Large model-based skill development method, skill calling method and device

Abstract

The invention provides a skill development method, a skill calling method and a skill calling device based on a large model and a computer storage medium. The skill development method comprises the steps that in response to a skill development instruction of a user, a specified skill template is selected, and the skill template comprises preset cue words, large model operation parameters, input and output formats and / or pre-configuration information; writing element information input by a user into the skill template to generate a skill version; and combining the skill version with a large model to generate a skill service. Through the skill development method, large model skill services of the same type are quickly developed by multiplexing the skill template extracted based on the business.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a skill development method based on a large model, a skill calling device, and a computer storage medium. Background Art

[0002] Currently, many large model development platforms provide prompt word development and debugging functions, but the structure and function are relatively simple, and the prompt word service is also called in a variety of ways, such as: Bot, application, etc. The principle of its implementation is ultimately based on the arrangement of prompt word content, relying on manual input of prompt word content, and debugging of various parameters. It does not comprehensively consider the business scenarios faced by the actual implementation of the business, and still tends to focus on prompt word content development.

[0003] There are still many pain points that cannot be solved in order to realize intelligent question and answer, factor extraction, rule judgment, content writing and other business applications based on the big model prompt word service. These pain points seriously affect the development and application of the business around the big model. For example, each similar function implemented based on the big model requires customized development. Although some platforms provide skill templates, they are basically writing prompt words, debugging prompt words, and tuning prompt words. The writing and optimization of prompt words are repeated to complete the final business vision. The above problems have led to the low development efficiency of the existing big model services, which cannot meet the business needs of users. Summary of the invention

[0004] In order to solve the above technical problems, the present application proposes a skill development method, a skill calling method, a skill calling device and a computer storage medium based on a large model.

[0005] In order to solve the above technical problems, the present application proposes a skill development method based on a large model, and the skill development method includes:

[0006] In response to a skill development instruction from a user, a designated skill template is selected, wherein the skill template includes preset prompt words, large model operation parameters, input and output formats, and / or pre-configured information;

[0007] Writing the element information input by the user into the skill template to generate a skill version;

[0008] The skill version is combined with the large model to generate a skill service.

[0009] Wherein, after combining the skill version with the large model to generate a skill service, the skill development method further includes:

[0010] The skill service is bound to the system application to provide a calling interface externally through the system application.

[0011] The step of combining the skill version with the large model to generate a skill service includes:

[0012] Call several large models supported by the system and run the skill versions respectively;

[0013] Display the running results of each large model for the skill version;

[0014] In response to the user's large model selection instruction, the corresponding large model is selected and combined with the skill version to generate a skill service.

[0015] Among them, the pre-configuration information includes verification rules for input and output parameters, and / or data processing rules; wherein the data processing rules are composed of regular expressions and scripting languages, and are used to process data generated during the skill calling process.

[0016] In order to solve the above technical problems, the present application also proposes a skill calling method based on a large model, and the skill calling method includes:

[0017] The caller sends a request message to the provider based on the access token.

[0018] The provider analyzes the system application, skill service and user problem based on the request information;

[0019] The provider runs the system application, calls the skill service to perform business processing on the user problem, and returns the business processing result to the caller;

[0020] Among them, the skill service is developed through the above-mentioned skill development method.

[0021] In order to solve the above-mentioned technical problems, the present application also proposes a skill calling device, which includes a memory and a processor coupled to the memory; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the skill development method and / or skill calling method as described above.

[0022] In order to solve the above technical problems, the present application also proposes a computer storage medium, which is used to store program data. When the program data is executed by a computer, it is used to implement the above-mentioned skill development method and / or skill calling method.

[0023] Compared with the prior art, the beneficial effects of the present application are as follows: the skill calling device selects a specified skill template in response to the user's skill development instruction, wherein the skill template includes preset prompt words, large model operation parameters, input and output formats, and / or pre-configured information; the element information input by the user is written into the skill template to generate a skill version; the skill version is combined with the large model to generate a skill service. Through the above-mentioned skill development method, by reusing the skill template extracted based on the business, the same type of large model skill service can be quickly developed. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] in:

[0026] Figure 1 It is a flowchart of an embodiment of a skill development method based on a large model provided by the present application;

[0027] Figure 2 It is a schematic diagram of the overall process of the skill development method based on the big model provided in this application;

[0028] Figure 3 This is a schematic diagram of an interface of an embodiment of a skill template provided by the present application;

[0029] Figure 4 This is a schematic diagram of the interface for creating skills from factor extraction using the factor extraction template provided by this application;

[0030] Figure 5 It is a flowchart of an embodiment of a skill calling method based on a large model provided by the present application;

[0031] Figure 6 This is a schematic diagram of the overall process of the skill calling method based on the large model provided by this application;

[0032] Figure 7 It is a flowchart of an embodiment of a method for requesting and obtaining an access token provided by the present application;

[0033] Figure 8 This is a schematic diagram of the overall process of requesting and obtaining an access token provided by this application;

[0034] Fig. 9 It is a structural diagram of an embodiment of a skill calling device provided by the present application;

[0035] Fig.10 It is a structural diagram of an embodiment of a computer storage medium provided by the present application. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0037] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can, for example, be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0038] This application relates to technical fields such as large model service applications, and in particular, a method and system for how to perform API (Application Programming Interface) encapsulation of large model services to form an interface service for external use. Specifically, the method proposed in this case is to have multiple sets of skill preset information built in for users to use and choose when creating new skills. Different templates correspond to different preset prompt words, input and output parameters, large model operation parameters, etc.; generate corresponding prompt word skills through templates or custom methods, and simultaneously provide users with convenient debugging and comparison tools for debugging preview and comparison of effects between multiple models, multiple parameters, and multiple versions; at the same time, it also provides a variety of data rule processing, provides regularization and processing of user questions, large model return results, and structured extraction results, and improves the final effect of prompt word services; and finally, it will provide a unified authorization access control mechanism at the API (Application Programming Interface) entrance to meet business call needs.

[0039] In view of the current situation where there are many large model development platforms, this application proposes a method for realizing large model skill development based on API encapsulation technology from the perspective of business application use, and proposes two processes of skill development and skill calling. This method greatly improves the problems of low reusability, difficulty in solving development and debugging prompts, and inconvenient multi-model comparison in the existing technology.

[0040] Among them, this application proposes a method for realizing efficient development of large model skills by using API encapsulation technology. This method is divided into two processes: skill development and skill call: skill development, that is, providing methods and means for skill development and release; skill call, that is, providing methods and means for using skill services developed by skills.

[0041] For details about the large model skill development process, please refer to Figure 1 and Figure 2 , Figure 1 is a flow chart of an embodiment of a skill development method based on a large model provided by the present application, Figure 2 It is a schematic diagram of the overall process of the skill development method based on the big model provided in this application.

[0042] The skill development method based on the big model of the present application is applied to the skill calling device, wherein the skill calling device of the present application can be a server, a terminal device, or a system composed of a server and a terminal device. Accordingly, the various parts included in the skill calling device, such as various units, sub-units, modules, and sub-modules, can all be set in the server, can all be set in the terminal device, or can be set in the server and the terminal device respectively.

[0043] Furthermore, the above-mentioned server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules, such as software or software modules used to provide distributed servers, or it can be implemented as a single software or software module, which is not specifically limited here.

[0044] like Figure 1 As shown, the specific steps are as follows:

[0045] Step S11: In response to the user's skill development instruction, a specified skill template is selected, wherein the skill template includes preset prompt words, large model operation parameters, input and output formats, and / or pre-configuration information.

[0046] In an embodiment of the present application, the skill calling device may be a provider of skill services, such as a skill service system, etc. The skill service system is described below as a provider.

[0047] The skill service system provides a skill development function. When a user selects the skill development function in the function interface provided by the skill service system, the skill service system can display the skill development interface according to the generated skill development instructions. Among them, the skill development interface can initialize one or a category of skill templates. For details, see Figure 3 , Figure 3 It is a schematic diagram of the interface of an embodiment of a skill template provided in this application.

[0048] like Figure 3 As shown, the skill service system provides skill preset information through skill templates for users to choose. Among them, the skill template of this application is a template and pre-configured information with preset prompt words, large model operation parameters, input and output formats, etc., which is summarized and refined from the prompt word service that has been implemented and is actually effective in the past. The pre-configured information includes but is not limited to the verification rules of input and output parameters, data processing rules and other data.

[0049] Among them, the data processing rules of the present application can be composed of regular expressions and scripting languages, which can enable text information to be replaced and processed into corresponding data content when it meets the rules. Data processing rules can be applied to user question preprocessing, large model parsing result post-processing, and structured data value processing. The specific processing method will be described in the skill call process. Among them, the skill service system of the present application provides a configuration page for quickly adding and reusing data processing rules. Similarly, the verification rules for input and output parameters can also be configured through the configuration page provided by the skill service system, which will not be repeated here.

[0050] For the convenience of management, the skill service system can differentiate skill templates according to the actual service usage scenarios for users to choose from, such as factor extraction templates, chat dialogue templates, etc.

[0051] After the user selects a skill template for a certain scenario on the skill development interface, the skill service system displays the general skill template for the scenario on the display interface, such as Figure 3 The element extraction template shown includes at least large model skill information, input parameter information, output parameter information, etc.

[0052] Step S12: Write the element information input by the user into the skill template to generate a skill version.

[0053] In an embodiment of the present application, after the user selects a skill template in the skill development interface, the skill template will display corresponding element information that needs to be supplemented based on the specific skill scenario.

[0054] by Figure 3Taking the feature extraction template shown as an example, when users choose this type of template, they only need to describe all the feature information that needs to be extracted, supplemented by examples of feature extraction to optimize the calling effect. Users no longer need to pay attention to the specific prompt word writing content, but only need to describe clearly the information related to the specific business.

[0055] The skill template of this application can reduce the content that users need to configure during the skill development process by pre-setting prompt words, related information of input and output parameters, etc. Users only need to enter personalized element information to fill in the gaps in the skill template and generate a skill version that can be run.

[0056] It should be noted that the skill template of the present application can also preset element information in advance, and the user can fine-tune the element information provided in the skill template to obtain the personalized customized skill version required by the user.

[0057] Please refer to Figure 4 , Figure 4 This is a schematic diagram of the interface for creating skills from element extraction using the element extraction template provided by this application. Figure 4 As shown, the element information that users need to enter for the element extraction template is mainly the extraction content field, including but not limited to: the time of the event, the location of the event, the user's main demand, etc. Furthermore, users can also enter element information in different languages, such as the English version of the above extraction content field corresponds to: time, location, mainbody, etc.

[0058] Finally, the skill service system will generate specific skill services based on the skill preset information selected by the user, including but not limited to: prompt words, large model call parameters, service input and output parameter information, data processing rules, etc. The generated information will be displayed on the page, and the user can fine-tune the specific information as needed to obtain the best effect.

[0059] Step S13: Combine the skill version with the large model to generate a skill service.

[0060] In the embodiment of the present application, the skill version is configured with relevant parameters of the big model, and the skill service system also needs to configure a suitable big model for the skill version generated in step S12 to obtain a skill service that can be directly called.

[0061] Specifically, the skill service system is equipped with large models of different scenarios, types, versions, model parameters, and structures. The skill service system needs to search for the best large model that matches the skill version from the large number of large models and obtain the best combination. The search method can use the existing debugging and comparison tools to provide a variety of large model switching options on the skill service development page, so that the skill service system can call multiple models at the same time, display the results together for effect comparison, and facilitate users to adjust parameters, preview the analysis results, and select the best combination.

[0062] Please refer to Figure 3 , Figure 3 The skill template interface also includes a debugging and preview interface, which provides user input functions and the parsing results of the user input statements running on the large model, so that users can directly view the current effect of the skill service from the debugging and preview interface. Visualization is achieved through the debugging and preview interface, and subsequent fine-tuning operations and large model optimal combination matching operations can be implemented.

[0063] Furthermore, after debugging the effect, the user can use the combination he thinks is the best to publish the skill service. When publishing the skill service, it is also necessary to bind the application created in the skill service system. When binding, you can choose the application's concurrency, fuse parameters and other flow control information for this skill service. The application can bundle user-defined large model parameters, including authorized access information for third-party large models. The skill service system will synchronize the skill information to the API service entrance for users to call, and will bundle the API service and application information to provide skill services in the form of http / websocket / sse.

[0064] In this application, the skill calling device responds to the user's skill development instruction and selects a specified skill template, wherein the skill template includes preset prompt words, large model operation parameters, input and output formats, and / or pre-configured information; writes the element information input by the user into the skill template to generate a skill version; and combines the skill version with the large model to generate a skill service. Through the above-mentioned skill development method, by reusing the skill template extracted based on the business, the same type of large model skill service can be quickly developed.

[0065] Regarding the large model skill calling process, the skill calling process is a series of steps for how to use the skill service released by the skill development process, including access security, data processing and other related steps.

[0066] Please refer to Figure 5 and Figure 6 , Figure 5 is a flow chart of an embodiment of a skill calling method based on a large model provided by the present application. Figure 6 It is a schematic diagram of the overall process of the skill calling method based on the large model provided in this application.

[0067] Step S21: The caller sends request information to the provider based on the access token.

[0068] In the embodiment of the present application, after the caller obtains the access token from the provider, it adds the access token to the request header information to request the skill service from the provider. The provider in the embodiment of the present application is the skill service system in the above embodiment, and the caller is any device or apparatus required by the skill service.

[0069] Step S22: The provider analyzes the system applications, skill services and user issues based on the request information.

[0070] In an embodiment of the present application, the provider receives the request information from the caller, and parses the relevant service call parameters, access tokens, etc. from the request information, and obtains detailed information about the skill service that needs to be called, such as user questions.

[0071] Step S23: The provider runs the system application and calls the skill service to perform business processing on the user's question, and returns the business processing result to the caller.

[0072] In an embodiment of the present application, the provider runs a system application to implement business processing of user issues by the skill service, that is, inputs the user issue into the big model in the skill service, and configures the parameters of the big model according to the skill version in the skill service to perform business processing on the user issue, and returns the output result of the big model to the caller as the business processing result of the user issue.

[0073] Specifically, the provider runs the system application and calls the skill version and big model in the skill service; the provider uses the big model running parameters in the skill version to run the big model, and inputs the user problem into the big model for business processing, and obtains the business processing result output by the big model. The provider returns the business processing result to the caller.

[0074] Furthermore, before the provider handles the user's problem, it is also necessary to verify the caller's permissions. Specifically, the provider verifies the caller's access token, parses the application information corresponding to the token, extracts the relevant parameters of the custom large model bundled with the application, and confirms that the caller has the relevant permissions to access this service. Among them, the caller's access token is generated by the provider, so the provider can verify whether it is the access token issued by itself through the identifier of the access token, thereby determining the caller's calling permissions.

[0075] In a specific implementation, the provider performs data processing on the "user question" in the request parameter. Regularizing the user question according to the data processing rules can help the big model better understand it. For example, "Query people who take the D7429 train to Beijing on the 28th of this month" can be regularized as "Query people who take the train (train number: D7429, boarding date: October 28, 2024) to Beijing". The data processing rules are configured during the skill development process and will not be repeated here.

[0076] Furthermore, the provider can also perform conventional string replacement operations on the results returned by the large model to remove meaningless characters for structured extraction, and then perform data processing on the results after string replacement. The results returned can be regularized according to data processing rules to correct some errors in large model recognition. For example, "Train number: D7429, flight departure time: October 28, 2024" can be regularized to "Train number: D7429, boarding date: October 28, 2024".

[0077] The provider performs structured extraction and regular string replacement operations on the large model to return the result, and then processes the extracted value result again to meet certain special format requirements of the caller. For example, "{"CCRQ": "2024 October 28"}", is regularized to "{"CCRQ": "2024 / 10 / 28"}". At the same time, the data rules are used to confirm whether the value content meets the specified requirements. For example, confirm whether the extracted data meets the ID card number rules. If not, a specified prompt needs to be returned.

[0078] Finally, the provider can return the business processing results according to the caller's requirements.

[0079] For details on how to request and obtain access tokens for skill services, please refer to Figure 7 and Figure 8 , Figure 7 is a flowchart of an embodiment of a method for requesting and obtaining an access token provided by this application. Figure 8 This is a schematic diagram of the overall process of requesting and obtaining an access token provided by this application.

[0080] Step S31: The caller generates token application information, wherein the token application information includes application identification, application key, request time, and / or application public and private keys.

[0081] In the embodiment of the present application, the skill service access token is to request and obtain the token using the application ID (Identity document), application key (secret), request time reqTime and the public and private keys of the application. Therefore, the caller needs to send the token application information composed of the above information to the provider.

[0082] Specifically, the token application information uses the application ID, application secret and request time reqTime to perform sha256 hash calculation to obtain the sign signature string, which is then combined with the application ID, request time reqTime and sign signature string to form a json string.

[0083] Among them, SHA256 (Secure Hash Algorithm 256-bit) is a secure hash algorithm.

[0084] Step S32: The caller uses the application private key to encrypt the token application information, obtains the request information, and uses the request information to request the provider to access the token interface.

[0085] In the embodiment of the present application, the caller uses the private key corresponding to the application to encrypt the json string, and the encrypted data is encoded using base64 to obtain the request string. The caller requests the system backend, i.e. the provider, to access the token interface through the request string.

[0086] Step S33: The provider uses the application public key to decrypt the request information, and generates an access token in response to successful decryption.

[0087] In an embodiment of the present application, the provider uses the application public key for decryption, and if the decryption fails, a token request failure is returned.

[0088] After decryption is successful, the provider first confirms that the request time reqTime and the request time allowed by the system do not exceed the time difference threshold. Then, the application ID, application secret and request time reqTime are used to recalculate the sign value and compare it with the sign value uploaded by the user to resist time attacks. If the time difference between the request time reqTime and the request time allowed by the system is within the time difference threshold and matches the sign successfully, an access token will be generated and returned. In other cases, a failure message will be returned and the abnormal behavior of the caller will be recorded.

[0089] It should be noted that each caller's token has a time validity period, and needs to be reapplied after the validity period expires.

[0090] Among them, if the comparison algorithm is not implemented well, the time difference of the returned result will be relatively large due to different data contents. Some hackers can use this time difference to calculate the matching password. ConstantTimeCompare will only take different time due to inconsistent length of the compared data, and will not change the time length due to different data. These are some security measures taken to resist side channel attacks.

[0091] Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0092] In order to implement the above-mentioned skill development method and / or skill calling method, the present application also proposes a skill calling device, for details, please refer to Fig. 9 , Fig. 9 It is a structural diagram of an embodiment of a skill calling device provided in this application.

[0093] The skill calling device 400 of this embodiment includes a processor 41, a memory 42, an input and output device 43 and a bus 44.

[0094] The processor 41, memory 42, and input / output device 43 are respectively connected to a bus 44. The memory 42 stores program data. The processor 41 is used to execute the program data to implement the skill development method and / or skill calling method described in the above embodiments.

[0095] In the embodiment of the present application, the processor 41 may also be referred to as a CPU (Central Processing Unit). The processor 41 may be an integrated circuit chip having the ability to process signals. The processor 41 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or the processor 41 may also be any conventional processor, etc.

[0096] This application also provides a computer storage medium, please continue to refer to Fig.10 , Fig.10 It is a structural diagram of an embodiment of a computer storage medium provided in the present application. The computer storage medium 600 stores a computer program 61. When the computer program 61 is executed by the processor, it is used to implement the skill development method and / or skill calling method of the above-mentioned embodiment.

[0097] When the embodiments of the present application are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0098] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A skill development method based on a large model, characterized in that: The skills development methods include: In response to a skill development instruction from a user, a designated skill template is selected, wherein the skill template includes preset prompt words, large model operation parameters, input and output formats, and / or pre-configured information; Writing the element information input by the user into the skill template to generate a skill version; The skill version is combined with the large model to generate a skill service.

2. The skill development method according to claim 1, characterized in that: After combining the skill version with the large model to generate a skill service, the skill development method further includes: The skill service is bound to the system application to provide a calling interface externally through the system application.

3. The skill development method according to claim 1, characterized in that: Combining the skill version with the big model to generate a skill service includes: Call several large models supported by the system and run the skill versions respectively; Display the running results of each large model for the skill version; In response to the user's large model selection instruction, the corresponding large model is selected and combined with the skill version to generate a skill service.

4. The skill development method according to claim 1, characterized in that: The pre-configuration information includes verification rules for input and output parameters, and / or data processing rules; wherein the data processing rules are composed of regular expressions and scripting languages, and are used to process data generated during the skill calling process.

5. A skill calling method based on a large model, characterized in that: The skill calling method comprises: The caller sends a request message to the provider based on the access token. The provider analyzes the system application, skill service and user problem based on the request information; The provider runs the system application, calls the skill service to perform business processing on the user problem, and returns the business processing result to the caller; Wherein, the skill service is developed through the skill development method described in any one of claims 1 to 4.

6. The skill calling method according to claim 5, characterized in that: The provider runs the system application, calls the skill service to perform business processing on the user problem, and returns the business processing result to the caller, including: The provider runs the system application and calls the skill version and the large model in the skill service; The provider uses the big model running parameters in the skill version to run the big model, and inputs the user question into the big model for business processing, and obtains the business processing result output by the big model; The provider returns the business processing result to the caller.

7. The skill calling method according to claim 6, characterized in that: The skill calling method further includes: The provider uses the data processing rules of the pre-configured information in the skill version to perform regular processing on the user question.

8. The skill calling method according to claim 5, characterized in that: The skill calling method further includes: The caller generates token application information, wherein the token application information includes an application identifier, an application key, a request time, and / or an application public and private key; The caller encrypts the token application information using the application private key to obtain request information, and uses the request information to request an access token interface from the provider; The provider decrypts the request information using the application public key, and generates an access token in response to successful decryption.

9. The skill calling method according to claim 8, characterized in that: Generating an access token includes: The provider obtains the time difference between the request time in the request information and the system allowed time; The provider calculates a verification string using the application identifier, application key, and request time in the request information; In response to the time difference being less than a preset time difference threshold and the verification string successfully matching the signature string of the token application information, the provider generates the access token.

10. A skill calling device based on a large model, characterized in that: The skill calling device includes a memory and a processor coupled to the memory; The memory is used to store program data, and the processor is used to execute the program data to implement the skill development method as described in any one of claims 1 to 4, and / or the skill calling method as described in any one of claims 5 to 9.

11. A computer storage medium, characterized in that: The computer storage medium is used to store program data, and when the program data is executed by a computer, it is used to implement the skill development method described in any one of claims 1 to 4, and / or the skill calling method described in any one of claims 5 to 9.

Images