Normalize text indentation levels

When the same line or the document as a whole mixes tabs and spaces with no clear pattern, the task fails with invalid_input instead of silently guessing. Optional target_unit lets you re-scale space-based levels (for example from two spaces to four) without changing the logical depth tree.

How base indent detection and level rewriting work

Optional target_unit applies only to space-style documents: levels are preserved and the output uses that many spaces per level, which is how you convert a two-space tree into a four-space tree in one call. Tab-style input ignores target_unit so tabs are not silently expanded.

Input contract, response fields, and error behaviour

Prefer sending the raw multi-line body rather than pre-collapsing newlines when you want true per-line indent semantics; a single long line has only one indent to normalize. POST a JSON body with a required string field text (aliases input, value, string, s, content, and body). The value must be a non-empty string after overall whitespace trim and must not exceed one hundred thousand characters so a single paste cannot exhaust Worker or browser memory. Optional target_unit is a positive integer from one through thirty-two; non-integers and out-of-range values fail validation. On success the response echoes input as the original string, text as the fully rewritten multi-line result joined with LF, indent_style as spaces, tabs, or none, indent_unit as the effective output unit size, detected_unit as the unit inferred before any target override, line_count as the number of lines including blanks, indented_lines as how many lines had a positive level, max_level as the deepest indent level, levels as the per-line integer depth array aligned with lines, lines as the ordered array of rewritten line strings, and message as a short English summary of style, unit, line count, and max level. Failed validation—missing text, non-string values, empty content, oversize payload, invalid target_unit, same-line tab/space mix, or document-level style mix—returns invalid_input and is never charged. Pricing is a flat $0.002 per successful request with unit item. When integrating, assert on text or lines for the cleaned content, optionally compare levels and indent_unit for regression checks, and treat invalid_input as a client-side style conflict for mixed pastes rather than a billing event.

When to use this tool and how it differs from related searches

Use this capability when the search intent is normalize text indentation levels, fix inconsistent indent online, convert ragged two-and-four space paste to a single unit, or detect tab versus space indent style before a commit. Keep trailing whitespace and interior alignment columns in mind—this tool only rewrites the leading indent run, not alignment spaces inside a line. It is not a full code formatter, not a language-aware pretty printer, not a blank-line deleter, and not a global String.trim that would destroy intentional structure. Round-trip checks for tests are easy: feed a three-line block with zero, two, and four leading spaces and expect indent_unit two with levels zero, one, and two; feed pure tab-indented nested blocks and expect indent_style tabs with integer levels matching tab counts; deliberately send a line that begins with a tab then spaces, or one tab-indented line next to a space-indented sibling, and confirm invalid_input mentioning mix of tabs and spaces. Prefer str.trim_lines when you only need to strip leading and trailing whitespace from every line without preserving a depth tree; prefer data.json_format when the payload is JSON that should be pretty-printed with a fixed indent; prefer this capability when humans already wrote hierarchical plain text or code-like blocks and you need one consistent base unit. Failed tasks that return invalid_input do not consume prepaid balance, which keeps form validation and CI negative tests cheap.