Text Length

Text Length (StringLength) counts the characters in a string and hands you the number as an integer. That's the whole node - and it's easy to write off as trivia until you need a workflow to measure its text. Then it's the thing that lets your graph check its own work.

The genuinely useful patterns are all about validation and batching. You're feeding a prompt into a model with a hard token/character limit - a quick sanity check that your generated prompt isn't about to get truncated is worth a node. You built a filename out of parts and want to catch it getting absurdly long before the filesystem does. You're batching text and want to route anything under a length threshold down one path (short prompts get embellished, long ones pass through). In each case the trick is the same: this node turns an invisible property - "how long is this?" - into a number, and numbers are what routing, comparison, and loop logic are made of.

How it works

The mechanism is Python's len() on the input string - a single input, one output:

• string - the text to measure. Multiline, so a prompt with embedded newlines counts them: "a\nb" is length 3, not 2, because the line break is a character. Worth remembering, since copied text loves to hide \n and \r\n in there.
• Output: an INT named length - the character count. Wire it into a compare node, a threshold check, a math node, or a loop counter.

There's no option for "word count," no "token estimate," no trimming - it's a raw character count, and anything fancier is up to the nodes you connect it to.

A working sketch: an LLM prompt-writer node upstream, and you want every prompt to land between 200 and 600 characters. StringLength on the output → compare the INT against 200 and 600 → branch to "regenerate" or "use as-is." That's three or four nodes of graph for a rule you'd otherwise police by hand every run.

Common issues

The two things that bite: newlines count as characters, which inflates counts on multiline inputs in ways that surprise people measuring against a single-line expectation; and it's characters, not tokens - a 512-character prompt is not a 512-token prompt, so don't use the number to guess token limits without knowing your tokenizer's ratio. Beyond that it's about how you use the number: StringLength alone is inert, so pair it with a comparison or threshold node - the pattern is "length → compare against a target → branch." If you find yourself subtracting or slicing based on the result, the sibling Substring node (StringSubstring) is the partner that does something with the measured text.