eunomia-bpf / func-call-for-command

Generate GPT function call API from command-line help info

GPT Func Call generator for Command-line

Your one-click solution to transform traditional command-line tools into natural language-driven powerhouses!

Ever wondered if your command-line tools could understand plain English? We did too, so we built Naturaliser. Built on the bedrock of OpenAI's GPT technology, Naturaliser turns your dense, jargon-filled help files into sleek, interactive commands that anyone can understand and execute.

Features ✨

• 🧠 Smart Function Call Generation: Provide your tool's help file, and GPT Func Call generator will do the rest. We'll generate the corresponding OpenAI GPT function calls and JSON descriptions in a flash, and the driven library in any language as well.
• 🛠 Python and Web App Support: Whether you're dealing with Python scripts or web apps, Naturaliser has got you covered.
• 🔄 Single-Click Operation: Simplify your workflow with our user-friendly interface. One click, and you're good to go!

Let's make command-line tools more accessible, user-friendly, and less intimidating.

Web tool

See https://ai-func-call-gen.vercel.app/ for details.

A function call generator for terminal commands, where the user provides the help file for the command and GPT generates the corresponding function call and Json description. Support python scripts or web apps.

Repo: https://github.com/eunomia-bpf/ai-func-call-gen

Python Script Usage

Firstly, you need to clone this repo and install the necessary python package:

git clone https://github.com/eunomia-bpf/func-call-for-command.git
pip install -r requirements.txt

Then, use command line interaction to generate function calls and JSON descriptions. You can provide command name such as python ./main.py -k you_openai_api_key -c uname. When using this method, make sure you have the relevant commands installed.

There is an example about this tools:

$python ./main.py -k you_openai_api_key -c uname

During this process, GPT returns a JSON description of the command, like the following:

{
    "name": "uname",
    "description": "Print certain system information",
    "parameters": {
        "type": "object",
        "properties": {
            "all": {
                "type": "boolean",
                "description": "Print all information"
            },
            ...
            "hardware-platform": {
                "type": "boolean",
                "description": "Print the hardware platform (non-portable)"
            },
            "operating-system": {
                "type": "boolean",
                "description": "Print the operating system"
            }
        },
        "required": []
    }
}

This will generate a uname-gpt.sh file, and then you can use the uname command in a much easier way, like the following:

$bash ./uname-gpt.sh print the kernel information
Run: uname --kernel-name  --kernel-release  --kernel-version 
Linux 5.15.90.1-microsoft-standard-WSL2 #1 SMP Fri Jan 27 02:56:13 UTC 2023

You can find the specifics of uname-gpt.sh in examples/ of this project.