docuhook 0.1.1

A pre-commit hook that uses AI to automatically generate Python docstrings.

DocuHook 🤖

[!NOTE] DocuHook is a pre-commit hook that automatically generates docstrings for your Python code using AI.

Key Features

• Automatic Docstring Generation: Forget about writing docstrings manually. DocuHook does it for you with every commit.
• Seamless Git Integration: Works in the background as a pre-commit hook without changing your natural workflow.
• Support for Fast AI Models: Configured to work with the fastest models (e.g., Groq, Claude Haiku) to avoid slowing down your work.
• Smart Code Analysis: Uses the ast module to precisely find functions that require documentation.

Installation

To get started, install the package using pip:

pip install docuhook

Usage

To enable DocuHook in your project, follow these 3 simple steps:

1. Configure pre-commit

In your .pre-commit-config.yaml file, add the following entry:

repos:
  - repo: https://github.com/bartoszborowski/docuhook
    rev: v0.1.0
    hooks:
      - id: docuhook

2. Install the hook

Run the following command to have pre-commit activate the hook in your repository:

pre-commit install

3. Set the API Key

DocuHook needs an API key to communicate with the language model. Set it as an environment variable. For example, for Groq:

export GROQ_API_KEY="sk-..."

Done! From now on, with every git commit, DocuHook will automatically analyze your code and add any missing docstrings.

Configuration

To use a different language model, you need to update the API endpoint and the request payload in the docuhook/ai_client.py file.

Specifically, you will need to modify the API_URL variable and the payload dictionary inside the generate_docstring function.

Here is the relevant code snippet from docuhook/ai_client.py:

# docuhook/ai_client.py

import os
import requests

# Ensure you have the correct environment variable for your API key
API_KEY = os.getenv("YOUR_API_KEY") 

# 1. Change this URL to your provider's API endpoint
API_URL = "https://api.groq.com/openai/v1/chat/completions"


def generate_docstring(code_to_document: str) -> str:
    headers = {"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"}

    prompt_instruction = """
    Your sole task is to process the Python code below.
1. Generate a complete, Google-style docstring for the provided function.
2. Return only the new docstring as raw text
3. Don't include anything like: Here is the Google-style docstring for the provided function
"""
    # 2. Adjust the payload to match the new API's requirements
    payload = {
        "model": "llama3-8b-8192",
        "messages": [
            {
                "role": "system",
                "content": "You are an expert in writing clean Python code. Your task is to generate documentation in the Google Style Docstrings format. Documentation has to be written in English",
            },
            {
                "role": "user",
                "content": f"{prompt_instruction}:\n\n```python\n{code_to_document}\n```",
            },
        ],
        "temperature": 0.1,
    }

    response = requests.post(API_URL, headers=headers, json=payload)

    if response.status_code == 200:
        return response.json()["choices"][0]["message"]["content"]
    else:
        return f"API Error: {response.status_code} - {response.text}"

License

This project is licensed under the MIT License. See the LICENSE file for more details.