A declarative remote service interface calling protocol, method and device

Abstract

The present invention discloses a declarative remote service interface calling protocol, method and device, which relates to the technical field of remote service interface calling. The service management center installs a remote service interface calling protocol, which is used to implement remote calling. The service management center makes strong type declarations for all interfaces and registers them with the service management center, so that all callers can implement direct calling without formatting agreements. The remote service interface of the present invention provides two capabilities: active registration and passive perception, which completely eliminates the calling mode of demand agreement, allowing the caller to complete the call of the service interface without an agreement. The repeated development cost of the caller is reduced to 0. The caller only needs to implement a set of data interaction protocol codes and complete the data format specification conversion to implement the call of all interfaces, which can be exposed to users more conveniently. Metadata description information is provided to enable the interface to have the ability to directly serve AI.

Description

Technical Field

[0001] The present invention relates to the technical field of remote service interface calling, and in particular to a declarative remote service interface calling protocol, method and device. Background Art

[0002] The complete execution of a service request usually includes two roles: the "caller" and the "called service". The called service is the service provider, which exposes some interfaces with established functions, such as providing some data operation interfaces for adding, deleting, modifying and querying, or some interfaces for data processing capabilities. The caller can be diverse, such as the front-end browser, APP, client page, and the upstream service of the called service can also make requests to the called service through API or SDK.

[0003] A basic premise for the above request to be successfully executed is that the caller and the called service need to agree on a specific data interaction protocol. That is, the caller must know what request / response data format a certain interface of the called service requires, and the called service parses the request according to the established data format and returns it in the agreed format.

[0004] Therefore, in this mode, if the caller is not clear about the request and response format of the called service interface, it cannot use the service normally.

[0005] A call to a remote service interface requires the caller to obtain three pieces of information in advance: the service address, the interface path, and the request and response format.

[0006] There is currently no solution that can actively perceive request responses and interface paths. For example, although gRPC technology can make data transmission between the caller and the called service more standardized and efficient through protoBuf, gRPC itself is not currently centralized and requires the caller and the service provider to make an advance agreement during the development phase.

[0007] The prior art has the following disadvantages:

[0008] 1. High call cost. The caller cannot directly obtain detailed information about the service interface, including the path and request response format, and can only agree with the service provider. Each time a new interface is requested, the corresponding request code needs to be rebuilt.

[0009] 2. It is difficult to upgrade the service interface. When the interface is changed or upgraded, even if only one parameter is changed, the caller may need to refactor the relevant code of the request process; and for the caller who cannot perceive the interface upgrade, there is a high probability that the call will fail.

[0010] 3. There is no condition for AI to call directly. There is no unified standard for the request and response format of each service interface, and AI cannot learn how to initiate requests and understand the meaning of parameters. Summary of the invention

[0011] In order to solve the above-mentioned shortcomings of the prior art, the present invention provides a declarative remote service interface calling protocol, method and device. The following technical solutions are adopted:

[0012] A declarative remote service interface calling method, a service management center installs a remote service interface calling protocol, the remote service interface calling protocol is used to implement remote calling, and the remote calling includes the following steps:

[0013] The service management center declares all interfaces with strong types and registers them with the service management center, allowing all callers to implement direct calls without formatting agreements;

[0014] The service management center sets a metadata manager, which is used to maintain the paths and data interaction methods of all service interfaces. The metadata manager defines type standards, which are used to standardize the format of request responses.

[0015] Active registration: The instance corresponding to the caller's remote service communicates and interacts with the registration center of the service management center when it starts, and performs the action of registering itself. The action of registering itself is: apply for a dedicated domain name, register the IP and port number of the remote service in the dedicated routing table to which the name belongs, and then register metadata. According to the exposed interface, the metadata is created through the metadata manager according to the metadata specification defined in the type standard;

[0016] The caller obtains the access address of all instances through the exclusive domain name of the service management center, obtains the interface access information of all instances from the metadata manager, and fills in the data value according to the data interaction protocol after obtaining the access address and metadata.

[0017] By adopting the above technical solution, the following preparations must be made to implement declarative remote service interface calls: a service registration center is used for service registration and discovery, and all services must ensure network interoperability. Through a service management center, a declarative remote service interface perception method, protocol and device are implemented. The specific implementation includes the following steps:

[0018] This service maintains the IP and port numbers of all remote services and all interface metadata information. This service can adopt existing solutions in the industry, such as ETCD, etc.

[0019] The metadata manager is responsible for maintaining the paths of all service interfaces and data interaction methods, and defines a set of type standards to regulate the format of request responses.

[0020] The metadata manager provides interfaces such as service registration and discovery as well as configuration changes, as follows:

[0021] Registration: When a remote service is started, it will register with the registration center, tell the registration center its IP and port number, and maintain the heartbeat of the service through periodic health checks.

[0022] Discovery: The caller can obtain the access address IP and port number of all remote service instances through the service name registered by the remote service in the service management center.

[0023] Heartbeat: The remote service instance periodically sends heartbeats to the service management center to indicate its survival status.

[0024] Metadata changes: Allow remote services to upload, query, and update interface metadata information.

[0025] Metadata publish / subscribe: After the updated interface metadata is published, services are allowed to subscribe at the same time.

[0026] Active interface registration capability: The called service declares all interfaces with strong types and registers them with the service management center, allowing all callers (services that can access the service management center) to implement direct calls without formatting agreements.

[0027] Passive interface perception capability: The called service provides an access path with all interface information. The interface information under this path can be identified and obtained by all protocols supported by crawlers. For example, the Robots protocol defines the interface that needs to be exposed to the outside in the interface description file, which is maintained by the called service itself. After the external caller crawls the interface description file, it can also implement direct calls without formatting agreements.

[0028] Optionally, the service interface exposed by the remote service has passive perception capabilities. The remote service provides an access path with all interface information. The interface information under the access path allows it to be identified and crawled by all protocols supported by the crawler. The crawled interface information is stored in a public interface repository for external query.

[0029] For example, the Robots protocol, the caller remote service obtains the exposed interface meta-information through the Robots protocol, and directly completes the call to the interface according to the interface description, parameter description and data interaction protocol.

[0030] Optional, strongly typed declaration rules for fields: types, supporting default values ​​and optional parameters.

[0031] Optional, field: type, write the interface declaration using the following syntax:

[0032] Boolean: Use boolean to define the Boolean type; numeric: Use number to define the numeric type; string: Use string to define the numeric type; array: Use array <elemtype>Define array type; Enumeration: use enum to define enumeration type; Composite class: use class{} to define a composite type; Function description: use # as function description, followed by description content.

[0033] By adopting the above technical solution, the called service interface will be defined according to the interface metadata specification, and the data will be transmitted using json, as follows: Prerequisite: The caller already has a copy of the interface metadata information before requesting the interface; Data transmission: Regardless of the request or response, the data must be filled in the format of <field><data value> and meet the following rules: Field name: The field name in the interaction data must be exactly the same as the field name in the interface metadata definition; Data type: The data type in the interaction data must be exactly the same as the type in the interface metadata definition, where the composite type corresponds to {} and the array type corresponds to [].

[0034] Optionally, the interface metadata defines the functions, affiliated services, and data interaction methods of each interface, where each field in the request and response belongs to a unique type in the field: type definition.

[0035] Optionally, all interface metadata will be saved in the distributed KV storage service through the metadata manager, and some interfaces that need to be exposed to the outside are maintained by the service itself under the set path.

[0036] By adopting the above technical solution, the called service declares all interfaces with strong types and registers them with the service management center, so that all callers can access the services of the service management center and implement direct calls without formatting agreements.

[0037] Optionally, the caller implements AI calling based on an AI device. After receiving the natural language description command, the AI ​​device vectorizes it through a vector embedding model, matches it to a vector database, exchanges the search results with a large language model to obtain accurate interface results, and finally calls it through a request script.

[0038] By adopting the above technical solution, services, interfaces and parameters can be marked with descriptive strings to achieve AI automated calling.

[0039] The AI ​​call directly calls the corresponding service interface through natural language description. The pre-preparation is as follows:

[0040] The AI ​​calling device can access the service management center or public interface warehouse and all services

[0041] A vector database is used to store description data.

[0042] A vector embedding model is used to parse the description information into vectors.

[0043] A large language model is used to process the caller's natural language instructions.

[0044] The specific implementation is as follows:

[0045] i. Prepare the corpus

[0046] Obtain the full amount of interface metadata information through the metadata manager of the service management center or the public interface repository, and convert the interface description information into vectors through the vector embedding model and store them in the vector database. ii. Prepare a general request script

[0047] This script implements the data interaction protocol for use by AI

[0048] iii. Large model matching

[0049] The user describes an action to be performed in natural language, and the AI ​​calling device completes the service interface call. The process is as follows:

[0050] 1. The user enters in the input box: "All data in xxxx / xxx / xxx needs to be compressed and stored in xxx / xxx / xxx";

[0051] 2. The input is vectorized through the vector embedding model;

[0052] 3. Use the vectorized input to find the top-N interfaces that best match the vector database;

[0053] 4. The best matching interface is selected by the large model, and the metadata information of the interface is passed to the request script;

[0054] 5. The script requests the corresponding interface.

[0055] A declarative remote service interface calling device comprises a remote service calling terminal, a server and a network communication module. The remote service calling terminal is connected to the server through the network communication module, and the server sets a service management center by using a remote service interface calling method.

[0056] Optionally, an AI device is also included, and the AI ​​device is communicatively connected to the server via a network communication module.

[0057] Optionally, the AI ​​device includes a radio microphone, a chip and a memory. The chip is communicatively connected to the server via a network communication module. The radio microphone and the memory are communicatively connected to the chip respectively. A vector embedding model, a vector database and a large language model are set in the memory.

[0058] By adopting the above technical solution, the receiving microphone is used to sense the caller's natural language instructions and store them in the memory. The chip calls the vector embedding model, vector database and large language model to process the natural language description commands, thereby forming metadata information of the calling interface and passing it to the request script. The script requests the corresponding interface to realize AI calling.

[0059] In summary, the present invention includes at least one of the following beneficial technical effects:

[0060] The present invention can provide a declarative remote service interface calling protocol, method and device, which can optimize the problems of the prior art and achieve the following effects:

[0061] 1. Active interface registration capability: The called service declares all interfaces with strong types and registers them with the service management center, allowing all callers (services that can access the service management center) to implement direct calls without formatting agreements.

[0062] 2. Interface passive perception capability: The interface exposed by the called service provides passive perception capability. The interface that needs to be exposed to the outside is defined in the interface description file and maintained by the called service itself. After the external caller crawls the interface description file, it can also implement direct calling without format agreement.

[0063] 3. AI calling capability: Services, interfaces and parameters can be marked with descriptive strings to achieve AI automated calling.

[0064] The remote service interface provides two capabilities: active registration and passive perception. It completely eliminates the calling mode of demand agreement, allowing the caller to complete the call of the service interface without an agreement. The caller's repeated development cost is reduced to 0. The caller, whether it is a web page, an APP or an upstream service, only needs to implement a set of data interaction protocol codes and complete the data format specification conversion to realize the call of all interfaces and expose them to users more conveniently. Providing metadata description information enables the interface to have the ability to directly serve AI. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a logical schematic diagram of a declarative remote service interface calling protocol, method and device of the present invention;

[0066] Figure 2 It is an AI calling logic diagram of a declarative remote service interface calling protocol, method and device of the present invention. DETAILED DESCRIPTION

[0067] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0068] The embodiment of the present invention discloses a declarative remote service interface calling protocol, method and device.

[0069] Reference Figure 1 and Figure 2 Embodiment 1, a declarative remote service interface calling method, the service management center installs a remote service interface calling protocol, the remote service interface calling protocol is used to implement remote calling, the remote calling includes the following steps:

[0070] The service management center declares all interfaces with strong types and registers them with the service management center, allowing all callers to implement direct calls without formatting agreements;

[0071] The service management center sets up a metadata manager, which is used to maintain the paths and data interaction methods of all service interfaces. The metadata manager defines type standards, which are used to standardize the format of request responses.

[0072] Active registration: The instance corresponding to the caller's remote service communicates with the registration center of the service management center when it starts up and performs the action of registering itself. The action of registering itself is: apply for a dedicated domain name, register the IP and port number of the remote service into the dedicated routing table of the name, and then register metadata. According to the exposed interface, the metadata is created through the metadata manager according to the metadata specification defined in the type standard;

[0073] The caller obtains the access address of all instances through the exclusive domain name of the service management center, obtains the interface access information of all instances from the metadata manager, and fills in the data value according to the data interaction protocol after obtaining the access address and metadata.

[0074] To implement declarative remote service interface calls, the following preparations must be done first: a service registration center is used for service registration and discovery, and all services must ensure network interoperability. Through a service management center, a declarative remote service interface perception method, protocol, and device are implemented. The specific implementation includes the following steps:

[0075] This service maintains the IP and port numbers of all remote services and all interface metadata information. This service can adopt existing solutions in the industry, such as ETCD, etc.

[0076] The metadata manager is responsible for maintaining the paths of all service interfaces and data interaction methods, and defines a set of type standards to regulate the format of request responses.

[0077] The metadata manager provides interfaces such as service registration and discovery as well as configuration changes, as follows:

[0078] Registration: When a remote service is started, it will register with the registration center, tell the registration center its IP and port number, and maintain the heartbeat of the service through periodic health checks.

[0079] Discovery: The caller can obtain the access address IP and port number of all remote service instances through the service name registered by the remote service in the service management center.

[0080] Heartbeat: The remote service instance periodically sends heartbeats to the service management center to indicate its survival status.

[0081] Metadata changes: Allow remote services to upload, query, and update interface metadata information.

[0082] Metadata publish / subscribe: After the updated interface metadata is published, services are allowed to subscribe at the same time.

[0083] Active interface registration capability: The called service declares all interfaces with strong types and registers them with the service management center, allowing all callers (services that can access the service management center) to implement direct calls without formatting agreements.

[0084] Passive interface perception capability: The called service implements the Robots protocol. The interfaces that need to be exposed to the outside are defined in the interface description file, which is maintained by the called service itself. After the external caller crawls the interface description file, it can also implement direct calls without formatting agreements.

[0085] In embodiment 2, the service interface exposed by the remote service has passive perception capability. The remote service provides an access path with all interface information. The interface information under the access path allows all protocols supported by the crawler to be identified and crawled. The crawled interface information is stored in a public interface repository for external query.

[0086] For example, the Robots protocol, the caller remote service obtains the exposed interface meta-information through the Robots protocol, and directly completes the call to the interface according to the interface description, parameter description and data interaction protocol.

[0087] Strong typing declaration rules are fields: types, supporting default values ​​and optional parameters.

[0088] Field: The type of the interface declaration is written using the following syntax:

[0089] Boolean: Use boolean to define the Boolean type; numeric: Use number to define the numeric type; string: Use string to define the numeric type; array: Use array <elemtype>Define array type; Enumeration: use enum to define enumeration type; Composite class: use class{} to define a composite type; Function description: use # as function description, followed by description content.

[0090] The called service interface will be defined in accordance with the interface metadata specification, and the data will be transmitted using json, as follows: Prerequisite: The caller already has a copy of the interface metadata information before requesting the interface; Data transmission: Regardless of the request or response, the data must be filled in the format of <field><data value> and meet the following rules: Field name: The field name in the interaction data must be exactly the same as the field name in the interface metadata definition; Data type: The data type in the interaction data must be exactly the same as the type in the interface metadata definition, where the composite type corresponds to {} and the array type corresponds to [].

[0091] The specific request response data format is as follows:

[0092] GET xxx.com / student

[0093] {"id":"E3601'}

[0094] {"'name":"Tina""id":"E3601""gender":"female""grades":

[0095] [{"item":"physics""score":87}

[0096] {"item":"'math""score":91}]}.

[0097] Embodiment 3, the interface metadata defines the function of each interface, the subordinate service and the data interaction mode of each interface, wherein each field in the request response belongs to the field: a unique type in the type definition.

[0098] In Example 4, all interface metadata will be stored in the distributed KV storage service through the metadata manager, and some interfaces that need to be exposed to the outside will be maintained by the service itself under a set path.

[0099] A specific setting path is: / apis / path; the specific format is as follows:

[0100] url==;endpoint==;method==;description==;

[0101] Req();

[0102] Resp();

[0103] url: interface path;

[0104] endpoint: service address;

[0105] method: request method;

[0106] description: detailed description of the interface;

[0107] req: request format;

[0108] resp: the format of the response;

[0109] The format of the request and response is filled in according to the standard data type; a complete interface definition example is as follows:

[0110] url== / student;

[0111] endpoint==xxx.com;

[0112] method==GET;

[0113] description=="Get detailed information of a student";

[0114] reg(id: string; #student's id);

[0115] resp(student: Stu; #student type);

[0116] classStu{#Definition of student combination typeid: string; #Student id

[0117] name: string; #Student's name

[0118] gender: Gender; #student gender grades: array <grade>;Students’ scores in each subject

[0119] };

[0120] enumGender{male, female}; #Definition of gender type

[0121] classGrade{#Definition of grades

[0122] score: number; #subject score

[0123] item: string; #subject name}.

[0124] In embodiment 5, the called service declares all interfaces with strong types and registers them with the service management center, so that all callers can access the services of the service management center and implement direct calls without formatting agreements.

[0125] In Example 6, the caller implements AI calling based on an AI device. After receiving the natural language description command, the AI ​​device vectorizes it through a vector embedding model, matches it to a vector database, exchanges the search results with a large language model to obtain accurate interface results, and finally calls it through a request script.

[0126] Services, interfaces and parameters can be marked with descriptive strings to enable AI automated calls.

[0127] The AI ​​call directly calls the corresponding service interface through natural language description. The pre-preparation is as follows:

[0128] The AI ​​calling device can access the service management center and all services

[0129] A vector database is used to store description data.

[0130] A vector embedding model is used to parse the description information into vectors.

[0131] A large language model is used to process the caller's natural language instructions.

[0132] The specific implementation is as follows:

[0133] i. Prepare the corpus

[0134] The metadata manager of the service management center obtains the full amount of interface metadata information, and converts the interface description information into vectors through the vector embedding model and stores them in the vector database;

[0135] ii. Prepare a general request script

[0136] This script implements the data interaction protocol for use by AI;

[0137] iii. Large model matching

[0138] The user describes an action to be performed in natural language, and the AI ​​calling device completes the service interface call. The process is as follows:

[0139] 1. The user enters in the input box: "All data in xxxx / xxx / xxx needs to be compressed and stored in xxx / xxx / xxx";

[0140] 2. The input is vectorized through the vector embedding model;

[0141] 3. Use the vectorized input to find the top-N interfaces that best match the vector database;

[0142] 4. The best matching interface is selected by the large model, and the metadata information of the interface is passed to the request script;

[0143] 5. The script requests the corresponding interface.

[0144] Embodiment 7, a declarative remote service interface calling device, includes a remote service calling terminal, a server and a network communication module, the remote service calling terminal is connected to the server through the network communication module, and the server uses the remote service interface calling method to set the service management center.

[0145] Embodiment 8 further includes an AI device, and the AI ​​device is connected to the server for communication via a network communication module.

[0146] In Example 9, the AI ​​device includes a radio microphone, a chip and a memory. The chip is connected to the server through a network communication module. The radio microphone and the memory are connected to the chip respectively. A vector embedding model, a vector database and a large language model are set in the memory.

[0147] The receiving microphone is used to sense the caller's natural language instructions and store them in the memory. The chip calls the vector embedding model, vector database and large language model to process the natural language description commands, thereby forming the metadata information of the calling interface and passing it to the request script. The script requests the corresponding interface to realize AI call.

[0148] Example 10, supplementary air ticket query scenario calling process:

[0149] 1. The AI ​​calling device obtains the full amount of interface information from the service management center, vectorizes it and stores it in the corpus;

[0150] 2. The user provides instructions to the AI ​​calling device through natural language description (including but not limited to voice, text, etc.): "What time is the flight from Beijing to Shanghai tomorrow?"

[0151] 3. The AI ​​calling device vectorizes the instruction and searches for the matching flight query interface in the corpus;

[0152] 4. The AI ​​calling device automatically fills in the keywords entered by the user according to the request format of the flight query interface: Beijing (place of departure), Shanghai (destination), tomorrow (xxxx-xx-xx);

[0153] 5. The AI ​​calling device makes a request to the interface according to the address of the interface;

[0154] 6. Return the flight information.

[0155] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.< / grade> < / elemtype> < / elemtype>

Claims

1. A declarative remote service interface calling method, characterized in that: The service management center installs a remote service interface calling protocol, which is used to implement remote calling. The remote calling includes the following steps: The service management center requires all called services to make strong type declarations for all public interfaces and register them with the service management center, allowing all callers to implement direct calls without formatting agreements. The service management center sets a metadata manager, which is used to maintain the paths and data interaction methods of all service interfaces. The metadata manager defines type standards, which are used to standardize the format of request responses. Active registration: When the instance corresponding to the called service is started, it communicates with the registration center of the service management center and performs the action of registering itself. The action of registering itself is: apply for a dedicated domain name, register the IP and port number of the remote service in the dedicated routing table to which the name belongs, and then register metadata. According to the exposed interface, the metadata is created through the metadata manager according to the metadata specification defined in the type standard; The caller obtains the access address of all instances through the exclusive domain name of the service management center, obtains the interface access information of all instances from the metadata manager, and after obtaining the access address and metadata, fills the data value of the call request according to the data interaction protocol.

2. A declarative remote service interface calling method according to claim 1, characterized in that: The service interface exposed by the remote service has passive perception capabilities. The remote service provides an access path with all interface information. The interface information under the access path allows it to be identified and crawled by all protocols supported by the crawler. The crawled interface information is stored in a public interface repository for external query.

3. A declarative remote service interface calling method according to claim 1, characterized in that: Strong typing declaration rules are fields: types, supporting default values ​​and optional parameters.

4. The declarative remote service interface calling method according to claim 1, characterized in that: Field: The type of the interface declaration is written using the following syntax: Boolean: Use boolean to define the Boolean type; numeric: Use number to define the numeric type; string: Use string to define the numeric type; array: Use array <elemtype> Define array type; Enumeration: use enum to define enumeration type; Composite class: use class{} to define a composite type; Function description: use # as function description, followed by description content.< / elemtype> 5. A declarative remote service interface calling method according to claim 4, characterized in that: The interface metadata defines the functions, affiliated services, and data interaction methods of each interface, where each field in the request and response belongs to a unique type in the field: type definition.

6. A declarative remote service interface calling method according to claim 5, characterized in that: All interface metadata will be saved in the distributed KV storage service through the metadata manager. At the same time, some interfaces that need to be exposed to the outside are maintained by the service itself under the set path.

7. A declarative remote service interface calling method according to claim 5, characterized in that: The caller implements AI calling based on AI device. After receiving the natural language description command, the AI ​​device vectorizes it through the vector embedding model, matches it to the vector database, exchanges the search results with the large language model to obtain accurate interface results, and finally calls it through the request script.

8. A declarative remote service interface calling device, characterized in that: It comprises a remote service calling terminal, a server and a network communication module. The remote service calling terminal is connected to the server through the network communication module. The server adopts the remote service interface calling method described in claim 1 to set up a service management center.

9. A declarative remote service interface calling device according to claim 8, characterized in that: It also includes an AI device, which is connected to the server through a network communication module.

10. A declarative remote service interface calling device according to claim 9, characterized in that: The AI ​​device includes a radio microphone, a chip and a memory. The chip is connected to the server through a network communication module. The radio microphone and the memory are connected to the chip respectively. A vector embedding model, a vector database and a large language model are set in the memory.

Images