Galileo

#+TITLE: Welcome to Project Galileo's source repository. #+AUTHOR: xtt28

Project Galileo is an intelligent AI agent that can interact with your computer and help you with day-to-day tasks, including browsing, searching files, and more.

** Get started on your own device

*** Cloning/downloading

Clone these sources by clicking the green Code button above the file list.

**** Building from source

If you build from source, you must have the Go toolchain installed on your system. Run ~go run .~ in your terminal to compile and run the software, or ~go build .~ to build a binary.

**** Pre-compiled releases

At the moment, there are no pre-compiled releases available.

*** OpenAI API

At the moment, this application depends on the OpenAI API. Create an OpenAI API key and ensure that your account has a sufficient amount of credits. In the following step, you will set up your OpenAI API key to be used by this application.

*** Configuration

You will need to create a configuration file before launching. The configuration file should be located in the following location:

• On Windows: ~%APPDATA%\io.github.xtt28.galileo\config.json~
• On macOS: ~~/Library/Application Support/io.github.xtt28.galileo/config.json~
• On Linux: ~~/.config/io.github.xtt28.galileo/config.json~

Sample configuration file contents: #+BEGIN_SRC json { "openAIKey": "sk-proj-......" } #+END_SRC

** Features of Project Galileo

*** General

• The user can have a conversation with the AI via a graphical desktop application.

*** Filesystem

• The agent can read text files in the filesystem.
  • Before the agent is allowed to read the contents of a file in the filesystem, the user's permission is asked first via a dialog box.
• The agent can create text files in the filesystem.
• The agent can view the contents of the user's documents folder.
• The agent can open any file in its default program on the OS at request.

*** Applications

• The agent can view a list of the user's installed desktop applications.
• The agent can open any of these applications at request.

*** Web & Realtime

• The agent can get the weather forecast in the user's current location.
• The agent can search the web for a specific query via DuckDuckGo.
• The agent can use a headless Chromium browser to access and read web pages.

** Technological details

• The application uses Fyne, a cross-platform GUI toolkit built for Go.
• The official OpenAI API wrapper for Go is used for chat completions. By default, the application uses GPT 4o-mini.
  • The OpenAI API tool calling feature is used to enable agent functionality.
• The Norwegian Meteorological Institute public API is used for getting real-time weather data.
• The agent uses headless Chromium via the ~chromedp~ package to access web pages.
• To search the web, the agent uses headless Chromium as described above with the HTML version of DuckDuckGo ([[https://html.duckduckgo.com/html]]).
• To get the user's location for weather forecast purposes, the public API [[https://ip-api.com/json]] is used.

** License

This software is licensed under GNU General Public License, v3.0 or later. See the LICENSE.md file for more details.