🧹 String Cleaning

Text coming out of anywhere automated - an LLM response, a scraped caption, a metadata field - is rarely clean enough to use as-is, and this node is a single stop for the whole category of "strip this junk out" problems rather than a one-trick fixer. It's a long list of toggles and text-pair fields, almost all off by default, so nothing happens to your text unless you deliberately switch something on.

The everyday case is the simplest one: collapse_sequential_spaces, strip_leading_spaces, strip_trailing_spaces, and strip_empty_lines cover the "my text came back with sloppy whitespace" problem that shows up constantly after copy-pasting or parsing something automated. Turn on whichever combination matches what you're seeing.

The most practically useful trick in here, though, is strip_between_start / strip_between_end, and it's worth calling out specifically because the tooltip's own example is exactly the situation you'll hit if you're running this pack's Groq LLM API node against a reasoning model: set the start tag to <think> and the end tag to </think>, and the node removes an entire chain-of-thought block - everything between those two markers - leaving just the actual answer. Reasoning models routinely emit their internal deliberation wrapped in tags like that before the real response, and if you're feeding that response straight into a prompt or a save file without cleaning it first, you're shipping the model's scratch work along with it. The same pair-based pattern (strip_inside_tags for symmetric pairs like (), [], {}; remove_text_before / remove_text_after for one-sided markers) covers related cases - stripping parenthetical asides, or cutting everything before/after a specific delimiter in structured text.

Beyond those, there's a longer tail of narrower tools for less common situations: strip_leading_symbols / strip_trailing_symbols for stray punctuation at line edges, strip_newlines and replace_newlines_with_period_space for collapsing multi-line text into one flowing line, strip_leading_custom / strip_trailing_custom / strip_all_custom for removing your own specific strings (a repeated "Chapter" prefix, a boilerplate suffix), and multiline_find / multiline_replace as a manual find-and-replace pair, matched line by line.

Output: a single cleaned_string, and because this is an output node it shows the result directly in the UI too - genuinely handy for iterating on which toggles you actually need without wiring a separate Preview Text node.

Installing it: ComfyUI Manager, search "ComfyUI-mnemic-nodes," or git clone https://github.com/MNeMoNiCuZ/ComfyUI-mnemic-nodes into custom_nodes, then restart. It's string processing, nothing more - no dependencies to fight with.

Two real gotchas, both stated directly in the tooltips: multiline_find and multiline_replace are matched by line position, so they must have the same number of lines or the pairing breaks - the same requirement applies to strip_between_start and strip_between_end, which must line up one-to-one. And because every toggle defaults to off, the most common confusion is plugging text in, getting the exact same text back, and assuming the node is broken - check that you've actually enabled something before troubleshooting further.