JP7641701B1 - Apparatus, method and program for generating technical documentation for software - Google Patents

Abstract

A method for generating technical documentation corresponding to source code managed by a code management system is provided.
[Solution] In response to a repository selection received from a user terminal 110, the device 100 receives repository information related to the repository from the code management system 130 (S201 to S203). The device 100 then extracts multiple API code candidates and one or more root file candidates from the repository information (S204). The device 100 then requests a generative AI model (k) to determine paths of the multiple API codes and one or more root files using the extracted file candidates (S205), and receives the determined paths (S206). For each API code, the device 100 requests a generative AI model (l) to generate API documentation corresponding to the API code, and receives and stores the generated technical documentation (S207 to S209).
[Selected figure] Figure 2

Description

The present invention relates to an apparatus, method, and program for generating technical documentation for software.

There are various methods for developing software. In one method, use cases are first created, then technical documentation is created that describes the more technical details, and the necessary coding is then done based on the technical documentation to create the source code.

In recent years, technology has been developed to automatically generate source code based on technical documentation. For example, the applicant has disclosed a technology that obtains table definitions and API definitions for multiple tables and APIs required by software, and generates source code in accordance with the table definitions and API definitions (see Patent Document 1).

JP 2023-031251 A

Users of services that use such automatic source code generation technology generally manage the generated source code in a code management system such as GitHub (registered trademark). If a change or addition of functionality becomes necessary while actually providing software using the source code, the source code is modified. When such modifications occur, the technical documentation held in the source code generation service no longer corresponds to the source code. Therefore, even if users wish to further use the service to expand functionality, source code that is inconsistent with modifications made outside the service will be generated.

In addition, for users who want to use source code generation services to change or add functions to existing software, they cannot use the services because the services do not hold technical documentation for the software in question.

The present invention has been made in consideration of these points, and aims to provide an apparatus, method, or program for generating technical documentation corresponding to source code managed by a code management system.

To achieve this objective, a first aspect of the present invention is a method for generating technical documentation corresponding to source code managed in a code management system, comprising the steps of receiving repository information regarding a repository in which the source code is managed, determining paths of multiple API codes and one or more root files from the repository information, and making a first request to a first AI model, which is a generative AI model, to generate API documentation corresponding to at least one of the multiple API codes, the first request including code in the one or more root files, and storing the generated one or more API documentation.

A second aspect of the present invention is the method of the first aspect, in which the determining step includes a step of making a second request to a second AI model, which is a generative AI model, to determine paths of the plurality of API codes and the one or more root files from the repository information.

A third aspect of the present invention is the method of the second aspect, in which the determining step further includes a step of extracting a plurality of API code candidates and one or more root file candidates from the repository information before making the second request.

The fourth aspect of the present invention is the method of the first aspect, in which the determination is made using the directory structure of the repository.

A fifth aspect of the present invention is a method according to any one of the first to fourth aspects, further comprising a step of transmitting API document viewing information to a user terminal for viewing at least one of the one or more API documents.

A sixth aspect of the present invention is a method for generating technical documentation corresponding to source code managed in a code management system, comprising the steps of receiving repository information regarding a repository in which the source code is managed, determining paths of a plurality of business logic codes from the repository information, making a first request to a first AI model, which is a generative AI model, to generate business logic corresponding to at least one of the plurality of business logic codes, and storing a plurality of business logic codes corresponding to the plurality of business logic codes.

A seventh aspect of the present invention is a program for causing a computer to execute a method for generating technical documentation corresponding to source code managed in a code management system, the method including the steps of receiving repository information regarding a repository in which the source code is managed, determining paths of multiple API codes and one or more root files from the repository information, and making a first request to a first AI model, which is a generative AI model, to generate API documentation corresponding to at least one of the multiple API codes, the first request including code in the one or more root files, and storing the generated one or more API documentation.

An eighth aspect of the present invention is a device for generating technical documentation corresponding to source code managed in a code management system, configured to receive repository information regarding a repository in which the source code is managed, determine paths of multiple API codes and one or more root files from the repository information, make a first request to a first AI model, which is a generative AI model, to generate API documentation corresponding to at least one of the multiple API codes, the first request including code in the one or more root files, and store the generated one or more API documents.

A ninth aspect of the present invention is a program for causing a computer to execute a method for generating technical documentation corresponding to source code managed in a code management system, the method including the steps of receiving repository information regarding a repository in which the source code is managed, determining paths of a plurality of business logic codes from the repository information, making a first request to a first AI model, which is a generative AI model, to generate business logic corresponding to at least one of the plurality of business logic codes, and storing a plurality of business logic codes corresponding to the generated one or more business logic codes.

A tenth aspect of the present invention is an apparatus for generating technical documentation corresponding to source code managed in a code management system, configured to receive repository information regarding a repository in which the source code is managed, determine paths of a plurality of business logic codes from the repository information, make a first request to a first AI model, which is a generative AI model, to generate business logic corresponding to at least one of the plurality of business logic codes, and store the generated business logic or business logics.

According to one aspect of the present invention, the path of required types of files such as API code and business logic code is determined from repository information related to the repository in which source code is managed, and technical documentation corresponding to at least one of those files is automatically generated using a generative AI model, thereby bridging the gap between software development and documentation creation.

FIG. 1 illustrates a system for generating technical documentation according to one embodiment of the present invention. FIG. 2 is a flow diagram of a method for generating technical documentation according to an embodiment of the present invention. FIG. 13 illustrates an example of a prompt that may be included as part of the code for a request k for determining a path of a technical document in accordance with an embodiment of the present invention. FIG. 2 illustrates an example of code for creating a prompt that is included as part of the code for a request l to a generative AI model (l) in accordance with one embodiment of the present invention. FIG. 5 is a diagram showing an example of a template “BUILD_API_FROM_CODE” used in FIG. 4 .

The following describes in detail an embodiment of the present invention with reference to the drawings.

Figure 1 shows a system for generating technical documentation according to one embodiment of the present invention. In order to create technical documentation for software, the device 100 communicates with a user terminal 110 used by a user, a code management system 120, and a platform 130 that provides a generative AI model via an IP network such as the Internet. The generative AI model is illustratively provided by the platform 130 that can communicate with the device 100, but it is also possible to run an application for providing the generative AI model on the device 100 so that the generative AI model is provided by the device 100. The device 100 can also provide a source code generation service to the user that generates source code corresponding to the technical documentation.

The device 100 includes a communication unit 101 such as a communication interface, a processing unit 102 such as a processor or CPU, and a storage unit 103 including a storage device or storage medium such as a memory or a hard disk, and can be configured by executing a program for performing each process or operation in the processing unit 102. The device 100 may include one or more devices, computers, or servers. The program may include one or more programs, and may be recorded in a computer-readable storage medium to form a non-transient program product. The program may be stored in a storage device or storage medium such as the storage unit 103 or a database 104 accessible from the device 100 via an IP network, and instructions included in the program may be executed in at least one processor of the processing unit 102. Data described below as being stored in the storage unit 103 may be stored in a storage device or storage medium such as the database 104, and vice versa.

First, the device 100 receives from the user terminal 110 a selection of a repository on the code management system 120 that manages the software for which technical documentation is to be created (S201). This selection can be made, for example, by displaying to a user who has logged in to a service provided by the device 100 a selection screen for one or more repositories associated with that user, and allowing that user to select one of them.

Next, the device 100 requests repository information related to the repository from the code management system 120 (S202) and receives the repository information (S203). The repository information may include all files in the selected repository, or may exclude some files. Here, an example has been described in which the device 100 acquires repository information in response to a user's selection of a repository, but the device 100 can also detect changes, such as updates to files in a repository associated with the user, and acquire repository information related to the repository in response to detecting the change.

The device 100 that has acquired the repository information then extracts multiple API code candidates and one or more route file candidates from the repository information as necessary (S204). As an example, the extraction can be performed by registering the paths of multiple files included in the acquired repository information in a vector database and identifying one or more files whose similarity to a specified term is equal to or greater than a predetermined threshold. When generating an API document, the term may be "API", "APIs", "Controller", "Controllers", "Route", or "Routes". When generating business logic, the term may be "Service", "Services", "Business Logic", "Business Logics", "Domain Logic", "Domain Logics", "Handler", or "Handlers". The larger the scale of software development, the greater the number of files that are not related to the technical document to be generated. Therefore, by extracting related file candidates and reducing the number of files to be processed thereafter, the speed at which technical documents are generated can be significantly increased. Here, "API document" refers to a document that describes the function and usage of an API.

Then, the device 100 requests (also called "request k") the AI model (also called "generative AI model (k)"), which is the generative AI model, to determine the paths of the multiple API codes and one or multiple root files using the extracted file candidates if file candidates have been extracted, or using the acquired repository information if no file candidates have been extracted (S205).

Figure 3 shows an example of an instruction called a prompt that is included as part of the code for request k to a generative AI model (k) according to one embodiment of the present invention. In this prompt, the variable {files} is set to the path of all files or multiple files included in the repository information. The prompt also describes how to specify the business logic code, API code, or root file, and output the path of each specified file together with the category name in a specified format. If the purpose is to generate API documentation and generation of business logic is not required, it is not necessarily necessary to specify the business logic code. Here, "business logic" refers to logic for processing data sent or received in communication with a user interface (UI) based on business requirements, and "logic" means a series of instructions.

The example in FIG. 3 is an example of a prompt written in natural language that is included as part of code written in the programming language TypeScript, and by executing this code, the OpenAI API can be called and the generative AI model (k) provided on the platform 130 can determine a path. The OpenAI API is an example, and other APIs may be used. The programming language TypeScript is an example, and other languages may be used. The code for making request k is stored in the storage unit 103, and the device 100 can acquire this and make request k by executing code that includes instructions obtained by setting required values to variables included in the code.

After receiving the determined path (S206), the device 100 requests (also called "request l") the AI model (also called "generative AI model (l)"), which is a generative AI model, to generate API documentation corresponding to the API code for each API code (S207). After the generative AI model (l) creates technical documentation (S208), the device 100 receives and stores the technical documentation (S209).

It is preferable that request l includes all root files whose paths have been determined. It is also possible to determine one or more root files associated with each API code and include only the determined root files in request l, but since the number of root files is not huge even without making such a determination, by including all codes contained in all root files in request l, it is possible to increase accuracy without significantly slowing down the generation speed of technical documentation.

Figure 4 shows an example of code for creating a prompt included as part of the code for request l to a generative AI model (l) according to one embodiment of the present invention. In this example, the API code is set in the variable "code", the code in the route file in the variable "route", the format of the response to request l in the variable "response_introduction", and the module to which the API provided by the API code belongs in the variable "module" in the template "BUILD_API_FROM_CODE". The variables "module" and "response_introduction" are not necessarily required. The variable "response_introduction" may also be set to the format of the response specified by the user. Figure 5 shows an example of the template "BUILD_API_FROM_CODE" used in Figure 4.

In FIG. 2, generative AI model (k) and generative AI model (l) are distinguished, but they may be the same generative AI model. In this specification, "AI model" refers to a machine learning model trained to predict an output for an input, and "generative AI model" refers to a large-scale language model (LLM) trained using text data to generate an output that is not included in the input. As a generative AI model, an LLM that applies a transformer architecture is particularly preferable, but it is expected that the name of the architecture will change as technology advances. Therefore, in this specification, "transformer architecture" includes an architecture that uses one or more features of the transformer architecture or an improvement thereof. In this specification, whether "generative AI models" are the same is determined by whether the type of generative AI model specified by the user is the same. In the case of Open AI API, for example, in the case of GPT-4, if the value of the variable "model" is the same, it is expressed as the same generative AI model. If the first generative AI model and the second generative AI model are not the same, they may be provided on the same platform 130. In the example of FIG. 2, request k is followed by request l. If generative AI model (k) and generative AI model (l) are the same generative AI model or are generative AI models provided on the same platform, these requests may be made by a single call to the API. Needless to say, request k and request l may each include multiple requests, and may include one or more processes performed by device 100 other than the request for the generative AI model.

Then, the device 100 transmits technical document viewing information for viewing at least one of the generated technical documents to the user terminal 110 (S210). The technical document viewing information can be, for example, an HTML format file, which the user terminal 110 can use to display a technical document viewing screen on the display screen. Also, an application installed on the user terminal 110 may be operated, and the technical document viewing screen may be displayed within the application using the received technical document viewing information. Note that the "viewing screen" may adopt various formats, such as a web page, modal window, or pop-up window, when displayed on a web browser, and may be one screen of the application when displayed within the application. In either case, any screen that includes an area for displaying the generated technical documents in a viewable manner corresponds to a viewing screen.

As such, automatic generation of technical documentation from source code bridges the gap between software development and documentation creation, and significantly reduces the manual work typically required to create accurate and up-to-date technical documentation. Since the technical documentation to be generated may be duplicated, it is not necessary to generate technical documentation for some files whose paths have been determined.

Please note that in the above embodiments, unless there is a statement of "only", such as "based only on", "only in response to", "only in the case of", or "with reference only", it is assumed in this specification that additional information may be taken into consideration. As an example, please note that the statement "do b when a" does not necessarily mean "do b always when a" or "do b immediately after a" unless expressly stated. Furthermore, the statement "each a constituting A" does not necessarily mean that A is composed of multiple components, but includes the case where the component is singular.

For the avoidance of doubt, even if there is an aspect of a method, program, terminal, device, server, or system (hereinafter "method, etc.") that performs an operation different from that described in this specification, each aspect of the present invention is directed to an operation identical to any of the operations described in this specification, and the existence of an operation different from that described in this specification does not cause the method, etc. to fall outside the scope of each aspect of the present invention.

Please note that the disclosure of this specification includes any combination of the above-mentioned embodiments of the present invention to the extent that they are not mutually exclusive.

In this specification, we have explained the flow of making request l after request k, but in cases where the generative AI model is not necessarily used to determine the path, the generative AI model (l) may be referred to as the first generative AI model and request l as the first request.

REFERENCE SIGNS LIST 100 Device 101 Communication unit 102 Processing unit 103 Storage unit 104 Database 110 User terminal 120 Source code management system 130 Platform

Claims (9)

1. A method for generating technical documentation corresponding to source code managed in a code management system, comprising:
receiving , by a computer, repository information regarding a repository in which the source code is managed;
The computer determines, from the repository information, paths of a plurality of API codes and one or more root files;
A step of the computer making a first request to a first AI model, which is a generative AI model, for at least one of the plurality of API codes to generate API documentation corresponding to the API code, the first request including a code in the one or more root files;
and storing , by the computer, the generated one or more API documents. 2. The method of claim 1 ,
The determining step includes:
The method includes making a second request to a second AI model, which is a generative AI model, to determine paths of the plurality of API codes and the one or more root files from the repository information. 2. The method of claim 1 ,
The determination is made using the directory structure of the repository. 4. A method according to any one of claims 1 to 3, comprising:
The method further includes a step of transmitting API document viewing information to a user terminal for viewing at least one of the one or more API documents. 1. A method for generating technical documentation corresponding to source code managed in a code management system, comprising:
receiving , by a computer, repository information regarding a repository in which the source code is managed;
determining , by the computer, from the repository information, paths of a plurality of business logic codes;
A step of the computer making a first request to a first AI model, which is a generative AI model, for at least one of the plurality of business logic codes to generate business logic corresponding to the business logic code;
and storing , by the computer, the generated one or more business logics. A program for causing a computer to execute a method for generating technical documentation corresponding to source code managed in a code management system, the method comprising:
receiving repository information regarding a repository in which the source code is managed;
determining a plurality of API codes and paths of one or more root files from the repository information;
making a first request to a first AI model, which is a generative AI model, to generate API documentation corresponding to at least one of the API codes, the first request including a code in the one or more root files;
and storing the generated one or more API documents. An apparatus for generating technical documentation corresponding to source code managed in a code management system, comprising:
receiving repository information regarding a repository in which the source code is managed;
determining a plurality of API codes and paths to one or more root files from the repository information;
making a first request to a first AI model, which is a generative AI model, to generate API documentation corresponding to at least one of the API codes, the first request including code in the one or more root files;
A storage device is configured to store the generated API document or documents. A program for causing a computer to execute a method for generating technical documentation corresponding to source code managed in a code management system, the method comprising:
receiving repository information regarding a repository in which the source code is managed;
determining a plurality of business logic code paths from the repository information;
making a first request to a first AI model, which is a generative AI model, for at least one of the plurality of business logic codes to generate business logic corresponding to the business logic code;
and storing the generated one or more business logics. An apparatus for generating technical documentation corresponding to source code managed in a code management system, comprising:
receiving repository information regarding a repository in which the source code is managed;
determining a path for a plurality of business logic codes from the repository information;
making a first request to a first AI model, which is a generative AI model, for at least one of the plurality of business logic codes to generate business logic corresponding to the business logic code;
The database is configured to store a plurality of generated business logics.

Images