Danbooru Artist CSV Manager

This node isn't a converter, and it isn't something you wire into your workflow. It's the pack's housekeeping tool - the thing that fetches the one file this entire pack can optionally depend on. That file is danbooru_artist.csv, and it exists because of an ambiguity that plagues every anime prompt: a plain artist tag like wlop looks exactly like an ordinary tag like 1girl. There's no way to tell them apart by eye, and there's no way for a converter to tell them apart either - unless it has a list of actual Danbooru artist names to check against.

That's what this CSV is. The pack uses it for "exact artist detection": recognizing that a tag is an artist's name so it can be rebadged into @wlop (SD(anima)) or artist:wlop (NAI3/NAI4) when you convert prompts. Without it, plain artist names stay plain, which is fine until you're moving a prompt to a format where the artist prefix matters.

How it works

The node does one of two things depending on the action dropdown:

• check only - looks at the file on disk and reports whether it exists and how many artist names loaded.
• download/update - fetches the CSV from csv_url (default https://huggingface.co/datasets/Laxhar/noob-wiki/resolve/main/danbooru_artist.csv), writes it to a temp file, then atomically renames it into place.

It's plain urllib under the hood - no auth, no huggingface_hub dependency, nothing exotic. The status output tells you what happened: artist csv ready, artist csv not found, or download failed: <reason> if the fetch blew up. artist_count is handy - it's the number of names actually parsed, so you can confirm the file is populated rather than silently empty.

The default save_path is the pack's own data/danbooru_artist.csv inside ComfyUI/custom_nodes/ComfyUI-Prompt-Format-Converter/. You can point it anywhere you like and use that same path in the converter nodes' artist_csv_path input.

Why it exists at all

The CSV is deliberately not in the git repo - it's large, and the pack's .gitignore ignores data/*.csv. So you will not find it in your fresh clone, and that's expected. Run this node once with download/update, read the status, and you're done. It's a utility node, not a graph node: you don't chain its outputs anywhere. The resolved_path string is there in case you want to feed it into a converter's artist_csv_path manually.

Install and gotchas

Installing the pack doesn't download the CSV - that's a separate, opt-in step, and a lot of users never need it. Reach for it only when you're converting plain SD artist tags to a prefixed format. If you already maintain your own artist list - some people curate one for their own LoRAs - skip the download entirely and just point artist_csv_path at your file in the converter node. The format the parser expects is a standard CSV with artist names as a column, same as the HF source. One warning worth repeating: if the node reports "found but no names were loaded", your file's rows don't match what the parser reads, so don't trust a nonzero file size - check artist_count instead.