Embeddr Load Artifacts (V2)
Pull a whole batch out of your library in one go
- artifact_ids
- images
- artifact_ids
- masks
Load Artifacts is the batch version of Load Artifact. Instead of one UUID, it grabs a set of images from your Embeddr library and hands you a ready-to-process batch. It's the node you reach for when you want to do something to a bunch of stuff: run a whole collection through an upscaler, compare a run of variations, or build a reference stack for style work. One node, a batch out.
Where it gets genuinely powerful is the two ways to pick what loads - a live browse, or an explicit ID list - because they make it both a "show me what's in here" tool and a "process exactly these" tool.
How it picks
collection- a combo dropdown populated live from your server (entries look like"42: my-refs", i.e.ID: Label), plus"All". Pick a collection and you browse it.sort_by-newestorrandom. Therandomoption is the sleeper hit: combined withseed(an int, default 0), it gives you a reproducible random sample from the collection. Set a seed, get a stable random batch; change it, get a different one. Great for "surprise me with 5 images" workflows.limit- how many to load, 1–50, default 5.artifact_ids- the override input. Connect a list of IDs (from Merge IDs, the search nodes, or Split IDs' upstream) and it overrides the collection entirely - the tooltip says so, and it means you can load a specific set without touching the dropdowns.auth_ticket- optional per-user auth, same as Load Artifact. Ignore it unless your setup uses tickets.
Outputs
images- the loaded batch as anIMAGElist.artifact_ids- the IDs of what you loaded, in order. Keep them in sync with the images: index them together and you always know which image came from where, which is exactly what lineage tracking wants.masks- masks for the batch, if the artifacts carry them.
Each artifact resolves through your server (same /resolve?variant=original&proxy=1 path as Load Artifact), so this is a network-bound node - a 50-image load is a 50-image download. Latency is your server and disk, not your GPU.
Install & gotchas
The pack story, once more, briefly: embeddr serve running, pack via ComfyUI Manager or a release extracted into custom_nodes, restart ComfyUI, requests is the only declared dependency.
The dropdowns are populated when the node is created, so if your backend was down at graph-build time, you'll only see "All" - reconnect and re-add the node (or restart ComfyUI) to refresh the lists. And mind the override: a stale artifact_ids connection silently ignores whatever collection you picked, which reads as "the collection dropdown is broken" until you notice the wire. If a load errors, the usual suspects apply: server reachable, IDs valid, and remember sort_by: random with a fixed seed is only reproducible if the collection hasn't changed underneath it.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| collection | COMBO | All | 1 options: All |
| sort_by | COMBO | newest | 2 options: newest, random |
| limit | INT | 51–50 | — |
| seed | INT | 0 | — |
| artifact_idsopt | EMBEDDR_ARTIFACT_ID | Optional list of IDs to load (overrides collection) | |
| auth_ticketopt | STRING | Ephemeral auth ticket for per-user access |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| images | IMAGE | — |
| artifact_ids | STRING | — |
| masks | MASK | — |