Small, dependable HTML cleaner/minifier with sensible defaults

Navigation

Verified details

These details have been verified by PyPI
Maintainers
Avatar for Nosible from gravatar.com Nosible

Unverified details

These details have not been verified by PyPI
Project links
Meta
Classifiers
Report project as malware

Project description

smol

smol-html

Small, dependable HTML cleaner/minifier with sensible defaults.

Motivation

Nosible is a search engine, which means we need to store and process a very large number of webpages. To make this tractable, we strip out visual chrome and other non-essential components that don’t matter for downstream tasks (indexing, ranking, retrieval, and LLM pipelines) while preserving the important content and structure. This package cleans and minifies HTML, greatly reducing size on disk; combined with Brotli compression (by Google), the savings are even larger.

Cleaning impact on size

📦 Installation

pip install smol-html

⚡ Installing with uv

uv pip install smol-html

Requirements

  • Python: 3.9
  • Dependencies:
    • beautifulsoup4>=4.0.1
    • brotli>=0.5.2
    • lxml[html-clean]>=1.3.2
    • minify-html>=0.2.6

Quick Start

Clean an HTML string (or page contents):

from smol_html import SmolHtmlCleaner

html = """
<html>
  <head><title> Example </title></head>
  <body>
    <div>  Hello <span> world </span> </div>
  </body>
</html>
"""

# All constructor arguments are keyword-only and optional.
cleaner = SmolHtmlCleaner()
cleaned = cleaner.make_smol(raw_html=html)

print(cleaned)

Customization

SmolHtmlCleaner exposes keyword-only parameters with practical defaults. You can:

  • Pass overrides to the constructor, or
  • Adjust attributes on the instance after creation.
from smol_html import SmolHtmlCleaner

cleaner = SmolHtmlCleaner()
cleaner.attr_stop_words.add("advert")  # e.g., add a custom stop word

Usage Examples

Minimal:

from smol_html import SmolHtmlCleaner

cleaner = SmolHtmlCleaner()
out = cleaner.make_smol(raw_html="<p>Hi <!-- note --> <a href='x'>link</a></p>")

Customize a few options:

from smol_html import SmolHtmlCleaner

cleaner = SmolHtmlCleaner(
    attr_stop_words={"nav", "advert"},
    remove_header_lists=False,
    minify=True,
)

out = cleaner.make_smol(raw_html="<p>Hi</p>")

Compressed Bytes Output

Produce compressed bytes using Brotli with make_smol_bytes.

  • By default, the compressed bytes are URL-safe Base64 encoded (base64_encode=True).
  • If you enable Base64, you must decode before Brotli-decompressing.
  • You can disable Base64 by passing base64_encode=False and decompress directly.

Default (Base64-encoded) output:

from smol_html import SmolHtmlCleaner
import base64
import brotli  # only needed if you want to decompress here in the example

html = """
<html>
  <body>
    <div>  Hello <span> world </span> </div>
  </body>
</html>
"""

cleaner = SmolHtmlCleaner()

# Get compressed bytes (quality 11 is strong compression)
compressed = cleaner.make_smol_bytes(raw_html=html, compression_level=11)

# Because Base64 is enabled by default, decode before decompressing
decoded = base64.urlsafe_b64decode(compressed)
decompressed = brotli.decompress(decoded).decode("utf-8")
print(decompressed)

# Or write Base64-encoded compressed output directly to a file
with open("page.html.br.b64", "wb") as f:
    f.write(compressed)

Disable Base64 and decompress directly:

from smol_html import SmolHtmlCleaner
import brotli

cleaner = SmolHtmlCleaner()
compressed_raw = cleaner.make_smol_bytes(
    raw_html="<p>Hi</p>",
    compression_level=11,
    base64_encode=False,
)
print(brotli.decompress(compressed_raw).decode("utf-8"))

Parameter Reference

To improve readability, the reference is split into two tables:

  • What it does and when to change
  • Types and default values

What It Does

Parameter What it does When to change
non_text_to_keep Whitelist of empty/non-text tags to preserve (e.g., images, figures, tables, line breaks). If important non-text elements are being removed or you want to keep/drop more empty tags.
attr_stop_words Tokens matched against id/class/role/item_type on small elements; matches are removed as likely non-content. Add tokens like advert, hero, menu to aggressively drop UI chrome, or remove tokens if content is lost.
remove_header_lists Removes links/lists/images within <header> to reduce nav clutter. Set False if your header contains meaningful content you want to keep.
remove_footer_lists Removes links/lists/images within <footer> to reduce boilerplate. Set False for content-heavy footers you need.
minify Minifies output HTML using minify_html. Set False for readability or debugging; use --pretty in the CLI.
minify_kwargs Extra options passed to minify_html.minify. Tune minification behavior (e.g., whitespace, comments) without changing cleaning.
meta lxml Cleaner option: remove <meta> content when True. Usually leave False; enable only for strict sanitation.
page_structure lxml Cleaner option: remove page-structure tags (e.g., <head>, <body>) when True. Rarely needed; keep False to preserve structure.
links lxml Cleaner option: sanitize/clean links. Leave True unless you need raw anchors untouched.
scripts lxml Cleaner option: remove <script> tags when True. Keep False to preserve scripts; usually safe to remove via javascript=True anyway.
javascript lxml Cleaner option: remove JS and event handlers. Set False only if you truly need inline JS (not recommended).
comments lxml Cleaner option: remove HTML comments. Set False to retain comments for debugging.
style lxml Cleaner option: remove CSS and style attributes. Set False to keep inline styles/CSS.
processing_instructions lxml Cleaner option: remove processing instructions. Rarely change; keep for safety.
embedded lxml Cleaner option: remove embedded content (e.g., <embed>, <object>). Set False to keep embedded media.
frames lxml Cleaner option: remove frames/iframes. Set False if iframes contain needed content.
forms lxml Cleaner option: remove form elements. Set False if you need to keep forms/inputs.
annoying_tags lxml Cleaner option: remove tags considered "annoying" by lxml (e.g., <blink>, <marquee>). Rarely change.
kill_tags Additional explicit tags to remove entirely. Add site-specific or custom tags to drop.
remove_unknown_tags lxml Cleaner option: drop unknown/invalid tags. Set False if you rely on custom elements.
safe_attrs_only Only allow attributes listed in safe_attrs. Set False if you need to keep arbitrary attributes.
safe_attrs Allowed HTML attributes when safe_attrs_only=True. Extend to keep additional attributes you trust.

Types and Defaults

Parameter Type Default
non_text_to_keep set[str] media/meta/table/br tags
attr_stop_words set[str] common UI/navigation tokens
remove_header_lists bool True
remove_footer_lists bool True
minify bool True
minify_kwargs dict {}
meta bool False
page_structure bool False
links bool True
scripts bool False
javascript bool True
comments bool True
style bool True
processing_instructions bool True
embedded bool True
frames bool True
forms bool True
annoying_tags bool True
kill_tags set[str] &#124; None None
remove_unknown_tags bool True
safe_attrs_only bool True
safe_attrs set[str] curated set

make_smol_bytes Options

Parameter Type Default
compression_level int 4
base64_encode bool True

Project details

Verified details

These details have been verified by PyPI
Maintainers
Avatar for Nosible from gravatar.com Nosible

Unverified details

These details have not been verified by PyPI
Project links
Meta
Classifiers

Release history Release notifications | RSS feed

0.2.0

0.1.7

This version

0.1.6

0.1.5

0.1.4

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

smol_html-0.1.6.tar.gz (2.2 MB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

smol_html-0.1.6-py3-none-any.whl (4.9 kB view details)

Uploaded Python 3

File details

Details for the file smol_html-0.1.6.tar.gz.

File metadata

  • Download URL: smol_html-0.1.6.tar.gz
  • Upload date:
  • Size: 2.2 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.11

File hashes

Hashes for smol_html-0.1.6.tar.gz
Algorithm Hash digest
SHA256 5499a615fe083a0b9938dbca3cf34ac05d84647f3cf9a4a157a936a284ad9d16
MD5 7cfb93a62cf4c1b8c55ed2e57c22e642
BLAKE2b-256 fe1411a116d897057d9622ceaf3d54a95f8eb796a44d1a486fff29d151aef768

See more details on using hashes here.

File details

Details for the file smol_html-0.1.6-py3-none-any.whl.

File metadata

  • Download URL: smol_html-0.1.6-py3-none-any.whl
  • Upload date:
  • Size: 4.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.11

File hashes

Hashes for smol_html-0.1.6-py3-none-any.whl
Algorithm Hash digest
SHA256 dd8acd8daf26ce11239cae8356f1ee520c15e444078be11f8e13d84c01f61e32
MD5 e7f247ba5a4aff50cd69681ff1f031be
BLAKE2b-256 c4341499895c2cee75b9b1a48e4f8ee74c7d34f9ce19df4cb1ab531bc48b0437

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page