Regex Filter (Multiline)

When a prompt list, LLM dump, or tag batch arrives with lines you don't want, this node lets you keep only the ones that match a regex - or, flipped, drop exactly those. It's the "filter this list for me" node, and it's surprisingly useful once you're generating prompt lists programmatically instead of hand-curating them.

The model here is simple and powerful: every line of your multiline input is tested against the pattern; matching lines survive, the rest are dropped. invert turns that around - non-matching lines survive. You get the filtered text plus a count, so you always know how many lines made the cut.

What it needs

• text - the multiline list to filter.
• pattern - the regex, in Python syntax. Empty pattern means no filtering - the input passes through with its original line count. This is a nice safety property: an empty pattern is "do nothing," not "keep nothing."
• invert - flip the match. false keeps lines that match; true keeps lines that don't. "Keep everything except lines that contain 'nsfw'", for example.

Two outputs: filtered_text (the surviving lines, still multiline) and count (how many survived).

What a real pattern looks like

Say you have a list of prompts and want only the ones mentioning a character or a style:

pattern: (?i)cyberpunk

That keeps every line containing "cyberpunk" case-insensitively. To keep lines without a word, tick invert. For a quick sanity check on what matched, the count output is your friend - a zero there means your pattern matched nothing, which is usually a regex bug, not an empty list.

Two failure behaviors worth knowing: an invalid regex (a pattern that doesn't compile) returns empty text and count 0 instead of crashing, which is safe but confusing; and the match is a search (regex.search), so the pattern doesn't need to anchor the whole line - it just needs to appear somewhere in it. Use ^...$ if you need full-line matching.

Install

Part of Vantage-Nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/vantagewithai/Vantage-Nodes.git
pip install -r requirements.txt

or ComfyUI Manager → search "Vantage-Nodes" → Install → restart.

Common issues

• Everything got filtered out - your pattern matched nothing. Test it on a single line first; the count output tells you the damage.
• Regex too loose - red matches "hundred," "cred," anything containing the letters. Anchor or word-bound it (\bred\b) for exact words.
• Case sensitivity surprises - Python regex is case-sensitive by default; add (?i) at the front or use re.IGNORECASE-style flags in the pattern.