WoT Code Generator

This repository contains the source code of the code generator tool developed for the thesis "Correctness Evaluation of LLM-based Protocol Code Generation for Web of Things Applications." It also provides a flexible and extensible CLI tool and a Web-based UI for generating protocol-specific client-side code based on Web of Things (WoT) Thing Descriptions (TDs).

Features

• Supports deterministic and AI-powered code generation (ChatGPT, Gemini, Llama)
• Multi-language and multi-library support for various IoT protocols
• Interactive and one-line command execution for ease of use
• Web-based user interface

This tool serves as a testbed for evaluating AI-assisted vs. deterministic code generation approaches in WoT applications, contributing to a better understanding of their correctness, adherence to standards, generalizability, developer experience, and overall perfomance.

Code Generator CLI

A command-line interface tool for generating protocol-specific code snippets from Web of Things Thing Descriptions (TDs).

Features

• Interactive and one-line command modes
• Deterministic Support for multiple programming languages and libraries
• AI-powered code generation for not yet available cases (ChatGPT, Gemini, Llama)
• TD input via file of text editor
• Multiple output options (console of file)

Installation Steps

1. Clone the repository using Git:

   git clone https://github.com/SergioCasCeb/code-generator.git

2. Navigate to the project directory:

   cd code-generator

3. Install all project dependencies:

   npm install

4. Link the project globally to make the generate-code command available on your system:

   npm link

   This allows you to run the generate-code command anywhere in your terminal.

Usage

The CLI can be used in two modes: interactive and one-line command

Interactive Mode

Run the CLI in interactive mode:

generate-code -i

The interactive mode guides you through the following steps:

1. Choose TD input method (file or text editor)
2. Select an affordance
3. Choose a form index (optional)
4. Select and operation
5. Choose between deterministic or AI-powered generation (If a protocol is not supported AI is selected by default)
6. Select programming language
7. Select library
8. Choose output type (console or file)

Finally, after generating the code, the interactive mode also returns the equivalent options to run the CLI in the one-line mode.

One-line Command Mode

For automated workflows, use the one-line command mode:

Options

• -i, --interactive: Run CLI in interactive mode
• -t, --td <path>: Path to the Thing Description file
• -a, --affordance <name>: Affordance to use
• -f, --form-index <index>: Form index to use
• -o, --operation <name>: Operation to perform
• -l, --language <name>: Programming language
• -L, --library <name>: Library to use
• --ai: Use AI generation
• --tool <name>: AI tool to use (ChatGPT, Gemini, Llama)
• -O, --output <type>: Output type (console or file)
• -h, --help: Display help information
• -V, --version: Display version number

Examples

Generate code utilizing the deterministic generator

generate-code --td ./calculator.td.jsonld -a result -f 0 -o readproperty -l javascript -L node-wot -O console

Generate code utilizing the AI generator

generate-code --td ./calculator.td.jsonld -a result -f 0 -o readproperty -l javascript -L axios --ai --tool chatgpt -O console

Supported formats

Input files

• JSON (.json)
• JSON-LD (.jsonld)
• Text (.txt)

Programming Languages and Libraries

Deterministic:

• JavaScript

  • Fetch
  • node-wot
  • modbus-serial

• Python

  • Requests

No restrictions when using AI generation.

Output

Console Output

The generated code is printed directly to the console in simple text format. AI generation adds markdown syntax to the code.

File Output

A new folder called generator-ouput is created within the project and the generated code is saved to a file with the following naming convention <affordance>_<operation>_<language>.<extension>

UI Installation

1. Navigate to the UI directory:

   cd code-generator/ui

2. Install all project dependencies:

   npm install

3. Build the project with Webpack:

   npm run build

4. Serve the dist folder with express:

   npm run serve

   This will allow you to utilize the UI, and output the generated code through the projects API.