Fragment Library

A Chrome extension for capturing text selections while reading online. Semantic search over everything you've saved. Local storage via a companion daemon.

Type Chrome Extension + Server
Search OpenAI Embeddings
Storage Local SQLite
Status Released · 2026
01 — Interface
chrome-extension://fragment-library/popup.html
Fragment Library
6 fragments (active)
Search semantically...
View
Active
Is Particle Physics Dead, Dying, or Just Hard? | Quanta Magazine active
Export Archive Delete
s had spent billions of euros building the 27-kilometer supercollider not only to confirm the Standard Model but also to supersede it by uncovering components of a more complete theory of nature.
quantamagazine.org/is-particle-physics-dead-dying-or-just-hard
2026-01-29T22:07:13
How to fix your entire life in 1 day — by DAN KOE active
Export Archive Delete
Goals determine how you see the world. Goals determine what you consider "success" or "failure."
letters.thedankoe.com/p/how-to-fix-your-entire-life-in-1
2026-01-18T21:16:41
02 — How it works
01
Select text

Highlight any passage on any page. Press CtrlShiftK (or K on Mac).

02
Embed & store

The fragment is sent to your local daemon, embedded via OpenAI, and stored in SQLite. Source URL and timestamp are attached automatically.

03
Search & export

Open the popup and search semantically — by meaning, not keywords. Export fragments to Markdown or directly into Obsidian.

03 — Architecture
Client
Chrome Extension
Capture, browse, search, export UI
Daemon
fragmentd
FastAPI · SQLite · REST API
Embeddings
OpenAI
Vector similarity search

The daemon runs locally or on a home server. Your OpenAI key stays on the server — never in the extension. All fragments are stored in a single SQLite file you can back up and own. The embedding provider is swappable by design.

04 — Details

Why

I kept reading things worth remembering and then losing them. Bookmarks don't capture the specific passage. Notes apps require too much friction. This captures the exact text with one keystroke and makes it searchable by meaning.

Privacy

Fragment text is sent to your own daemon (localhost or home server) and to OpenAI for embedding only. No other services. No analytics. The SQLite database is yours to keep, back up, or delete.

Obsidian export

Fragments can be exported as Markdown notes directly into an Obsidian vault. Supports both daily digest mode (inbox triage) and curated export (keepers only). Designed for a capture-first, curate-later workflow.

Requirements

Chrome (or Chromium). Python 3.10+ for the daemon. An OpenAI API key for embeddings (or use the local stub for offline testing).

05 — Install

Get Fragment Library

Chrome extension + fragmentd daemon. Clone, configure, and run — takes a few minutes.