Organizing

Every link has a numeric #id (shown on each result from /recent and /find). The commands below act on a link by its id.

Favorites

/star <id>        — toggle the ⭐ on a link

Collections

Group links under a free-form name:

/collection <id> <name>     — add link to a collection (e.g. /collection 12 rust)
/collection <id>            — clear its collection (omit the name)

Lists

Filter your library without a keyword search:

/favorites      — your starred links
/in <name>      — everything in a collection (e.g. /in rust)

These reuse the same result cards as /find, so each still shows its #id, category, and tags.

Categories

Every link gets an AI category on save (it shows on each result card). The category is reuse-or-coin: the model reuses a category already in your library when one fits and only coins a new short lowercase one when nothing does, so the taxonomy stays consistent as it grows.

/recategorize     — re-derive every link's category from its stored text

• /recategorize backfills the smarter taxonomy across your whole library. It works from each link’s stored text, so it never re-fetches the web. Run it once after categories get smarter to tidy older saves.

Categories are surfaced in the dashboard, where you can filter by category (and edit a link’s category inline).

Auto-collections & duplicates

Two commands use the semantic embeddings to tidy your library:

/organize     — cluster similar links and assign suggested collection names
/dupes        — list near-duplicate / heavily-overlapping links to review

• /organize groups links by similarity, has the model name each group (e.g. password-vaults, dev-interviews, cs-fundamentals), and assigns that collection to the uncollected links in each group — it never overwrites a collection you set. Reversible: change any via the command or the dashboard. View a group with /in <name>.
• /dupes lists pairs of links that look like duplicates or strong overlaps (e.g. two Bitwarden servers) with a similarity %. It only reports — you decide whether to delete one (in the dashboard). Exact-URL duplicates are already blocked at save time.

Collections fill in automatically

You don’t have to run /organize to get collections. When you save a link, it’s auto-filed into the nearest matching collection by semantic similarity (the same vectors and threshold /organize uses) — but only if it isn’t already in one, so a collection you set by hand is never overridden. /organize is still there for bulk grouping after the fact.

Sharing — your public List

Most of your library is private. Mark individual links public to publish them to List, a clean public page of your hand-picked links:

/share <id>     — toggle a link onto (or off) your List

List lives at /list on your Worker — e.g. https://linkbrain.<you>.workers.dev/list — grouped by collection. A link shows 🌐 on its result card when it’s public. Everything else stays private; only links you explicitly /share are ever served.

Export

Get your whole library as a file, sent straight to your Telegram chat:

/export          — Markdown (default)
/export csv      — CSV (spreadsheet-friendly)
/export json     — JSON (full structured data)

The export covers every link (private ones too) with its url, title, summary, tags, note, category, and collection.