browser-ai is a simple Desktop app that works as a browser, but uses Google Gemini URL Context tool to create AI-generated summaries of the websites you visit, alongside displaying the actual website itself.
Note
This is a uv project, so make sure you have it installed, or install it before getting started with browser-ai.
You first need to clone this repository:
git clone https://github.com/AstraBert/browser-ai
cd browser-aiNow, you will have to spin up a virtual environment for the application to work. In order to do so, please run:
uv syncWhen executing this command, you might run into some dependency incompatibilities, especially if you are on a Linux or MacOS machine. In that case, please refer to the posix-specific installation guide for the project.
Rename the .env.example file to .env:
mv .env.example .envAnd then set the GOOGLE_API_KEY to a valid Google API Key that has access to Gemini models.
Launch the application from your terminal, by typing:
uv run main.pyYou will see a window popping-up on your Desktop: it's your browser app!
You just have to type a URL in the box above the browser, and click the Go button, and you'll see the website popping up on your screen, as well as (after some seconds) a summary of that same website generated by Gemini.
The browser application itself is served through Toga, a multi-platform, python-native framework to design simple and intuitive desktop and web applications: it leverages the WebView widget to visualize website, and the Label one to display AI-generated summaries.
In order to create AI summaries, browser-ai uses the latest version of Gemini-2.5-Flash, coupling it with the URL context tool, that allows Gemini to understand the content of website, and serving it through LlamaIndex.
Contributions are more than welcome! Follow the contribution guidelines to make sure your contribution is compliant with the repo's requirements :)
This is an open-source project distributed under an MIT License.