Important

The source code for this repository was generated by claude.ai [Sonnet4.6] **This is strictly to review codes generated by AI providers, which may follow with modifications in the future.

Warning

This project is provided as-is without warranty of any kind. Review the code before use in any environment you care about. Use at your own risk.

NoteSearch

A zero-dependency fuzzy file search engine for your notes and code snippets.
Built with pure Node.js — no npm install required.

Quick Start

# Basic: search your current directory
node server.js

# Point at your notes folder
node server.js --dir ~/notes

# Custom port + specific folder
node server.js --dir ~/Documents/notes --port 8080

# Only search certain file types
node server.js --dir ~/notes --ext md,txt,py

Then open http://localhost:3131 in your browser.

---

Features

• Real-time fuzzy search as you type (80ms debounce)
• Quoted phrase search — wrap terms in "quotes" for exact phrase matching
• Mixed search — combine quoted phrases with fuzzy tokens in a single query
• Exact + fuzzy matching with smart scoring
  • Quoted phrases must match exactly — non-matching lines are excluded entirely
  • Exact substring matches score highest among unquoted tokens
  • Fuzzy matches score by consecutive character runs
  • Filename matches are weighted 1.5× over content matches
• Inline code snippets with 3 lines of context above/below each match
• Syntax-aware color-coded file type badges (JS, Python, Shell, Markdown, etc.)
• File type filters auto-generated from your index
• Auto re-index every 30 seconds + watches for file changes
• Offline font support — run download-fonts.js once to serve fonts locally
• Copy path button on each result for quick access in your editor
• Zero dependencies — pure Node.js built-ins only

---

Offline Fonts Setup

By default, NoteSearch loads fonts from Google Fonts CDN. To make it fully offline, run the font downloader once while you have internet access:

node download-fonts.js

This fetches all WOFF2 font files into a fonts/ directory and writes a fonts/fonts.css file with local paths. From that point on, server.js automatically detects and serves fonts locally — no server restart needed.

The startup log will confirm which mode is active:

Fonts: local (offline ready)
# or
Fonts: Google Fonts CDN — run node download-fonts.js to go offline

For your git repository, add the binary font files to .gitignore but commit fonts/fonts.css so collaborators know the setup has been done. Anyone cloning the repo just runs node download-fonts.js once.

fonts/*.woff2
fonts/*.woff
fonts/*.ttf

---

Search Syntax

Syntax	Behavior
docker run	Fuzzy match both words independently
"docker run"	Exact phrase — only lines containing docker run literally
"docker run" container	Exact phrase required + fuzzy boost for container
"useState" "useEffect"	Both phrases must appear on the same line

Quoted phrases are hard requirements — a file with no exact match for a quoted term will not appear in results at all, regardless of other tokens.

---

Command Line Options

Flag	Default	Description
--dir	$NOTES_DIR or cwd	Directory to search
--port	3131	Port to serve on
--ext	See below	Comma-separated file extensions

Default extensions:

md, txt, js, ts, py, sh, bash, json, yaml, yml, env, conf,
cfg, ini, toml, rb, go, rs, c, cpp, h, java, cs, php, html,
css, sql, r, lua, vim, zsh, fish

---

Environment Variable

Set NOTES_DIR to avoid typing --dir every time:

# In your .bashrc / .zshrc
export NOTES_DIR=~/notes

# Then just run:
node server.js

---

API Endpoints

Endpoint	Description
GET /	The search UI
GET /api/search?q=<query>	JSON search results
GET /api/stats	Index stats + extension breakdown
GET /api/reindex	Trigger a manual re-index
GET /fonts/<file>	Serves local font files (after running download-fonts.js)

---

How Scoring Works

1. Quoted phrase (score ~110–120): Exact substring required
   • Disqualifies the line entirely if the phrase is not found
   • Bonus if the match starts at a word boundary
2. Exact token match (score ~110–120): Unquoted token found as substring
   • Same bonus for word-boundary alignment
3. Fuzzy token match (score 1–40): All characters found in order
   • Consecutive character runs add bonus points
   • Higher ratio of matched chars to text length = higher score
4. Filename matches are multiplied by 1.5× over content matches
5. Nearby duplicate matches within 5 lines are de-duplicated (highest score kept)
6. Results sorted by score, top 50 shown

---

Live Re-indexing

The server keeps your index fresh automatically:

• File watcher (fs.watch) triggers a re-index within 500ms of any change
• Polling fallback re-indexes every 30 seconds regardless (useful on Linux where recursive fs.watch has limited platform support)
• Files over 1MB are skipped to keep search fast

---

Tips

• Short queries work well — arr sort finds files about array sorting
• Paste partial snippets — the first few chars of a code snippet are enough
• Pin with quotes — "function fetchUser" zeroes in on exact definitions
• Mix and match — "git stash" apply branch finds stash-related lines and ranks higher those mentioning apply or branch
• Use type filters — click .py or .sh to narrow results to one language
• ESC clears the search box instantly