Method, apparatus and device for function call of large model based on permission awareness, and medium

Abstract

The application discloses a large model function calling method and device based on permission awareness, equipment and medium, and relates to the technical field of function calling.The method comprises the following steps: registering a plurality of candidate calling functions; establishing a role function mapping relationship table according to the candidate calling functions; obtaining a target role corresponding to a user, and obtaining an initial to-be-called function set according to the role function mapping relationship table and the target role; obtaining a separate authorized function corresponding to the user, and obtaining a target to-be-called function set according to the separate authorized function and the initial to-be-called function set; and generating a function calling instruction according to a preset user intention and the target to-be-called function set.In this way, the user permission is bound to the large model function calling, the function automatic calling is realized according to the user demand, the function that the user has the permission to call and cannot call is avoided, and therefore the large model function calling efficiency and data confidentiality are improved, and the user experience is improved.

Description

Technical Field

[0001] This invention relates to the field of function call technology, and in particular to a method, apparatus, device and medium for large-scale function calls based on permission awareness. Background Technology

[0002] With the development of artificial intelligence technology, large-scale language models are widely used in scenarios such as intelligent customer service, office assistants, and business automation. However, when existing AI assistants are integrated with enterprise application systems, they may access system functions that users do not have permission to access, leading to data leaks or unauthorized operations. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to overcome the shortcomings of the prior art and provide a method, apparatus, device and medium for calling large model functions based on permission awareness, for binding user authorization permissions with calling functions, and further performing intelligent decision-making and calling, so as to realize a secure, compliant and personalized automated service for calling large model functions.

[0004] This invention provides the following technical solution:

[0005] Firstly, this invention proposes a permission-aware large-model function call method, including:

[0006] Register multiple candidate call functions; establish a role-function mapping table based on each candidate call function;

[0007] Obtain the target role corresponding to the user, and obtain an initial set of functions to be called based on the role function mapping table and the target role;

[0008] Obtain the individual authorization function corresponding to the user, and obtain the target set of functions to be called based on the individual authorization function and the initial set of functions to be called;

[0009] Function call instructions are generated based on the preset user intent and the target set of functions to be called.

[0010] In one embodiment, obtaining the target set of functions to be called based on the individual authorization function and the initial set of functions to be called includes:

[0011] Determine whether the individual authorized function exists in the initial set of functions to be called;

[0012] If so, the initial set of functions to be called is deduplicated according to the individual authorization function to obtain the target set of functions to be called;

[0013] If not, the individual authorized function and the initial set of functions to be called are merged to obtain the target set of functions to be called.

[0014] In one embodiment, generating function call instructions based on a preset user intent and the target set of functions to be called includes:

[0015] Generate candidate function invocation strategies based on the preset user intent and the target set of functions to be invoked;

[0016] The target function is determined based on the candidate function calling strategy, and the function calling instruction is generated based on the target function.

[0017] In one embodiment, generating a candidate function invocation strategy based on the preset user intent and the target set of functions to be invoked includes:

[0018] Determine whether the number of functions in the target set of functions to be called is greater than a preset threshold;

[0019] If so, then according to the preset user intent, a set of candidate functions to be called is selected from the target set of functions to be called, and the set of candidate functions to be called is used as the candidate function calling strategy;

[0020] If not, the target set of functions to be called will be used as the candidate function calling strategy.

[0021] In one embodiment, the step of filtering the set of candidate functions to be called from the target set of functions to be called according to the preset user intent includes:

[0022] Calculate first vector data based on the preset user intent, and calculate multiple second vector data based on the target set of functions to be called;

[0023] The similarity between each second vector data and the first vector data is calculated to obtain the candidate similarity corresponding to each second vector data.

[0024] Based on the similarity values, multiple target similarities are determined from the candidate similarities. The set of candidate functions to be called is generated according to the target function to be called corresponding to the second vector data of each target similarity.

[0025] In one embodiment, registering multiple candidate call functions includes:

[0026] Build a unified function registry;

[0027] In the unified function registration center, each candidate call function is registered according to the preset function identifier, preset function name, preset function description information, preset permission tag, and the application to which the preset function belongs.

[0028] In one embodiment, establishing a role function mapping table based on each of the candidate calling functions includes:

[0029] Identify multiple preset roles and at least one corresponding candidate function call;

[0030] Establish the role function mapping table based on each preset role and its corresponding at least one candidate calling function.

[0031] Secondly, this invention proposes a permission-aware large-scale model function call device, comprising:

[0032] Establish a module for registering multiple candidate call functions; and create a role-function mapping table based on each candidate call function.

[0033] The acquisition module is used to acquire the target role corresponding to the user, and to acquire an initial set of functions to be called based on the role function mapping table and the target role;

[0034] The determination module is used to obtain the individual authorization function corresponding to the user, and to obtain the target set of functions to be called based on the individual authorization function and the initial set of functions to be called;

[0035] The generation module is used to generate function call instructions based on the preset user intent and the target set of functions to be called.

[0036] Thirdly, the present invention proposes a computer device including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the permission-aware large model function call method as described in the first aspect.

[0037] Fourthly, a computer-readable storage medium storing a computer program that, when executed by a processor, implements the permission-aware large model function call method as described in the first aspect.

[0038] This invention discloses a method, apparatus, device, and medium for invoking large model functions based on permission awareness. The method involves registering multiple candidate functions; establishing a role-function mapping table based on each candidate function; obtaining the target role corresponding to the user; obtaining an initial set of functions to be invoked based on the role-function mapping table and the target role; obtaining a unique authorization function corresponding to the user; obtaining a target set of functions to be invoked based on the unique authorization function and the initial set of functions to be invoked; and generating a function invocation instruction based on a preset user intent and the target set of functions to be invoked. This binds user permissions to large model function invocation, enabling automatic function invocation based on user needs. It avoids the system failing to identify functions the user has permission to invoke, and invoking functions the user does not have permission to invoke, thereby improving the efficiency of large model function invocation, data confidentiality, and user experience. Attached Figure Description

[0039] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope of protection of the present invention. In the various drawings, similar components are numbered similarly.

[0040] Figure 1 This embodiment shows a flowchart of the permission-aware large model function call method proposed in this embodiment;

[0041] Figure 2 This illustration shows another flowchart of the permission-aware large model function call method proposed in this embodiment;

[0042] Figure 3 This illustration shows another flowchart of the permission-aware large model function call method proposed in this embodiment;

[0043] Figure 4 This illustration shows another flowchart of the permission-aware large model function call method proposed in this embodiment;

[0044] Figure 5 This illustration shows another flowchart of the permission-aware large model function call method proposed in this embodiment;

[0045] Figure 6 A schematic diagram of the structure of the permission-aware large model function call device proposed in this embodiment is shown.

[0046] Explanation of reference numerals in the attached diagram:

[0047] 600 - Permission-aware large model function call device; 601 - Establish module; 602 - Obtain module; 603 - Determine module; 604 - Generate module. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0049] The components of the embodiments of the invention described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0050] In the following, the terms “comprising,” “having,” and their cognates, which may be used in various embodiments of the invention, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as excluding, firstly, the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more features, numbers, steps, operations, elements, components, or combinations thereof.

[0051] Furthermore, the terms "first," "second," and "third" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0052] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of the invention pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of the invention.

[0053] Example 1

[0054] This disclosure provides a permission-aware method for calling large model functions, which binds user authorization permissions to the called functions and further performs intelligent decision-making and calling to achieve a secure, compliant, and personalized automated service for calling large model functions.

[0055] Please see Figure 1 The permission-aware large model function call method includes steps S101 to S104, and each step is described in detail below.

[0056] Step S101: Register multiple candidate calling functions; establish a role function mapping table based on each candidate calling function.

[0057] In this embodiment, in the application scenario of a large language model (LLM), multiple candidate call functions are registered, and role permissions are associated based on all candidate call functions to establish a role-function mapping table. The role-function mapping table includes each role and its corresponding callable candidate functions.

[0058] Step S102: Obtain the target role corresponding to the user, and obtain the initial set of functions to be called based on the role function mapping table and the target role.

[0059] In this embodiment, all target roles corresponding to the user are obtained, and all target roles are matched according to the role function mapping table to obtain candidate call functions corresponding to each target role, so as to form an initial set of callable functions corresponding to the user role.

[0060] Step S103: Obtain the individual authorization function corresponding to the user, and obtain the target set of functions to be called based on the individual authorization function and the initial set of functions to be called.

[0061] In this embodiment, the individual authorized function corresponding to the user is obtained, and the individual authorized function and the initial set of functions to be called are integrated to obtain the target set of functions to be called. The target set of functions to be called contains all callable functions associated with the user, which can prevent the system from being unaware of the user's actual calling permissions. Among them, the individual authorized function is a specific function permission granted to the user alone, exceeding the default permissions of the role.

[0062] Step S104: Generate function call instructions based on the preset user intent and the target set of functions to be called.

[0063] In this embodiment, a target set of functions to be called is matched with a preset user intent to generate a function call instruction that matches the user intent. The function call instruction includes the optimal function corresponding to the preset user intent in the target set of functions to be called. The call instruction is generated based on the calling permissions of the optimal function for the user or user role to achieve correct function invocation. When the user intent involves system functions that the user does not have permission to access, even if a corresponding function can be matched from the target set of functions to be called, it is necessary to determine whether the user has permission to call the function. This avoids calling system functions that the user does not have permission to access, which could lead to data leakage or illegal operations. Correspondingly, binding user calling permissions to functions can also avoid the inability to perceive the user's actual permissions, which could lead to the generated call request being rejected by the system and affecting the user experience.

[0064] It should be noted that the preset user intent can be natural language input by the user about their needs into the system, or intent keywords extracted by the system based on natural language.

[0065] In one specific embodiment, step S101 includes: constructing a unified function registration center; and registering each of the candidate call functions in the unified function registration center according to a preset function identifier, a preset function name, a preset function description, a preset permission tag, and the application to which the preset function belongs.

[0066] In this embodiment, a unified function registry is established within the application system to centrally manage all callable functions or methods provided by the system. Each candidate function registration includes the following information: a preset function ID (unique identifier within the system); a preset function name (the name used in the program); a preset description (function description, input / output parameters, and a calling example); a preset permission tag (a tag used for permission system identification); and a preset application to which the function belongs (used to distinguish different application systems). Here, the application system refers to a low-code platform or a contract management system.

[0067] In one specific embodiment, step S101 includes: determining a plurality of preset roles and at least one corresponding candidate calling function; and establishing a role-function mapping table based on each preset role and at least one corresponding candidate calling function.

[0068] In this embodiment, multiple system preset roles are defined, such as "financial specialist", "department manager" and "system administrator"; each preset role is assigned at least one candidate function that it can call; and a role function mapping table is established by integrating each preset role and its corresponding at least one candidate function.

[0069] Please see Figure 2 In one specific embodiment, step S103 includes steps S1031 to S1033, and each step is described in detail below.

[0070] Step S1031: Determine whether the individual authorized function exists in the initial set of functions to be called.

[0071] In this embodiment, each function in the initial set of functions to be called is traversed, and each row of data is converted into hashmap data, with the key being the function identifier and the value being the function object, in order to determine whether there is a single authorized function in the initial set of functions to be called.

[0072] Step S1032: If yes, then the initial set of functions to be called is deduplicated according to the individual authorization function to obtain the target set of functions to be called.

[0073] In this embodiment, if there is a single authorized function in the initial set of functions to be called, the single authorized function in the initial set of functions to be called is deduplicated, and the deduplicated initial set of functions to be called and the single authorized function are integrated to obtain the target set of functions to be called.

[0074] Step S1033: If not, merge the individual authorized function and the initial set of functions to be called to obtain the target set of functions to be called.

[0075] In this embodiment, if there is no individually authorized function in the initial set of functions to be called, the initial set of functions to be called and the individually authorized function are directly integrated to obtain the target set of functions to be called.

[0076] Please see Figure 3 Step S104 includes steps S1041 to S1042, and each step is described in detail below.

[0077] Step S1041: Generate a candidate function calling strategy based on the preset user intent and the target set of functions to be called.

[0078] In this embodiment, the target set of functions to be called is initially screened according to the preset user intent in order to generate a candidate function calling strategy that conforms to the preset user intent.

[0079] Step S1042: Determine the target function according to the candidate function calling strategy, and generate the function calling instruction according to the target function.

[0080] In this embodiment, the optimal target function is determined by combining the business context, function description and calling parameters according to the candidate function calling strategy, and a function calling instruction is generated according to the target function to complete the function call.

[0081] Please see Figure 4 Step S1041 includes steps S401 to S403, and each step is described in detail below.

[0082] Step S401: Determine whether the number of functions in the target set of functions to be called is greater than a preset threshold.

[0083] In this embodiment, it is determined whether the number of functions in the target set of functions to be called is greater than a preset threshold, thereby filtering out situations where the number of functions is too large and the data processing pressure is too high.

[0084] Step S402: If yes, then according to the preset user intent, filter the set of candidate functions to be called from the set of target functions to be called, and use the set of candidate functions to be called as the candidate function calling strategy.

[0085] In this embodiment, if the number of functions in the target set of functions to be called is greater than a preset threshold, a set of candidate functions to be called with a smaller number of functions is selected from the target set of functions to be called using a preset user intent, and the set of candidate functions to be called is used as a candidate function calling strategy, thereby reducing the data processing pressure.

[0086] Step S403: If not, use the target set of functions to be called as the candidate function calling strategy.

[0087] In this embodiment, if the number of functions in the target set of functions to be called is less than or equal to a preset threshold, the target set of functions to be called is directly used as the candidate function calling strategy.

[0088] Please see Figure 5 Step S402 includes steps S4021 to S4023, and each step is described in detail below.

[0089] Step S4021: Calculate first vector data according to the preset user intent, and calculate multiple second vector data according to the target set of functions to be called.

[0090] In this embodiment, the preset user intent is converted into a vector to obtain the first vector data; each function in the target set of functions to be called is converted into a corresponding second vector data. This allows the correlation between the user intent and the function call to be obtained through the vector data.

[0091] Step S4022: Calculate the similarity between each of the second vector data and the first vector data to obtain the candidate similarity corresponding to each of the second vector data.

[0092] In this embodiment, the similarity between each second vector data and the first vector data is calculated to obtain the candidate similarity corresponding to each second vector data.

[0093] Step S4023: Based on the similarity value, determine multiple target similarities from the candidate similarities, and generate the candidate function set according to the target function to be called corresponding to the second vector data of each target similarity.

[0094] In this embodiment, the candidate similarities corresponding to each second vector data are sorted from high to low according to the similarity value. The second vector data corresponding to the N candidate similarities with the largest similarity values ​​are selected, and the target functions to be called corresponding to these N second vector data are used as a set of candidate functions to be called.

[0095] This embodiment proposes a permission-aware large-model function invocation method that registers multiple candidate invocation functions; establishes a role-function mapping table based on each candidate function; obtains the target role corresponding to the user; and obtains an initial set of functions to be invoked based on the role-function mapping table and the target role; obtains the individual authorization function corresponding to the user; and obtains a target set of functions to be invoked based on the individual authorization function and the initial set of functions to be invoked; finally, it generates a function invocation instruction based on the preset user intent and the target set of functions to be invoked. In this way, user permissions are bound to large-model function invocation, enabling automatic function invocation based on user needs. This avoids the system failing to identify functions that the user has permission to invoke, or invoking functions that the user does not have permission to invoke, thereby improving the efficiency of large-model function invocation, data confidentiality, and user experience.

[0096] Example 2

[0097] Furthermore, this disclosure provides a permission-aware large model function call apparatus 600, please refer to [link to relevant documentation]. Figure 6 ,include:

[0098] Module 601 is established to register multiple candidate call functions; a role-function mapping table is established based on each candidate call function.

[0099] The acquisition module 602 is used to acquire the target role corresponding to the user, and to acquire an initial set of functions to be called based on the role function mapping table and the target role;

[0100] The determination module 603 is used to obtain the individual authorization function corresponding to the user, and to obtain the target set of functions to be called based on the individual authorization function and the initial set of functions to be called;

[0101] The generation module 604 is used to generate function call instructions based on the preset user intent and the target set of functions to be called.

[0102] In one embodiment, the determining module 603 is further configured to determine whether the separate authorized function exists in the initial set of functions to be called; if so, the initial set of functions to be called is deduplicated according to the separate authorized function to obtain the target set of functions to be called; if not, the separate authorized function and the initial set of functions to be called are merged to obtain the target set of functions to be called.

[0103] In one embodiment, the generation module 604 is further configured to generate a candidate function calling strategy based on the preset user intent and the target set of functions to be called; determine a target function based on the candidate function calling strategy; and generate the function calling instruction based on the target function.

[0104] In one embodiment, the generation module 604 is further configured to determine whether the number of functions in the target set of functions to be called is greater than a preset threshold; if so, then according to the preset user intent, a set of candidate functions to be called is selected from the target set of functions to be called, and the set of candidate functions to be called is used as the candidate function calling strategy; if not, the target set of functions to be called is used as the candidate function calling strategy.

[0105] In one embodiment, the generation module 604 is further configured to: calculate first vector data according to the preset user intent; calculate multiple second vector data according to the target function set to be called; calculate the similarity between each second vector data and the first vector data to obtain candidate similarity corresponding to each second vector data; determine multiple target similarities from each candidate similarity based on the similarity value; and generate the candidate function set to be called according to the target function corresponding to the second vector data corresponding to each target similarity.

[0106] In one embodiment, the establishment module 601 is further configured to construct a unified function registration center; in the unified function registration center, each of the candidate call functions is registered according to a preset function identifier, a preset function name, a preset function description information, a preset permission tag, and the application to which the preset function belongs.

[0107] In one embodiment, the establishment module 601 is further configured to determine multiple preset roles and their corresponding at least one candidate calling function; and to establish the role function mapping relationship table according to each preset role and its corresponding at least one candidate calling function.

[0108] The apparatus provided in this disclosure can execute the steps of the permission-aware large model function call method provided in Embodiment 1. To avoid repetition, the steps will not be repeated.

[0109] This embodiment proposes a permission-aware large-model function invocation device that registers multiple candidate invocation functions; establishes a role-function mapping table based on each candidate function; obtains the target role corresponding to the user; and obtains an initial set of functions to be invoked based on the role-function mapping table and the target role; obtains a unique authorization function corresponding to the user; and obtains a target set of functions to be invoked based on the unique authorization function and the initial set of functions to be invoked; and generates a function invocation instruction based on a preset user intent and the target set of functions to be invoked. In this way, user permissions are bound to large-model function invocation, enabling automatic function invocation based on user needs. This avoids the system failing to identify functions that the user has permission to invoke, or invoking functions that the user does not have permission to invoke, thereby improving the efficiency of large-model function invocation, data confidentiality, and user experience.

[0110] Example 3

[0111] Furthermore, this disclosure provides a computer device including a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the permission-aware large model function call method described in Embodiment 1.

[0112] The device provided in this embodiment can execute the steps of the permission-aware large model function call method provided in Embodiment 1. To avoid repetition, these steps will not be repeated.

[0113] Example 4

[0114] This disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the permission-aware large model function call method described in Embodiment 1.

[0115] In this embodiment, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0116] The computer-readable storage medium provided in this embodiment can implement the permission-aware large model function call method provided in Embodiment 1. To avoid repetition, it will not be described again here.

[0117] In all examples shown and described herein, any specific values ​​should be interpreted as merely exemplary and not as limitations; therefore, other examples of exemplary embodiments may have different values.

[0118] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0119] The above-described embodiments are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for calling large-scale model functions based on permission awareness, characterized in that, include: Register multiple candidate functions for invocation; Establish a role function mapping table based on each of the candidate call functions; Obtain the target role corresponding to the user, and obtain an initial set of functions to be called based on the role function mapping table and the target role; Obtain the individual authorization function corresponding to the user, and obtain the target set of functions to be called based on the individual authorization function and the initial set of functions to be called; Generate function call instructions based on the preset user intent and the target set of functions to be called; The registration of multiple candidate call functions includes: Build a unified function registry; In the unified function registration center, each candidate call function is registered according to the preset function identifier, preset function name, preset function description information, preset permission tag, and the application to which the preset function belongs. The step of obtaining the target set of functions to be called based on the individual authorization function and the initial set of functions to be called includes: Determine whether the individual authorized function exists in the initial set of functions to be called; If so, the initial set of functions to be called is deduplicated according to the individual authorization function to obtain the target set of functions to be called; If not, the individual authorized function and the initial set of functions to be called are merged to obtain the target set of functions to be called.

2. The method for calling large model functions based on permission awareness according to claim 1, characterized in that, The step of generating function call instructions based on preset user intent and the target set of functions to be called includes: Generate candidate function invocation strategies based on the preset user intent and the target set of functions to be invoked; The target function is determined based on the candidate function calling strategy, and the function calling instruction is generated based on the target function.

3. The method for calling large model functions based on permission awareness according to claim 2, characterized in that, The step of generating a candidate function invocation strategy based on the preset user intent and the target set of functions to be invoked includes: Determine whether the number of functions in the target set of functions to be called is greater than a preset threshold; If so, then according to the preset user intent, a set of candidate functions to be called is selected from the target set of functions to be called, and the set of candidate functions to be called is used as the candidate function calling strategy; If not, the target set of functions to be called will be used as the candidate function calling strategy.

4. The method for calling large model functions based on permission awareness according to claim 3, characterized in that, The step of filtering the set of candidate functions to be called from the target set of functions to be called according to the preset user intent includes: Calculate first vector data based on the preset user intent, and calculate multiple second vector data based on the target set of functions to be called; The similarity between each second vector data and the first vector data is calculated to obtain the candidate similarity corresponding to each second vector data. Based on the similarity values, multiple target similarities are determined from the candidate similarities. The set of candidate functions to be called is generated according to the target function to be called corresponding to the second vector data of each target similarity.

5. The method for calling large model functions based on permission awareness according to claim 1, characterized in that, The step of establishing a role function mapping table based on each of the candidate calling functions includes: Identify multiple preset roles and at least one corresponding candidate function call; Establish the role function mapping table based on each preset role and its corresponding at least one candidate calling function.

6. A permission-aware large-scale function call device, characterized in that, include: Create a module to register multiple candidate functions for invocation; Establish a role function mapping table based on each of the candidate call functions; The acquisition module is used to acquire the target role corresponding to the user, and to acquire an initial set of functions to be called based on the role function mapping table and the target role; The determination module is used to obtain the individual authorization function corresponding to the user, and to obtain the target set of functions to be called based on the individual authorization function and the initial set of functions to be called; The generation module is used to generate function call instructions based on the preset user intent and the target set of functions to be called; The establishment module is also used to construct a unified function registration center; in the unified function registration center, each candidate call function is registered according to a preset function identifier, a preset function name, a preset function description information, a preset permission tag, and the application to which the preset function belongs; The determining module is further configured to determine whether the separate authorized function exists in the initial set of functions to be called; if so, the initial set of functions to be called is deduplicated according to the separate authorized function to obtain the target set of functions to be called; if not, the separate authorized function and the initial set of functions to be called are merged to obtain the target set of functions to be called.

7. A computer device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, implements the permission-aware large model function call method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the permission-aware large model function call method as described in any one of claims 1 to 5.

Images