Search Wikidata for entities matching a text query, returning candidate Q-ids with labels and descriptions.

Parse an IPv4 or IPv6 address and classify it — version, normalized form and RFC range (unicast, private, loopback, linkLocal, multicast, uniqueLocal, ipv4Mapped, …). Detects IPv4-mapped IPv6.

Levenshtein edit distance (insertions, deletions, substitutions) between two strings, plus a normalized similarity in [0,1]. Operates on Unicode code points.

Parse an A1 range like "A1:C10" into its start and end cells (normalized so start is the top-left) and its extent — row count, column count and total cell count.

Name the interval between two notes (`to` at or above `from`) — its size in semitones, its diatonic number and its quality (perfect/major/minor/augmented/diminished). Correct spelling matters: C→F# is an augmented fourth, C→Gb a diminished fifth.

Guess which chain an address belongs to from its shape and encoding, without requiring the caller to know it up front. Returns the chain, format and address type, or nulls when nothing matches.

Parse an SMPTE timecode (HH:MM:SS:FF, or HH:MM:SS;FF for drop-frame) at a given frame rate into its components, total frame count and real-time seconds. Drop-frame is auto-detected from a ';' separator unless dropFrame is set. Out-of-range or non-existent drop-frame values are a clean 400.

Test whether a RE2 pattern matches anywhere in the input. Returns { matched }. RE2 guarantees linear-time evaluation, so this is safe against catastrophic backtracking. An invalid pattern is a clean 400; no match is a successful 200.

Return the ISO weekday (1=Monday..7=Sunday) and its name for a Gregorian date.

Parse an expression and return the free variables it references, without evaluating it — so an agent can discover which inputs a formula needs before supplying values. A syntax error is a clean 400.

Return both the fully-expanded (eight 4-hex groups) and canonical compressed (RFC 5952) forms of an address. IPv4 is returned unchanged in both fields.

Add or subtract from a timecode, either another timecode (b) or a raw frame count (frames), at a given frame rate. Uses frame-exact SMPTE math (including drop-frame). A result below zero is rejected as a clean 400.

Convert between a spreadsheet column letter and its 1-based index in either direction. Provide exactly one of `letter` (e.g. "AA") or `number` (e.g. 27); the response always contains both.

Split the input into parts on each match of a RE2 pattern. An optional limit caps the number of parts (0 = unlimited). Returns the parts array and its count.

Build an A1-style cell reference from 1-based row and column numbers, optionally marking the row and/or column absolute ($).

Replace matches of a RE2 pattern with a replacement string. Replaces all matches by default (set all=false for first-only). The replacement may reference capture groups with $1, $2 or ${name}. Returns the resulting string.

Guess the character set of a base64-encoded byte buffer: sniffs a byte-order mark, then falls back to ASCII / UTF-8 validation and a UTF-16 NUL-pattern heuristic. Returns a best-effort charset with a confidence label — a low-confidence guess is still a successful 200.

Transpose a note by a (possibly negative) number of semitones. Result is spelled chromatically with sharps, or flats when `preferFlats` is true.

Bitwise AND across two or more non-negative integers (given as decimal strings, evaluated as arbitrary-precision BigInt). Returns the result in decimal, hex and binary.

Apply a Unicode normalisation form (NFC, NFD, NFKC or NFKD) to a string. Reports whether the text changed and its length in UTF-16 code units and code points.