sitemix 0.2.6

Opinionated CLI wrapper around trafilatura for LLM-oriented webpage/site dumps.

sitemix

sitemix is a small, opinionated CLI that turns a webpage or a small website into a single LLM-oriented dump file.

Core extraction is powered by trafilatura.

License and dependency note

• sitemix code: MIT (see LICENSE)
• trafilatura: Apache 2.0
• sitemix is an opinionated wrapper around trafilatura; MIT applies only to the sitemix codebase.

What it does

• sitemix page URL: extract one page into Markdown (default), JSON, or XML.
• sitemix site URL: crawl a small site politely and produce one unified dump file.
• Outputs are deterministic and structured for RAG/LLM ingestion.

Quickstart

python -m venv .venv
source .venv/bin/activate
pip install -e .
sitemix --help

pipx usage:

pipx install sitemix

CLI examples

Single page:

sitemix page "https://example.com/post" \
  --format md \
  --min-text-chars 400

Site crawl:

sitemix site "https://example.com" \
  --max-pages 200 \
  --delay 1.0 \
  --concurrency 2 \
  --format md

URLs from stdin:

printf '%s\n' "https://example.com/a" "https://example.com/b" | sitemix site "https://example.com" --no-sitemap

Sitemap path or URL:

sitemix site "https://example.com" --sitemap ./urls.txt
sitemix site "https://example.com" --sitemap https://example.com/sitemap.xml

Write to stdout:

sitemix page "https://example.com" --stdout

Politeness defaults

• Robots respected by default (--ignore-robots to override)
• Delay + jitter between requests
• Same-host crawl by default (--include-external to override)
• Query-heavy URLs skipped by default (--allow-query-heavy to override)
• Basic binary URL skipping

See docs/politeness.md.

Output formats

• Markdown: strict delimiters for each page (--- SITEMIX_PAGE ---, --- SITEMIX_TEXT ---, --- END_SITEMIX_PAGE ---)
• JSON: single object with tool/run metadata and pages
• XML: single <sitemixDump> root

See docs/format.md.

Development commands

pip install -e .[dev]
ruff check .
pytest

Homebrew distribution

A formula file lives at packaging/homebrew/sitemix.rb and is updated by release automation.

Automated release flow:

1. Bump project.version in pyproject.toml and merge to main.
2. .github/workflows/release-pypi.yml publishes to PyPI (trusted publishing).
3. .github/workflows/release-homebrew.yml renders a formula from PyPI metadata and updates shaypal5/homebrew-tap.

Required setup:

• GitHub Environment pypi configured for trusted publishing.
• Repository secret HOMEBREW_TAP_GITHUB_TOKEN with push access to shaypal5/homebrew-tap.

Manual maintenance command:

python scripts/render_homebrew_formula.py --version X.Y.Z

Install with:

brew tap shaypal5/tap
brew install sitemix

Limitations

• Not a massive web crawler; intended for small sites (hundreds of pages, not millions).
• Extraction quality depends on source HTML and trafilatura behavior.
• JS-heavy pages without server-rendered content may produce short output.

Ethics

Use only where you have legal and ethical right to fetch and process content. Respect robots, rate limits, and site terms.