qui_workflows

Automation workflows for qui — a qBittorrent automation manager. These workflows manage the complete torrent lifecycle: tagging, maintenance, share limits, and cleanup.

21 automations organized by function, designed for a hardlink-aware setup with cross-seed support.

Requirements

• qui instance with API access
• qBittorrent with hardlink detection enabled (save path and hardlink target on the same filesystem)
• curl and python3 (for the export script)

Quick Start

Import

Import individual JSON files through the qui web UI, or use the API:

curl -X POST "http://your-qui:7474/api/instances/${QUI_INSTANCE_ID:-1}/automations" \
  -H "X-API-Key: YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d @tagging/Tag\ -\ tracker\ name.json

The id field in each JSON is from the source instance and will be reassigned on import. Sort order and conditions are preserved.

Export (update from live instance)

# Set your qui URL and API key
export QUI_URL="http://your-qui:7474"
export QUI_API_KEY="your-api-key"

# Run the export script
./scripts/export.sh

The export script fetches all automations from the API, strips instance-specific fields (instanceId, createdAt, updatedAt), and writes individual JSON files to the categorized directories. When adding new automations, update the FILE_MAP in the script.

Structure

qui_workflows/
├── tagging/           # Tracker name, noHL, stalledDL, tracker issue
├── maintenance/       # Resume, delete unregistered, recheck missing
├── limits/            # Share limits and speed limits by category
├── cleanup/           # Delete rules by category with guards
├── scripts/
│   └── export.sh      # Export automations from live instance
└── README.md

Automations

Limits and cleanup automations are interleaved by sort order (limits at even positions, cleanup at odd) so that each category's cleanup rule runs immediately after its limits rule.

All limit and cleanup automations run on a 1-hour interval (intervalSeconds: 3600), except HL-remove-limits which uses qui's default interval.

Tagging (sort 1-6)

Sort	Name	Interval	Action
1	Tag: tracker name	30min	Auto-tag with tracker display name
4	noHL	6h	Tag noHL when no external hardlinks and completed 90+ minutes ago (excludes MAM, cross-seeds)
5	Tag: stalledDL	30min	Tag stalledDL for stalled downloads
6	Tag: tracker issue	30min	Tag issue for errored/down/missing files

Maintenance (sort 2-8)

Sort	Name	Interval	Action
2	Resume Incomplete	6h	Resume stopped torrents that aren't complete
3	Resume: min seed failsafe	30min	Resume any non-downloading, non-active torrent seeding < 15 days; also resets share limits to unlimited (re-applied by later limit automations)
7	Delete: unregistered	15min	Delete unregistered torrents (90min age guard, tracker not down)
8	Recheck: missing files	default	Built-in qui action — rechecks torrents with missing files (empty conditions is intentional)

Limits (sort 9-20)

Sort	Name	Ratio	Seed Time	Upload Cap
9	HL-remove-limits	unlimited	unlimited	—
10	noHL-movies-limits	33	15 days	—
12	noHL-tv-limits	unlimited	21 days	—
14	TL-limits	unlimited	365 days	~25 MB/s (25390 KiB/s)
16	TLnoHL-limits	unlimited	15 days	~25 MB/s (25390 KiB/s)
18	noHL-catchall-limits	15	35 days	—
20	noHL-xseed-limits	unlimited	15 days	—

Cleanup (sort 11-21)

Sort	Name	Guard	Delete Trigger
11	noHL-movies-cleanup	seed >= 15d	ratio >= 33 OR seed >= 15d
13	noHL-tv-cleanup	seed >= 15d	ratio >= 3 OR seed >= 21d
15	TL-cleanup	—	seed >= 365d
17	TLnoHL-cleanup	seed >= 15d	ratio >= 1.25 OR seed >= 12d
19	noHL-catchall-cleanup	seed >= 15d	ratio >= 15 OR seed >= 35d
21	noHL-xseed-cleanup	—	seed >= 15d

Execution Model

Sort Order & Last-One-Wins

All matching automations fire in sortOrder sequence. For conflicting share limits, the last matching automation wins — earlier limits get overwritten by later ones.

This is why limits are ordered from most specific to least specific:

1. HL-remove-limits (hardlinked → unlimited)
2. Category-specific limits (movies, TV, TL)
3. Catchall (everything else)

Delete Isolation

Delete automations cannot combine with other actions. Each delete rule is a separate automation. Share limits + speed limits can combine in a single automation.

Category Routing

Torrents are routed to rule groups based on conditions. Routing uses HARDLINK_SCOPE directly (not the noHL tag), so it applies immediately without waiting for the 90-minute tagging grace period.

Condition	Group
HARDLINK_SCOPE = outside_qbittorrent + not TL + not cross-seed	HL (remove limits, seed forever)
CATEGORY CONTAINS movie + not hardlinked + not TL + not cross-seed	Movies
CATEGORY CONTAINS tv + not hardlinked + not TL + not cross-seed	TV
TAGS CONTAINS TorrentLeech + hardlinked	TL
TAGS CONTAINS TorrentLeech + not hardlinked	TLnoHL
CATEGORY CONTAINS cross + not hardlinked	Cross-seed
Not hardlinked + not MAM + not TL + not movie + not tv + not cross	Catchall

Special Handling

• MyAnonamouse (MAM): Excluded from noHL tagging and catchall — seeds forever
• TorrentLeech (TL): Speed-limited to ~25 MB/s upload (25390 KiB/s), separate HL/noHL rules
• Cross-seeds: Dedicated rules — seed 15 days then delete. Cross-seeds with hardlinks are naturally exempt (treated as hardlinked)
• All deletes: Use deleteWithFilesPreserveCrossSeeds mode — deletes the torrent but preserves files if a cross-seed instance references them. Exception: unregistered uses deleteWithFiles (no cross-seed check since the torrent is already dead on the tracker)

Resume Failsafe

The "Resume: min seed failsafe" (sort 3) catches torrents that get paused or stopped prematurely. It resumes any torrent that is not downloading, not active, and has been seeding less than 15 days.

Important: Cleanup guards should be <= the failsafe threshold to prevent stuck torrents in the gap between guard and failsafe.

Conditions Schema

The tag and tags fields in tagging automations contain identical conditions — this is a qui schema requirement (both must be present and match). The conditions.schemaVersion field is always "1".

Time Reference

Condition values (seconds)

Used in SEEDING_TIME, ADDED_ON_AGE, COMPLETION_ON condition fields:

Seconds	Human
900	15 minutes
1800	30 minutes
5400	90 minutes
21600	6 hours
86400	1 day
1036800	12 days
1296000	15 days
1814400	21 days
3024000	35 days
31536000	365 days

Share limit values (minutes)

Used in shareLimits.seedingTimeMinutes:

Minutes	Human
21600	15 days
30240	21 days
50400	35 days
525600	365 days

Warning: Condition fields use seconds while seedingTimeMinutes uses minutes. Mixing these up is an easy mistake that results in dramatically wrong behavior.

Known Differences from TRaSH Workflows

This setup uses category-based routing (movies/tv/TL/catchall) while TRaSH's workflows use tracker-tier-based routing (Tier 1/2/3). Neither approach is wrong — they reflect different philosophies.

Key differences:

• No tier system: We route by category + tracker tag rather than tracker tier
• No upload speed throttling for non-TL: TRaSH throttles Tier 2 (2 MB/s) and Tier 3 (1 MB/s) after ratio > 2.0. We only throttle TL at 25 MB/s
• No stalled download cleanup: TRaSH deletes stalled downloads < 9.5% progress after 4h
• No problem cross-seed detection: TRaSH tags and deletes cross-seeds that are stalled/missing after 24h

Improvement Candidates

Issues identified during review (documented, not yet applied):

1. noHL-movies-cleanup redundant OR: Guard seed >= 15d + trigger ratio >= 33 OR seed >= 15d — the seed branch is always true when the guard passes, making ratio irrelevant. Every noHL movie deletes at exactly 15d.

2. TLnoHL-cleanup guard/trigger mismatch: Guard is seed >= 15d but OR fallback is seed >= 12d. The 12d branch can never fire because the 15d guard blocks it. Either lower the guard to seed >= 12d (earliest delete at 12d) or raise the OR fallback to seed >= 15d (earliest delete at ratio OR 15d).

3. Consider adding stalled download cleanup and problem cross-seed detection per TRaSH patterns.

Related

• baker-scripts/StarrScripts — includes qui-xseed.sh for cross-seed automation
• TRaSH-Guides/qui_workflows — TRaSH's tier-based reference implementation

Contributors

Disclaimer

These automations are provided as-is with no warranty. They will delete torrents and files based on the configured conditions. Always review the JSON conditions before importing and test with dryRun: true first. The authors are not responsible for any data loss resulting from the use of these workflows.

License

MIT