Internal link suggester for Markdown and HTML articles

What it does

Finds internal linking opportunities in a folder of blog or documentation articles. Your files are never modified.

• Builds TF-IDF vectors for every article and computes which ones are related. Japanese and Chinese text is handled with character bigrams plus katakana and kanji runs, so no morphological analyzer is needed
• Reads existing links (Markdown links and <a href>) and only suggests pairs that are not linked yet
• Offers anchor text candidates taken from the target's title, headings and key terms that already appear in the source text, and shows the sentence where the link fits
• Lists orphan pages (no inbound internal link) and the best pages to link from
• Outputs Markdown, CSV and JSON

How to use

python scripts/link_suggest.py ./content --base-url https://example.com/blog/ --out ./link-report

Ask your agent to suggest internal links for your content folder; it reviews the candidates and presents a per-article plan, and can edit only the agreed sentences if you ask.

Requirements

• Python 3 (standard library only)

Caveats

• Relatedness is based on shared vocabulary. The agent and you decide whether a link actually helps readers.
• Menu and template links are not part of the article files, so a page may show as orphan here while being linked on the live site.
• Thousands of articles take longer (400 articles took about 5 seconds in testing).