Gemi

Gemi is a free, open-source AI agent powered by Google's Gemini 2.5 Flash. It can browse your files, read them, write new ones, and even run Python scripts to help you with coding tasks.

It comes with a friendly terminal interface (TUI) so you can chat with it like a normal app, and everything it does stays inside a safe sandbox folder so it can't touch the rest of your computer.

• List files and folders in your project
• Read file contents to understand your code
• Write or update files for you
• Run Python scripts and report back the results

You'll need two things before you start:

1. Python 3.13 or newer —> download it from python.org if you don't have it.
2. A free Gemini API key —> get one in a couple of clicks at Google AI Studio.

Then follow these steps:

git clone https://github.com/andrefetch/Gemi.git
cd Gemi

pip install -r requirements.txt

Create a file named .env in the Gemi folder and paste your key inside it like this:

GEMINI_API_KEY=your_key_here

(Replace your_key_here with the key you copied from Google AI Studio.)

The easiest way to start is the chat interface:

./gemi.sh

Then just type what you'd like Gemi to do and press Enter. Press Cntrl+C whenever you want to quit.

Prefer a quick one-off command instead of the chat window? You can ask Gemi a single question straight from your terminal:

gemi "list the files in this project"

Add --verbose if you'd like to see exactly what Gemi is doing behind the scenes.

The TUI has a built in theme switcher that allows user to install themes such as: catppuccin. And also provides a keymapping system to see what every keybind does.

Note: Gemi does not have a memory system and can't keep memory of past conversations.

Google-Genai (As the model for the agent)
Rich, Textual & UC-Micro-Py: TUI Libraries
Argparse: for parsing arguments to pass

Gemi is open source and contributions are very welcome. If you'd like to add a feature or fix something, feel free to open a Pull Request, thanks for helping out!