Gextto user manual¶

This manual explains how to use Gextto: from installation to the first download, from the series you follow every week to looking after the library. It is organised by task — add a series, find out why an episode did not arrive, make a backup — rather than by page of the interface. The full list of every setting, field and button is in the appendices at the end.

The UI is available in Italian, English, German, French, Spanish and Polish; this manual uses the English labels and is also the manual for German, French, Spanish and Polish. The Italian version is MANUAL.it.md. The Manual entry in the interface opens this document online, in the chosen language; Changelog opens the release notes. Neither is built into the program: they are updated without updating Gextto.

[!IMPORTANT] Start in test mode (dry-run): Gextto searches and decides, but downloads nothing. Check paths, sources and one test title before switching to real downloads. The web interface is an administration panel and is open by default: keep it on a trusted network or, for access from outside, turn on the login and use HTTPS (see Access and security).

Contents¶

Part I — Getting started

What Gextto is and how it works
Installing Gextto
The initial setup
The first download, in test mode and then for real
Finding your way around the interface

Part II — Everyday use

Following a TV series
Following a movie
Searching and downloading by hand
Watching the downloads
Where the files end up
Comics
eD2k and Kad with aMule

Part III — Configuring

Sources: feeds and indexers
Quality, scores and upgrades
The torrent engine
Speed and network
Library, NAS and space
Notifications
Integrations with other programs
Access and security

Part IV — Keeping everything in order

Checking that everything works
Looking after the library
Backup and restore
Updating Gextto
Using Gextto from the terminal (TUI)

Part V — When something goes wrong

Troubleshooting
Glossary

Appendices — Reference

A. Configuration, setting by setting
B. Integrations
C. Maintenance

Part I — Getting started¶

1. What Gextto is and how it works¶

Gextto is a single service that runs on an always-on computer (a mini PC, a server, a NAS): you tell it which series and movies you are interested in, and it searches for them, picks the best version according to your rules, downloads it, gives it a tidy name and puts it in your library. You control it from the browser (http://<host>:5000) or from the terminal. No other components are needed: the torrent engine, gx-torrent, is included and Gextto starts it by itself.

The path of an episode¶

Every download follows the same road. Knowing it tells you where to look when something goes wrong.

Followed title. You add a series or a movie to the library, with a quality profile (resolution, language, subtitles, audio).
Search cycle. At regular intervals Gextto reads the feeds and queries the indexers (Jackett, Prowlarr, indexarr…) for the titles you follow. Each result is a release.
Choice. Every release goes through a series of checks: is it the right title? does it meet quality and language? is an equal or better copy already in the library? Among the valid ones, the release with the highest score wins.
Download. The chosen release goes to the torrent engine (or to aMule, for eD2k). The Download page shows its progress.
Into the library. When the download finishes, Gextto renames the file with your template and copies, moves or links it into the series or movie folder. If the library held a worse version, that one goes to the trash.
Seeding. The torrent keeps sharing until it reaches the seeding limits, then it leaves the session.

The words you will use¶

Word	What it means
Release	a result found by a feed, an indexer or a web engine
Cycle	one automatic pass of searching and choosing
Test mode (dry-run)	Gextto searches and decides, but starts no download
Torrent session	the torrents the engine is still handling (downloading or seeding)
Library	the final, tidy folders with series and movies (often on a NAS)
Archive	the local catalogue of the releases seen in the feeds, searchable
Trash	where replaced or removed files go, before they are deleted for good
Upgrade	replacing a file in the library with a better version

Keep these four sets apart: torrent session, library, archive and trash. Cleaning one does not touch the others: removing a torrent from the session does not delete the file from the library, and emptying the archive does not delete any video.

2. Installing Gextto¶

You need Linux with systemd, on x86_64 or aarch64. Gextto is written in pure Go and compiles nothing on the server.

With the installer (recommended)¶

curl -fsSL https://raw.githubusercontent.com/buzzqw/gextto/main/install.sh | sudo bash

The installer downloads the latest published version, verifies its SHA-256 checksum and installs it:

the program in /opt/gextto, the data in /var/lib/gextto;
the gextto.service service, started right away;
the base packages under the right names for Debian, Ubuntu, Fedora, openSUSE and Arch and, when the distribution provides them, ffmpeg, mediainfo and mkvtoolnix (used to read the real quality of the files, check their integrity and clean the audio tracks). --no-media-tools skips them.

The service runs as the user who runs the installer, so it can write into your media folders. For a separate account use --user NAME, or --user gextto for the isolated system account without login (with --media-group to give it access to other users' media).

Other useful options:

Option	Effect
`--channel stable` / `--release <tag>`	installs the stable version or a specific release instead of the latest
`--local-archive <file>`	installs from an archive already downloaded, without network
`--no-media-tools`	does not install ffmpeg, mediainfo and mkvtoolnix
`GEXTTO_REPO`, `GEXTTO_RELEASE`	choose another repository or another release

An optional /etc/gextto/gextto.env file, when present, is loaded as the service's environment variables.

Without root, from a checkout¶

make build
GEXTTO_DATA_DIR="$HOME/gextto-data" \
  GEXTTO_LISTEN=127.0.0.1:5000 \
  scripts/install-user-service.sh
systemctl --user status gextto.service

This is a user systemd service. If it must stay on after you log out, run loginctl enable-linger "$USER".

With Docker¶

Gextto, gx-torrent, Jackett, Prowlarr, FlareSolverr and aMule together in one composition: see Docker: full stack.

Checking that it runs¶

Open http://<host>:5000. If the page does not open:

systemctl status gextto            # system installation
systemctl --user status gextto     # user service
/opt/gextto/gexttod --version      # installed version

Gextto accepts one instance per data folder: if another Gextto already uses the same folder, or port 5000 is taken, the new process stops at once and says so in the log, without touching the databases.

3. The initial setup¶

The first time you open the interface the Initial setup starts, in five steps. You can skip it and reopen it whenever you like from /?view=setup; it no longer appears on an installation that already has titles or an archive.

Access. Choose a password. You can leave access from the home network free and ask for it only from outside.
Folders. Set library, downloads, temporary folder and trash. Browse opens the server's folders (and creates new ones); Verify checks, as the service user, that each folder exists and is writable, and shows the disk type (NAS included), the free space and whether it sits on the same disk as the library.
Sources. An indexer (Prowlarr, Jackett or indexarr) and the TMDB key, recommended, for titles, posters and episodes. The TVDB key, with its PIN if you have one, also works as an alternative.
First title. Search for a series or a movie and add it.
Activate. Turn on the automatic cycle and choose whether to leave test mode (Really download). For the first round, keep test mode on.

The folders and their role¶

Folder	What it is for	Notes
Downloads	the data of the torrents in the session	do not delete it by hand while a torrent is active
Temporary	incomplete data and metadata	must be writable; can be on a fast disk
Library	the final files, sorted by series and movie	usually on the NAS
Trash	replaced or removed files	emptied according to the retention
Watched folders	`.torrent` and `.magnet` files to import	optional

Two tips that avoid most problems:

Test the permissions as the service user, not as yourself. A folder you can see from your shell may not be visible to the systemd service.
Put downloads and library on the same filesystem if you can: Gextto can then use hardlinks, and a file still seeding enters the library at once without taking space twice (see Where the files end up).

4. The first download, in test mode and then for real¶

Make one change at a time: if you change paths, quality and sources together, a problem becomes hard to reproduce.

Check the folders in Configuration → Archive and space and in Configuration → Library and renaming.
Add a single source in Configuration → Sources and press Verify. Add the others only once this one works (see Sources).
Add a test title: a series with a single season, or a movie.
Start a cycle from the Dashboard (Automatic search, only the domain you care about: Series, Movies or Comics).
Read the Log: you should see the cycle, the sources queried and, for every release, why it was accepted or rejected.
Run a manual search for the same title and open Why not this one? on a result (see Searching and downloading by hand).
Switch to real downloads: reopen the last step of the initial setup (/?view=setup&step=5; the DRY-RUN badge at the top reminds you that you are in test mode), tick Really download and press Complete. It takes effect at once, without a restart; in the TUI the same switch is under Settings. A Gextto started with --dry-run on the command line stays in test mode until it is restarted.

Checklist before real downloads¶

[ ] http://<host>:5000 opens;
[ ] Health reports no permission or space problems;
[ ] the temporary folder is writable;
[ ] the library (or NAS) is mounted and writable by the service user;
[ ] at least one source answers Verify;
[ ] a manual search returns sensible releases;
[ ] the test-mode cycle produced no unexpected errors.

5. Finding your way around the interface¶

Group	Pages	What they are for
Overview	Dashboard	the state of everything at a glance, search, cycles
Download	Download, Log	the transfers in progress; the live log
Library	TV Series, Movies	the titles you follow, their episodes and what is missing
Discovery	Explore, Archive, Comics, eD2k	finding new titles, browsing the releases seen, comics, aMule
System	Configuration, Integrations, Maintenance, Health, Blocklist, Manual, Changelog	settings, connected services, clean-ups, diagnosis

The top bar¶

Add torrent, on every page: opens the form for a magnet, a URL or a .torrent file (see Adding a torrent or a magnet).
Refresh: reloads the page's data at once.
The download and upload speeds (torrents plus aMule) and CPU and RAM per process: Gextto, gx-torrent and, when active and local, aMule. They refresh every 5 seconds.
Theme, text size and Font. The Theme dialog has ready-made themes (Dark, Light, Cream, Nord, Dracula, Solarized, Gruvbox, High contrast and more, in light and dark variants); Customize builds your own theme from four colours (background, panels, text, accent) and warns you when the contrast is too low. Saved themes live on the server and follow you to every device; the active theme, text size and font remain a choice of each browser.
The interface language.
Donate, bottom right (on a phone it is in the More menu), and Update available when a new version is out.

From a phone¶

Below 900 pixels wide the interface becomes compact: scrollable menu at the top, metrics in two columns, full-width dialogs, tables that scroll inside their panel. On the Downloads page a torrent name wraps in full instead of stopping after the first characters: on a phone there is no mouse tooltip to show the rest. Every action stays reachable by touch. Theme, text size, Font, language and Donate are in the More menu, so the top bar keeps only the title, ＋ and Refresh. In the TV series and Movies lists every title is a card: name and status on one line, Pause and Delete side by side below. On the Dashboard, when nothing was transferred in the last 30 minutes, one line says so instead of the bandwidth chart. Pages travel compressed and scripts stay in the browser cache, so the interface stays usable on a slow network too.

Accessibility¶

The interface works from the keyboard and with assistive technologies: controls have accessible names; dialogs manage focus, Tab, Esc and the return to the control that opened them; sortable headers work from the keyboard and announce the sort order; updates, errors and notifications use live regions; the layout holds up on narrow viewports, zoom and 320-pixel reflow. The automated checks (axe-core and Playwright) run from the checkout with cd uiweb/end2end && npm ci && npm run test:a11y. They are not a WCAG conformance statement: that also needs a manual review. Details in ACCESSIBILITY.md.

Part II — Everyday use¶

6. Following a TV series¶

Adding a series¶

Open Explore, search for the title and pick the right result: check year, network and poster, because titles with the same name are common. You can also use TV Series → Add.
Choose the Quality profile. The profile gives the series its quality, minimum audio, language, subtitles, exclusions and «Allow upgrades»: the form has no other fields for these values, so they cannot contradict each other. Profiles are prepared in Configuration → Quality and upgrades (see Quality profiles).
Add, if needed, Extra exclusions for this series only: words that reject a release when they appear in its name. They are added to the profile's.
Choose the seasons to follow. Ranges are allowed.
Check the folder: it is already filled with library folder + title. With «Create folder» on it is created when you add the series; untick it to type it yourself. «Subfolders per season» puts the episodes in Season 01, Season 02…
Save and open the series page.

In folder and file names Gextto removes the characters that Windows, exFAT and SMB shares reject (\ / : * ? " < > |) and a trailing dot or space: «Title: Subtitle» becomes «Title Subtitle». The library stays readable from any computer.

[!TIP] Keep the seasons that have not aired yet enabled: the calendar and the search for missing episodes will keep working when they arrive.

The series page¶

At the top you find poster, plot, cast (linked to TMDB or TVDB), year, network, status, completeness and the library path. The actions:

Action	When to use it
Search missing	to search for all missing episodes now, without waiting for the cycle
Scan archive	you copied files into the series folder by hand and want Gextto to recognise them
Update from TMDB (or TVDB)	new episodes are out, or titles and dates changed
Rename preview	see the Old → New names before renaming; Force rename also recomputes the files that look correct already
Edit	change seasons, folder, aliases, IDs, profile
Download history	reconstruct what happened to each episode

The most important fields of the page:

Field	Effect
Quality profile	applies a profile's values to the series. The profile that matches the current values is preselected; if none matches you see Current values, which changes nothing when saved
Seasons	which seasons are followed
Alias	other names under which the series appears in releases
Exclusions	words that reject a release
TMDB ID / TVDB ID	identify the series in the two services; a wrong ID is fixed here. They are independent: saving one does not clear the other

Minimum audio and Allow upgrades are not on the page: they come from the profile and are changed with Edit selected.

The episodes¶

Episodes are grouped by season. On each one you can:

🔍 search for it: the search opens in a panel below the episode and tells you how far it got (how many sources it is querying, how many answered, when it moves on to eD2k);
📜 see its history: every event of that download (see An episode did not arrive);
copy the magnet, ignore it (it will no longer be searched; re-enabling it makes it a candidate again), force it, delete it;
↻ Redownload: moves the current file to the trash and searches for the episode again, even when it would not normally count as missing (for example because a pack covers the season or later episodes are already there).

Missing episodes¶

The Missing tab of TV Series gathers every episode that has aired and is not in the library, grouped by series and season:

one click on an episode searches for it, first in the local archive and then online, and shows the results with Download;
Search missing searches for every missing episode of a series in the background: the tab shows the progress, you can cancel, and at the end it lists the results with Download. It downloads nothing on its own, and the list stays open after each download: you can take more than one without running the search again;
Filter by series narrows the list.

Not missing: episodes that have not aired yet, episodes that already have a download in progress (even added by hand or from a watched folder), and the episodes of a season whose pack is downloading. If the download ends in an error, the episode goes back to the missing list.

If no copy of an episode matches the profile, the list offers the only copy in the requested language with the 🆘 last resort badge: it is the one Gextto would take on its own, and an upgrade will replace it as soon as a copy within the profile comes out.

You do not have to do it all by hand: in the automatic cycles Gextto searches for missing episodes on its own, first in the local archive and then online, and — if you enable it — asks eD2k once per season when torrents find nothing. Search missing only saves you waiting for the next cycle.

Season packs and single episodes¶

A season pack can fill many episodes, but the comparison with the library stays episode by episode: a pack does not replace an episode you already have in equal or better quality.
When a single episode and a pack containing it have the same score, the release covering more episodes wins; then a REMUX; then the one with known seeds.
Gextto does not download an old episode again when you already have later ones, unless it is a recognised gap, a real upgrade or your explicit request (↻ Redownload).
With pack inspection on (Configuration → Seeding and completion → Pack inspection, off by default) Gextto reads the pack's file list as soon as it arrives: if the pack adds nothing it is dropped, if only some files are useful only those are downloaded. It applies only to packs chosen by the cycle: a pack you start by hand is never dropped.
If a download stays stuck at 0 B/s or waiting for metadata longer than the time set in Configuration → Quality and upgrades → Stuck downloads, the search can start an alternative for the same episode (or season) without removing the stuck one: if it recovers, the library keeps the better copy anyway.
If a pack is given up because it was stuck, the episodes already complete still enter the library and only the missing ones are searched again (in the log: 🛟 … already complete, added to the library before giving up).

7. Following a movie¶

Adding a movie¶

Search for the movie in Explore and check year and original title.
Add it and choose the Quality profile; add, if needed, exclusions for this movie only.
Open its page and press Search now to see the results without waiting for the cycle.
Before queuing a borderline release, open Why not this one?.

A movie is recognised by title and year, and punctuation does not count: a release matches the same way with or without a colon. When one entry's name is contained in another's, the more specific one wins. Do not add the same movie twice under different spellings: fix the metadata of the one you already have.

The alternative titles TMDB knows for the movie also count — the international one and those used in other countries: a release published under another name still arrives, and the page's archive matches are searched with those titles too. The indexer searches use the names that matter: the monitored name, the original title (the film's original language), the title in the language chosen for the film and the English one. The other titles TMDB lists are not queried one by one — they would multiply the requests — but they still count for recognising releases and for the archive search. Working titles are excluded, because they sometimes belong to another film.

The comparison is the one Sonarr and Radarr make: the whole title counts, not a word found anywhere in the name. A release is recognised when its name opens with the film title or with one of its alternative titles, word by word, ignoring accents, capitals and punctuation; the tag an indexer puts before the title (a language such as "ITA", a resolution, a group in square brackets) is skipped. A one-word character name — which TMDB lists among the alternative titles — therefore no longer takes a release of another film. Leading articles do not count, and at the same length the film's own name wins over an alternative title. The page's archive matches show only the releases Gextto would really take.

The movie page¶

It shows poster, plot, cast, the best results found in the sources and the archive matches, each with Why not this one?. Under the plot the Alternative titles (TMDB) line opens the list of names the film was released under abroad: they are the ones Gextto uses to recognise releases and to search the archive, and they stay saved with the movie. If the film was added by hand and TMDB has not been asked yet, the line says not read yet: Aggiorna TMDB reads them at once, together with the metadata. From the page you can edit the movie, download it again or search for it now.

In the editor:

Language takes one code or several separated by commas (ita,eng), even beyond the entries of the list;
Subtitles with Subtitles required: Yes rejects a release without those subtitles, No treats them only as a preference;
Allow upgrades and minimum audio are changed with Edit selected.

The Movies page has two tabs, Monitored and Downloaded, with sortable columns. In the Monitored tab, next to active or paused, a ✓ means a version of the movie has already been downloaded: hovering over the mark shows the path of the library file, or the archived title when the move was not recorded.

List or posters¶

Above the TV Series and Movies tables, the List / Posters switch changes the view: Posters shows one card per title with cover, quality, language and status. The choice applies to both pages and is remembered.

Editing several titles at once¶

To change the same values on many titles — for example moving ten movies to 2160p or pausing several series — use the List view:

Tick the rows. The box in the header ticks every visible row: combine it with Filter the list… to take, for example, every title in 720p.
Press Edit selected. Tick only the fields to change (quality, minimum audio, language, subtitles, exclusions) and type the new value. A ticked, empty field is cleared; an unticked field stays as it is. Allow upgrades and Status change only if you pick a value other than «Leave as is».
Start from a profile copies a profile's values into the fields, except the exclusions, which belong to each title (tick them only if you want to replace them).
Press Apply to the selected titles.

Seasons, folder, aliases and the other values of the titles are not touched. A movie's mandatory requirements (subtitles, required languages) follow the value they qualify: changing language or subtitles updates the requirement too.

8. Searching and downloading by hand¶

The automatic cycle does almost everything on its own. A manual search lets you see what is out there before you change a setting, pick a specific release, or download something you do not follow.

Where to search¶

From	What it searches
Dashboard, search box	local archive, indexers, web engines; eD2k if you tick eD2k
Download, Search for a release	the same, without leaving the downloads page
Explore	titles on TMDB (trending, popular, upcoming) to add; generic release search
Archive	the releases seen in the feeds, instantly; online only if you tick Also search the web
🔍 on an episode, Search now on a movie	that exact title

Reading the results¶

Type a title (or a technical query) and wait for the sources to answer; slow ones are flagged at the timeout.
Compare title, source, quality, size and seeds/peers.
Narrow down with Filter the already loaded results: it works on the results received and does not search again.
Open Why not this one? on the interesting results.
Press Queue (or Download) when you are convinced: the results stay in the list, so you can add others without running the search again.

A manual search also shows releases the cycle would not pick, precisely so you can inspect them: seeing one does not mean it would be downloaded.

«Why not this one?»¶

It is a read-only diagnosis of a release: it queues nothing and changes nothing. It shows the outcome, the score with its parts and every check:

Passed — the check passed;
Blocking — enough on its own to reject the release;
Informational — depends on the rest of the cycle (other candidates, delay, free space, gaps to fill).

At the end comes the comparison with the library: first download, possible upgrade, or a copy no better than the one you have. The checks run in this order: followed title, blocklist and duplicates, global and source filters, release sanity (burnt-in subtitles, absurd size), the title's quality/language/subtitles/exclusions, comparison with the library and active downloads. The score only chooses among the releases that passed every check.

Adding a torrent or a magnet¶

Add torrent, in the top bar of every page, takes a magnet, a .torrent URL or a .torrent file. You can set the save folder, Download now (outside the queue) and Do not rename. With gx-torrent there are also:

Skip check: adds the torrent on files you already have, without checking them first; it starts seeding at once and every piece is verified the first time a peer asks for it. Use it only for data you know is complete;
sequential download and first/last piece of the files;
space preallocation, which can be turned off for a single torrent.

A magnet with so= downloads only the listed files; with ws= it adds web seeds. An address that answers with a magnet is added as a magnet.

You can also copy .torrent or .magnet files into a watched folder (Configuration → Sources → Watched folders): Gextto waits until the file stops changing, adds it and then removes it or renames it .imported.

The Archive¶

Archive is the catalogue of every release seen in the feeds, including titles you do not follow. The search starts by itself as you type (from three characters; for shorter ones press Enter) and is instant. -word excludes the titles containing it. The same torrent appears only once. From each row you can queue (⬇️), open Why not this one? (🔍), copy the magnet (🧲) or delete the row (🗑️).

Also search the web extends the search to indexers and online engines: slower, it starts only when you press Search.
Also search eD2k (with aMule active) adds the eD2k results below.
The Seen in feeds tab groups releases by title, with the number of versions and the best resolution and score: expand a group to queue one.

The releases of a feed removed from Sources (or commented out with # in the list) stay in the catalogue, like those of a switched-off indexer: their .torrent links are no longer opened, so Gextto no longer reports the error of a feed that is gone. Putting the feed back makes them usable again on their own; rows archived before this version behave as before.

eD2k links¶

In Explore and on the eD2k page you can paste one or more ed2k:// links (even mixed with text): Gextto checks them, removes duplicates and hands them to aMule with Add all or one by one.

9. Watching the downloads¶

The Download page shows everything being transferred: torrents, HTTP comic downloads and aMule eD2k files (with the eD2k badge), in a single list that refreshes every 5 seconds.

Reading the list¶

Each row shows name, status, progress, speed, estimated time, peers and ratio. The swarm (🌐 seeders / leechers) is the one reported by trackers and DHT: with 0 leechers there is nobody to upload to. Sort with Sort by (for the whole list at once), filter by name and by tag.

Status	What it means	What to do
Queued	registered but not started yet	wait: the queue will start it
Downloading	transferring	check speed and peers
Stalled	no new bytes for too long; paused	wait for the automatic retry
Seeding	complete, sharing	let it finish according to the seeding limits
Error	the engine reported an error	read the reason before removing it
Archived	the file is in the library	nothing; do not delete the source while it seeds
📌 pin	forced outside the queue	Unlock pin on its page puts it back in the queue

Acting on downloads¶

On one row: pause/resume, recheck, remove.
On several rows: tick the torrents and use the selection bar: Pause, Resume, Check again, Tag (assign, create or remove a tag), Remove from session.
Tools: Clean completed (removes torrents that finished seeding), Orphaned files…, Refresh every 5 s and the link to Removal after seeding.

Remove takes the torrent out of the session and can ask whether to delete the data too. Removing a torrent from the session does not remove the file from the library. A torrent with infinite seeding is never removed by Clean completed.

A torrent's details¶

Click the name: the details open above the Download history, so the list stays visible. The tabs:

Tab	What you find
General	torrent data, copy magnet, Export .torrent, reannounce, pin, restart, mark as failed, Super seeding, web seeds
Tracker	add trackers (`tier\|url` per line) or remove one
Content	the files, with priority (Skip, Normal, High, Maximum) and, with gx-torrent, ▶ for streaming
Peers	the connected peers
Pieces	the piece map: green downloaded, yellow in progress, grey missing
Limits	ratio, seeding days and bandwidth of that torrent: -1 uses the global value, 0 removes the limit
Storage	Move storage to another folder
History	every event of that download, from the choice to the library

Hovering over a value or a button shows what it means. If you move the storage to a folder that already holds the data, Gextto rechecks and reuses it instead of downloading it again. Every save shows a notice at the top saying what was saved; if something fails the notice is red, gives the reason and the tab stays as it was, with the values you typed.

Streaming and media server¶

With gx-torrent each file has ▶: it opens HTTP streaming, and the pieces of the part you are watching are downloaded first. You can start watching before the download ends.

The same engine works as a DLNA/UPnP media server (Configuration → Torrent engine → DLNA/UPnP server, off by default): TVs and consoles on the home network see the completed media and the library folders, with the posters and thumbnails already there (in the Jellyfin, Kodi and Emby format) and the .srt/.vtt subtitles with the same name as the video. It has no authentication, so it answers only devices on the local network.

Bandwidth¶

The Bandwidth button, top right of the page, shows what limits the speed right now and lets you make a temporary exception to the rule set in Configuration → Speed and network:

Auto — Gextto measures the line and adapts;
Limit download / Limit upload — a value from the list, or No limit;
Duration — 1, 3, 6, 12 hours or Until I remove it;
Custom values… — download and upload in KiB/s and duration in minutes;
Remove the exception — back to the rule at once.

The icon tells you what applies: ⚡ an exception (with the minutes left), ⏰ a window of the schedule, otherwise the rule. With eD2k active aMule follows the same limits. See also Speed and network.

Stalled torrents¶

When a torrent adds no bytes for a while it becomes stalled: Gextto pauses it, so it does not take a queue slot, and every so often resumes and reannounces it. If it does not restart before the final limit, it is given up and another release is searched for. Having connected peers without receiving data does not reset the timer.

The times are in Configuration → Seeding and completion → Stalled torrents:

Setting	Default
Consider stalled after	60 minutes
Retry stalled torrents every	60 minutes
Remove stalled torrents after	14 days (`0` = never)

In the log: is stuck at (with the reason), is still stuck at each retry, is downloading again when it restarts, Gave up on when it is given up. The Dashboard lists stuck downloads under To check, with the next retry and when they will be given up.

Seeding limits: per-source rules and idle seeds¶

When a torrent stops sharing is decided by Gextto, with the same rule for gx-torrent and qBittorrent. For each limit (ratio, time, inactivity) the first level that sets it applies:

the limit of the single torrent (Limits tab);
the first matching seed rule (Configuration → Seeding and completion → Seed rules);
the global limit.

A seed rule names a source (part of the indexer's name is enough, case does not matter: mircrew also matches mircrew indexerr) and/or a category (TV Series, Movies, Comics, Other). This way a private tracker's rules apply only to its torrents. An empty field uses the global value. A torrent from Jackett's aggregate search carries no tracker name: category or global apply to it.

Infinite seeding. 0 means infinite, for ratio as for time, and one of the two being infinite is enough for seeding never to end: the torrent is not stopped even for inactivity and is not moved into the library (a copy or a hardlink goes there instead). To stop seeding, ratio and time must both be greater than zero. The same applies to eD2k.

Stop an idle seed after stops a torrent that has sent no data for that many minutes. Only time actively seeding counts (not queued or paused), and the count restarts at every Gextto restart: a restart can only make seeding longer.

At the end of seeding the torrent follows its path: archiving into the library and, if enabled, removal from the session.

The history¶

Download history, at the bottom of the page, lists finished downloads: name (with the NAS badge when the file is archived), type, season and episode, folder rule, score, library path and date.

10. Where the files end up¶

What happens when a download finishes¶

Gextto recognises the content, renames it with your template and puts it in the series or movie folder. The behaviour depends on the shape of the torrent and on whether a library is configured:

Content	With a library configured	Without a library
One episode, as a single file	archived and renamed according to the series rules	stays in the downloads folder
One episode inside a folder	the video enters the library, renamed; the folder stays for seeding and then goes to the trash	the folder stays intact in Downloads
A season pack	the episode videos (recognised by `S01E02`, `1x02`…) enter the library; the pack stays for seeding and then goes to the trash	the pack stays intact in Downloads

Subtitles, NFO files and images of the pack stay in the source folder. A source folder is never moved or deleted while it is seeding, and it goes to the trash only when the trash is configured and removal after seeding is on; without a trash, source files are never deleted.

Copy, move or hardlink¶

Move (Move episodes/packs to the archive): at the end of seeding the file is moved into the library.
Hardlink: while the file seeds, Gextto tries to create in the library a second name for the same data. The file shows up in the library at once and does not take space twice. Deleting one of the two names does not touch the other: the data disappears only when both are deleted.
Copy: if a hardlink is not possible, Gextto copies and says so once in the log, with the reason.

A hardlink works only if downloads and library are on the same filesystem. Two folders of the same NAS mounted separately (for example nas:/volume1/Downloads and nas:/volume1/Video as two mounts) are not enough: mount a common root (nas:/volume1). If you use programs that edit videos in place (some taggers, mkvpropedit), turn off Configuration → Seeding and completion → Hardlink instead of copy while seeding: they would change the seeding file too.

Upgrades¶

When a better version of a file you have arrives, the new one enters the library and the old one goes to the trash. An equal or worse version is dropped (in the log: discarded as inferior). To freeze a title set Allow upgrades to No with Edit selected.

File names¶

With Rename episodes on, the name is chosen in Configuration → Library and renaming → Name composition, with four formats and a preview:

Format	Episodes	Movies
Basic (default)	series, season, episode and title	`Title (Year)`
Standard	plus resolution and codec	`Title (Year) [resolution][codec]`
Complete	plus audio, HDR and language	`Title (Year) [resolution][audio channels][HDR][codec]`
Personalized	your template, with the tokens `{Serie}`, `{Stagione}`, `{Episodio}`, `{Titolo}`, `{Source}`, `{Gruppo}`, `{Risoluzione}`, `{VideoCodec}`, `{Audio}`, `{Canali}`, `{HDR}`, `{Lingue}`	the movie template (below)

With Personalized movies have a template of their own, in the field Custom template for movies: the tokens are {Titolo}, {Anno}, {Source}, {Gruppo}, {Risoluzione}, {VideoCodec}, {Audio}, {AudioCodec}, {Canali}, {HDR} and {Lingue}, with their own buttons and preview. The default, {Titolo} ({Anno}) [{Risoluzione}], is the name movies had before, so a library already renamed keeps its names. A token with no value (an unknown HDR, a missing year) disappears together with its brackets.

The technical tags are not taken from the release name alone: before renaming, Gextto reads the file with mediainfo (or, when missing, ffprobe) and uses the real resolution, codec, audio, channels, HDR/Dolby Vision and languages. A tag already in the name stays unless the file contradicts it. A file holding several consecutive episodes gets the range (S01E01-E02). Without an episode title (no TMDB or TVDB key) the name uses «Episodio N» (the placeholder is always Italian).

Importing a file you already have¶

For a video Gextto did not download — an episode recovered by hand, an old season, a movie from another disk — use Maintenance → Import file:

type the path or press Browse (one click on a video picks it, Select takes the whole folder);
press Analyze: Gextto proposes series or movie, season, episode and destination folder;
correct it if needed and press Import now.

The file is renamed, moved into the library and registered, and Jellyfin or Plex are refreshed. If a file with the same name already exists the import stops: tick Replace the existing file to send the old one to the trash.

If instead you copied files straight into a series folder, Scan archive on the series page is enough.

11. Comics¶

In Comics → Add comic type the title and press Find.
Pick the exact GetComics result and confirm. Gextto saves that post and uses it for the download, without confusing it with an issue of the same name.
A title with a number (for example «Title #41») follows the series from that number on: later issues are downloaded, earlier ones, other series and collections are not.

The page also has the list of followed comics, link extraction from a post, the weekly pack settings and the history with resend, delete and force. A weekly pack is downloaded only once; one found before its links were published is checked again at every cycle and queued as soon as the link appears.

Direct HTTP downloads show up in Download like the others. In CBZ files downloaded directly Gextto adds a ComicInfo.xml with series, number and year, which Komga, Kavita and readers use to sort the issues; a file that already has one, comics from torrents (they are seeding) and CBR files are not touched. An issue that arrives as RAR is saved as .cbr.

12. eD2k and Kad with aMule¶

Gextto can also use the eD2k/Kad network through aMule: amuled does the network work, Gextto drives it through external connections (EC), the same protocol as amulecmd. It is optional and off by default; with aMule off, torrents do not change.

Connecting aMule¶

Install aMule (amule-daemon on Debian/Ubuntu) and enable external connections in amule.conf:

[ExternalConnect]
AcceptExternalConnections=1
ECAddress=127.0.0.1
ECPort=4712
ECPassword=<md5 of the password>

ECPassword is the MD5 of the password: printf '%s' 'secret' | md5sum. 2. Start amuled. 3. In Configuration → eD2k and Kad turn on eD2k and Kad through aMule and enter address, port and the password in clear. 4. Open aMule's TCP and UDP ports on the router: you get a High ID and a Kad not behind a firewall, which means more sources.

Gextto must see aMule's incoming folder at the same path (it does when they run on the same machine): that is where it imports completed files from. Use a dedicated folder, because aMule shares everything in it. With aMule in Docker, mount the downloads folder at the same path as on the host, set INCOMING_DIR and TEMP_DIR there and, for UPnP, use network_mode: host.

The eD2k page¶

aMule — version, server, High/Low ID, Kad status, speeds, upload queue, bytes exchanged; the Connect, Disconnect, Start/Stop Kad and Leave the server buttons. If aMule does not answer, the page says what to check.
eD2k servers — the server list, with Connect; from here you also download a new server.met or a nodes.dat.
Search eD2k and Kad — pick the network and search; each result has Download, and already known marks a file aMule already has. The filter above the results (words, minimum sources, minimum size, hide known) does not search again.
Paste eD2k links — one or more ed2k:// links, then Download with aMule.
aMule log, aMule folders (with Create the folders when they are missing) and Shared files (bytes uploaded, requests, queued clients, links to copy; Reload shared files after manual changes).

What happens to an eD2k download¶

eD2k downloads appear in Download with the eD2k badge, with pause, resume and 🗑️ (on a running download it cancels it; on a completed one it only removes it from aMule's list). ⓘ opens the details, with sources, hash, link and priority in aMule.

When aMule completes a video, Gextto waits until the file is stable in the incoming folder, then treats it like a completed torrent: same quality check, renaming, library, MediaInfo, notification. With Move episodes/packs to the archive on the file is moved (and aMule stops sharing it); off, it enters with a hardlink or a copy and stays shared. A file that is not a video stays in the incoming folder. With infinite seeding the file is copied and stays shared, with the «seed ∞» badge.

eD2k as a fallback¶

With Use eD2k when torrents find nothing on, the search for missing episodes asks eD2k/Kad when torrents and archive have no results: one query per season, lasting as long as eD2k search duration. The results appear with the eD2k/Kad source and Download hands them to aMule.

The complete guide — installation, Docker, common problems — is in amule.md.

Part III — Configuring¶

The Configuration page has four areas: What to look for (General, Sources, eD2k and Kad, Quality and upgrades), How to download (Torrent engine, Speed and network, Queues and performance, Seeding and completion), Where to store (Library and renaming, Archive and space) and System (Automatic maintenance, Notifications, Access and services, Diagnostics and updates).

How to use it:

the sections are in the left column (on a phone, in the Section drop-down); the search box finds a setting by name or by technical key;
each row shows name, description, unit and special values (for example «0 = never»); settings that depend on another one appear only when they are needed;
a value changed and not saved is marked not saved and the Unsaved changes bar appears, with Save all and Discard changes; leaving the page with pending changes asks for confirmation;
↺ Default puts the default value back; Modified only shows only what you changed;
settings the active torrent engine does not use are gathered under Not used by the active engine; rare ones under For experts.

This part explains the choices that matter. Every setting, one by one, is in Appendix A.

13. Sources: feeds and indexers¶

Gextto finds releases in three ways: RSS feeds (read at every cycle, they fill the local archive), Torznab indexers (queried for the titles you follow) and web engines. Everything is set in Configuration → Sources.

Adding Jackett or Prowlarr¶

In Jackett (or Prowlarr) enable at least one indexer and copy the API key.
In Gextto press + Jackett (or add a row to the Torznab indexers list and pick the type).
Enter only the base address, for example http://jackett:9117 or http://prowlarr:9696, and the API key. Do not add the Torznab path: Gextto builds it.
Save and press Verify.
Check Health → API status.

Gextto reads the list of indexers from the manager and queries them one by one, so each has its own timeout and health and the slowest does not hold up the others. If the list cannot be read, it falls back to the aggregate search. In the results the source shows as jackett:TrackerName.

A switched-off indexer is not queried, and the .torrent links of the releases already collected from it are no longer opened: the releases stay in the archive catalogue — nothing is deleted — but Gextto does not try to download them and no longer reports the error of the service you turned off. Switching the indexer back on makes them usable again.

Adding indexarr¶

The same, with type indexarr and base address http://indexarr:8788. indexarr is queried in one go: it merges its sources' duplicates itself and says which site each result comes from. Its eD2k sources return ed2k:// links, which Gextto hands to aMule instead of the torrent engine. When a download completes, Gextto tells indexarr, which marks the row as downloaded on its Download page. There is nothing to configure.

FlareSolverr¶

Sites behind Cloudflare need a working FlareSolverr: enter its URL in Configuration → Sources and press the test button. With a proxy configured, FlareSolverr is not used as a fallback, because it would download outside the tunnel.

Reachable does not mean working¶

An indexer can answer and still refuse the request (wrong API key, no indexer enabled in Jackett). That is why Health tells them apart:

Status	Meaning	What to do
OK	answers and accepts requests	nothing
API error	answers, but refuses the API key or the configuration	copy the API key again, check the enabled indexers and the base URL. Restarting Gextto does not help
Unreachable	no answer within the timeout	check DNS, container, port, firewall

Behind a manager the table also shows each of its indexers, so a tracker in error or paused is visible even when the manager answers.

Backoff: when a source fails¶

A source that keeps failing is put in backoff: for a while Gextto does not query it, with longer and longer pauses. If an indexer answers 429 Too Many Requests with Retry-After, the pause lasts at least the time asked. The state is in Health → Provider status, with level, expiry and last error. Once you have fixed the cause press Reset on that source; do not use it to hide a wrong API key, or the backoff will start again.

Gextto warns you by itself (log and notification) when a source fails in three of the latest searches, when a source's .torrent links fail on two different releases in the same cycle (usually an expired login) and when everything is back to normal.

Copies of the same release¶

The same torrent often arrives from several sources. Gextto merges them into one release when they have the same infohash or — when a source does not give it — when name and size match. When in doubt it does not merge: a different group, resolution, language, PROPER or episode stays separate. It uses the best copy of the release and keeps the others in reserve: if a .torrent link does not work, it tries the same release from another source at once.

14. Quality, scores and upgrades¶

How Gextto chooses¶

Filters. A release that does not meet the title's requirements (quality, language, required subtitles, exclusions) or the sanity checks is rejected.
Score. Among the remaining releases the one with the highest score wins. The score adds up the weights of resolution, source, codec, audio, HDR/Dolby Vision, fixes (PROPER, REPACK) and group.
Comparison with the library. If you already have the file, the new release replaces it only if it is a real quality improvement.

The sanity checks are automatic and cannot be configured: burnt-in subtitles (HC) and sizes that make no sense for the resolution are rejected. The rejections are visible in the log at DEBUG level.

Quality profiles¶

A profile is a named template: quality, minimum audio, language, subtitles, exclusions and «Allow upgrades». They are managed in Configuration → Quality and upgrades → Quality profiles.

Quality is picked from a list (360p … 2160p, a minimum with «+», the 720p-1080p range, or Any).
Language and Subtitles offer the common choices (ita, eng, ita,eng, multi, any) but also take other codes (fre, ita,eng,fre). Empty means Gextto's default language.
Until you save the list, Gextto generates a profile for each combination already in the library; the first is Default. Two profiles cannot share a name.

Profiles are templates, not rules. When you pick a profile for a title, its values are copied into the title. Changing the profile later does not change the titles already created: to update many at once use Edit selected. The profile list is also what Jellyseerr sees; each profile keeps its number even if you rename it, because Jellyseerr remembers the number.

Trying before changing¶

At the top of Quality and upgrades:

Examples with the current weights — the scores of a few typical releases;
Impact on the library → Check now — how many episodes in the library would be replaced with the current rules. Use it after changing weights or thresholds: a high number means the next cycles will download a lot again;
Test a release — paste a release name: you see the score item by item and whether it would replace the copy in the library, with the reason.

None of them downloads anything.

The weights¶

Below are the weights, in closed panels: Video (4K, Full HD, HD, SD, codec, source, HDR/DV), Audio (Lossless, DTS/DD+, AC3, AAC/Opus, MP3, channels, Atmos), Release fixes and Custom groups (a bonus or a penalty for a release group, written like the release's final tag, for example TBK). Higher means more preferred. After a weight change the library scores are recalculated by themselves.

Language is not a bonus: it is a requirement of the title, so you can prefer English as much as Italian. Preferred subtitles give a small bonus. A freeleech release gets a separate bonus only when choosing among candidates, never for an upgrade.

When a file is replaced¶

Setting	Effect
Min score difference for upgrade	how much better the new release must be. Only quality counts: size, subtitles and swarm health are not enough
Stop upgrading above this score	above this score the file is no longer replaced, except by a REPACK or PROPER
Replace versions already archived	turns on upgrades of files already in the library

Before replacing a file Gextto reads both the new one and the one in the library with ffprobe: if the name promised more than is there (a «2160p» upscaled from 1080p, an «HDR» without HDR) the download is dropped and blocklisted (in the log: Upgrade not confirmed by MediaInfo). A library file that does not declare its source or group is not considered worse without proof.

Delay before downloading¶

In Configuration → General → Delay before downloading you can make Gextto wait a few minutes after finding a release, to give a better version time to show up. Skip the delay from this score starts excellent releases at once. The titles waiting are in the Title queue panel of Health.

15. The torrent engine¶

Gextto always owns the database, queue, scores, renaming and library; the engine only does the transfer. It is chosen in Configuration → Torrent engine, and the Download page and the automations are identical with either. The interface header shows the engine in use with its version.

gx-torrent (default)¶

The included engine, in pure Go. Gextto starts and supervises it in a separate process; there is nothing to install.

It stays on when Gextto restarts (for example for an update): transfers are not interrupted. It is restarted only if its program or options changed, never during a file move. If Gextto stays off, it stops by itself after 15 minutes.
It has its own web page at its listen address (default 127.0.0.1:8890, this computer only): magnets, pause, check, queue, IP filter, log. To open it on the network use 0.0.0.0:8890 and a token.
The disk cache sizes itself from free memory and disk type. The IP filter from a URL is refreshed once a day.
With a proxy configured, feeds, indexers and .torrent downloads go through the tunnel too; services on the local network stay direct.
It supports streaming, DLNA, sequential download, v2 torrents and magnets with so=, ws= and dht=.
In Configuration → Torrent engine → gx-torrent there are three native qBittorrent options: recheck data when a download completes (verify the files once before seeding them), anonymous mode (hide the client version from peers and public trackers, not the IP) and announce to all trackers in a tier (ask every tracker of a tier together). The first two are off by default, the third is on. After changing them, restart Gextto to apply them to the managed daemon.

If the engine cannot stay up (6 abnormal starts in 10 minutes), Gextto switches by itself to qBittorrent-nox and, as a last resort, to a recovery state with transfers stopped and the interface reachable. An update of gx-torrent that does not start is first rolled back to the previous version. You get a notification in every case.

qBittorrent-nox¶

Gextto can drive a qBittorrent-nox (from 4.6 to 5.3) through its Web API: set URL, user, password, category and tag. If the two programs see the folders under different paths (Docker, NAS, another machine) you need path mappings local=remote: Propose reads them from qBittorrent. If a path cannot be translated the activation is refused, so a file cannot end up in the wrong folder.

In managed mode Gextto downloads qBittorrent-nox, installs it next to itself, starts and stops it with the service, updates it and restarts it if it crashes. The full steps are in qbittorrent-nox.md.

Switching engine¶

A torrent belongs to one engine at a time: switching is a migration, not two clients on the same data. The engine panel has a reachability and path test: use it before switching. The state of the active engine (queue, port, DHT, encryption, proxy, cache, disk I/O) is in Health → Torrent engine.

On a small machine¶

Gextto runs well on a mini PC or a NAS. At rest you do not notice it; during a transfer the cost is the engine's, which gives memory back when the download ends. If resources are tight:

turn on the dynamic queue and Don't count stalled torrents in active slots (Configuration → Queues and performance);
do not set a cache limit: the automatic value is almost always the best.

16. Speed and network¶

The speed rule¶

In Configuration → Speed and network, How the speed is decided offers three rules:

Auto — Gextto measures the line (the ping to public servers) and raises or lowers the speed so as not to saturate it. If you know your line's real speed, enter it in Auto — maximum line speed: Auto finds the right point faster.
Fixed limits — one cap for download and one for upload.
Hourly schedule — one row per day, with Off, Limit (your values) or Auto, plus start and end time. Outside the hourly windows: Auto decides what applies outside the windows.

Peers on your home network are not limited, unless you turn on Apply the limits to peers on the local network too. For a temporary exception use the Bandwidth button in Download (see Bandwidth).

Ports, VPN, encryption¶

Ports: open the listen port on the router; UPnP and NAT-PMP can do it by themselves. Test ports only checks that the port opens on this computer, not the forwarding on the router.
VPN interface (killswitch): binds torrent traffic to one network interface (tun0, wg0); if the VPN drops, the traffic stops instead of going out in the clear. Applied after a restart.
Encryption and IP filter: in Speed and network → Encryption and IP filter.

RAM disk¶

To spare writes on a slow disk you can keep incomplete data in RAM (Configuration → Queues and performance → RAM disk). Gextto lists the available tmpfs and, if there is none, creates /dev/shm/gextto. Its content is lost when the machine restarts and the folder has to be created again.

17. Library, NAS and space¶

A library on a NAS¶

Mount the NAS before Gextto starts (for example with a systemd unit or fstab).
Give the service user read and write access.
Set a prudent minimum free space (Archive and space → Disk space): below that threshold Gextto starts no new downloads.
If you copied existing files, run Scan archive.
After the first download check the actual path in the Download history.

If the NAS is not mounted, do not replace the path with a local folder to make it work: the files would end up in the wrong place. Fix the mount and let the download wait: failed moves are retried after 1, 2, 4, 8, 16 and 32 minutes. An unmounted NAS deletes nothing: an empty mount point is treated as a missing volume and the episodes stay registered.

Trash¶

The trash (Archive and space → Trash) receives the files replaced by an upgrade, the packs at the end of seeding and the duplicates. It is emptied according to the retention; Clean trash in Maintenance empties it now. Without a trash configured Gextto never deletes source files by itself.

Language and metadata¶

Default language (Library and renaming): the language series and movies get unless you choose another.
Metadata: TMDB is the default source of titles, episodes and posters; TVDB is used when only its key is set. The keys go in Access and services → Metadata services.
Remove audio tracks in other languages: after the import it keeps only the audio in the chosen language, without re-encoding, and checks the copy before replacing the file. It needs ffprobe, ffmpeg and mkvmerge.

18. Notifications¶

Gextto can notify you through Telegram, e-mail, webhook, Discord, Slack, Gotify, ntfy, Pushover, Apprise or your own program. It is set in Configuration → Notifications.

Setting up a channel¶

Fill in the channel's fields (an empty field turns the channel off).
In the Events group choose what to receive.
Press Send test notification: it uses every configured channel and shows an error at once.
Try a real completion only with an unimportant download.

Which events¶

Group	What it covers
Download notifications	a download starts, completes or fills a missing episode
Error notifications	a download fails, a torrent or a comic goes into error, the service stops on an error
Health and engine notifications	problems with health, the cycle, the indexers, logins, the torrent engine, and the return to normal
Maintenance notifications	restarts, completed backups, audio track clean-up

Completion notifications carry size, duration and average speed. For a season pack they tell the new episodes apart from those that replaced a worse version. Anti-spam — minimum interval avoids receiving the same event several times; Message language chooses the language, independently of the interface's.

Gextto also warns you about restarts (and abnormal stops, at the next restart: a dead process cannot report its own crash), fatal errors, gx-torrent problems and the switch to another engine.

Webhook: which URL and which format¶

The Webhook channel sends events to an external service. Pick the Format, paste the URL and put the credential in Webhook token, not in the URL.

Service	Webhook URL	Format	Credential
Your own endpoint	your URL	Gextto (signed JSON)	«Webhook secret» signs the body (HMAC SHA-256, header `x-gextto-signature`)
Discord	the channel's webhook URL	Discord	none
Slack	the incoming webhook URL	Slack	none
ntfy	`https://ntfy.sh/<topic>`	ntfy	«Webhook token» = Bearer token, if the topic is protected
Gotify	`https://gotify.example/message`	Gotify	«Webhook token» = application token
Pushover	`https://api.pushover.net/1/messages.json`	Pushover	«Webhook token» = application token, «Webhook user» = user key

Each service also has dedicated fields (Discord, Slack, Gotify, ntfy, Pushover, Apprise, Program), so several channels can be on at once. Every channel receives the same event, and a slow one does not hold up the others. The Program receives the data in the GEXTTO_* variables and runs without a shell.

19. Integrations with other programs¶

No integration is needed to download and archive. Turn them on one at a time and always use the test button. Fields and actions are in Appendix B.

Jellyseerr (in place of Sonarr and Radarr)¶

Gextto answers Jellyseerr's requests as Sonarr and Radarr would:

in Jellyseerr add a Sonarr server and a Radarr server with Gextto's address;
as the key use Gextto's API key (Configuration → Access and services);
pick quality profile and folder from the lists Gextto exposes.

A requested series is followed only for the seasons asked; if you already follow it, the seasons are added. A movie already there is not duplicated. Jellyseerr's «search now» is not handled: the title is searched by the normal cycle. It is not full Sonarr and Radarr compatibility: unhandled requests answer 404.

Jellyfin and Plex¶

Enter the server URL (reachable from Gextto's computer) and a token: Gextto asks for a library refresh when a new file arrives.

Simkl, calendar, hooks¶

Simkl: complete the PIN flow, check the account and then import the watchlist.
iCal calendar: upcoming episodes in a calendar you can add to your phone.
Event hooks: a program run on events such as download_started, torrent_completed, season_pack_completed, torrent_error. The fields take placeholders ({title}, {path}, {series}, {episode}…) and receive the GEXTTO_* variables. It runs without a shell (no pipes or &&: use a script), with a default timeout of 60 seconds and at most four hooks at once. Try it first with a harmless program that writes a log.
gx-torrent metrics and events: GET /api/v1/metrics for Prometheus and GET /api/v1/events for Home Assistant (see API.md).

20. Access and security¶

The interface is an administration panel: whoever opens it can change everything. By default it is open, because Gextto is meant for a trusted home network.

Home only: leave it as it is, or restrict port 5000 with the firewall.
From outside: in Configuration → Access and services turn on Require login and set a password. With No login from the local network (the default) you are not asked at home. Put Gextto behind an HTTPS reverse proxy or a VPN.
Scripts, TUI, calendar: they use the API key (X-Api-Key header or ?apikey=). Generate creates a random value, Copy copies it.
gx-torrent on the network: if you open its page on the LAN, set its token.

Regardless of the login, Gextto refuses changes that the browser declares as coming from another site: a malicious page cannot reconfigure it through your browser. The full model is in SECURITY.md; the keys in Access keys.

Part IV — Keeping everything in order¶

21. Checking that everything works¶

The Dashboard¶

This is the page to look at every day. From the top:

In session — torrents downloading, stuck and seeding, the bandwidth of the last 30 days and the traffic chart of the last 30 minutes (download and upload, with the peak and the limit in force; with eD2k, aMule as dashed lines);
To check — stuck downloads (with reason and next retry), pending moves to the library, warnings and errors. A warning stays while the download is in the session and disappears by itself when it leaves with nothing pending; Clean marks them as read. When all is well it says «Nothing to check»;
Manual search and Automatic search — the next run, the outcome of the last cycle and the buttons to start one (everything, or only Series, Movies, Comics);
Background operations — renames, scans, MediaInfo, integrity check, searches for missing episodes: with the percentage and Cancel;
Upcoming, latest downloads, recent activity and latest found.

The DRY-RUN or ACTIVE badge at the top tells you whether Gextto really downloads.

Health¶

Health (🩺) answers «does everything work?»:

process, system and service state, folder permissions, disks;
API status and Provider status: each indexer with its state and backoffs (see Sources);
Searches: the latest cycles, with duration, releases examined, downloads started, errors and sources that did not answer;
Title queue: the titles held by the delay, and for how long;
Failed imports: how many per day over the last 7 days (a growing number points to a disk or NAS refusing the moves);
Torrent engine and, with eD2k on, eD2k and Kad (aMule).

The log¶

Log shows the live log, with filter, number of lines and follow/pause. The lines are in English, in readable sentences. Every cycle prints a SOURCE REPORT with the outcome of each source and the decisions taken. The file is <data>/gextto.log (rotated at 5 MB); the engine's is <data>/gx-torrent/gx-torrent.log, also readable from Maintenance → gx-torrent log.

The history of a download¶

Every log line about a download ends with an acquisition ID (acq: 7f3a2c), the same from the choice of the release to its arrival in the library. Type it in the log filter to see only that download. The same lines are kept for 180 days as history: History tab in the torrent's details, 📜 on every episode, Download history on the series page. They stay readable after log rotation and after the torrent is removed.

Charts and activity¶

Charts shows CPU, RAM, network, disk and daily usage; Activity the recent events.

22. Looking after the library¶

All these actions are in Maintenance (full list in Appendix C). Long operations run in the background: you can keep using Gextto and cancel them.

[!TIP] Always use the preview first, where there is one. Operations that move files send them to the trash, they do not delete them.

Renaming¶

Rename preview on a series, or Rename all for the whole library: first you see Old → New, then you apply.
Rename folder contents: pick any folder; Gextto recognises series and movies from the names, matches them against TMDB/TVDB and proposes new names. You can pick an alternative result, accept or reject each proposal and then Apply selected. Conflicts are left unchanged; subtitles, images and NFO files follow the new name.
Restore source: puts the [WEB-DL], [BluRay]… token back in names that lost it, taking it from the original release. If it is unknown, the file does not change; if the folder already holds a file with the new name, the episode is skipped and counts as an error, and nothing is overwritten.

Duplicates¶

Duplicate preview and Clean duplicates find clearly worse videos left next to the better version in the same folder (the old 480p next to the new 1080p) and move them to the trash. At equal resolution the version in the preferred language is kept. Files of torrents in the session are protected.

Real quality and file integrity¶

Refresh MediaInfo reads with ffprobe resolution, codec, HDR, audio and languages of the library files. This data is used to compare upgrades with the file as it really is, not with its name. New files are read when the download finishes; for the others there is also the automatic backfill.
File integrity check (Configuration → Automatic maintenance) decodes the videos a little at a time, in the background, and finds truncated or damaged files. The Check status panel lists the damaged or unreadable files with the reason; Redownload next to an episode takes it out of the library, blocklists that release and looks for a different one. With File integrity check — re-download damaged files on it does so by itself. It needs ffprobe and ffmpeg.

Night maintenance¶

In Configuration → Automatic maintenance → Night maintenance pick a window (default 01:00–07:00) for the heavy jobs: database optimisation and checks, the report of files gone from disk, audio clean-up of the files already in the library (if you enable it), integrity check and MediaInfo. Each job starts only when Gextto is idle, one at a time. The panel shows what is running or why it is waiting; Run now starts a pass at once.

Database clean-ups¶

Housekeeping keeps the database tidy (cycle statistics, old errors, expired backoffs, VACUUM when needed). It does not touch the library, the downloads or the Archive.
Release archive cleanup (Automatic maintenance) removes old releases from the catalogue, if you enable it. It touches no file and no download.
Archive SQL maintenance (in Maintenance) runs ready-made operations on the catalogue: releases older than 6 months or 1 year, in a language or without it, empty titles, duplicates. It first shows the preview with the number of rows, then Delete permanently. Make a backup before a large clean-up.

What not to confuse¶

Set	What it holds	What cleans it
Torrent session	the torrents in the engine	Remove, Clean completed, end of seeding
Library	your videos	Clean duplicates, upgrades (to the trash)
Trash	the files taken out of the library	retention, Clean trash (final)
Archive	the catalogue of the releases seen	Release archive cleanup, Archive SQL maintenance
Download history	the list of finished downloads	the history retention

23. Backup and restore¶

What a backup contains¶

A gextto-backup-<date>_<time>.zip file in <data>/backups with:

the four databases: series, episodes, movies and downloads (gextto_series.db), settings and indexers (gextto_config.db), feed archive (gextto_archive.db), comics (gextto_comics.db);
the other configuration files of the data folder, gx-torrent's state included;
the external folders you list in Files and folders to include.

It does not contain the videos, the torrent session, the IP filter, the caches, the logs, the trash or /etc/gextto/gextto.env: keep a copy of that yourself.

Scheduling backups¶

In Maintenance → Backup choose the retention and when to make them (by hand, every N hours or at a fixed time). You can also send them to FTP (with Test FTP, which tells you which step fails), to a synced folder or to Telegram. The Dashboard button makes a backup at once.

Restoring¶

Install Gextto with the same options as before, or stop the service: sudo systemctl stop gextto.
Set the current data folder aside: sudo mv /var/lib/gextto /var/lib/gextto.old.
Extract the backup and give it to the service user:

sudo install -d /var/lib/gextto
sudo unzip /path/gextto-backup-<date>.zip -d /var/lib/gextto
sudo chown -R <user>:<group> /var/lib/gextto

Put /etc/gextto/gextto.env back if you had one, then sudo systemctl start gextto.
Open Health. With gx-torrent the downloads in progress are added again from the saved magnets, and the data already downloaded is rechecked and reused.

A restore does not move the videos. If the library changed place in the meantime, fix the paths or run Scan archive before turning the searches back on.

24. Updating Gextto¶

From the interface¶

Maintenance → Updates shows the installed version, the channel, the latest published version and what is new. Gextto checks by itself every 6 hours (Update check interval); Check for updates checks now. When a new version is out:

press Update: Gextto backs up the databases and asks the gextto-update system service for the update;
wait: Gextto finishes the copies to the library, restarts and the panel tells you by itself;
at the first visit a Gextto updated pop-up shows what is new.

If the new version does not start, the previous one comes back by itself. At the bottom of the page you find the exact version: the third number is the build (1.2.2065), with the commit it was built from.

From the command line¶

sudo /opt/gextto/gexttod --update

It downloads the archive, verifies its checksum (mandatory), replaces program and engine and restarts the service, keeping the previous versions (gexttod.prev, gx-torrent.prev) and putting them back if the new ones do not start. Data and configuration are not touched, and neither is the systemd unit: your customisations stay. Options: --channel stable, --release <tag>, --archive <file> (offline), --force, --no-restart.

On a source installation: git pull and make build.

Other commands¶

Command	Effect
`gexttod --version`	installed version, build and channel
`gexttod --help`	usage summary
`gexttod --config <file>`	uses a different configuration file
`gexttod --dry-run`	starts in test mode until the next restart
`gexttod tui`	opens the terminal interface

25. Using Gextto from the terminal (TUI)¶

The TUI is a tabbed interface for the terminal, handy over SSH even from a phone. It uses only the service's API, so it works from another computer too.

/opt/gextto/gexttod tui
/opt/gextto/gexttod tui --url http://host:5000 --lang en
GEXTTO_URL=http://host:5000 GEXTTO_API_KEY=<key> /opt/gextto/gexttod tui

The tabs are Status, Torrents, Log, Health, Archive, Missing, Blocklist, Library and Maintenance.

Key	Action
`1`–`9`, `Tab`	change tab
`r`	refresh
`?`	help
`/`	filter (in the Log)
`f`	follow or freeze the log
`Enter`	open the details
`p`	pause or resume
`y`	copy the link
`d`	remove an eD2k download from aMule

On a narrow terminal (under 60 columns) the interface becomes compact. The full guide is in tui.md.

Part V — When something goes wrong¶

26. Troubleshooting¶

An episode did not arrive: where to look¶

Start from the series page and press 📜 on the episode:

What you see	What it means	What to do
No events	Gextto never downloaded it	🔍 on the episode and Why not this one?: is the release missing from the sources, or was it rejected by filters, quality, language or delay? Check Health too
📥 Downloading and then nothing	the download is running or stuck	open the torrent in Download: «⏸️ is stuck» means few people are sharing; «❌ Gave up» that it was given up and another version will be searched for
🎉 downloaded but no 📁	downloaded but not in the library	look for «still cannot be moved» or «Could not move»: usually the NAS is unreachable, permissions or space. The retries resume by themselves
discarded as inferior	the library already had a version at least as good	nothing: this is the expected behaviour
📁 added / ♻️ updated	it is in the library	if the name is not the expected one, see A file is not renamed below

For the full detail open Show in the log from the torrent: the log is filtered on that download alone.

Nothing is downloaded¶

Does the badge at the top say DRY-RUN? You are in test mode (see chapter 4).
Is the Automatic cycle on (Configuration → General)?
Is the series or movie active, and are the seasons followed?
Are the sources OK in Health?
Does a manual search find releases? If so, open Why not this one?: it is usually a quality or language filter that is too strict.
Is there enough free space compared with the minimum set?

A release is visible but not chosen¶

Open Why not this one? and look at the Blocking checks: followed title, filters, sanity, quality/language, comparison with the library. If all pass, look at the Informational ones: a better candidate, the delay, free space or gaps to fill can change the choice.

A source does not answer or shows «API error»¶

Check Health → API status and press Verify on the source.
Jackett: base URL (http://host:9117), API key copied again, at least one indexer enabled. Do not add the /api/v2.0/indexers/all/results/torznab/api path.
indexarr: base URL (http://host:8788) and type indexarr.
Sites behind Cloudflare: a working FlareSolverr is needed.
After fixing it, Reset the source's backoff.

A torrent is stalled¶

It is paused on purpose, out of the active queue: it will be resumed at the next retry. You can also Resume or Restart it by hand. Look at the swarm in the row: with 0 seeders it will probably not restart, and Gextto will look for another version. The times are in Stalled torrents.

The download is complete but not in the library¶

Look in the log for move, permissions and space. Temporary errors (NAS unreachable, disk full, timeout) are retried after 1, 2, 4, 8, 16 and 32 minutes; only after the last one the torrent goes into error. A move interrupted by a restart resumes by itself about a minute later. Do not delete the source files until the library path shows up in the history.

A file is on the NAS but Gextto counts it as missing¶

Check the file name (it must contain season and episode, for example S01E02), the series path and the permissions, then press Scan archive.

An episode in the library cannot be downloaded again¶

Use ↻ Redownload on the episode: it moves the file to the trash and searches for the episode again even if a pack covers the season or you already have later episodes. A special SxxE00 does not cover the season.

A file is not renamed¶

Check the Library and renaming settings and the TMDB key (or TVDB and its PIN): without metadata the episode title becomes «Episodio N». Having mediainfo installed helps.

A watched folder does not import the file¶

The file must have a .torrent or .magnet extension. Gextto waits until size and date are stable, then retries errors by itself. Check the log lines about the watched folder.

The interface does not open¶

Check the service: systemctl status gextto (or systemctl --user status gextto).
If the log says the data folder is already in use or the port is taken, another instance is running: stop it and restart.
After a restart Gextto can take about a minute to be ready.

The FTP backup fails¶

Use Test FTP: it tells you which step fails (connection, login, path, upload, removal) and writes it in the log.

Where the logs are¶

Gextto: <data>/gextto.log (rotated at 5 MB × 4), or the Log page.
gx-torrent: <data>/gx-torrent/gx-torrent.log, or Maintenance → gx-torrent log.
aMule: eD2k → aMule log.
Updates: /var/log/gextto-update.log.

27. Glossary¶

Term	Meaning
Alias	another name under which a series appears in releases
Archive	the local catalogue of the releases seen in the feeds
Backoff	the growing pause in requests to a failing source
Blocklist	the releases that will not be downloaded again
Cycle	one automatic pass of searching and choosing
Dry-run	test mode: searches and decides, but does not download
Forced gap	the explicit request to download an episode again (↻ Redownload)
Gap	a missing episode among the ones you have
Hardlink	a second name for the same data, without taking space twice
Indexer	a service that searches releases on one or more sites (Jackett, Prowlarr, indexarr)
Library	the final folders with series and movies
Placeholder	the temporary row of a download not yet complete
Quality profile	a template of quality, language, subtitles and exclusions
Release	a result found by a feed, an indexer or a web engine
Season pack	a torrent with a whole season
Seeding	sharing a torrent after it completes
Stalled	a torrent that has received no data for too long
Torrent session	the torrents still handled by the engine
Trash	where files taken out of the library go before deletion
Upgrade	replacing a file with a better version

Appendix A. Reference — Configuration¶

The page is split into four areas, each with its sections: What to look for (General, Sources, Quality and upgrades), How to download (Torrent engine, Speed and network, Queues and performance, Seeding and completion), Where to store (Library and renaming, Archive and space) and System (Automatic maintenance, Notifications, Access and services, Diagnostics and updates). On a computer the sections are in the left column, with the number of settings; on a phone you pick them from the Section list. The page address follows the open section, so it can be bookmarked. Links to the old tabs (Acquisition, Advanced, Translations) open the section that took over their content (Translations opens Diagnostics and updates). Entering Configuration hides the main menu bar; the Menu button at the top left brings it back over the content (Esc closes it).

When a section gathers several groups of settings (for example Search cycle and Missing episodes under Automatic maintenance), the group title stays stuck to the top while you scroll, with a coloured accent bar.

Under the section title there is the group index (clickable chips) to jump straight to a group. The Modified only checkbox, next to Show technical keys, hides every setting still at its default value: handy to review only what you changed. Rows highlight on hover.

How to read a row:

Name and description on the left; the control in the middle with its unit (seconds, minutes, GB, KiB/s…) and, below it, the meaning of the special values (for example “0 = never”).
Yes/no options are switches; times use the time picker. In Speed and network, How the speed is decided picks the rule: Auto (Gextto measures the line and adapts), Fixed limits (the two global limits) or Hourly schedule. With the schedule you get the table with one row per weekday, where every day has one choice among Off, Limit (your values, the only ones using the speed fields) and Auto, plus start and end, and Outside the hourly windows: Auto, which says whether Auto or the fixed limits apply outside the windows. The older single-window settings no longer appear on the page: until the table is saved it shows their values and they stay in use.
Options that depend on a switch or a choice (for example the speed schedule or RAM disk fields) appear only when they matter, even before saving.
A value different from the default shows changed, the Default value and the ↺ Default button, which puts the default back in the field (it still has to be saved).
The technical key (e.g. refresh_interval) is shown with Show technical keys; the search finds it anyway.
Settings the active torrent engine does not use are collected, closed, in Not used by the active engine (or in the panel of the inactive engine): they stay saved and become active again when you switch engine. With qBittorrent-nox this includes the network, protocol, queue, cache and RAM disk options only gx-torrent reads: qBittorrent keeps its own configuration. Rare options are in the closed For experts panel.

Each row is saved with its own Save. A field changed and not yet saved is marked not saved and the Unsaved changes bar appears at the bottom, with Save all and Discard changes. Changing section or leaving the page with pending changes asks for confirmation. The editors (feeds, filters, rules, hooks) have their own Save button and also show not saved until you save them.

What to look for¶

General¶

When and how often Gextto looks for new releases and missing episodes.

Search cycle

Setting	What it does
Automatic cycle (search and download)	The master switch of the cycle: when off, Gextto neither searches nor downloads on its own. On by default; with dry-run on it searches without downloading.
Automatic search interval	Interval between automatic series/movie searches, in seconds (21600 = 6 hours). (Unit: seconds)
Online title search in the cycle	Search titles on indexers during the cycle: auto (only when no feeds are configured), yes (always), no (never, use only feeds and the local archive).
Maximum release age	Ignore releases older than N days (0 = no limit). Releases without a date are treated as published today. (Unit: days; 0 = no limit)

Missing episodes

Setting	What it does
Search for missing episodes	Enable filling recognised gaps (missing episodes) from available releases.
Missing episodes per series in each cycle	Maximum gaps to search per series in one cycle (0 = unlimited). (0 = unlimited) Shown only while «Search for missing episodes» is on.
Deep search interval	How many hours between targeted live indexer searches for gaps. (Unit: hours) Shown only while «Search for missing episodes» is on.
Deep searches per cycle	Maximum number of live searches per cycle. The episodes never searched online come first, one per series in rotation (each pass starts after the series served last), then the ones searched longest ago. When the torrents have nothing, aMule answers once per season: one search fills every missing episode of that season, as Search missing does. Shown only while «Search for missing episodes» is on.
Last resort: accept a resolution outside the profile	On by default. When a missing episode has no copy within the quality profile, Gextto takes the one in the requested language even with a resolution outside the profile or unknown; this also covers copies found on eD2k, which often do not state the resolution in the name. Language, subtitles, audio and exclusions always apply. The download carries the «🆘 last resort» badge in Downloads, the log and the notification say so, and the upgrade replaces it as soon as a copy within the profile appears. Shown only while «Search for missing episodes» is on.

Delay before downloading

Setting	What it does
Delay for series	Delays the start of series downloads by this many minutes. 0 starts immediately. (Unit: minutes; 0 = none)
Delay for movies	Delays the start of movie downloads by this many minutes. 0 starts immediately. (Unit: minutes; 0 = none)
Skip the delay from this score	If a release reaches at least this score, the configured delay is ignored. (0 = never)

Sources¶

Where releases come from: RSS feeds, web engines, filters and watched folders.

Feeds and blacklist

Setting	What it does
Blacklist (one word per line)	Forbidden words, one per line: releases containing them are discarded.
Feed pages to read	How many listing pages to read per feed (3 is a good compromise).
Disable feeds after days of errors	After how many days without a success the feed is disabled automatically with a `#` in front of its line. It only fires when another feed worked on the same day; `0` turns the rule off.

Editors

Setting	What it does
RSS feeds	List of RSS feeds read on every cycle (one URL per line). Put `#` at the start of a line to exclude it without deleting it: `# https://…` or `#https://…`.
Web engines	Web search engines used by gap-filling when feeds and indexers find nothing.
Excluded content filters	Releases containing these words or scripts (e.g. [porno]) are excluded.
Source filters	Keywords to accept or reject for a single source, enabled per row.
Watched folders	The .torrent and .magnet files found in the given folders are downloaded and added to the session automatically (recursive, delete after).

Below the editor there is the Disabled feeds frame: it lists the lines with # and says whether they were disabled by hand or automatically, with the date and the days of failure. The full activity — state, last success, last error, last results and attempts/successes/failures/releases taken over 1, 7 and 30 days — is on the Health → Feed activity page, with the last-24-hours summary. The panel tells feeds that share host and path apart through their parameters (for example ext.to's filter=u=…), masks credentials and, until a first attempt, says “waiting for the first cycle”. Removing the # and saving enables a feed again.

Many feeds can point at the same site (the uploader listings of ext.to or TorrentGalaxy). Gextto reads them spaced apart, not all together, and when the site answers “too many requests” it leaves that site alone: the pause starts at ten minutes and grows each time it happens again, up to six hours, while the other feeds of the same site wait for a later cycle. The pause clears by itself; you do not have to disable the feeds.

eD2k and Kad¶

Connection to aMule, search on Kad and the eD2k servers, automatic fallback and stalled downloads.

eD2k and Kad go through aMule (see eD2k and Kad with aMule): Gextto drives an amuled that is already running, through external connections (EC). It is off by default: without your consent Gextto does not contact the network.

Setting	What it does
eD2k and Kad through aMule	Connects Gextto to aMule: search, `ed2k://` links, downloads in Downloads and the eD2k page. Off by default.
aMule — address	Machine running `amuled`, usually `127.0.0.1`. Gextto must see aMule's incoming folder at the same path.
aMule — external connections (EC) port	`ECPort` in `amule.conf`, 4712 by default.
aMule — external connections (EC) password	The password in clear; aMule keeps its MD5 in `ECPassword`. Not displayed.
aMule folders	The incoming (download) and temporary folders are shown read-only: they belong to aMule and are changed in aMule/amuled. After the change restart Gextto (or reopen this page) so it uses the new path.
eD2k search network	Kad (default, no server needed), Kad + eD2k servers (both searches, results merged), eD2k servers (global) (every server) or eD2k server (connected). Applies to the manual search; the automatic fallback always searches Kad and the servers. The same choice is on the eD2k page.
Use eD2k when torrents find nothing	A second, separate switch: when the torrent search for the episode and the season produces nothing, Gextto searches the same episode on eD2k/Kad (with the series name and every alias). It applies both to the deep gap-filling pass and to Search missing and single-episode search. Off by default.
Minimum sources for the fallback	eD2k source floor for accepting a result automatically. Manual search shows everything. (Unit: sources; 0 = no minimum)
eD2k search duration	How many seconds aMule collects results before Gextto shows them. (Unit: seconds)
Give up on a stalled eD2k download after	An eD2k download chosen by Gextto that receives no data for this many days is cancelled in aMule, the release is blocklisted and the episode is searched again. Downloads added by hand and paused ones stay. (Unit: days; 0 = never)

The automatic fallback only accepts a result that matches the episode exactly, has a recognised quality, a plausible size and at least the source floor. In every other case the result stays visible in manual search and the gap stays open: a missing episode is better than a wrong file in the library. If aMule is unreachable, torrents continue unaffected.

On the Sources page the eD2k and Kad entry shows the connection sensor: open (connected to a server with a high ID), limited (firewall/NAT: low ID or Kad only), not connected, aMule unreachable or unknown. When it is limited, open aMule's TCP and UDP ports on the router.

Quality and upgrades¶

How releases are compared and when a better version replaces the one in the library.

The page starts with what is needed most often and leaves the weights last:

Upgrades and replacement — the rules that decide when a better release replaces the one in the library (table below).
Examples with the current weights — reference scores (1080p WEB-DL, 1080p H.265 DD+, 2160p Dolby Vision) and the stop threshold. Impact on the library — Check now replays the releases already seen in the feeds with the current rules and weights and says how many archived episodes would be replaced, with a few examples. Use it after changing the weights or thresholds: a high number means the next cycles would re-download a lot. It downloads nothing; series with “Allow upgrades” set to No do not count.
Test a release — paste a release name: Gextto shows the score item by item, the copy already in the library with its score and whether it would replace it, with the reason. It downloads nothing.
Quality profiles — the templates of quality, language, subtitles and exclusions (see Quality profiles); the list is the one exposed to Jellyseerr.
Video, Audio, Release fixes — the weights, in closed cards: open them only to change the preferences. For each: higher = more preferred.
Custom groups.

The Custom groups panel adds a bonus or a penalty to a release group (the name must match the final tag of the release, e.g. TBK; case does not matter). The library scores are recomputed automatically when you change the weights or when a Gextto update changes the scoring rules, so a change is never mistaken for an upgrade; Maintenance → Recalculate scores is still there to force it.

An upgrade by score needs a real quality gain: resolution, source, codec, audio, HDR/DV, fixes and group must be worth at least the minimum difference more. File size, subtitles and swarm health help choose between equivalent releases, but on their own they never re-download an episode or a movie of the same quality. The same rule holds for season packs and after the download: a file of the same quality does not replace the archived copy just because its score is higher. When the archived copy does not declare the source or the group (for example a file imported or renamed with the title only), those fields are not credited to the release found: without proof that the copy is worse, it is not replaced. ffprobe fills resolution, codec, audio and channels when the name has lost them.

The rule that prefers a WEB-DL source to a WEBRip or HDTV one follows the weights too: it holds only while the WEB-DL Source weight is higher. Give WEB-DL and WEBRip the same weight and Gextto treats them as the same quality: a WEB-DL no longer replaces a WEBRip (HDR, resolution, REPACK/PROPER and the score still count).

MediaInfo check before entering the library. When a completed download is about to replace an archived episode or movie, Gextto reads with ffprobe the new file, still in the download folder, and the copy in the library, before moving, linking or copying anything. Resolution, video codec, audio, channels and HDR10/HLG are taken from the real streams, not from the name. If the name promised more than there is (a «2160p» that is an upscaled 1080p, an «HDR» without HDR, a «DDP 5.1» that is stereo AAC) and the real quality is not an upgrade, the download is discarded and blocklisted and the library copy stays as it is; the log explains it with «Upgrade not confirmed by MediaInfo». Source, Dolby Vision and REPACK/PROPER stay as named, because ffprobe does not tell them apart reliably. Without ffprobe, or with the archived copy unreachable, the decision taken before the download stands. Season packs do not go through this check.

Upgrades and replacement

Setting	What it does
Min score difference for upgrade	Minimum score difference to replace a file with a better upgrade. Only quality counts: size, subtitles and swarm health are not enough to reach it.
Stop upgrading above this score	Upgrade ceiling: once the library file scores at least this much it is no longer replaced by better releases; only a REPACK or PROPER, which fixes a defective release, is still accepted. With the default scores: 1080p WEB-DL H.264 ≈ 1280, 1080p WEB-DL H.265 DD+ ≈ 1480, 2160p WEB-DL ≈ 2480. It adds to the minimum difference: that one says by how much a release must improve, this one up to where. To stop upgrades for a single title set “Allow upgrades” to No with Edit selected (or in the profile you add it with). “Why not this one?” says when a release is skipped because of this ceiling. (0 = always upgrade: no ceiling)
Replace versions already archived	Replaces lower versions already archived with better upgrades.
Min score difference for cleanup	Minimum score difference to replace an existing file with a better one (cleanup).

Stuck downloads

Setting	What it does
Try an alternative after (stall)	If a download has been stuck for at least these minutes — at 0 B/s or still waiting for its file list (metadata, typical of dead magnets) — the search may start an alternative version of the same episode or movie. The stuck download stays in the session and may still recover; archive deduplication will then keep the best copy. (0 = never)
Score drop for the alternative	How many points below the stuck release an alternative may score: only an adequate alternative is started, never a poor version. The stuck download is not removed. (0 = disabled)

How to download¶

Torrent engine¶

Which engine downloads the torrents and how Gextto reaches it.

Torrent engine

Setting	What it does
Torrent engine	Active torrent engine (gx-torrent or qBittorrent-nox).

gx-torrent

Setting	What it does
gx-torrent — Web API URL	gx-torrent API URL (e.g. http://127.0.0.1:8890); Gextto drives the daemon there.
gx-torrent — listen address (LAN)	Listen address of the managed daemon. Default `127.0.0.1:8890`: the web page and API stay on this host only. To expose them on the LAN use `0.0.0.0:8890` and set `gxtorrent_token` (without a token Gextto starts the daemon in `-insecure`). The port is aligned with the URL.
gx-torrent — automatic management (cache and queue)	Turns on gx-torrent self-management: dynamic queue and adaptive cache (sized on available memory, active downloads/seeds and storage type). On by default. Turn it off to set cache and slots by hand.
gx-torrent — proxy (socks5:// or http://)	Proxy for peers, HTTP trackers, web seeds, the IP filter list, RSS feeds, Torznab indexers and .torrent downloads; localhost and the local network stay direct. With a proxy, DHT and UDP trackers are disabled and FlareSolverr is not used as a fallback (not shown).
gx-torrent — announced client identity	What peers and trackers see: `gx-core` with the real version (default) or a qBittorrent, Transmission, Rain or anacrolix/torrent preset. It changes only the announced identity, not the protocol: some private trackers only accept whitelisted clients, check their rules.
gx-torrent — announced client version	Version used with the preset; empty uses the preset's default.
gx-torrent — recheck data when a download completes	Verifies the files of a finished download once, before seeding them: a disk that returns a wrong block is caught at once instead of being seeded for hours. It costs one full read of the data. Applies to downloads completed after it is turned on.
gx-torrent — anonymous mode	Hides the client version from peers and public trackers: the peer id loses it, the handshake sends no version and public trackers receive no User-Agent. Private trackers keep the configured agent. It does not hide your IP address: a VPN does that.
gx-torrent — announce to all trackers in a tier	Contacts every tracker of a multi-tracker tier at once and merges the peers, as qBittorrent does (on by default). Turn it off to contact one tracker per tier and move to the next on failure.
gx-torrent — request timeout	Timeout in seconds for HTTP requests to gx-torrent. (Unit: seconds)
gx-torrent — polling interval	Minimum interval in milliseconds between torrent status reads (the daemon owns the queue). (Unit: ms)
gx-torrent — DLNA/UPnP server	Enables the daemon's DLNA/UPnP media server: the torrents' completed media is visible to TVs and consoles on the local network. Local network only, no authentication; off by default. Works with the gx-torrent engine.
gx-torrent — DLNA: HTTP bind	Address and port the media server listens on (e.g. `:8200` for all interfaces, `192.168.1.10:8200` for one). Empty = `:8200`. Discovery uses UDP 1900 on the same interface.
gx-torrent — DLNA: server name	Name shown to the DLNA players on the network. Empty = `gx-torrent`.
gx-torrent — DLNA: library folders	Folders (e.g. the archive on the NAS) the media server shows beside the torrents, browsable by folder; separate them with `;`. Empty = automatic: the archive folder and the folders holding the series archives. Read only; the trash, hidden folders and NAS system folders (`@eaDir`, `#recycle`) are excluded.

qBittorrent-nox

Setting	What it does
qBittorrent-nox — downloaded and updated by Gextto	Gextto downloads the latest qBittorrent-nox release itself, installs it in the application folder (next to gexttod), starts/stops it with the service and updates it (with backup and rollback). The engine in use is still chosen in “Torrent engine”: this option does not change it.
qBittorrent-nox — Web API URL	qBittorrent-nox Web UI URL (e.g. http://127.0.0.1:8080).
qBittorrent-nox — user	qBittorrent-nox Web UI username.
qBittorrent-nox — password	qBittorrent-nox Web UI password (not shown).
qBittorrent-nox — category	Category applied to torrents added to qBittorrent-nox.
qBittorrent-nox — tag	Tag applied to torrents added to qBittorrent-nox.
qBittorrent-nox — request timeout	Timeout in seconds for HTTP requests to qBittorrent-nox. (Unit: seconds)
qBittorrent-nox — polling interval	Interval in milliseconds between torrent status reads. (Unit: ms)
qBittorrent-nox — path mapping	Needed only when qBittorrent sees the folders under different paths (Docker, NAS, another host): one line per path, local=remote form. Propose reads qBittorrent's download and temporary folders and fills the lines; check them and press Save. Same machine with the same folders — or managed mode —: leave it empty.

Actions

Setting	What it does
Apply engine	Checks the configured engine and tells whether a restart is needed (Torrent engine panel).
Install / Optimize qBittorrent-nox, qBittorrent-nox status, Connection test	In the qBittorrent-nox panel: install or update the managed binary, show where it is installed and whether an update exists, check the connection.

Speed and network¶

Bandwidth limits, time slots, ports and protocols of the torrent session.

Speed limits

Setting	What it does
How the speed is decided	The speed rule: Auto (Gextto measures the line and adapts, no cap), Fixed limits (the two limits below) or Hourly schedule (windows per day, each with limits or Auto). In Downloads, from the Bandwidth menu, you can always make a temporary exception, which beats the rule. It is not a setting of its own: it turns «Speed schedule enabled» and «Outside the hourly windows: Auto» on or off.
Global download limit	Global download limit in KiB/s (0 = unlimited). (Unit: KiB/s; 0 = unlimited) Shown with «Fixed limits» and «Hourly schedule».
Global upload limit	Global upload limit in KiB/s (0 = unlimited). (Unit: KiB/s; 0 = unlimited) Shown with «Fixed limits» and «Hourly schedule».
Speed schedule enabled	Set from «How the speed is decided» and not shown on the page; still available through the API.
Outside the hourly windows: Auto	With the hourly schedule: outside the windows the speed is Auto instead of the fixed limits. Shown only with «Hourly schedule».
Schedule — per day	One choice per day: Off (no window), Limit (the window with your download and upload values) or Auto (the window in which Gextto measures the line and adapts). Only «Limit» enables the two speed fields; the times stay editable. A window past midnight belongs to the day it starts on. (a «Limit» day has its own values (0 = unlimited); «Auto» lets Gextto measure the line and adapt; a window past midnight belongs to the day it starts on) Shown only while «Speed schedule enabled» is on.
Schedule — start time	Schedule start time (HH:MM). Not shown on the page: replaced by the «Schedule — per day» table, which shows its value until saved.
Schedule — end time	Schedule end time (HH:MM). Not shown on the page: replaced by the «Schedule — per day» table, which shows its value until saved.
Schedule — days	Active days: 0=Mon … 6=Sun (e.g. 0,1,2,3,4). (no day = never active) Not shown on the page: replaced by the «Schedule — per day» table, which shows its value until saved.
Schedule — download	Download limit in KiB/s during the schedule. (Unit: KiB/s; 0 = unlimited) Not shown on the page: replaced by the «Schedule — per day» table, which shows its value until saved.
Schedule — upload	Upload limit in KiB/s during the schedule. (Unit: KiB/s; 0 = unlimited) Not shown on the page: replaced by the «Schedule — per day» table, which shows its value until saved.
Auto respects latency	With the speed on Auto, Gextto measures the ping to Google and Cloudflare every 2 seconds (8.8.8.8, 8.8.4.4, 1.1.1.1, 1.0.0.1; the lowest time is used). While the ping is good it raises download and upload; when it rises it cuts 15% and goes down to 128/64 KiB/s. It remembers the speed at which the ping degraded and settles just below it, trying to pass it once a minute. If the ping stays good at full speed the cap is removed; when only background traffic (DHT) is left, the cap is dropped. The first time it applies a cap it writes it to the log at once; after that it writes at most one line a minute, only when the speed changes by at least 10% and the trend confirms it. On an asymmetric line upload counts as load from a tenth of the capacity (between 32 and 256 KiB/s) and is capped before it saturates; if it starts loading during the control phase the cap starts at once. The measured ping appears next to the ↓ Download speed in the top bar. On by default.
Auto — what measures the latency	`both` (default): the ping decides, the router counters are a diagnostic and a brake. `ping`: ping only, the router is not read. `router`: old behaviour, yield only to the other devices the router sees. In `both` and `router`, when the router sees another device active, Auto takes no more room: the caps stay where they are and do not drop. This is how it respects a call or a game, which build no queue and so do not move the ping; in `ping` this cannot happen by construction, because the router is not read.
Auto — host to measure latency	Host (or host:port) to measure the latency to with a TCP connection every 2 seconds, only while Auto is on. Empty = Google (8.8.8.8, 8.8.4.4) and Cloudflare (1.1.1.1, 1.0.0.1), port 443; the lowest time is used, so a distant or slow server cannot fake the measurement.
Auto — line maximum, download	The real maximum download speed of the line, in KiB/s (e.g. 9000 for 9 MB/s). 0 = unknown: Auto discovers it. If you enter it, Auto never goes above it and finds the right point sooner.
Auto — line maximum, upload	The real maximum upload speed of the line, in KiB/s (e.g. 800). 0 = unknown: Auto discovers it. If you enter it, Auto never goes above it and finds the right point sooner.
Apply the limits to peers on the local network too	Off (default, recommended): the peers of your home or office network (private addresses 10.x, 172.16–31.x, 192.168.x, link-local, loopback and local IPv6 fc00::/7) download and upload at full speed, as in qBittorrent and libtorrent: the speed limits are meant for the Internet line. On: they follow the global and per-torrent limits too. With a proxy configured every peer follows the limits (behind a proxy gx-torrent cannot tell where a peer is). gx-torrent only (qBittorrent has its own option). Changing it restarts gx-torrent.

Ports and interfaces

Setting	What it does
Min port	Minimum torrent session port (requires service restart).
Max port	Maximum torrent session port (requires service restart).
Listen interfaces	Where the engine accepts connections: 0.0.0.0:6881-6891 for all interfaces, 127.0.0.1:6881 local only, or wg0:6881/tun0:6881 for a VPN. The suggested value works in most cases.
Outgoing interface	VPN killswitch: interface used for all outgoing BitTorrent traffic.
Test ports	Checks that the peer port is listening and that the router forwards it (the message says "open", "listening but not forwarded" or "closed"). Available with the gx-torrent engine.

Peer and tracker discovery

Setting	What it does
DHT	Enable the DHT network to find peers without a tracker.
PEX	Peer Exchange: exchange peers with other clients.
LSD	Local Service Discovery: finds peers on the local network.
UPnP	Opens router ports automatically with UPnP.
NAT-PMP	Opens router ports automatically with NAT-PMP (PCP, RFC 6887, is tried too).
uTP	Enable the uTP (UDP) protocol in addition to TCP.
Holepunching (BEP 55)	Opens a direct connection to peers behind NAT through a relaying peer (BEP 55). Requires uTP.
DHT bootstrap nodes	Initial DHT nodes (host:port separated by comma).

Encryption and IP filter

Setting	What it does
Encryption	Encryption policy: 0 disabled, 1 enabled, 2 forced.
Apply IP filter	Apply the IP filter to HTTP trackers as well. It is off by default: the bundled list contains Cloudflare ranges, so trackers hosted there (many private ones) would stay unreachable and the Trackers tab would show "tracker IP is blocked". Peer filtering is always on.
IP filter (file/URL)	Local file or URL of the IP blocklist. The default is the eMule Security list (`https://upd.emule-security.org/ipfilter.gz`), in P2P format and updated every day.

Actions

Setting	What it does
IP filter → Load / update now	Downloads the list from the configured URL (or uses the local file) and applies it to the active engine right away, gx-torrent included.

Queues and performance¶

How many torrents stay active, connections, RAM disk and cache.

Automatic management

Setting	What it does
Adapt values over time	Adjusts over time the values the system chose (the minimum free space, for instance) when conditions change. It never touches values you set by hand; turn it off to keep them fixed.

Queue and slots

Setting	What it does
Automatic queue and resource management	Automatically adjusts how many torrents are active based on load.
Minimum dynamic download slots	Minimum number of dynamic downloads. The queue changes by at most one at a time. Shown only while «Automatic queue and resource management» is on.
Maximum dynamic download slots	Maximum number of dynamic downloads. Consecutive consistent samples are required before increasing the queue. Shown only while «Automatic queue and resource management» is on.
Don't count stalled torrents in active slots	Torrents that are not transferring data do not consume an active slot.
Active downloads	Base value for active downloads; with the dynamic queue it is adapted at runtime.
Active seeds	Base value for active seeds; with the dynamic queue it drops to 1 when downloads are queued.
Active torrents limit	Base value for the active torrent limit; with the dynamic queue it becomes max(base, downloads + seeds + 2).

Download mode

Setting	What it does
Sequential download	Download files sequentially instead of in scattered pieces.
First/last piece of files	Download the start and end of every file first in all new torrents, then continue normally: useful to watch a video while it downloads. Works with every engine; torrents already running are not changed.
Preallocate disk space	Reserves all disk space up front before starting the download.

Connections

Setting

What it does

Automatic connections (sized from the machine)

On (default): Gextto lets the engine decide how many connections to keep for each torrent from the machine's cores and RAM, and trims the ones on torrents that have made no progress for a couple of minutes despite many peers. With many torrents active together the engine also keeps a total for the whole machine (75 connections per core), shared among the torrents with downloads first; a torrent on its own always gets all of its connections. The value below stays the ceiling. Turn it off to use a fixed number. gx-torrent only.

Max connections per torrent

Ceiling on the connections per torrent (-1 = unlimited, the default). With automatic connections on it is the most the engine may use; with it off it is the fixed limit. It applies to every torrent without a limit of its own (Limits tab of the detail) and takes effect at once, running torrents included; gx-torrent only.

RAM disk

Setting	What it does
Use the RAM disk	Downloads to RAM the torrents that fit the threshold; larger ones go to disk.
RAM disk folder	RAM disk to use for in-progress downloads, if available. Shown only while «Use the RAM disk» is on.
Maximum size per torrent	Maximum size of a single torrent allowed on the RAM disk (GB). (Unit: GB) Shown only while «Use the RAM disk» is on.
Free margin to keep	Free space to leave on the RAM disk once the download completes (GB). (Unit: GB) Shown only while «Use the RAM disk» is on.
Minimum free space	Minimum free space in bytes required to use the RAM disk. 0 = use the configured margin. (Unit: bytes; 0 = use the margin) Shown only while «Use the RAM disk» is on.

The choice is made as early as possible: a torrent whose size is already known and exceeds the threshold goes straight to disk; one with an unknown size (as with most magnets) stays on disk until the metadata arrives, and is then moved onto the RAM disk only if the real size fits. An oversized file never touches the RAM disk. The rule applies to every download — automatic and manually added — and to every engine that uses the RAM disk; qBittorrent, which does not support it, always downloads to disk.

Gextto also checks the machine's available memory: a tmpfs lives in RAM, so a 6 GB mount can look almost empty while the system has little memory left. Before starting, the RAM disk's usable space is capped by the available memory (MemAvailable): a download goes to the RAM disk only if both the tmpfs space and the system memory can hold it, otherwise it stays on disk like oversized files. When the memory cannot be read, only the tmpfs space counts, as before.

For experts

Setting	What it does
Disk cache	Disk cache size in blocks (-1 = automatic). (Unit: blocks; -1 = auto)
Cache expiry	Seconds of inactivity after which a block leaves the cache. (Unit: seconds)

Actions

Setting	What it does
Optimize	Computes cache and buffers from the RAM.
Optimize for the system	Proposes the values that fit this hardware (memory, disks, network share, RAM disk), one row each with the reason; nothing is written until you apply the selected rows. A value you saved by hand is never proposed again.

Seeding and completion¶

How long to seed, what to do when a download finishes, how to inspect downloading packs and how to handle stalled torrents.

Seeds

Setting	What it does
Global seed ratio	Upload/download ratio after which seeding stops. 0 = infinite: this alone makes the seed never end and the file is not moved, even with a time limit. (0 = infinite)
Maximum seeding	Primary seeding limit in days; when greater than 0 it takes precedence over the limit in minutes. With days and minutes at 0 the time is infinite. (Unit: days; 0 = use the limit in minutes)
Maximum seeding (fallback)	Seeding limit in minutes, used only when Maximum seed (days) is 0. With days and minutes at 0 the time is infinite: the seed never ends and the file is not moved. (Unit: minutes)
Stop an idle seed after	Stops a seeding torrent that has sent no data for this many minutes (only time spent actively seeding counts). Like the other limits, the end of the seed is followed by archiving or removal. With qBittorrent the value is also sent as `inactiveSeedingTimeLimit`. (Unit: minutes; 0 = off)

Editor

Setting	What it does
Seed rules	Seeding limits (ratio, days, inactivity) by source/indexer and category; the first matching rule applies, an empty field uses the global limit, 0 removes the limit.

When a download completes

Setting	What it does
Remove completed items after seeding	On: at the end of seeding the completed torrent is removed from the session (like “Clean completed”). Off: at the end of seeding the torrent stays in the list as Completed and you remove it with “Clean completed”. It does not affect where files are moved.
Move episodes/packs to the archive (do not copy)	On: the downloaded source is moved into the library and removed from the download folder (for torrents at the end of the seed; for eD2k at completion, because seeding is managed by aMule). Off: the source is copied or linked and kept for sharing.
Hardlink instead of copy while seeding	On (default): a file that keeps seeding enters the library as a hardlink, without using the space twice; when downloads and library are on different filesystems it is copied. See “Hardlinks instead of copies”.

Pack inspection

Setting	What it does
Inspect downloading packs and discard useless ones	On: as soon as the engine knows the file list, Gextto checks whether at least one episode is really wanted (missing from the library or a real upgrade). If none is, the torrent is removed with its partial data and blocklisted, so it is not downloaded again. Off (default): the pack is downloaded in full and decided only when the download ends.
In useful packs, download only the files that are needed	On (with inspection): when the pack holds only some useful episodes, the engine downloads only those files and skips the others (priority 0); subtitles and files without an episode number stay active. Off: the useful pack is downloaded in full.

Stalled torrents

Setting	What it does
Consider stalled after	After this many minutes without progress the torrent is considered stalled. (Unit: minutes)
Retry stalled torrents every	Interval between reannounce attempts for stalled torrents. (Unit: minutes)
Remove stalled torrents after	After this period without progress, the torrent is removed automatically. Set 0 to completely disable automatic removal due to stalling. (Unit: minutes; 0 = never)
Remove torrents without seeders after	Stall with zero seeders (dead swarm): removed after this shorter period, and the episode is searched again right away. Set 0 to use the general threshold. (Unit: minutes; 0 = use the general threshold)
Let a release without seeders back in after	A release with an empty swarm — abandoned during the download or refused before starting because nobody was sharing it — stays blocklisted for this period, then it is back among the candidates: a 4K that has found seeders in the meantime can be downloaded as an upgrade, and the log stops repeating the same line at every search. A release that comes back and is still empty stays away four times as long, and twelve times as long the third time, up to three months; the count is forgotten after six months, and a download that succeeds starts it over. The copy already in the library stays meanwhile. Set 0 to keep it blocked forever. (Unit: days; 0 = forever)

Where to store¶

Library and renaming¶

How files are named and organized in the library.

Rename

Setting	What it does
Rename episodes	Renames downloaded files using TMDB metadata (or TVDB without a TMDB key).
Movies as single files (flatten folders)	If Yes, a movie that arrives inside a torrent folder is moved to the movies folder as a single file, with its subtitles and artwork. If No, it stays in its folder.
Check renamed files every	How many hours between checks that archived/renamed files are still present. (Unit: hours)
Remove audio tracks in other languages	If Yes, after importing Gextto keeps only the audio tracks in the chosen language and removes the others, without re-encoding the video: quality, HDR and sync stay those of the original file. A file is touched only when it has more than one audio track and at least one is tagged in the chosen language. Before replacing it Gextto checks the copy: the same tracks in the same order, with the same codec, language and flags (a subtitle that was off stays off), the same duration, the same sync, and every packet of the video, of the kept audio and of the subtitles identical to the original, byte for byte. For MKV files it tries mkvmerge and, when the copy fails the check, makes it again with ffmpeg; when neither passes, the file stays as it is and the log says so. When the default track was among the removed ones, the track in the chosen language becomes the default. The temporary copy is written in the same folder: when there is not enough free space the file stays as it is. On network shares (NFS, SMB) the permissions are not touched: the share decides them. Files imported while the option is on are cleaned; those waiting or in progress when Gextto stops are cleaned after the restart. When the library file is a hardlink of a torrent still seeding, the cleaned copy is a new file: while seeding lasts the space is used twice. The option can be enabled only when ffprobe, ffmpeg and mkvmerge are installed; otherwise the row stays blocked with a warning. Every cleaned file sends a notification on the configured channels (Telegram, webhook, email) with the file, the removed tracks and the space saved.
Audio language to keep	Choose from the list (Italian, English, French…) or type the two or three letter code, for example ita or eng. The spellings of the same code are equivalent: it, ita, it-IT and Italian for Italian, pl and pol for Polish, and so on for every language in the list. Tracks declared in another language are removed; for untagged tracks, «Audio tracks without a language» decides.
When several tracks are in the chosen language	Keep them all, only the best one (more channels and bitrate) or only the first.
Audio tracks without a language	What to do with audio tracks that declare no language (no tag, und, mul…), which are often the original audio. Keep them (default): they stay in the file next to the tracks in the chosen language. Remove them: they are removed with the other languages. Leave the file untouched: a file with even one of them stays as it is.

Metadata languages

Setting	What it does
Default language	Default preferred language for series and movies (e.g. ita, eng).
TVDB language	Preferred language for TVDB metadata (e.g. ita, eng).
TMDB language	Language used for TMDB metadata (e.g. it-IT, en-US).

Editors

Setting	What it does
Name composition	Template editor with tokens and preview to compose file names.
Tag → folder rules	Associates a torrent tag with a temporary and a final folder.

Archive and space¶

Working folders, trash and minimum disk space.

Folders

Setting	What it does
Archive folder	Default archive folder for content without a dedicated path.
Download folder	Default download folder for every engine.
Temporary folder	Temporary folder for in-progress downloads.
Copy .torrent files to	Copy the .torrent files of downloads here (empty = no copy).

Trash

Setting	What it does
Trash folder	Folder where replaced/duplicate files are moved (if left empty, uses the trash subfolder in the data folder).
What to do with replaced files	What to do with replaced files: move to trash or delete.
Trash retention	Retention days for non-forced cleanups; 0 deletes the whole trash content. Manual UI actions always empty the trash immediately. (Unit: days; 0 = empty everything)

Disk space

Setting	What it does
Minimum free space to download	Minimum free space (GB) on the download folder: below this threshold the cycle does not start downloads and gx-torrent pauses the running ones, which resume by themselves when space returns. (Unit: GB; 0 = no cycle check, gx-torrent keeps its default of 1 GiB)

System¶

Automatic maintenance¶

Periodic cleanup of technical data, MediaInfo analysis and orphaned files.

Housekeeping

Setting	What it does
Periodic housekeeping enabled	Enables periodic cleanup of technical data and history.
Housekeeping — interval	Interval between two automatic housekeeping runs, in hours. (Unit: hours) Shown only while «Periodic housekeeping enabled» is on.
Housekeeping — search-cycle statistics kept	Number of search cycle statistics to keep. Shown only while «Periodic housekeeping enabled» is on.
Housekeeping — seen in feeds	Deletes feed-seen release history rows older than N days. (Unit: days; 0 = never) Shown only while «Periodic housekeeping enabled» is on.
Housekeeping — download history	Deletes download history rows of removed torrents older than N days. (Unit: days; 0 = keep forever) Shown only while «Periodic housekeeping enabled» is on.
Housekeeping — error cards	Deletes error torrent entries older than N days (minimum 1). (Unit: days) Shown only while «Periodic housekeeping enabled» is on.
Housekeeping — gap search log	Deletes the missing-episode search log older than N days. (Unit: days; 0 = never) Shown only while «Periodic housekeeping enabled» is on.
Housekeeping — upgrade backups	Deletes backups of files replaced by upgrades older than N days. (Unit: days; 0 = never) Shown only while «Periodic housekeeping enabled» is on.

Release archive cleanup

Setting	What it does
Automatic archive cleanup	Enables automatic archive cleanup on every housekeeping pass: it removes only release-catalogue rows older than the maximum age. Downloaded files, the library, download history and active torrents are untouched.
Archive — maximum age	Maximum age of archive releases, in days (0 = no limit). (Unit: days; 0 = no limit) Shown only while «Automatic archive cleanup» is on.
Archive — keep at least N entries	Minimum number of recent releases to always keep, even when older than the maximum age. Shown only while «Automatic archive cleanup» is on.

Night maintenance

Setting	What it does
Night window — start	Start time (HH:MM, 01:00 by default) of the window in which the heavy jobs run: one job at a time and only while Gextto is idle, that is no download transfers more than 100 KB/s, no download is being archived, the library check and the integrity check are not running and the system load is low. Seeding does not count: it goes on all night.
Night window — end	End time (HH:MM, 07:00 by default). After it no new job starts; the one in progress is completed. An end before the start makes the window cross midnight; the same start and end mean the whole day.
Integrity check and MediaInfo analysis only in the window	Yes (default): the integrity check and the MediaInfo analysis run at their interval but only inside the window and while Gextto is idle. No: they run at their interval at any hour, as before. The manual commands (Check now, Refresh MediaInfo) always start at once.
Database maintenance every night	Yes (default): once a night the databases are optimised (`PRAGMA optimize`) and compacted when at least 20% or 64 MB of them is free space; once a week a full check (`integrity_check`) reads every page and every index, and a problem sends a notification at once, without waiting for the summary.
Clean the audio of files already in the library	Applies Remove audio tracks in other languages, with the same language and choices, to the files already archived: one at a time, in the window, with the same checks as the cleanup after an import. Every file is examined once (Gextto remembers its size and date); a file still seeding through a hardlink waits for the end of the seed, so the cleaned copy does not use the space twice, and a file written in the last hour waits for the next night. A file that fails the check for a passing reason (a tool interrupted, a disk that does not answer) stays as it is and is tried again after a week; a file that cannot be cleaned without changing it — the track check does not match — is remembered and not tried again unless the file changes (size or date). Off by default; it works only with the audio cleanup on.

Night maintenance — panel

Under the options, in the same group, the Night maintenance status panel shows the window, what is running or why it waits, and the outcome of each job.

Item	What it does
Run now	Runs a maintenance pass at once, even outside the window; it still waits until Gextto is idle.
Database	Last optimisation, total size of the databases and space freed, last integrity check and the problems found.
Audio cleanup of the library	Files checked out of the total, files cleaned, space saved, files left as they were and the last check not passed.
Library	The last pass of the library check (correct files, renamed, duplicates in the trash, errors: the check runs by itself every Check renamed files every hours) and the night report of the files the database knows but the disk no longer has. A folder that cannot be read, for example a NAS that is not mounted, is listed apart and does not count as missing files.
Mark as missing	Removes from the archive the episode whose file is gone, so the search downloads it again. It does nothing when the file is back or its folder cannot be read.

When the window closes, the log sums up the night in one line; a notification on the configured channels, in the maintenance group, also arrives when the night found database problems, missing files, cleaned files or files left as they were, or freed at least 64 MB.

MediaInfo analysis

Setting	What it does
Automatic MediaInfo backfill	Periodically analyses with ffprobe the files already present that have no MediaInfo yet.
MediaInfo backfill — interval	Minutes between two MediaInfo backfill passes. (Unit: minutes) Shown only while «Automatic MediaInfo backfill» is on.
MediaInfo backfill — files per run	Maximum number of files analysed in each MediaInfo pass. Shown only while «Automatic MediaInfo backfill» is on.

File integrity check

Setting	What it does
Automatic file integrity check	Periodically probes the library files with `ffprobe` and `ffmpeg`, one file at a time, to find the damaged ones.
File integrity check — interval	Minutes between two check passes. Shown only while «Automatic file integrity check» is on.
File integrity check — files per pass	Maximum number of files verified in each pass.
File integrity check — quick check for 4K files	For 4K files it reads the whole file but decodes only the keyframes: the check is much faster and still finds truncated or unreadable files, but it may miss a small defect between one keyframe and the next. It is on by default.
File integrity check — re-download damaged files	As soon as the check finds a damaged file it blocklists the release, moves the file out of the library and searches at once for a different release of the same episode or movie. It is off by default: without it the replacement stays manual, with Redownload.

File integrity check — panel

Under the options, in the same group, the panel shows the state and the files it found.

Item	What it does
Start check / Resume check	Starts the pass or resumes it from the first file not yet verified; files already checked and unchanged are skipped. It works with the automatic pass off too.
Stop	Stops the check after the current file; the position stays saved.
Start over	Starts again from the first file, clears the previous results and rechecks the whole library.
Damaged / Unreadable	The file could not be decoded, or it could not be read within the allotted time: an unreadable file is not treated as broken and is retried automatically in a later pass.
Redownload	Moves the damaged file out of the library, blocklists the old release and searches the same episode for a different release. With File integrity check — re-download damaged files on, Gextto does it by itself as soon as a file is reported as damaged.

Orphaned files

Setting	What it does
Move orphaned data in the temp folder to the trash	Every 30 minutes, moves to the trash the files and folders in the download temp folder that belong to no listed torrent and have not changed for at least the given number of days. Nothing is deleted: they stay in the trash until you empty it.
Orphaned data — idle for at least	How many days an item in the temp folder must stay untouched, with no listed torrent, before it is moved to the trash. (Unit: days) Shown only while «Move orphaned data in the temp folder to the trash» is on.

Notifications¶

Telegram, email, webhooks, Discord, Slack, Gotify, ntfy, Pushover, Apprise and programs to run on events. «Send test notification» sends a message on every configured channel.

Telegram

Setting	What it does
Telegram enabled	Send notifications to Telegram.
Telegram bot token	Telegram bot token (from @BotFather). Shown only while «Telegram enabled» is on.
Telegram chat ID	ID of the chat/channel where notifications are sent. Shown only while «Telegram enabled» is on.

Email

Setting	What it does
Email enabled	Send notifications by email.
SMTP	SMTP server as host:port (e.g. smtp.gmail.com:587). Shown only while «Email enabled» is on.
Email sender	Sender address for notification emails. Shown only while «Email enabled» is on.
Email recipient	Email recipients (comma-separated). Shown only while «Email enabled» is on.
Email password	SMTP password/app password (not shown). Shown only while «Email enabled» is on.

Webhook

Setting	What it does
Webhook URL	Webhook URL where events are sent.
Webhook secret	HMAC secret to sign webhook requests.
Webhook format	Format of the payload sent to the webhook URL (Gextto, Discord, Slack, ntfy, Gotify, Pushover).
Webhook token	Provider credential (ntfy, Gotify, Pushover); not shown. Not needed for Gextto, Discord and Slack.
Webhook user (Pushover)	Pushover only: the user key. Not needed for the other formats.

Discord

Setting	What it does
Discord webhook URL	Webhook URL of the Discord channel. Empty does not use Discord.

Slack

Setting	What it does
Slack webhook URL	Slack incoming webhook URL. Empty does not use Slack.

Gotify

Setting	What it does
Gotify URL	Address of the Gotify server (e.g. https://gotify.example); notifications are posted to `/message`.
Gotify token	Application token of Gotify, needed together with the URL.

ntfy

Setting	What it does
ntfy URL	Address of the ntfy server (e.g. https://ntfy.sh).
ntfy topic	Topic the notifications are published to.
ntfy token	Bearer token, only when the topic is protected.

Pushover

Setting	What it does
Pushover token	Application token of Pushover.
Pushover user key	User or group key of your Pushover account.

Apprise

Setting	What it does
Apprise URL	Address of an Apprise API server (e.g. http://192.168.1.10:8000); notifications are posted to `/notify`.
Apprise key	Optional key of one Apprise configuration; empty uses the default one.

Program

Setting	What it does
Program	Executable run on every event, with the data in the `GEXTTO_*` environment variables.
Program arguments	Arguments passed to the program, separated by spaces; quotes keep together the ones with spaces.

Events

Setting	What it does
Download notifications	Notifies when a download starts, completes or fills a missing episode.
Error notifications	Notifies when a download fails or a torrent or comic goes into error.
Health and engine notifications	Notifies about health, search cycle, indexer, provider login and torrent engine problems.
Maintenance notifications	Notifies about restarts, completed backups and audio-track cleanup.

Anti-spam

Setting	What it does
Anti-spam — minimum interval	Within this many minutes the same event is not sent again. `0` disables the anti-spam window; the maximum is `10080` (seven days).

Message language

Setting	What it does
Message language	Language of the notifications; empty follows the interface language.

Editors

Setting	What it does
Event hooks	Runs a program on selected events (name, events, program, arguments, timeout).

Access and services¶

Login, API keys and credentials of external services.

Gextto is meant for a trusted LAN: access is open by default. When you reach it from outside (reverse proxy, port forwarding, VPN) you can require a login from clients that are not on the local network.

Login

Setting	What it does
Require login	Off (default): no checks. On: clients outside the local network must log in or use the API key. Until a password or a key is set everything stays open (and the log says so).
No login from the local network	On (default): 127.0.0.1, 192.168.x.x, 10.x.x.x, 172.16–31.x.x and local IPv6 addresses need no login. Behind a reverse proxy the real client address forwarded by the proxy counts (`X-Forwarded-For`, `X-Real-IP`, `Forwarded`), so a client coming from the Internet through the proxy still has to log in. Shown only while «Require login» is on.
Username	Login username (default `admin`). Shown only while «Require login» is on.
Password	Stored only as a bcrypt hash; changing it closes every open session. Leave the field empty to keep it. Shown only while «Require login» is on.

Writes from another origin. Regardless of the login, Gextto refuses changing requests that the browser declares to come from another site (Origin/Referer, Sec-Fetch-Site: cross-site): a malicious page cannot reconfigure Gextto (settings, event hooks, torrent actions) through the browser of someone who has the interface open. Reads (GET) and clients without a browser (curl, the TUI) are untouched.

API keys

Setting	What it does
API key (scripts, TUI, calendar)	For non-browser access: header `X-Api-Key: <key>` or `?apikey=<key>` in the URL. The TUI reads it from the `GEXTTO_API_KEY` variable.
gx-torrent — access token (page and API on the LAN)	Under the Access tab: shared secret required by the page and the API; mandatory when the daemon listens on the network (not shown).

Both secrets are generated from the row itself: Generate creates a long random value and saves it, Copy puts it in the clipboard. A new gx-torrent token needs the daemon restarted to be applied. What they are and how to use them is explained in Access keys.

Metadata services

Setting	What it does
TMDB API key (recommended)	Free TMDB API key: the main source of titles, posters and metadata. When set, it takes precedence over TVDB.
TVDB API key (alternative to TMDB)	TheTVDB v4 API key: without the TMDB key it provides all the metadata.
TVDB subscriber PIN	TheTVDB subscription PIN, asked at login only by “user-supported” keys.

Metadata: TMDB or TVDB¶

Gextto uses metadata to know how many episodes each season has (missing episodes, season packs, completeness), the series status, the calendar, episode titles in file names, posters and the series and movie pages. One of the two keys is enough:

TMDB (recommended) — free key from themoviedb.org → Settings → API. When it is set Gextto always uses TMDB, even if the TVDB key is set too.
TVDB (alternative) — v4 key from thetvdb.com → Dashboard → API keys. Without the TMDB key it provides every feature listed above except the trending and popular lists of Explore, which exist only on TMDB. “User-supported” keys require a paid TheTVDB subscription: enter its PIN in TVDB subscriber PIN. Episode titles follow the TVDB language (e.g. ita); an untranslated episode keeps its original title.

Without either key Gextto still downloads, but without the data above: missing episodes stop at the last one already downloaded and files get generic titles (“Episodio N”). TMDB and TVDB use different ids: a series added with TVDB keeps its TVDB id, and if you later set the TMDB key Gextto finds the series on TMDB by name.

A browser session lasts 30 days; /logout ends it. After five wrong passwords from the same address further attempts are blocked for a minute and each failure is logged. If you lock yourself out, start Gextto with the environment variable GEXTTO_AUTH_DISABLE=1: access control is off while the variable is set and you can fix the settings. For access from outside always use HTTPS (reverse proxy with a certificate, or a VPN), otherwise the password travels in clear text.

Diagnostics and updates¶

Detailed logs and the update check.

Diagnostics

Setting	What it does
Debug (detailed logs)	Enables detailed logs and periodic diagnostics for debugging.

Updates

Setting	What it does
Check for updates	Downloads the latest build information from GitHub on a timer (how often is set by «Update check interval») and, when a newer one exists, shows «Update available». It sends no data and never updates on its own: the update only starts from Maintenance.
Update check interval	How many hours apart Gextto checks whether a new build is out (1 to 720, default 6). The check downloads only a small file from GitHub; «Check for updates» in Maintenance checks at once.

Appendix B. Reference — Integrations¶

The page is grouped — Services (Simkl), Media server (Jellyfin, Plex), Sources (indexers, FlareSolverr, source check) and Links (iCal calendar, browser handlers) — with the group title stuck to the top while you scroll.

Simkl¶

Field / action	What it does
Client ID	Client ID of the Simkl app.
Calendar days	How many days ahead to show in the Simkl calendar.
Watchlist status	Status assigned to imported series: Plan to watch, Watching or Completed.
Mark as watched	Marks downloaded episodes as watched on Simkl.
Start access / Confirm	Starts the PIN flow and confirms the code shown by Simkl.
Revoke	Revokes access and removes the stored token.
Import watchlist	Imports the series from the Simkl watchlist into the library.
Watchlist / Calendar	Read-only tables with the watchlist and the upcoming releases.

Jellyfin¶

Field / action	What it does
Jellyfin URL	Jellyfin server URL (e.g. `http://127.0.0.1:8096`).
Jellyfin API key	API key generated in Jellyfin → Dashboard → API Keys.
Jellyfin — path mapping	Only when Jellyfin sees the library under different paths (e.g. Docker): one line per folder, `gextto_path=jellyfin_path`.
Test connection	Checks that Jellyfin responds.
Refresh library	Asks Jellyfin to refresh the whole library.

Plex¶

Field / action	What it does
Plex URL	Plex server URL (e.g. `http://127.0.0.1:32400`).
Plex token	`X-Plex-Token` used to access the library.
Plex — path mapping	Only when Plex sees the library under different paths (e.g. Docker): one line per folder, `gextto_path=plex_path`.
Test connection	Checks that Plex responds.
Refresh library	Asks Plex to refresh the whole library.

After each import Gextto asks Jellyfin and Plex to rescan only the folder that changed (the season or movie that just arrived), not the whole library: the NAS is spared a full scan per episode and the file shows up in seconds. For Plex the folder must be inside one of the folders of one of its libraries. When the targeted request is not possible (unknown path, server refusal) Gextto asks for the full refresh as before. When the server runs in Docker and sees the files under another path, fill in the path mapping.

iCal calendar¶

/feed/calendar.ics is a calendar to subscribe to from Thunderbird, Google Calendar, Apple Calendar or a phone (link under Integrations → iCal calendar). Gextto was born for an Italian audience, where series often arrive months or years after the original broadcast, so the calendar tells three things apart:

Entry	What it means
📥 Series S01E04	episode or movie that arrived in the library, on the day Gextto downloaded it (last 30 days). The only date that says when it is really available.
📺 Series S02E03 · Title	original broadcast (TMDB, or TVDB) of the current season, from the last week to the next two months. It is not the date of a localised release; ✓ when the episode is already in the library.
🎬 Movie	release of a monitored movie in your country (TMDB: digital, then home video, then cinema; the country follows the TMDB language, e.g. `it-IT`). When that country has no date yet the original one is shown, marked “(original release)”. With only the TVDB key the movie's first release is shown, always marked “(original release)”.

A TMDB or TVDB key is required; the calendar is rebuilt at most every 30 minutes. With access control on (see Access) add ?apikey=<key> to the address.

Torznab indexers¶

Field / action	What it does
Name	Indexer label (e.g. `jackett` / `prowlarr` / `indexarr`).
Base URL	Base URL of the service; Gextto appends the Torznab path.
API key	Indexer API key.
Type	Auto-detected, Prowlarr, Jackett or indexarr.
Enabled	Enables or disables the indexer.
Test	Tests the indexer (for Jackett and indexarr it uses `t=caps`).

A direct Torznab indexer (not managed by Jackett, Prowlarr or indexarr), for example the mircrew-indexer service on the same machine or the LAN, is added here with its base address (e.g. http://192.168.1.10:9118) or one ending in /api: Gextto uses <base>/api, the same rule as gx-torrent, so the address to type is identical in both. Automatic searches have a 100-second budget per title: an indexer behind Prowlarr may wait up to 90 seconds (Prowlarr serves one search at a time per indexer), a direct indexer or one behind Jackett up to 45, indexarr up to 90 (one search per title over the whole manager). A lane whose queue would outlast the budget is skipped without counting as a failure, so one slow tracker does not stop the cycle. When an indexer on a local or private-network address drops the connection or does not answer, Gextto does not retry through FlareSolverr (Cloudflare does not protect a LAN service): the error stays visible as it is. For mircrew-indexer, Health → Provider status shows whether the service answers and whether its login to the MirCrew forum worked; if it did not, log in again from the service's web UI.

Gextto reads the indexer list and its capabilities from the manager (the Torznab caps; for Prowlarr the list of its indexers, re-read every 10 minutes) and sends only the parameters that the single indexer declares. For example Jackett's "all" aggregate does not accept the TMDB id in series and movie searches, and Prowlarr's public indexers cannot search by id: in these cases the search uses title, season and episode instead of failing or coming back empty. If an indexer still rejects an id, the search is repeated right away without it; the error shown in Health carries the indexer's own explanation (for example "HTTP 400: Torznab 100: Invalid API Key") instead of the bare HTTP code.

FlareSolverr¶

Field / action	What it does
URL	URL of the FlareSolverr service used to bypass Cloudflare.
Test FlareSolverr	Checks that FlareSolverr responds.

Event hooks (Configuration → Notifications)¶

Field / action	What it does
Name	Hook label.
Enabled	Enables or disables the hook.
Events	Events that trigger the hook (empty = all).
Program	Executable to run (without a shell).
Arguments	Arguments with placeholders `{title}`, `{hash}`, `{path}`, `{series}`, `{episode}`; each placeholder stays a single argument even with spaces. The same values are exposed as `GEXTTO_*` variables.
Timeout (s)	Timeout in seconds (default 60, maximum 24 hours; `0` = default).

Browser handlers and source check¶

Item	What it does
Magnet / Torrent manager	Downloads the scripts to open magnets and `.torrent` files directly in Gextto.
Magnet / Torrent `.desktop`	`.desktop` versions for Linux desktop integration.
`install.sh`	Downloads the installation script.
Check sources / Refresh	Checks feeds, indexers and web engines; with a query it also measures results, without changing settings.

Browser extension (qBittorrent client)¶

Gextto answers a subset of the qBittorrent API (/api/v2) on the address of its own interface, so an extension written for qBittorrent — Torrent Control for Firefox, for example — sends magnets and .torrent files to Gextto: the torrent enters the queue and the active engine downloads it (gx-core by default), as if it had been added from the interface.

In the extension, configure a server with these values:

Field	Value
BitTorrent client	qBittorrent (current version)
Server address	`http://<gextto address>:5000/` (the interface port, not the gx-torrent one)
Username / Password	the ones of the protected access, when it is on; otherwise they can stay empty
Download folder	empty to use the configured folder, or a path the engine accepts
Label / Category	empty: categories, tags and a content layout other than Original are refused with an explicit error, because in Gextto they would change the queue tags and the folders on disk

When the protected access is on, the extension logs in with the same credentials as the interface and receives a session cookie valid only for /api/v2: the rest of the interface stays protected.

Appendix C. Reference — Maintenance¶

Quick actions¶

Action	What it does
Backup now	Immediately creates a database backup snapshot.
Clean trash	Empties the trash completely (forced action: `trash_retention_days` does not apply).
Recalculate scores	Recomputes the score of archived releases with the current weights.
Scan archives	Re-reads the archive folders and updates the library.
Update MediaInfo	Analyses with `ffprobe` the archived files that have no MediaInfo.
Rename all	Renames every archived file using the configured format.
Housekeeping	Cleans technical data and history without touching the library.
Restart service	Restarts the Gextto daemon.

Rename folder content¶

Item	What it does
Folder / Browse	Type or pick the folder to analyse from the server.
Scan and propose	Recursively analyses the videos, recognises series/movies with TMDB/TVDB and proposes new names (preview only).
Per-row result	Selects an alternative TMDB/TVDB match.
Accept / reject	Approves or discards the single proposal.
Accept all proposals	Approves every proposal at once.
Apply selected	Renames the selected files that really exist with a valid destination.
Rename progress	Progress of the background operation.

Library¶

Item	What it does
Video duplicates — Preview duplicates	Lists inferior duplicates (lower resolution) without deleting anything.
Video duplicates — Clean duplicates	Moves the inferior copies found to the trash.
Database optimisation — VACUUM / ANALYZE	Compacts the databases and refreshes the query planner statistics.
Database optimisation — Refresh sizes	Re-reads size and status of the database files.
Database optimisation — Check integrity and indexes	Checks integrity, foreign keys and indexes (including the archive FTS).
RAM disk	Shows writable `tmpfs`/`ramfs` paths, lets you select one, create a dedicated one and refresh the list.

Cleanups¶

Item	What it does
Trash — Open trash	Shows the trash items so you can delete them one by one.
Trash — Empty trash	Deletes every trash item and shows a notice with the removed files and the freed space.
Database cleanup — Cycles to keep	Number of cycle statistics to keep.
Database cleanup — Error days	Keeps error torrent entries for at least N days before removing them.

Log¶

Item	What it does
gx-torrent log	Latest lines of the torrent engine log (`gx-torrent/gx-torrent.log`), with text filter, warnings/errors only, line count and a choice between the active file and the rotations.

Backup¶

Item	What it does
Backup now	Immediately creates a backup archive, without waiting for the interval or the scheduled time. The same command is in Quick actions. While it works it shows a notice and stays disabled.
Save settings	Saves the backup settings of this section (retention, schedule, FTP, cloud folder, extra paths, Telegram): it does not create an archive.
Backups to keep	Number of local ZIP files to keep in the `backups` folder.
Interval (hours)	Automatic backup interval in hours (0 disables the interval).
Time (HH:MM)	Local daily time of the automatic backup; when set it takes precedence over the interval.
FTP host	FTP server host/address (empty = no FTP upload).
FTP user	FTP server user.
FTP password	FTP password (not shown; empty keeps the stored one).
FTP path	Remote destination folder for the ZIP.
Cloud folder	Local path of a folder already synced by a cloud service.
Files and folders to include	Absolute paths of files or folders to add to the ZIP (one per line). They end up in the archive's `extra/` folder with their original path; paths that no longer exist are skipped.
Send to Telegram	Sends a copy to Telegram using the configured bot.
Test FTP	Checks connection, login, remote path, a test upload and removal.
Available backups / Verify	Table of snapshots with label, name, size and date; the Verify button checks the archive, shows a notice while it works and confirms when the ZIP is intact.

Gextto user manual¶

Contents¶

Part I — Getting started¶

1. What Gextto is and how it works¶

The path of an episode¶

The words you will use¶

2. Installing Gextto¶

With the installer (recommended)¶

Without root, from a checkout¶

With Docker¶

Checking that it runs¶

3. The initial setup¶

The folders and their role¶

4. The first download, in test mode and then for real¶

Checklist before real downloads¶

5. Finding your way around the interface¶

The menu¶

The top bar¶

From a phone¶

Accessibility¶

Part II — Everyday use¶

6. Following a TV series¶

Adding a series¶

The series page¶

The episodes¶

Missing episodes¶

Season packs and single episodes¶

7. Following a movie¶

Adding a movie¶

The movie page¶

List or posters¶

Editing several titles at once¶

8. Searching and downloading by hand¶

Where to search¶

Reading the results¶

«Why not this one?»¶

Adding a torrent or a magnet¶

The Archive¶

eD2k links¶

9. Watching the downloads¶

Reading the list¶

Acting on downloads¶

A torrent's details¶

Streaming and media server¶

Bandwidth¶

Stalled torrents¶

Seeding limits: per-source rules and idle seeds¶

The history¶

10. Where the files end up¶

What happens when a download finishes¶

Copy, move or hardlink¶

Upgrades¶

File names¶

Importing a file you already have¶

11. Comics¶

12. eD2k and Kad with aMule¶

Connecting aMule¶

The eD2k page¶

What happens to an eD2k download¶

eD2k as a fallback¶

Part III — Configuring¶

13. Sources: feeds and indexers¶

Adding Jackett or Prowlarr¶

Adding indexarr¶

FlareSolverr¶

Reachable does not mean working¶

Backoff: when a source fails¶

Copies of the same release¶

14. Quality, scores and upgrades¶

How Gextto chooses¶

Quality profiles¶

Trying before changing¶

The weights¶

When a file is replaced¶

Delay before downloading¶

15. The torrent engine¶

gx-torrent (default)¶

qBittorrent-nox¶

Switching engine¶

On a small machine¶