[0012] Processing the N tasks based on the target task chain includes: processing the N tasks based on Q target agents associated with the target task chain.

[0013] In an embodiment of the present application, a target task chain is associated with Q agents out of M agents; the M agents are agents associated with a large model, each with different processing capabilities; therefore, each agent can handle different tasks. The process of determining a target task chain based on N tasks is to determine Q target agents based on the N tasks, and then determine the task chains corresponding to the Q agents as the target task chain, ensuring that the N tasks can be processed sequentially by each target agent in the target task chain, thereby obtaining the target output content corresponding to the input content.

[0014] In combination with the first aspect and the above implementations, in certain implementations of the first aspect, determining a target task chain from multiple task chains associated with the large model based on the N tasks includes:

[0015] Perform keyword matching or semantic matching on N tasks to determine the similarity between the N tasks and each task chain in multiple task chains;

[0016] The task chain with the highest similarity among the multiple task chains is determined as the target task chain.

[0017] In an embodiment of the present application, keyword matching or semantic matching is performed on the N tasks corresponding to the input content, the similarity between the N tasks and the pre-configured task chains is determined, and the task chain with the highest similarity (i.e., matching degree) is determined as the target task chain. Since the agent in the task chain with the highest similarity to the N tasks is able to better process the N tasks corresponding to the input content, the task chain with the highest similarity is determined as the target task chain to ensure that the target task chain matches the input content.

[0018] In combination with the first aspect and the above implementations, in some implementations of the first aspect, the method further includes:

[0019] Embed interfaces for external databases and external query tools that call large models in multiple task chains;

[0020] Process N tasks based on the target task chain, including:

[0021] If the N tasks include a task of calling an external database and / or an external query tool, the external database and / or the external query tool is called through the interface embedded in the target task chain.

[0022] In an embodiment of the present application, an interface for calling an external database and an external query tool is embedded in the task chain; if N tasks include a task that calls an external database and / or an external query tool, the external database and the external query tool can be called directly according to the interface, thereby improving the processing capability of the task chain (i.e., the processing capability of the large model), so that the large model can combine the external database and the external query tool to more efficiently generate the target output content corresponding to the input content.

[0023] In combination with the first aspect and the above implementations, in some implementations of the first aspect, the method further includes:

[0024] Record the execution status of each task node in the target task chain, including the executed state, unexecuted state and failed state;

[0025] If the execution status of the first task node in the target task chain is an execution failure state, the task associated with the first task node is re-executed.

[0026] In an embodiment of the present application, each of the N tasks is a task node. If a first task node in the target task chain fails to execute, the task associated with the first task node is re-executed. Compared to re-executing the entire target task chain upon execution failure, this solution can re-execute the task at the failed task node, avoiding the re-execution of tasks that have not failed, improving the execution efficiency of the target task chain, and thus improving the efficiency of generating the target output content.

[0027] In combination with the first aspect and the above implementations, in certain implementations of the first aspect, when processing the N tasks based on the target task chain, the method further includes:

[0028] The execution order of the target task chain, the time consumption of each task node, and the execution status of each task node are displayed in the display interface, where the display interface is the client display interface of the automobile enterprise platform.

[0029] In an embodiment of the present application, the execution order of the target task chain, the time consumption of each task node and the execution status of each task node are displayed in the display interface of the client, ensuring that the user can quickly understand the generation process of the target output content in the display interface of the automobile enterprise platform, so as to facilitate business personnel and R&D teams to monitor the processing process corresponding to the target task chain, thereby improving the controllability of large models in the automobile enterprise platform.

[0030] In combination with the first aspect and the above implementations, in some implementations of the first aspect, the method further includes:

[0031] The different processing capabilities of the large model are encapsulated to obtain M agents, where each agent in the M agents has a different processing capability;

[0032] Determine at least two agents among the M agents corresponding to different services;

[0033] At least two agents are sorted to obtain multiple task chains associated with a large model.

[0034] In the embodiments of this application, the different processing capabilities of a large model are encapsulated to obtain M agents. Because each of the M agents has different processing capabilities, each agent can handle different tasks. At least two agents are sorted according to the business of the automotive platform (for example, vehicle condition inquiry, fault consultation, service appointment, etc.) to obtain multiple task chains associated with the large model; this ensures that the task chains can handle the business of the automotive platform.

[0035] In combination with the first aspect and the above implementations, in certain implementations of the first aspect, sorting at least two agents to obtain multiple task chains associated with a large model includes:

[0036] Determining input content and output content of each of at least two intelligent agents;

[0037] Based on the data format of the input and output content of each agent, at least two agents are sorted to obtain multiple task chains associated with the large model;

[0038] Among them, among at least two intelligent agents, the data format of the output data of the previous intelligent agent among the two adjacent intelligent agents is the same as the data format of the input content of the next intelligent agent.

[0039] In the embodiments of this application, the output data format of the preceding agent in two adjacent agents is the same as the input data format of the succeeding agent. Following this principle, at least two agents are sorted according to the data formats of their input and output content, resulting in multiple task chains associated with a large model. Different task chains correspond to different businesses within the automotive company platform. This ensures that task chains can be pre-arranged based on automotive company businesses, so that upon detecting user input, the corresponding target task chain can be directly matched based on the user input.

[0040] In a second aspect, a large model-based interactive device is provided. The device is applied to a vehicle enterprise platform equipped with a large model. The device includes:

[0041] The acquisition module is used to obtain the user's input content and parse the input content to obtain N tasks, where N is an integer greater than 0;

[0042] The processing module is used to determine the target task chain from multiple task chains associated with the large model based on N tasks; process the N tasks based on the target task chain to generate target output content corresponding to the input content.

[0043] In combination with the second aspect, in certain implementations of the second aspect, the processing module is specifically used to: determine Q target agents from the M agents associated with the large model based on N tasks; determine the task chain corresponding to the Q target agents as the target task chain; and process the N tasks based on the Q target agents associated with the target task chain.

[0044] In combination with the second aspect and the above-mentioned implementation methods, in certain implementation methods of the second aspect, the processing module is specifically used to: perform keyword matching or semantic matching on N tasks, determine the similarity between the N tasks and each task chain in multiple task chains; and determine the task chain with the highest similarity among the multiple task chains as the target task chain.

[0045] In combination with the second aspect and the above-mentioned implementation methods, in some implementation methods of the second aspect, the processing module is also used to: embed interfaces for calling external databases and external query tools of large models in multiple task chains; if N tasks include tasks for calling external databases and / or external query tools, call the external database and / or external query tool through the interface embedded in the target task chain.

[0046] In combination with the second aspect and the above-mentioned implementation methods, in some implementation methods of the second aspect, the processing module is also used to: record the execution status of each task node in the target task chain, the execution status includes the executed status, the unexecuted status and the execution failure status; if the execution status of the first intelligent agent in the target task chain is the execution failure status, re-execute the task associated with the first intelligent agent.

[0047] In combination with the second aspect and the above-mentioned implementation methods, in some implementation methods of the second aspect, the processing module is also used to: display the execution order of the target task chain, the time consumption of each task node, and the execution status of each task node in the display interface.

[0048] In combination with the second aspect and the above-mentioned implementation methods, in some implementation methods of the second aspect, the processing module is also used to: encapsulate the different processing capabilities of the large model to obtain M intelligent agents, wherein each of the M intelligent agents has different processing capabilities; determine at least two intelligent agents among the M intelligent agents corresponding to different businesses; sort at least two intelligent agents to obtain multiple task chains associated with the large model.

[0049] In combination with the second aspect and the above implementations, in some implementations of the second aspect, the processing module is specifically configured to: determine input content and output content of each of the at least two agents;

[0050] Based on the data format of the input and output content of each agent, at least two agents are orchestrated to obtain multiple task chains associated with the large model;

[0051] Among them, among at least two intelligent agents, the data format of the output data of the previous intelligent agent among the two adjacent intelligent agents is the same as the data format of the input content of the next intelligent agent.

[0052] In a third aspect, a vehicle is provided, comprising a memory and a processor, wherein the memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, so that the vehicle executes the method in the above-mentioned first aspect or any possible implementation of the first aspect.

[0053] In a fourth aspect, a computer program product is provided, comprising: a computer program code, which, when executed on a computer, enables the computer to execute the method in the first aspect or any possible implementation of the first aspect.

[0054] In a fifth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed, it implements the method in the above-mentioned first aspect or any possible implementation of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a schematic flow chart of an interactive method based on a large model provided in an embodiment of the present application;

[0056] Figure 2 This is a schematic diagram of the process of encapsulating and calling an intelligent agent provided in an embodiment of the present application;

[0057] Figure 3 is a schematic flow chart of another large model-based interaction method provided in an embodiment of the present application;

[0058] Figure 4 This is a schematic diagram of the structure of an interactive device based on a large model provided in an embodiment of the present application;

[0059] Figure 5 It is a structural schematic diagram of a vehicle provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] The following will clearly and thoroughly describe the technical solutions in this application in conjunction with the accompanying drawings. In the description of the embodiments of this application, unless otherwise specified, " / " means or, for example, A / B can mean A or B: "and / or" in the text is only a description of the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more than two.

[0061] In the following, the terms "first" and "second" are used for descriptive purposes only and should not be understood to imply or suggest relative importance or implicitly indicate the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features.

[0062] Because the interaction method of the large models in the automobile enterprise platform in the existing technology can only handle a single response to a single question (for example, when a user asks "What should I do if the fuel consumption is too high", the model directly gives the answer), the large model cannot cope with scenarios that require multiple rounds of interaction and multi-task collaboration. Therefore, it is difficult to complete more complex systematic tasks in the automotive field (including but not limited to: vehicle maintenance recommendations, owner interaction analysis and maintenance process planning, etc.).

[0063] In addition, for more complex systematic tasks, the model needs to switch between different data sources and tasks, and a single-round interaction is difficult to achieve the connection and coherence between multiple tasks. For example, for vehicle maintenance recommendations, it is necessary to first collect vehicle fault codes and vehicle age information, and then conduct step-by-step analysis based on the collected information. A single question and answer (i.e., a single-round interaction) is difficult to complete information and logical deduction. For maintenance process planning, it is necessary to integrate variable information such as user time, parts inventory, and the location of maintenance stores. A single-round interaction is difficult to cover all the required variables. For owner interaction analysis, after the owner asks a question, it is necessary to sequentially implement tasks such as intent recognition, vehicle status retrieval, comprehensive analysis, and response generation. Obviously, the single-round interaction of the model in the existing technology cannot achieve automatic connection between multiple tasks.

[0064] In view of this, the present application provides an interaction method, interaction device and vehicle based on a large model. The method can parse the user's input content in the automobile enterprise platform, match the target task chain associated with the large model according to the N tasks obtained from the analysis, and process the N tasks according to the target task chain to generate target output content, so that the large model can cope with scenarios that require multiple rounds of interaction and multi-task collaboration, ensuring that the large model can handle complex business in the vehicle field, thereby improving the intelligence of model interaction.

[0065] The following combination Figure 1 The method flow chart in the figure provides a detailed description of the large model-based interactive method provided in this application.

[0066] Figure 1 This is a schematic flowchart of an interactive method based on a large model provided in an embodiment of the present application.

[0067] For example, Figure 1 The method 100 shown can be executed by the automobile enterprise platform, or can be executed by the server side of the automobile enterprise platform. The automobile enterprise platform represents a comprehensive technology platform built by the automobile enterprise based on artificial intelligence big model technology, which has multi-dimensional intelligent processing capabilities and can serve all aspects of the entire life cycle of the automobile.

[0068] It should be noted that the car company platform includes a client and a server. The client refers to the device or software that the user directly interacts with, which is used to receive user input and display information (for the car company platform, the client can be an application in the car system); the server refers to the computer or program that provides services to the client, which is used for data storage, logical processing and resource scheduling. The big model is installed on the server side of the car company platform (also known as the cloud server of the car company platform). The user enters content in the client application, and the client packages the user's input content into a request and sends it to the server through the Application Programming Interface (API) so that the big model on the server side processes the request. The server side sends the content generated after the big model is processed to the client application for display.

[0069] like Figure 1 As shown, the large model-based interaction method 100 includes S110 to S130 , which will be described in detail below.

[0070] S110: Obtain user input content, and parse the input content to obtain N tasks.

[0071] Exemplarily, the user's input content is the content entered by the user in the application of the client of the car enterprise platform (for example, the application of the car system). The client packages the user's input content into a request and sends it to the server through the application program interface so that the server can obtain the user's input content. By parsing and disassembling the user's input content, N tasks corresponding to the input content are obtained, where N is an integer greater than 0. Among them, the process of parsing the input content to obtain N tasks can be understood as a process of identifying the intent of the input content and disassembling the intent into multiple step-by-step execution actions.

[0072] In one implementation, the user's input content is preprocessed; the user's intention is matched based on a preset dictionary and grammatical rules; and the matched intention is decomposed into N executable actions (i.e., N tasks), wherein the preprocessing includes data cleaning and feature extraction of the user's input content.

[0073] Exemplarily, when performing data cleaning, if the user input content is voice, the voice is converted into normalized text. If the user input content is text, the text corresponding to the user input content is normalized (for example, normalization includes removing punctuation, correcting typos and replacing synonyms). For example, the user input content is "Help me find a nearby charging station", and the content is cleaned and converted into "charging station". When performing feature extraction, the user input content is extracted for keywords through a deep learning model, such as keywords such as "navigation" and "temperature". Alternatively, a keyword thesaurus for the vehicle field can be pre-set to determine whether the user input content matches the keywords in the thesaurus. If the keyword is matched, the matched keyword is determined as the keyword of the user input content.

[0074] For example, user intentions are matched based on preset dictionaries and grammatical rules. For example, "navigate to address A" is identified as "route planning intention"; "mobile phone battery is too low" is identified as "in-vehicle charging port query intention" or "turn on in-vehicle wireless charging intention."

[0075] For example, if the user input is "My vehicle has fault A. How do I get it repaired?", the input is converted into the intent "Maintenance process planning." This intent is then broken down into multiple tasks, including scheduling a repair, detecting vehicle fault codes, querying nearby repair shops, selecting a time, and confirming an appointment. By executing these tasks sequentially, the output corresponding to the user input is obtained.

[0076] For example, the automobile enterprise platform (i.e. the automobile enterprise's big model capability platform) refers to a comprehensive technology platform built by the automobile enterprise based on artificial intelligence big model technology, which has multi-dimensional intelligent processing capabilities and can serve all aspects of the entire life cycle of the automobile.

[0077] Application scenarios for the automotive platform include, but are not limited to, smart cockpit interaction, autonomous driving assistance, and connected car services. Specifically, smart cockpit interaction refers to the use of natural language processing to enable voice assistants, scenario-based dialogue, and understanding of user intent, thereby accurately executing vehicle control commands and providing personalized services to users. Autonomous driving assistance combines computer vision, reinforcement learning, and other technologies to enhance perception and planning capabilities in complex road conditions. Connected car services provide services such as in-car entertainment, intelligent navigation, intelligent question-and-answer (Q&A), remote vehicle control, and after-sales forecasting based on user profiling and scenario analysis.

[0078] S120 , determining a target task chain from multiple task chains associated with the large model according to the N tasks.

[0079] The multiple task chains associated with the large model are pre-designed for different business scenarios on the automotive platform. This facilitates the subsequent detection of user input and allows the target task chain to be directly identified from these pre-designed task chains based on the user input. The following details the process of identifying multiple task chains associated with the large model.

[0080] Specifically, the different processing capabilities of the large model are encapsulated to obtain M intelligent agents, wherein each of the M intelligent agents has different processing capabilities; at least two intelligent agents among the M intelligent agents corresponding to different businesses are determined; and at least two intelligent agents are sorted to obtain multiple task chains associated with the large model.

[0081] For example, the different processing capabilities of the large model are encapsulated to obtain reusable task units (i.e., agents); encapsulation refers to hiding the complex implementation details within the system and exposing only a unified interface that complies with industry or protocol standards to facilitate external calls, thereby improving modularity and enhancing maintainability. The automotive platform provides a standard definition for each agent. The definition of the agent includes input parameters, output parameters, the agent call model, and data such as the agent's prompt template (Prompt template) and permission control. For example, the agents associated with the large model include a "car model recommendation agent", which is defined as receiving basic user information and car purchase intentions, calling the large model to generate recommendation text, and returning structured suggestions with confidence. By encapsulating the large model, each agent can be reused, scheduled, and managed as a separate module.

[0082] It should be noted that the prompt word template is used to automatically convert user input into appropriate prompt words, automatically fill in variables, and then input them into the large model. The user does not need to manually configure the prompt words, thereby improving the quality of the prompt words and ensuring the output effect of the large model.

[0083] For example, the different processing capabilities of the large model are encapsulated to obtain M agents, each with different processing capabilities. The different businesses within the automotive platform are pre-split, each business being divided into at least two executable actions (or at least two tasks). Based on the at least two executable actions, at least two agents corresponding to the business are determined. The at least two agents are sorted to obtain task chains corresponding to the business. The multiple task chains corresponding to the pre-designed multiple businesses are determined as the multiple task chains associated with the large model.

[0084] For example, if the user input is "There is an abnormal noise when braking, how to deal with it", the corresponding prompt words can be: "Fault clues: User description: {There is an abnormal noise when braking}; Data characteristics: {Brake pressure fluctuation frequency 20Hz, brake pad thickness 3mm}; Output maintenance plan and indicate whether it needs immediate processing."

[0085] In this embodiment, the different processing capabilities of a large model are encapsulated to obtain M agents. Because each of the M agents has different processing capabilities, each agent can handle different tasks. At least two agents are sorted according to the business of the automotive platform (for example, vehicle condition inquiry, fault consultation, service appointment, etc.) to obtain multiple task chains associated with the large model, ensuring that the business of the automotive platform can be processed through the task chains.

[0086] Optionally, the input content and output content of each of the at least two intelligent agents are determined; based on the data format of the input content and output content of each intelligent agent, the at least two intelligent agents are sorted to obtain multiple task chains associated with the large model; wherein, among the at least two intelligent agents, the data format of the output data of the previous intelligent agent in the two adjacent intelligent agents is the same as the data format of the input content of the latter intelligent agent.

[0087] For example, between two adjacent agents in a task chain, the output of the preceding agent serves as the input of the succeeding agent; therefore, the data format of the preceding agent's output needs to be the same as the input of the succeeding agent to ensure smooth data transmission throughout the chain. For example, a task chain includes an intent recognition agent and a vehicle status query agent. The intent recognition agent outputs the user's intent to query the vehicle's status, and the vehicle status query agent requires input of the user's intent and the vehicle's identification code. The user's intent comes from the intent recognition agent's output, and the vehicle's identification code comes from the contextual input information.

[0088] It's understood that the output data format of the preceding agent in two adjacent agents is the same as the input data format of the succeeding agent. Following this principle, at least two agents are sorted according to the data formats of their input and output content, resulting in multiple task chains associated with the large model. Different task chains correspond to different businesses within the automotive platform, ensuring that task chains are pre-arranged based on automotive business needs. This allows the system to directly match the user input to the corresponding target task chain upon detection.

[0089] For example, if we break down the businesses on an automotive company's platform, we find the following agents: Agent 1 {input content A; output content C}; Agent 2 {input content B; output content A}; and Agent 3 {input content C; output content D}. Based on the input and output data of each agent, the task chain corresponding to this business is: {Agent 2, Agent 1, Agent 3}. The process flow of the business chain is as follows: Agent 2 receives content B as input and outputs content A; Agent 2 inputs content A into Agent 1, which then outputs content C based on content A; Agent 1 inputs content C into Agent 3, which then outputs content D based on content C. Content D is the final output.

[0090] It should be noted that the above are examples of intelligent agents obtained by decomposing the business, as well as the input and output of each intelligent agent. They are only used to describe the process of sorting task chains based on the input and output of the intelligent agent. This application does not limit the specific task chain arrangement process.

[0091] Optionally, after pre-segmenting different services within the vehicle enterprise platform to generate at least two tasks, the corresponding agents are sorted according to the order in which the tasks are to be executed, resulting in task chains corresponding to the different services. For example, in the vehicle condition inquiry service, the multiple tasks generated include intent recognition, vehicle status retrieval, and response generation. Based on the order in which the tasks are split, the task chain corresponding to the vehicle condition inquiry service is determined to be {intent recognition agent, vehicle status retrieval agent, response generation agent}.

[0092] It should be noted that the above embodiment illustrates the process of determining multiple task chains associated with a large model. After determining multiple task chains, based on the tasks corresponding to the user input content, the task chain corresponding to the input content can be directly determined from the multiple task chains (i.e., the task chain can be reused). The process of determining a target task chain from multiple task chains is further described below.

[0093] In one implementation, keyword matching or semantic matching is performed on N tasks to determine similarities between the N tasks and each task chain in multiple task chains; and the task chain with the highest similarity among the multiple task chains is determined as the target task chain.

[0094] Exemplarily, the keyword matching process includes: establishing a correspondence between preset keywords and intelligent agents, and storing the preset keywords corresponding to the intelligent agents; extracting keywords for N tasks, and comparing the keywords with the preset keywords corresponding to the intelligent agents in the task chain to determine the similarity between the keywords of the N tasks and the preset keywords of the intelligent agents in the task chain; and determining the task chain with the highest similarity as the target task chain.

[0095] Exemplarily, the semantic matching process includes: converting the task corresponding to the input content into a semantic vector; calculating the similarity between the semantic vector and the semantic vector of the intelligent agent in each task chain, which is the similarity between N tasks and each task chain in multiple task chains; and determining the task chain with the highest similarity as the target task chain.

[0096] In an embodiment of the present application, keyword matching or semantic matching is performed on the N tasks corresponding to the input content, the similarity between the N tasks and the pre-configured task chains is determined, and the task chain with the highest similarity (i.e., matching degree) is determined as the target task chain. Since the agent in the task chain with the highest similarity to the N tasks is able to better process the N tasks corresponding to the input content, the task chain with the highest similarity is determined as the target task chain to ensure that the target task chain matches the input content.

[0097] In one implementation, the large model is associated with M agents, each of the M agents has different processing capabilities, and the target task chain is associated with Q target agents among the M agents, where M≥Q>1, and M and Q are integers;

[0098] According to N tasks, a target task chain is determined from multiple task chains associated with the large model, including: according to the N tasks, Q target intelligent agents are determined from M intelligent agents associated with the large model; the task chains corresponding to the Q target intelligent agents are determined as the target task chain; the N tasks are processed based on the target task chain, including: the N tasks are processed based on the Q target intelligent agents associated with the target task chain.

[0099] It is understood that the correspondence between tasks and target agents can be a one-to-one relationship, a one-to-many relationship, or a many-to-one relationship, depending on the processing requirements of the tasks. Specifically, if one task among N tasks requires multiple target agents to process, the correspondence between tasks and target agents is one-to-many. If multiple tasks among N tasks require one agent to process, the correspondence between tasks and target agents is many-to-one. If different tasks among N tasks require different target agents to process, the correspondence between tasks and target agents is one-to-one.

[0100] For example, N tasks include converting fuel consumption data into a line graph, converting fault statistics into a pie chart, and converting driving behavior data into a heat map. The M agents associated with the large model include a data visualization agent (used to convert different data into graphs / charts for data visualization). This data visualization agent performs the above three tasks, so the correspondence between the tasks and the target agents is many-to-one. The N tasks include a driving plan recommendation task, which requires identifying the user's driving style and generating a driving method based on the vehicle parameters and driving style. The driving plan recommendation task requires a behavior analysis agent (used to identify the user's driving style) and a driving plan generation agent (used to generate a driving plan based on the vehicle parameters and driving style), so the correspondence between the task and the target agent is one-to-many.

[0101] In an embodiment of the present application, the target task chain is associated with Q agents out of M agents; the M agents are agents associated with the large model, and each agent has different processing capabilities; therefore, each agent can handle different tasks. The process of determining the target task chain based on N tasks is to determine Q target agents based on the N tasks, and then determine the task chains corresponding to the Q agents as the target task chain. It is ensured that the N tasks can be processed in sequence by each target agent in the target task chain, so that the target output content corresponding to the input content can be obtained.

[0102] S130 , processing the N tasks based on the target task chain to generate target output content corresponding to the input content.

[0103] Exemplarily, each agent in the target task chain calls the large model in sequence to process N tasks and obtain output content corresponding to the input content.

[0104] It can be understood that the intelligent agent is an encapsulation of the capabilities of the large model, which is used to organize, guide and manage the use of the model. The intelligent agent cannot replace the large model, but rather calls the large model to implement the reasoning process and obtain the output content.

[0105] In one implementation, the above method also includes: embedding interfaces for calling external databases and external query tools of large models in multiple task chains; if N tasks include tasks for calling external databases and / or external query tools, calling the external database and / or external query tools through the interface embedded in the target task chain.

[0106] Among them, external databases and query tools include the vehicle real-time status query interface of the Telematics Service Provider (TSP) platform, the vehicle maintenance history database, and the customer service knowledge query tool.

[0107] For example, in a vehicle business scenario, content generation may require integrating external data sources and query tools. For example, a fault query agent needs to access the vehicle's real-time status through an interface between an external database and a query tool, and then perform fault queries based on that real-time status. A maintenance planning agent needs to call upon an external data source and query tool interface to retrieve historical vehicle maintenance information from a maintenance history database, and then perform maintenance planning based on that historical information.

[0108] In an embodiment of the present application, by embedding an interface for calling an external database and an external query tool in the task chain, if N tasks include tasks that call an external database and / or an external query tool, the external database and the external query tool can be called according to the interface, thereby improving the processing capability of the task chain (that is, the processing capability of the large model), so that the large model can combine the external database and the external query tool to more efficiently generate the target output content corresponding to the input content.

[0109] In some specific embodiments, the method also includes: recording the execution status of each task node in the target task chain, the execution status including the executed status, the unexecuted status and the execution failed status; if the execution status of the first task node in the target task chain is the execution failed status, re-executing the task associated with the first task node.

[0110] For example, each of the N tasks is a task node. If a first task node in the target task chain fails, the task associated with the first task node is re-executed. Compared to re-executing the entire target task chain upon failure, this solution re-executes the task at the failed task node, avoiding re-execution of tasks that did not fail, improving the execution efficiency of the target task chain and thus improving the efficiency of generating the target output content.

[0111] For example, the target task chain is {intent recognition agent; vehicle status query agent; response generation agent}. When generating the target response, each agent in the target task chain sequentially calls the master model to execute its corresponding task. If, during task execution, the vehicle status query agent experiences an abnormal interruption in the task chain while querying the vehicle status and reports an error, after repairing the problem, execution restarts from the node where the vehicle status is queried, rather than from the intent recognition node.

[0112] Optionally, when processing N tasks based on the target task chain, the method also includes: displaying the execution order of the target task chain, the time consumed by each task node, and the execution status of each task node in a display interface, and the display interface is the client display interface of the car company platform (that is, the display interface of the client application).

[0113] For example, when each agent in a target task chain processes N tasks, the display interface shows the execution order of the target task chain, the time taken for each task node, and the execution status of each task node. This ensures that users can quickly understand the generation process of the target output content on the display interface, making it easier for business personnel and R&D teams to monitor the processing of the target task chain, thereby improving the controllability of large models within the automotive platform.

[0114] In the above embodiment, the user's input content in the car company platform is parsed, and the target task chain associated with the large model is matched according to the N tasks obtained from the parsing, and the N tasks are processed according to the target task chain to generate the target output content. Compared with the existing technology, when processing complex tasks (complex tasks refer to tasks with a large processing flow composed of multiple tasks), the user needs to enter prompt words multiple times so that the model can process them in sequence according to the processing flow of the complex tasks. This solution can determine the target task chain from multiple pre-configured task chains based on the user's input content, and automatically execute the N tasks corresponding to the input content according to the target task chain, so that the large model can cope with scenarios that require multiple rounds of interaction and multi-task collaboration, ensuring that the large model can handle complex business in the vehicle field, thereby improving the intelligence of model interaction.

[0115] Figure 2 The method 200 shown is a schematic diagram of the method for encapsulating and calling an intelligent agent provided in an embodiment of the present application.

[0116] For example, Figure 2 As shown, the capabilities of the large model are disassembled, and S201 is executed to encapsulate the intelligent agent; by encapsulating the intelligent agent into reusable task units, each intelligent agent (such as recommendation, recognition, response, etc.) provides a calling interface to the outside world as an independent task unit, which facilitates the subsequent orchestration, calling and management of the intelligent agent.

[0117] Exemplarily, the agent is encapsulated and S202 to S205 are executed, including defining the input and output of the agent, configuring the prompt word template of the agent, configuring the ability of the agent to call the model, and orchestrating through the agent process orchestration engine. Among them, defining the input and output of the agent includes defining the input data of the agent, the data format of the input data, the output parameters and the data format of the output parameters, etc. By configuring the prompt word template of the agent, when the input content is detected, the input content is converted into a standard prompt word according to the prompt word template, thereby ensuring the output result. The ability to configure the agent to call the model can be understood as defining the task that the agent can achieve, and the task is achieved by the ability of the agent to call the large model. The agent is orchestrated through the agent process orchestration engine. When orchestrating the agent, S206 is executed to orchestrate the agent and S207 is set to set the process control logic, wherein orchestrating the agent means allowing developers to connect multiple agents in order to form a task chain (such as Figure 2 As shown, Agent 1 is connected to Agent 2, and Agent 2 is connected to Agent n. Setting the process control logic includes sequential logic (i.e., executing the agent's task nodes in sequence), conditional logic (i.e., executing the next task node when a specific condition is met), or loop logic (i.e., looping through the task nodes in the task chain when a condition is met).

[0118] Exemplarily, the encapsulation and arrangement of the intelligent agent is achieved through S201 to S207, and the calling of the intelligent agent is achieved through S208 to S209, including calling external tools through interfaces and generating output content through the task execution engine.

[0119] For example, the interfaces for external tools embedded in the task chain include: a real-time vehicle status query interface, an auto insurance damage assessment system, a vehicle maintenance history database, and a customer service knowledge query tool. Each agent can configure the calling method, authentication mechanism, and parameter mapping relationship to ensure seamless collaboration between the platform and external tools during execution. The results returned by the external tools can be used as input for the agents in subsequent task nodes, forming a closed-loop chain. For example, a "troubleshooting agent" can read vehicle fault codes through the external tool interface and automatically trigger an agent to generate repair suggestions based on the fault codes.

[0120] For example, a unified task execution engine is built into the car company platform. The task execution engine is used to schedule, run and manage task flows composed of multiple intelligent agents. The task execution engine supports full-process monitoring of the task chain, recording the execution time, request response logs, context variable changes and error status of each task node. If a task is abnormally interrupted during execution, the task execution engine can automatically restore the interrupted state and continue running. It supports execution timeout control and custom alarm mechanisms (for example, if the execution time of a task node reaches the preset time, a custom alarm mechanism will be used to prompt) to ensure the stable operation of the task chain.

[0121] Optionally, after the task execution engine generates output content, the output content is displayed. The output of all intelligent agents on the platform supports structured returns, such as JSON format or key-value fields, etc., which is convenient for client applications (for example, applications in the car computer, work order system programs) to perform secondary processing and display. The platform's system also integrates task process visualization tools to display the task execution path, time consumption of each node, call status and output preview in real time, which is convenient for business personnel and R&D teams to conduct joint debugging and monitoring. At the same time, it supports exporting task processes as graphical reports to help teams analyze effects, tune models and reproduce problems, thereby improving the engineering feasibility and transparent controllability of large models in car companies.

[0122] For example, combined Figure 2 Take the flowchart in the example below: For example, after a car owner enters a question in the car's computer application, the client application sends the question to the server, initiating the entire orchestration process. First, the intent recognition agent determines that the car owner is inquiring about a fault. Then, the preliminary fault classification agent infers that the problem may be a powertrain anomaly. The process engine orchestrates the following steps: retrieving vehicle operating data (such as engine status and accelerometer data), searching the vehicle's recent maintenance history (to determine whether any maintenance is overdue or abnormal), and invoking the fault knowledge base search plug-in to query for cases similar to the acceleration failure. After aggregating the retrieved data, the comprehensive analysis agent determines the possible cause. The diagnostic suggestion generation agent (or response generation agent) outputs a structured recommendation, such as "Preliminary diagnosis: severe carbon deposits on the throttle valve. Cleaning is recommended at a repair shop." The task execution engine ensures monitoring and logging of the entire process. Finally, the structured results are output to the application and presented to the car owner.

[0123] Figure 3 This is a schematic flowchart of another large model-based interaction method provided in an embodiment of the present application. Figure 3 The method 300 shown may be executed on a vehicle manufacturer's platform, or may be executed by a server of the vehicle manufacturer's platform.

[0124] like Figure 3As shown, the large model-based interaction method 300 includes S301 to S311 , which will be described in detail below.

[0125] S301, encapsulate the different processing capabilities of the large model to obtain M intelligent agents.

[0126] For example, encapsulation refers to hiding the complex implementation details within a system and exposing only a unified, industry- or protocol-compliant interface for easy external access. Each of the M agents corresponds to a different processing capability of the larger model, enabling different processing tasks.

[0127] Optionally, when encapsulating the large model to obtain an agent, a standard definition is provided for each agent, and the definition of the agent includes input parameters, output parameters, an agent calling model, and data such as the agent's prompt word template and permission control.

[0128] S302: Determine at least two agents corresponding to different services among the M agents based on different services.

[0129] For example, according to different business scenarios of the automobile enterprise platform, each business is split into at least two executable actions (or at least two tasks), and at least two intelligent agents corresponding to different businesses among the M intelligent agents are obtained based on the at least two executable actions.

[0130] S303: Sort at least two agents to obtain multiple task chains associated with the large model.

[0131] For example, different businesses within the automotive company platform are pre-split to obtain at least two tasks. The corresponding agents are then sorted according to the order in which the tasks are to be executed, resulting in task chains corresponding to the different businesses. Alternatively, at least two agents are sorted based on the data format of their input and output content, resulting in multiple task chains associated with a large model; wherein, for at least two adjacent agents, the output data format of the preceding agent is the same as the input data format of the succeeding agent.

[0132] S304: Check whether the user input is detected; if so, execute S305.

[0133] Exemplarily, it is determined whether the user input is detected; if the input is detected, the input is parsed to obtain N tasks. Optionally, if the user input is not detected, subsequent steps are not performed.

[0134] S305: Parse the input content to obtain N tasks.

[0135] Exemplarily, when the user's input content is detected, the input content is subjected to intent recognition, and the recognized intent is broken down into multiple tasks to be executed step by step.

[0136] Exemplarily, the user's input content is preprocessed; the user's intention is matched based on a preset dictionary and grammatical rules, and the matched intention is decomposed into N executable actions (i.e., decomposed into N tasks), wherein the preprocessing includes data cleaning and feature extraction of the user's input content.

[0137] Optionally, the implementation of S304 to S305 can refer to Figure 1 The relevant description of S110 is not repeated here.

[0138] S306 , performing keyword matching or semantic matching on the N tasks to obtain similarities between the N tasks and multiple task chains.

[0139] Exemplarily, a correspondence between preset keywords and intelligent agents is established, and the preset keywords corresponding to the intelligent agents are stored; keywords are extracted for N tasks, and the keywords are compared with the preset keywords corresponding to the intelligent agents in the task chain to determine the similarity between the keywords of the N tasks and the preset keywords of the intelligent agents in the task chain.

[0140] Exemplarily, the task corresponding to the input content is converted into a semantic vector; the similarity between the semantic vector and the semantic vector of the agent in each task chain is calculated, and the similarity is the similarity between N tasks and each task chain in the multiple task chains.

[0141] S307: Determine the task chain with the highest similarity among the multiple task chains as the target task chain.

[0142] Exemplarily, the task chain with the highest similarity among multiple task chains is determined as the target task chain, including: determining the task chain with the highest similarity between the keywords in the task chain and the keywords of N tasks as the target task chain; or determining the task chain with the highest similarity between the semantic vectors in the task chain and the semantic vectors of N tasks as the target task chain.

[0143] S308, calling the large model in sequence according to the execution order of each agent in the target task chain.

[0144] Exemplarily, the large model is called sequentially according to the execution order of each agent in the target task chain; wherein, the output content of the previous agent of two adjacent agents in the task chain is used as the input content of the next agent, thereby realizing the binding of context variables during the processing of the target task chain to ensure the transmission of context information flow.

[0145] S309: Is there a task node in the target task chain that has failed to execute? If so, execute S311; if not, execute S310.

[0146] Exemplarily, determine whether there is a task node that has failed to execute in the target task chain; if there is a task node that has failed to execute, re-execute the task corresponding to the task node; if there is no task node that has failed to execute, output the target output content obtained after executing the target task chain.

[0147] S310: Output target output content obtained after executing the target task chain.

[0148] For example, if there is no task node that failed to execute, the target output content obtained after executing the target task chain is output; wherein, the target output content supports structured return, such as JSON format or key-value fields, etc., to facilitate upper-level applications (for example, applications in the vehicle computer, work order systems) to process and display.

[0149] S311, re-execute the task corresponding to the task node.

[0150] For example, if there is a task node that has failed to execute, the task corresponding to the task node is re-executed to avoid repeated execution of tasks that have not failed, thereby improving the execution efficiency of the target task chain and thus improving the generation efficiency of the target output content.

[0151] In an embodiment of the present application, the capabilities of the large model are encapsulated to obtain standardized intelligent agent units, so that the capabilities of the large model can be visualized, reusable, and iterative in the form of intelligent agents, thereby lowering the threshold for business access. When a new business needs to be accessed, the encapsulated intelligent agent unit can be directly called for orchestration, and there is no need to retrain the model for the new business. By orchestrating the process of the intelligent agent, the model is no longer just a static "question and answer interface", but a collection of intelligent agents that can complete multi-step logical reasoning, tool calls, and information integration, and more efficiently realize the model interaction in multi-step car business scenarios such as owner interaction, after-sales service, and risk judgment. In addition, the collaborative processing of the intelligent agent execution process and the external system database or tools (for example, TSP platform, vehicle-mounted data platform) improves the interaction efficiency and interaction quality of the model.

[0152] Combined with the above Figures 1 to 3 The interactive method based on the large model provided by the embodiment of the present application is described in detail; Figure 4 and Figure 5 The device embodiments of the present application are described in detail. It should be understood that the devices in the embodiments of the present application can execute the various methods of the aforementioned embodiments of the present application, that is, the specific working processes of the following various products can refer to the corresponding processes in the aforementioned method embodiments.

[0153] Figure 4This is a structural diagram of an interactive device based on a large model provided in an embodiment of the present application.

[0154] For example, Figure 4 As shown, the interactive device 400 based on the large model includes:

[0155] The acquisition module 410 is used to acquire input content from the user and parse the input content to obtain N tasks, where N is an integer greater than 0;

[0156] The processing module 420 is configured to determine a target task chain from a plurality of task chains associated with the large model according to the N tasks; and process the N tasks based on the target task chain to generate target output content corresponding to the input content.

[0157] Optionally, as an embodiment, the processing module 420 is specifically used to: determine Q target agents from the M agents associated with the large model based on N tasks; determine the task chain corresponding to the Q target agents as the target task chain; and process the N tasks based on the Q target agents associated with the target task chain.

[0158] Optionally, as an embodiment, the processing module 420 is specifically used to: perform keyword matching or semantic matching on N tasks, determine the similarity between the N tasks and each task chain in multiple task chains; and determine the task chain with the highest similarity among the multiple task chains as the target task chain.

[0159] Optionally, as an embodiment, the processing module 420 is also used to: embed interfaces for calling external databases and external query tools of large models in multiple task chains; if N tasks include tasks for calling external databases and / or external query tools, call the external database and / or external query tool through the interface embedded in the target task chain.

[0160] Optionally, as an embodiment, the processing module 420 is also used to: record the execution status of each task node in the target task chain, the execution status including the executed status, the unexecuted status and the failed execution status; if the execution status of the first agent in the target task chain is the failed execution status, re-execute the task associated with the first agent.

[0161] Optionally, as an embodiment, the processing module 420 is also used to: display the execution order of the target task chain, the time consumption of each task node, and the execution status of each task node in the display interface, wherein the display interface is the client display interface of the automobile enterprise platform.

[0162] Optionally, as an embodiment, the processing module 420 is also used to: encapsulate the different processing capabilities of the large model to obtain M intelligent agents, wherein each of the M intelligent agents has different processing capabilities; determine at least two intelligent agents among the M intelligent agents corresponding to different businesses; sort at least two intelligent agents to obtain multiple task chains associated with the large model.

[0163] Optionally, as an embodiment, the processing module 420 is specifically used to: determine the input content and output content of each of the at least two intelligent agents; based on the data format of the input content and output content of each intelligent agent, arrange the at least two intelligent agents to obtain multiple task chains associated with the large model; wherein, among the at least two intelligent agents, the data format of the output data of the previous intelligent agent in the two adjacent intelligent agents is the same as the data format of the input content of the latter intelligent agent.

[0164] It should be noted that the above-mentioned interactive device based on the large model is embodied in the form of a functional unit. The term "module" here can be implemented in the form of software and / or hardware, and is not specifically limited to this.

[0165] For example, a "module" may be a software program, a hardware circuit, or a combination of the two that implements the aforementioned functionality. The hardware circuit may include an application specific integrated circuit (ASIC), an electronic circuit, a processor (e.g., a shared processor, a dedicated processor, or a group of processors) and memory for executing one or more software or firmware programs, combined logic circuits, and / or other suitable components that support the described functionality.

[0166] Therefore, the units of each example described in the embodiments of this application can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0167] Figure 5 It is a structural schematic diagram of a vehicle provided in an embodiment of the present application.

[0168] Exemplarily, vehicle 500 includes a processor 510 , a memory 520 , and executable program code 530 .

[0169] Exemplarily, the vehicle 500 includes one or more processors 510, which can support the vehicle 500 in implementing the large model-based interaction method in the method embodiment. The processor 510 can be a general-purpose processor or a special-purpose processor. For example, the processor 510 can be a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, such as discrete gates, transistor logic devices, or discrete hardware components.

[0170] For example, the processor 510 can be used to control the vehicle 500, execute software programs, and process data of the software programs. The vehicle 500 can also include a communication unit to implement signal input (reception) and output (transmission).

[0171] Exemplarily, the vehicle 500 may include one or more memories 520 on which executable program code 530 is stored. The executable program code 530 can be executed by the processor 510 to generate instructions so that the processor 510 executes the large model-based interaction method described in the above method embodiment according to the instructions.

[0172] Optionally, data may be stored in the memory 520. Optionally, the processor 510 may read data stored in the memory 520. The data may be stored at the same storage address as the executable program code 530, or may be stored at a different storage address from the executable program code 530.

[0173] Exemplarily, the processor 510 and the memory 520 may be provided separately or integrated together, for example, integrated on a system on chip (SOC) of the terminal device.

[0174] Exemplarily, the memory 520 can be used to store relevant programs of the large model-based interaction method provided in the embodiment of the present application, and the processor 520 can be used to call the executable program code 530 stored in the memory 520 when controlling the vehicle to execute the large model-based interaction method of the embodiment of the present application; for example, obtain the user's input content, parse the input content to obtain N tasks, where N is an integer greater than 0; based on the N tasks, determine the target task chain from multiple task chains associated with the large model; process the N tasks based on the target task chain to generate target output content corresponding to the input content.

[0175] The present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the large model-based interaction method of any of the aforementioned embodiments.

[0176] Among them, computer-readable storage media may include, but are not limited to, any type of disk, including floppy disks, optical disks, Digital Video Discs (DVDs), Compact Disc Read-Only Memory (CD-ROMs), microdrives and magneto-optical disks, Read-Only Memory (ROMs), Random Access Memory (RAMs), Erasable Programmable Read-Only Memory (EPROMs), Electrically Erasable Programmable Read-Only Memory (EEPROMs), Dynamic Random Access Memory (DRAMs), Video Random Access Memory (VRAMs), flash memory devices, magnetic or optical cards, nanosystems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.

[0177] The present application also provides a computer program product. When the computer program product is run on a computer, it enables the computer to execute the above-mentioned related steps to implement an interactive method based on a large model in the above-mentioned embodiment.

[0178] In addition, the vehicle provided in the embodiments of the present application can specifically be a chip, component or module, and the vehicle may include a connected processor and memory; wherein the memory is used to store instructions, and when the vehicle is running, the processor can call and execute instructions to enable the chip to execute a large model-based interaction method in the above-mentioned embodiment.

[0179] Among them, the vehicle, computer-readable storage medium, computer program product or chip provided in this application are all used to execute the corresponding large-model-based interaction method provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding large-model-based interaction method provided above, and will not be repeated here.

[0180] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0181] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0182] The above content is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. An interactive method based on a large model, characterized in that: The method is applied to a vehicle enterprise platform equipped with a large model, and includes: Obtain user input content, and parse the input content to obtain N tasks, where N is an integer greater than 0; Determining a target task chain from a plurality of task chains associated with the large model according to the N tasks; The N tasks are processed based on the target task chain to generate target output content corresponding to the input content.

2. The method according to claim 1, characterized in that The large model is associated with M agents, each of the M agents has a different processing capability, and the target task chain is associated with Q target agents among the M agents, where M≥Q>1, and M and Q are integers; Determining a target task chain from a plurality of task chains associated with the large model according to the N tasks includes: Determine the Q target agents from the M agents associated with the large model according to the N tasks; Determine the task chains corresponding to the Q target agents as the target task chains; The processing of the N tasks based on the target task chain includes: The N tasks are processed based on the Q target agents associated with the target task chain.

3. The method according to claim 1, characterized in that Determining a target task chain from a plurality of task chains associated with the large model according to the N tasks includes: Performing keyword matching or semantic matching on the N tasks to determine similarities between the N tasks and each task chain in the plurality of task chains; The task chain with the highest similarity among the multiple task chains is determined as the target task chain.

4. The method according to claim 1, wherein Also includes: Embedding interfaces for calling external databases and external query tools of the large model in the plurality of task chains; The processing of the N tasks based on the target task chain includes: If the N tasks include a task of calling the external database and / or the external query tool, the external database and / or the external query tool is called through the interface embedded in the target task chain.

5. The method according to any one of claims 1 to 4, characterized in that The method further comprises: Recording the execution status of each task node in the target task chain, wherein the execution status includes executed status, unexecuted status and execution failure status; If the execution status of the first task node in the target task chain is the execution failure status, re-execute the task associated with the first task node.

6. The method according to any one of claims 1 to 4, characterized in that When processing the N tasks based on the target task chain, the method further includes: The execution order of the target task chain, the time consumption of each task node, and the execution status of each task node are displayed in a display interface, wherein the display interface is a client display interface of the automobile enterprise platform.

7. The method according to claim 1, characterized in that The method further comprises: Encapsulating the different processing capabilities of the large model to obtain M intelligent agents, wherein each intelligent agent in the M intelligent agents has a different processing capability; Determine at least two agents among the M agents corresponding to different services; The at least two agents are sorted to obtain a plurality of task chains associated with the large model.

8. The method according to claim 7, characterized in that The step of sorting the at least two agents to obtain a plurality of task chains associated with the large model includes: Determining input content and output content of each of the at least two agents; sorting the at least two agents based on the data formats of the input content and the output content of each agent to obtain a plurality of task chains associated with the large model; Among the at least two intelligent agents, the data format of the output data of the first intelligent agent among the two adjacent intelligent agents is the same as the data format of the input content of the second intelligent agent.

9. An interactive device based on a large model, characterized in that: The device is applied to a vehicle enterprise platform equipped with a large model, and includes: An acquisition module is used to obtain user input content and parse the input content to obtain N tasks, where N is an integer greater than 0; A processing module is used to determine a target task chain from multiple task chains associated with the large model according to the N tasks; and process the N tasks based on the target task chain to generate target output content corresponding to the input content.

10. A vehicle, characterized in that: The vehicle comprises: a memory for storing executable program code; A processor is configured to call and run the executable program code from the memory, so that the vehicle executes the method according to any one of claims 1 to 8.

Images