How Search Works

How search works under the hood

The honest explainer of what happens between typing a query and getting files.

1. The engines

Ten trackers are queried in parallel:

SolidTorrents · The Pirate Bay (real API, not a scraper) · 1337x · TorrentsCSV · Nyaa · BitSearch · BitMusic (music-only) · Torlock · LimeTorrents · Archive.org

The status bar shows per-engine counts, so you can see which tracker is having a mood that day without digging through logs.

2. Deduplication

Torrents are matched by infohash. The same release on three sites collapses into a single row labeled like TPB + 1337x, keeping the highest-seeded copy. No more five near-identical rows for one movie.

3. Ranking

Smart sort ranks by relevance first, seeders second. The point:

ubuntu 24.04 should give you actual ISOs — not "unrelated spam • 9999 seeders".

Those seed counts are optimistic, especially on Torlock, which is exactly why the ranking doesn't trust them blindly.

4. Bait filter

Before you ever see a result, known spam signals are dropped:

• [REAL]
• Full Version
• Direct Download!!1
• and friends

Yes, it's personal. No, there's no setting to turn it off.

5. Category badges

Icons (🌐 📦 🎵 🎬 …) are guessed from the filename, not from metadata — so they're wrong sometimes. That's fine; they're a hint, not a promise.

6. The Real-Debrid step

If a release is already in RD's cache, it shows a ⚡ instant badge before you click. c flips to cached-only mode. The pipeline for anything you pick:

magnet → added to RD → cached/ready → unrestricted links → download

Your IP talks to the trackers (search) and to RD (everything heavy). The actual file data comes from RD's CDN — never from the swarm.

7. The local database

Everything worth remembering lives in ~/.odyssey/settings.db:

• encrypted RD token
• search history
• download log
• blacklist
• result cache
• preferences

It's outside the repo on purpose, so nothing sensitive ever reaches GitHub.