Airium

Airium is the most natural, pure-Python HTML generator with no runtime dependencies. It lets you describe an HTML document with ordinary Python code, without introducing a template language or a separate template file. The result is an HTML string or UTF-8 encoded bytes that can be returned from a web endpoint, written to a file, or passed to another system.

Airium is designed around these principles:

• Natural Python syntax: use with blocks for nesting, function arguments for attributes, and normal Python control flow for conditions and loops.
• A 1:1 representation of HTML structure: Python indentation and context managers mirror the resulting DOM tree, so the code is easy to read and refactor.
• No templates required: keep the view logic and its markup together in one Python module instead of switching between Python and a template language.
• Precise IDE assistance: the package includes generated typing stubs for standard HTML elements and their attributes, while retaining a dynamic escape hatch for custom HTML.
• Bidirectional translation: convert existing HTML back into starter Python code with the reverse translator.
• Readable output by default: generate pretty-printed HTML, or choose minified output when the document size matters.

Airium intentionally focuses on generating HTML, not validating it. It does not check whether a tag is valid HTML, whether an attribute belongs to a tag, or whether the final tree satisfies browser standards. This keeps the API small and flexible, and means that mistakes are passed through to the output instead of being silently rewritten. Validate the generated document separately when validation is important.

Key features:

• simple and straightforward
• template-less: write HTML directly in Python
• DOM structure represented by Python indentation and context managers
• reverse translator from HTML to Python
• pretty (default) or minified HTML output

Type checking

Airium ships a py.typed marker and public HTML typing helpers. Projects using Airium can run mypy normally after installing the package:

python -m pip install mypy airium
python -m mypy your_project

TagName describes standard HTML element names and HTMLAttributes provides common attribute-name completion while still allowing Airium's dynamic get_tag_ escape hatch for custom elements and attributes.

Standard tags with tag-specific stubs provide keyword completion in supported IDEs. For example, a.img( exposes image attributes such as src, alt, width, height, loading, and srcset. The img element is an HTML void element, so it must be called directly rather than used as with a.img():.

The element and attribute signatures are generated by tools/generate_html_stubs.py. Run it after updating the HTML vocabulary:

python tools/generate_html_stubs.py

The generated signatures include global HTML attributes and the standard element-specific attributes for every supported HTML element. aria-*, data-*, custom attributes, and custom elements remain available through Airium's dynamic keyword and get_tag_ escape hatches.

How it works

An Airium instance is both a document builder and a callable text writer. Accessing an attribute on it, such as a.div or a.custom_element, invokes Airium.__getattr__ and creates a tag factory dynamically. There is no fixed runtime list of allowed element names: any attribute name can be used as a tag name. The generated factory accepts positional attribute fragments and keyword attributes.

Use a context manager when an element has children:

from airium import Airium

a = Airium()
with a.section(id="intro", **{"data-page": "home"}):
    a.h1(_t="Welcome")
    a("Text can also be written directly.")

This produces:

<section id="intro" data-page="home">
  <h1>Welcome</h1>
  Text can also be written directly.
</section>

Use _t for text content and klass (or class_) for the HTML class attribute, because class is a Python keyword:

a.input(type="email", required=True)
a.div(klass="card", _t="A card")

Void elements such as img, input, br, and meta are called directly and must not be used as context managers. Python expressions work naturally around the markup, so loops and conditions can generate repeated or optional elements.

The dynamic API also means Airium can emit custom elements and attributes:

with a.my_widget(**{"data-state": "ready", "aria-label": "Status"}):
    a.span(_t="Ready")

would produce:

<my_widget data-state="ready" aria-label="Status">
  <span>Ready</span>
</my_widget>

For standard elements, the bundled .pyi stubs add exact tag and attribute completion in supported IDEs and help static type checkers catch spelling and value errors. Dynamic access remains available when the HTML vocabulary is custom or ahead of the generated stubs.

Generating HTML code in python using airium

Basic HTML page (hello world)

from airium import Airium

a = Airium()

a('<!DOCTYPE html>')
with a.html(lang="pl"):
    with a.head():
        a.meta(charset="utf-8")
        a.title(_t="Airium example")

    with a.body():
        with a.h3(id="id23409231", klass='main_header'):
            a("Hello World.")

html = str(a)  # casting to string extracts the value
# or directly to UTF-8 encoded bytes:
html_bytes = bytes(a)  # casting to bytes is a shortcut to str(a).encode('utf-8')

print(html)

Prints such a string:

<!DOCTYPE html>
<html lang="pl">
  <head>
    <meta charset="utf-8" />
    <title>Airium example</title>
  </head>
  <body>
    <h3 id="id23409231" class="main_header">
      Hello World.
    </h3>
  </body>
</html>

In order to store it as a file, just:

with open('that/file/path.html', 'wb') as f:
    f.write(bytes(html))

Simple image in a div

from airium import Airium

a = Airium()

with a.div():
    a.img(src='source.png', alt='alt text')
    a('the text')

html_str = str(a)
print(html_str)

<div>
    <img src="source.png" alt="alt text"/>
    the text
</div>

Table

from airium import Airium

a = Airium()

with a.table(id='table_372'):
    with a.tr(klass='header_row'):
        a.th(_t='no.')
        a.th(_t='Firstname')
        a.th(_t='Lastname')

    with a.tr():
        a.td(_t='1.')
        a.td(id='jbl', _t='Jill')
        a.td(_t='Smith')  # can use _t or text

    with a.tr():
        a.td(_t='2.')
        a.td(_t='Roland', id='rmd')
        a.td(_t='Mendel')

table_str = str(a)
print(table_str)

# To store it to a file:
with open('/tmp/airium_www.example.com.py') as f:
    f.write(table_str)

Now table_str contains such a string:

<table id="table_372">
  <tr class="header_row">
    <th>no.</th>
    <th>Firstname</th>
    <th>Lastname</th>
  </tr>
  <tr>
    <td>1.</td>
    <td id="jbl">Jill</td>
    <td>Smith</td>
  </tr>
  <tr>
    <td>2.</td>
    <td id="rmd">Roland</td>
    <td>Mendel</td>
  </tr>
</table>

Chaining shortcut for elements with only one child

New in version 0.2.2

Having a structure with large number of with statements:

from airium import Airium

a = Airium()

with a.article():
    with a.table():
        with a.thead():
            with a.tr():
                a.th(_t='Column 1')
                a.th(_t='Column 2')
        with a.tbody():
            with a.tr():
                with a.td():
                    a.strong(_t='Value 1')
                a.td(_t='Value 2')

table_str = str(a)
print(table_str)

You may use a shortcut that is equivalent to:

from airium import Airium

a = Airium()

with a.article().table():
    with a.thead().tr():
        a.th(_t="Column 1")
        a.th(_t="Column 2")
    with a.tbody().tr():
        a.td().strong(_t="Value 1")
        a.td(_t="Value 2")

table_str = str(a)
print(table_str)

<article>
  <table>
    <thead>
      <tr>
        <th>Column 1</th>
        <th>Column 2</th>
      </tr>
    </thead>
    <tbody>
      <tr>
        <td>
          <strong>Value 1</strong>
        </td>
        <td>Value 2</td>
      </tr>
    </tbody>
  </table>
</article>

Options

Pretty or Minify

By default, Airium builds HTML code indented with spaces and with line breaks using line feed (\n) characters. These defaults can be changed while creating an Airium instance:

a = Airium(
    base_indent='  ',  # str
    current_level=0,  # int
    source_minify=False,  # bool
    source_line_break_character="\n",  # str
)

minify

That's a mode when size of the code is minimized, i.e. contains as less whitespaces as it's possible. The option can be enabled with source_minify argument, i.e.:

a = Airium(source_minify=True)

In case if you need to explicitly add a line break in the source code (not the <br/>):

a = Airium(source_minify=True)
a.h1(_t="Here's your table")
with a.table():
    with a.tr():
        a.break_source_line()
        a.th(_t="Cell 11")
        a.th(_t="Cell 12")
    with a.tr():
        a.break_source_line()
        a.th(_t="Cell 21")
        a.th(_t="Cell 22")
    a.break_source_line()
a.p(_t="Another content goes here")

Will result with such a code:

<h1>Here's your table</h1><table><tr>
<th>Cell 11</th><th>Cell 12</th></tr><tr>
<th>Cell 21</th><th>Cell 22</th></tr>
</table><p>Another content goes here</p>

Note that the break_source_line cannot be used in context manager chains.

indent style

The default indent of the generated HTML code has two spaces per each indent level. You can change it to \t or 4 spaces by setting Airium constructor argument, e.g.:

a = Airium(base_indent="\t")  # one tab symbol
a = Airium(base_indent="    ")  # 4 spaces per each indentation level
a = Airium(base_indent=" ")  # 1 space per one level
# pick one of the above statements, it can be mixed with other arguments

Note that this setting is ignored when source_minify argument is set to True (see above).

There is a special case when you set the base indent to empty string. It would disable indentation, but line breaks will be still added. In order to get rid of line breaks, check the source_minify argument.

indent level

The current_level integer can be set to a non-negative value. Airium then starts indentation at the given level offset.

line break character

By default, just a line feed (\n) is used for terminating lines of the generated code. You can change it to different style, e.g. \r\n or \r by setting source_line_break_character to the desired value.

a = Airium(source_line_break_character="\r\n")  # windows' style

Note that the setting has no effect when source_minify argument is set to True (see above).

Using Airium with web frameworks

Airium works with any framework that accepts a string or bytes as an HTTP response. The following FastAPI example is a complete HTML endpoint with inline CSS and JavaScript. Inline assets are useful for small pages, prototypes, and self-contained responses; larger applications can generate link and script elements pointing to static files instead.

from fastapi import FastAPI
from fastapi.responses import HTMLResponse

from airium import Airium

app = FastAPI()


@app.get("/", response_class=HTMLResponse)
def home() -> HTMLResponse:
    a = Airium()
    a("<!DOCTYPE html>")

    with a.html(lang="en"):
        with a.head():
            a.meta(charset="utf-8")
            a.meta(
                name="viewport",
                content="width=device-width, initial-scale=1",
            )
            a.title(_t="Airium + FastAPI")
            a.style(
                _t="""
                :root { font-family: system-ui, sans-serif; }
                body { margin: 2rem auto; max-width: 42rem; padding: 0 1rem; }
                button { cursor: pointer; padding: .6rem 1rem; }
                .message { color: #176b3a; }
                """,
            )

        with a.body():
            with a.main():
                a.h1(_t="Hello from FastAPI")
                a.p(
                    id="message",
                    klass="message",
                    _t="This page was generated with pure Python.",
                )
                a.button(id="toggle-message", type="button", _t="Toggle message")

            a.script(
                _t="""
                const button = document.querySelector("#toggle-message");
                const message = document.querySelector("#message");
                button.addEventListener("click", () => {
                    message.hidden = !message.hidden;
                });
                """,
            )

    return HTMLResponse(content=bytes(a))

Run it with:

pip install fastapi uvicorn airium
uvicorn main:app --reload

The endpoint returns the generated document directly; no template loader, template directory, or HTML file is needed.

Airium can also be used with frameworks like Flask or Django. It can replace template engines, reducing code-file scatter and keeping related presentation and application logic together.

Here is an example of using airium with django. It implements reusable basic_body and a view called index.

# file: your_app/views.py
import contextlib
import inspect

from airium import Airium
from django.http import HttpResponse


@contextlib.contextmanager
def basic_body(a: Airium, useful_name: str = ''):
    """Works like a Django/Ninja template."""

    a('<!DOCTYPE html>')
    with a.html(lang='en'):
        with a.head():
            a.meta(charset='utf-8')
            a.meta(content='width=device-width, initial-scale=1', name='viewport')
            # do not use CSS from this URL in a production, it's just for an educational purpose
            a.link(href='https://unpkg.com/@picocss/pico@1.4.1/css/pico.css', rel='stylesheet')
            a.title(_t=f'Hello World')

        with a.body():
            with a.div():
                with a.nav(klass='container-fluid'):
                    with a.ul():
                        with a.li():
                            with a.a(klass='contrast', href='./'):
                                a.strong(_t="⌨ Foo Bar")
                    with a.ul():
                        with a.li():
                            a.a(klass='contrast', href='#', **{'data-theme-switcher': 'auto'}, _t='Auto')
                        with a.li():
                            a.a(klass='contrast', href='#', **{'data-theme-switcher': 'light'}, _t='Light')
                        with a.li():
                            a.a(klass='contrast', href='#', **{'data-theme-switcher': 'dark'}, _t='Dark')

                with a.header(klass='container'):
                    with a.hgroup():
                        a.h1(_t=f"You're on the {useful_name}")
                        a.h2(_t="It's a page made by our automatons with a power of steam engines.")

            with a.main(klass='container'):
                yield  # This is the point where main content gets inserted

            with a.footer(klass='container'):
                with a.small():
                    margin = 'margin: auto 10px;'
                    a.span(_t='© Airium HTML generator example', style=margin)

            # do not use JS from this URL in a production, it's just for an educational purpose
            a.script(src='https://picocss.com/examples/js/minimal-theme-switcher.js')


def index(request) -> HttpResponse:
    a = Airium()
    with basic_body(a, f'main page: {request.path}'):
        with a.article():
            a.h3(_t="Hello World from Django running Airium")
            with a.p().small():
                a("This bases on ")
                with a.a(href="https://picocss.com/examples/company/"):
                    a("Pico.css / Company example")

            with a.p():
                a("Instead of a HTML template, airium has been used.")
                a("The whole body is generated by a template "
                  "and the article code looks like that:")

            with a.code().pre():
                a(inspect.getsource(index))

    return HttpResponse(bytes(a))  # from django.http import HttpResponse

Route it in urls.py just like a regular view:

# file: your_app/urls.py
from django.contrib import admin
from django.urls import path

import your_app

urlpatterns = [
    path('index/', your_app.views.index),
    path('admin/', admin.site.urls),
]

The resulting web page on my machine looks like this:

Reverse translation

Airium is equipped with a transpiler [HTML -> py]. It generates python code out of a given HTML string.

Using reverse translator as a binary:

Ensure you have installed [parse] extras. Then call in command line:

airium http://www.example.com

That will fetch the document and translate it to python code. The code calls airium statements that reproduce the HTML document given. It may give a clue - how to define HTML structure for a given web page using airium package.

To store the translation's result into a file:

airium http://www.example.com > /tmp/airium_example_com.py

You can also parse local HTML files:

airium /path/to/your_file.html > /tmp/airium_my_file.py

You may also try to parse your Django templates. I'm not sure if it works, but there will be probably not much to fix.

Using reverse translator as python code:

from airium import from_html_to_airium

# assume we have such a page given as a string:
html_str = """\
<!DOCTYPE html>
<html lang="pl">
  <head>
    <meta charset="utf-8" />
    <title>Airium example</title>
  </head>
  <body>
    <h3 id="id23409231" class="main_header">
      Hello World.
    </h3>
  </body>
</html>
"""

# to convert the html into python, just call:

py_str = from_html_to_airium(html_str)

# airium tests ensure that the result of the conversion is equal to the string:
assert py_str == """\
#!/usr/bin/env python
# File generated by reverse AIRIUM translator (version 0.3.2).
# Any change will be overridden on next run.
# flake8: noqa E501 (line too long)

from airium import Airium

a = Airium()

a('<!DOCTYPE html>')
with a.html(lang='pl'):
    with a.head():
        a.meta(charset='utf-8')
        a.title(_t='Airium example')
    with a.body():
        a.h3(klass='main_header', id='id23409231', _t='Hello World.')
"""

Transpiler limitations

so far in version 0.2.2:

• result of translation does not keep exact amount of leading whitespaces within <pre> tags. They come over-indented in python code.

This is not however an issue when code is generated from python to HTML.

• although it keeps the proper tags structure, the transpiler does not chain all the with statements, so in some cases the generated code may be much indented.

• it's not too fast

Installation

If you need a new virtual environment, call:

virtualenv venv
source venv/bin/activate

Having it activated - you may install airium like this:

pip install airium

In order to use reverse translation - two additional packages are needed, run:

pip install airium[parse]

Then check if the transpiler works by calling:

airium --help

Enjoy!