Automatically fills in all GitHub repository topics to the max using AI to guess which topics are the best.

I found myself wanting to fill each of my repos’ topics as much as I could, but it came at the cost of slowly and sequentially thinking about what keywords to choose. So of course I will take hours of my time to shave off a few seconds of creative thought.

A Python program that uses GitHub and OpenAI’s APIs, wiring them together to make an AI model generate a list of topics (the tags/keywords you see on the repository page) based on the repo’s metadata: i.e., title, description, existing tags, programming languages, and of course the contents of the legendary README.md file.

Final topics are ordered as: existing topics first, then AI-generated ones. If you already have 5 topics, those five remain; the app only appends new topics until it reaches the maximum (GithubConn.MAX_TOPIC_COUNT). You can change this limit programmatically if you prefer to cap the list at N topics.

1. Clone the repository and enter it.

   git clone https://github.com/Urpagin/gh-topicgen.git
   cd gh-topicgen

2. Before running, populate the two API tokens you’ll need. Copy the .env.example file to .env and follow the instructions inside.

   cp .env.example .env

3.A [Linux/macOS] Create a Python virtual environment and activate it (optional).

python3 -m venv .venv
source .venv/bin/activate

3.B [Windows] Create a Python virtual environment and activate it (optional).

py -m venv .venv
.\.venv\Scripts\activate

1. Install the app (editable mode).

   pip install -e .

2. Run it to generate relevant topics for a repository.

ghtopicgen --visibility public --regex 'YourUserName/YourRepoTitle'

To quickly get the hang of the app, skim the Examples section.

$ python -m gh_topicgen -h
usage: ghtopicgen [-h] [--gh-token GH_TOKEN] [--openai-token OPENAI_TOKEN] [--model MODEL] [--prompt PROMPT] [--regex REGEX] [--ignore-case]
                  [--env-file PATH] [--no-input] [--take-my-money] --visibility {public,private,all}

Manage tokens/config.

optional arguments:
  -h, --help            show this help message and exit
  --gh-token GH_TOKEN, -g GH_TOKEN
                        GitHub token (overrides env; aka GH_TOKEN)
  --openai-token OPENAI_TOKEN, -o OPENAI_TOKEN
                        OpenAI token (overrides env; aka OPENAI_API_KEY)
  --model MODEL, -m MODEL
                        Which OpenAI model to use (https://platform.openai.com/docs/models)
  --prompt PROMPT, -p PROMPT
                        Path to AI system prompt file.
  --regex REGEX, -r REGEX
                        Regex for repo names (matches against full_name, e.g., 'Urpagin/slpcli').
  --ignore-case, -i     Case-insensitive regex.
  --env-file PATH, -e PATH
                        Path to .env to load first (default: auto-detect)
  --no-input            Do not prompt; fail if tokens are missing
  --take-my-money       Special model spec: GPT-5 Thinking with high effort. If set, any --model value is ignored.
  --visibility {public,private,all}, -v {public,private,all}
                        What type of repositories to use.

Used to generate this repo’s topics:

ghtopicgen -v public -m gpt-5 -r 'Urpagin/gh-topicgen'

Generate the topics of all your repos (private + public) using the best model currently:

ghtopicgen -v all --take-my-money

Generate the topics of all your repos using the fastest and cheapest model (default model):

ghtopicgen -v all -m gpt-5-nano

The system prompt given to the model is stored in system_prompt.txt. Feel free to tweak it as you wish.

Made with compatibility in mind, avoiding highly modern 3.11/3.12-only features.

Compatible with Python ≥ 3.9.

$ vermin --no-tips src/gh_topicgen/*
Minimum required versions: 3.9
Incompatible versions:     2

• The PyGithub library — interacting with the GitHub REST API
• The official OpenAI Python library — interacting with the OpenAI REST API

Every file in the repository falls under the MIT License.

MIT © 2025 Urpagin.