Repository navigation
FR: Cross seed to a specific instance #1807
Replies: 3 comments 1 reply
|
I think this is a good use-case of the move/export torrent PR I have open. I've been running it for over a month now to do this. Yes it still gets injected in the same instance, but then I am using the automations to move it to the instance I want it to be in. My automation action does the export and tags it for removal on the source. Then a second automation monitors for the tag to remove on the source instance. I dont want to say the cross-seed instance functionality would be hard to develop, but I think it leaves room for a lot of misconfiguration/errors where xseeds end up on an instance without files and just start re-downloading. The move automation has this fully gated so you cant move a torrent unless it succeeds/fully re-checks. |
Agent BriefCategory: enhancement Current behavior: Desired behavior: Edge cases that need defined behavior:
Key interfaces:
Acceptance criteria:
Out of scope:
Why a human: A destination instance that cannot see the source data turns a cross-seed into a full re-download, the risk raised in the thread. Whether qui gates that, how it verifies reachability, and which configuration level to expose (global, per-flow, or per-rule) are product-scope calls. |
Uh oh!
There was an error while loading. Please reload this page.
Problem / Context
Currently, when qui finds a cross-seed match, the resulting torrent is always injected into the same qBittorrent instance where the source files were found. There is no way to route cross-seed injections to a different instance.
This means users who want a dedicated
qbit-crossinstance — separate fromqbit-radarr— cannot use it as a cross-seed destination via qui's automated flows. The cross-seed always ends up back on the source instance.Use case
A common multi-instance setup:
qbit-radarr→ library-managed contentqbit-cross→ dedicated cross-seed instanceToday, if a match is found for a torrent on
qbit-radarr, the cross-seed is injected intoqbit-radarr. The user has no way to tell qui: "find the match onqbit-radarr, but inject the cross-seed intoqbit-cross".Proposed solution
Add a target instance option to cross-seed injection, separate from the source instance where the match was found. This would allow rules like:
This could be configured at several levels (in order of increasing flexibility):
qbit-radarr, inject intoqbit-cross")Even option 1 alone would cover the majority of use cases.
Interaction with Hardlink Mode
When hardlink mode is enabled, the link tree must be on the same filesystem as the source files. A cross-instance injection would still work if:
qbit-cross) has filesystem access to the same pathsqui already supports per-instance hardlink configuration and multi-filesystem base directories, so the building blocks are in place.
Current Behavior (API evidence)
The issue is also observable via the
/cross-seed/applyAPI endpoint. When callingapplywith a specificinstanceIdstarget, qui returnsno_matchbecause it looks for the source files on the target instance itself, rather than searching across all instances:The source files exist on
qbit-radarr(a different instance), but sinceqbit-crossdoesn't own them, qui cannot complete the cross-seed.This confirms that source lookup and injection target are currently tightly coupled: qui only injects into an instance if that same instance already holds the source files.
Technical notes
Target instancesfield (which instances receive cross-seeds from RSS). Extending this concept to the other discovery methods (Library Scan, Completion Search) would be a consistent and natural evolution.All reactions