bookmarkly

organize your messy browser bookmarks using local AI.

i had 800+ bookmarks spread across random folders and the bookmarks bar. this tool reads your browser's bookmark export, figures out what each site is, and spits out a clean organized file you can import back. everything runs locally — your bookmarks never leave your machine.

before

after

terminal output

how it works

1. domain matching — a lookup table of ~200 popular sites handles the bulk instantly (github -> Dev & Code, youtube -> Video & Music, etc)
2. ollama AI — anything the table doesnt recognize gets batched and sent to a local ollama model for classification
3. dedup + merge — strips duplicate URLs and collapses tiny folders so you end up with a clean set of ~10-14 categories

most bookmark collections are 70-80% well-known sites, so the domain matcher does the heavy lifting and ollama only handles the long tail.

quickstart

# 1. export bookmarks from your browser
#    chrome:  chrome://bookmarks > ... > Export bookmarks
#    firefox: Ctrl+Shift+O > Import and Backup > Export Bookmarks to HTML
#    edge:    edge://favorites > ... > Export favorites

# 2. make sure ollama is running (skip if using --no-ai)
ollama serve
ollama pull llama3.2

# 3. run it
python bookmarkly.py -i bookmarks.html

# 4. import the output back into your browser
#    output file will be bookmarks_organized.html

usage

python bookmarkly.py -i bookmarks.html                          # basic
python bookmarkly.py -i bookmarks.html -o clean.html            # custom output
python bookmarkly.py -i bookmarks.html -m mistral               # different model
python bookmarkly.py -i bookmarks.html --max-folders 10         # limit folders
python bookmarkly.py -i bookmarks.html --no-ai --preview        # domain-only, dry run
python bookmarkly.py -i bookmarks.html --detail                 # show bookmarks per folder

flags

flag	default	what it does
-i / --input	(required)	path to exported bookmarks html
-o / --output	auto	output file path
-m / --model	llama3.2	which ollama model to use
--max-folders	14	cap on number of bookmark folders
--batch-size	15	bookmarks per ollama request
--no-ai	off	skip ollama, domain matching only
--preview	off	show results without writing file
--detail	off	list bookmarks in each folder

folder categories

sorts into these folders (intentionally under 15):

• Dev & Code — github, stackoverflow, package managers, cloud platforms
• News & Blogs — news sites, medium, substack
• Social — twitter, reddit, discord, linkedin
• Video & Music — youtube, netflix, spotify, gaming
• Shopping — amazon, ebay, etsy, retail
• Money & Finance — banking, investing, crypto
• Learning — coursera, udemy, arxiv, universities
• Work & Productivity — notion, slack, google workspace, AI tools
• Tech & Science — tech news, science sites
• Health — medical sites, fitness trackers
• Travel — maps, booking, airlines
• Design & Art — figma, canva, dribbble, stock photos
• Reference & Docs — wikipedia, MDN, documentation
• Misc — everything else

requirements

• python 3.8+
• ollama (optional but recommended)

no pip dependencies. stdlib only.

tested with

• chrome, firefox, edge, brave bookmark exports
• ollama models: llama3.2, llama3.1, mistral, phi3, gemma2
• bookmark collections from 50 to 2000+ items

license

MIT