/**
 * A byte count written the way a person would say it: `800 B`, `2.6 KB`,
 * `123 KB`, `20.5 GB`.
 *
 * ── Why this exists ─────────────────────────────────────────────────────────
 *
 * Every one of the 294 shipped packs is smaller than a quarter of a megabyte —
 * the largest is 248 KB and 120 are under 10 KB. Printed in megabytes to one
 * decimal, as `pack list` did, the whole catalogue collapsed into three
 * strings: `0.0 MB` for 194 packs, `0.1 MB` for 53, `0.2 MB` for the last 47.
 *
 * A size that cannot tell two packs apart is not a size, it is a decoration;
 * and `0.0` actively misinforms, because it reads as "nothing" when the honest
 * answer is "three kilobytes".
 *
 * The rules are the ones people already read without noticing:
 *
 *   - below a kilobyte, whole bytes — `800 B`, never `0.8 KB`
 *   - below a hundred of a unit, one decimal — `2.6 KB` distinguishes packs
 *     that `3 KB` does not
 *   - at a hundred and above, whole numbers — `123 KB`, because a tenth of a
 *     kilobyte there is noise
 *
 * ── Why the arithmetic looks like this ──────────────────────────────────────
 *
 * All five implementations must produce the same string for the same number:
 * a shared CLI fixture compares their output byte for byte, so a size that
 * differs in the last digit is a five-way parity failure. Hence integers
 * throughout — no float division, no `toFixed`, and no reliance on how a
 * language happens to round a half.
 */
export declare function humanBytes(bytes: number): string;
//# sourceMappingURL=human-bytes.d.ts.map