Basic Features

• 🔌 Seamless from CLI or code (e.g., gw.awg.find_cable() is gway awg find-cable)
• ⛓️ CLI chaining: proj1 func1 - proj2 func2 (implicit parameter passing by name)
• 🧠 Sigil-based context resolution (e.g., [result-context-environ|fallback])
• ⚙️ Automatic CLI generation, with support for *, *args and **kwargs
• 🧪 Built-in test runner and self-packaging: gway test and gway release build
• 📦 Environment-aware loading (e.g., clients and servers .env files)

Examples

from gway import gw

result = gw.awg.find_cable(meters=30, amps=60, material="cu", volts=240)
print(result)

# Basic cable sizing
gway awg find-cable --meters 30 --amps 60 --material cu --volts 240

# With conduit calculation
gway awg find-cable --meters 30 --amps 60 --material cu --volts 240 --conduit emt

# Chain cable calculation and echo the result
gway awg find-cable --meters 25 --amps 60 - print --text "[awg]"

> gway -r website

> gway recipe register-gwr

Online Help & Documentation

Browse built-in and project-level function documentation online at:

📘 https://arthexis.com/gway/help

• Use the search box in the top left to find any callable by name (e.g., find_cable, resource, start_server).
• You can also navigate directly to: https://arthexis.com/gway/help?topic=<project-or-function>

This is useful for both the included out-of-the-box GWAY tools and your own projects, assuming they follow the GWAY format.

Installation

Your chosen Installation method will depend on how you intend to use GWAY:

1. If you intend to contribute to GWAY at some point or want to access the latest updates from the open source community, you should Install from Source. As a plus, you get everything in the basic ecosystem from the get go.
2. If you want to use GWAY for a private use, such as work for a company or customer that prevents you from sharing your code, or you alredy have an open-source repo but want a second personal one, Install via PyPI.

You may also install them either way and just experiment with what each mode offers. For example, PyPI install allows you to easily use GWAY within Google Colab or other IPython/Jupyter projects.

Install via PyPI:

pip install gway

# TODO: Explain configuration of the local GWAY_ROOT directory

Install from Source:

git clone https://github.com/arthexis/gway.git
cd gway

# Run directly from shell or command prompt
./gway.sh        # On Linux/macOS
gway.bat         # On Windows

When running GWAY from source for the first time, it will auto-install dependencies if needed.

To upgrade to the latest version from source:

./upgrade.sh     # On Linux/macOS
upgrade.bat      # On Windows

This pulls the latest updates from the main branch and refreshes dependencies.

To make GWAY available from any directory (requires root access):

sudo ln -s "$HOME/gway/gway.sh" /usr/local/bin/gway

Project Structure

Here's a quick reference of the main directories in a typical GWAY workspace:

After placing your modules under projects/, you can immediately invoke them from the CLI with:

gway project-dir-or-script your-function argN --kwargN valueN

By default, results get reused as context for future calls made with the same Gateway thread.

Overview

• Views are simply Python functions in a project (e.g. projects/web/site.py) named according to a pattern (by default, view_{name}).
• The web.app.setup function registers views from one or more projects and sets up all routing and static file handling.
• The web.server.start-app function launches your site on a local server using Bottle (or FastAPI, for ASGI).
• All configuration can be scripted using GWAY recipes (.gwr files) for full automation.

Minimal Example

Suppose you want to create a very simple website:

# projects/mysite.py

def view_hello():
    return "<h1>Hello, World!</h1>"

def view_about():
    return "<h2>About This Site</h2><p>Powered by GWAY.</p>"

def view_user(*, user_id=None):
    if user_id:
        # We have a user_id, so greet the user
        return f"<h1>Welcome {user_id}</h1>"
    else:
        # No user_id, so render a form to collect it
        return '''
        <form method="get" action="">
            <label for="user_id">Enter User ID:</label>
            <input type="text" id="user_id" name="user_id" required />
            <button type="submit">Submit</button>
        </form>
        '''

Note that these views don't need to be decorated and you don't have to return the entire HTML document. You also don't have to specify http methods or where the variables come from (they can be read from a form or passed as a query param.)

Then in your own recipe:

# recipes/my-website.gwr
web app setup --project mysite --home hello
web app setup --project web.navbar
web server start-app --host 127.0.0.1 --port 8888
forever

Navigate to http://127.0.0.1:8888/mysite/hello or /mysite/about to see your views, including a handy navbar. Press Ctrl+D or close the terminal to end the process.

The forever function keeps the above apps and servers running forever.

Composing Sites from Multiple Projects

You can chain as many projects as you want; each can define its own set of views and home page:

# recipes/my-website.gwr
web app setup --home readme
    --project web.cookie
    --project web.navbar --home style-changer
    --project vbox --home uploads
    --project conway --home game-of-life --path games/conway

web server start-app --host 127.0.0.1 --port 8888
until --file VERSION --pypi

The above example combines basic features such as cookies and navbar with custom projects, a virtual upload/download box system and Conway's Game of Life, into a single application.

The above recipe also shows implicit repeated commands. For example, instead of writing "web app setup" multiple times, each line below that doesn't start with a command repeats the last command with new parameters.

The until function, as used here, will keep the recipe going until the package updates in PyPI (checked hourly) or a manual update ocurrs. This is appropriate for self-restarting services such as those managed by systemd or kubernetes.

How It Works

• web.app.setup wires up each project, registering all views (functions starting with the given prefix, default view_).
• You call setup multiple times to configure each project. The project/function name can be skipped on repeat lines.
• Each project can declare a "home" view, which becomes the landing page for its route.
• Static files are served from your data/static/ directory and are accessible at /static/filename.
• The routing system matches /project/viewname to a function named view_viewname in the relevant project.
• Query parameters and POST data are automatically passed as keyword arguments to your view function.

View Example with Arguments

# projects/vbox.py

def view_uploads(*, vbid: str = None, timeout: int = 60, files: int = 4, email: str = None, **kwargs):
    """
    GET: Display upload interface or create a new upload box.
    POST: Handle uploaded files to a specific vbid.
    """
    ...

This view can be accessed as /vbox/uploads and will receive POST or GET parameters as arguments.

Recipes make Gway scripting modular and composable. Include them in your automation flows for maximum reuse and clarity.

Keep a Goal in Mind by Making it a TODO

Before you start writing any code, even if you have already planned the feature in an external system, write a # TODO comment explaining the required changes in as much detail as you need to define it unambiguously.

TODOs should live with the code they intend to affect. They warn that things are going to soon be different. It allows a good feature to be noted instead of lost to priorities. Keeping the tasks in the code itself as TODOs is superior to using an external tool. Those systems should look at the code as the source of truth to determine what TODOs need to be accomplished and make them more visible instead.

However, if a TODO affects the entire project, put it at the top of the file instead.

You may write the TODO and then dispatch it in the same coding session, or it may live on for many commits until its time comes. Or maybe you change your mind and delete the TODO. You get a space, close to the code, where you can see the effects of what you intend to integrate next.

On Comments and the Code that Binds Them

Comments and code should be like DNA — two strings that reflect each other.

This reflection creates a form of internal consistency and safety. When code and its comments are in alignment, they mutually verify each other. When they diverge, the inconsistency acts as a warning sign: something is broken, outdated, or misunderstood.

Treat comments not as annotations, but as the complementary strand of the code itself. Keep them synchronized. A mismatch is not a small issue — it's a mutation worth investigating.

The Holy Hand Grenade of Antioch Procedure

If there is not only one good way to do it, then you should have three.

Five is right out.

One way implies clarity. Two implies division. Three implies depth. Five implies confusion, and confusion leads to bugs. When offering choices — in interface, design, or abstraction — ensure there are no more than three strong forms. The third may be unexpected, but it must still be necessary.

Beyond that, you're just multiplying uncertainty. This same principle applies to other aspects of coding. A simple function fits a single IDE screen. A complex one may span three. Five means: refactor this.

License

MIT License