VRGDG_ArchiveLlmBatchFolders

VRGDG_ArchiveLlmBatchFolders is the cleanup step in the VRGDG LLM scene-prompt pipeline. It takes a trigger string, and returns that same trigger plus a details string describing what it just did. There are no model files, no math, and no generation - it's the node that files away last run's batch folders so the next run doesn't trip over them.

Here's the context that makes it make sense. The pack's music-video workflow splits your story into scene groups and has an LLM write prompt files for each scene into folders - by default a subfolder called llm_batches (that's the output_subfolder default on VRGDG_LLM_PromptBatcher). Run that once and you get a pile of scene prompt files. Run it again and you want a fresh pile, not a merge with stale prompts from the previous take. This node archives the old batch folders out of the way.

How it fits

You wire it after the batch-producing part of the workflow (or before a new run starts), trigger it with any string - a signal that the batch phase finished - and it does the folder chore while passing the trigger through unchanged. The details output is the log line: what got archived, and where. That pass-through design means you can drop it into an execution chain without breaking the data flow; it observes and tidies rather than transforms.

It's the kind of node that only makes sense if you're running the "unlimited length" music-video workflow on repeat, which is exactly who it's for. If you load one of those graphs and wonder why there's a node dedicated to moving folders around - that's why. Long multi-session runs accumulate state, and state you can't see is how you get a scene 3 that's suddenly using stale lyrics from two versions ago.

Installing and using it

Same pack, same install as everything else here:

cd ComfyUI/custom_nodes
git clone https://github.com/vrgamegirl19/comfyui-vrgamedevgirl

or ComfyUI Manager → search "vrgamedev", then restart. Requirements (kornia, librosa, imageio) are shared with the rest of the pack.

The honest warning is that this node is only as good as your habit of running it. Archive folders never happens on its own, so if you're chaining many LLM batch runs, make sure this node is actually in the queue - otherwise the workflow just keeps appending to the same folders and the archive step is a paperweight. On the flip side, if you only ever do single-shot scene prompts and never re-run, you don't need it. It's discipline as a node, which is a nicer way to get discipline than a sticky note.