PeopleSearchParseResponse
The typed line read into search fields.
ready is false while fields has no name, @handle or phone — then questions[0].ask
is NAME and the client asks before searching.
fields objectrequired
What the user knows about the person. Every key is optional; a search needs one
of name / username / phone. state / country are the place LEVELS a refinement
question resolved — separate slots, never glued onto location.
Input policy (also CreateRequest.name / username / state and the typed text of
run and parse). maxLength counts Unicode code points. The server first normalizes
every value: NFC; invisible characters removed (zero-width, bidi controls, BOM, soft
hyphen, tag characters; ZWJ / ZWNJ kept); a line break becomes a space (; in info /
work / interests / life, one newline in text), any other control character a
space; { } [ ] < > ` \ $ ^ * " and their full-width forms become a space;
whitespace collapsed and trimmed, empty = unset. Then name needs a letter; username
is at most 3 distinct handles [a-z0-9._-]{1,50} (a link is read down to its handle,
@ dropped, lowercased), sent on as "h1, h2"; phone is one number (digits, spaces,
+ ( ) - . /, dashes; a + only in front or right after a leading ( — (+34) 612…;
at most 15 digits; full-width digits, +, brackets and dashes and the … of smart
punctuation are folded to their ASCII forms first, and the cap counts the folded
number); gender other than male / female is dropped; language is a code
(pt-BR) or a name. No patterns or enums here on purpose: normalization runs first, so
a pasted invisible character is removed, not a 400. maxLength itself is checked on the
value as sent, before normalization: a value over its cap only because of invisible
characters is still a 400, and phone's 40 applies both to the value as sent and to the
folded number.
Errors: 400 with GeneralError.message = <code>:<field> (the field's name or JSON
pointer, e.g. input_too_long:/fields/name). Stable codes: input_too_long,
input_empty (a blank text), input_no_letters, username_too_many_handles,
username_bad_handle, phone_bad_format, language_bad; input_invalid for any
other schema violation. A pointer names only properties declared here and array
indexes — any other key (a map key) is *. A body key that differs from a declared
property only in case (Username) is input_invalid:<pointer to that property>; a body
that passes the schema but does not decode (a number beyond its integer type) is
input_invalid:/. A body over 128 KiB is 413 body_too_large. The rejected value is
never echoed back or logged.
Possible values: <= 100 characters
@handle, with or without the @; up to 3 handles or profile links, comma or space separated
Possible values: <= 200 characters
A full number ALWAYS makes the report a phone lookup, wherever it arrives (at create or in a later round): the number ALONE is searched — a name and facts sent with it are kept on the report, not searched — and the results are masked from that round on. There is no combined "name + phone" search. The last two digits alone are not a phone number: such a value is dropped and never masks anything.
Possible values: <= 40 characters
Possible values: <= 100 characters
Possible values: <= 64 characters
Possible values: <= 64 characters
Anything else, free text — the pieces of several turns, ; -separated.
Possible values: <= 500 characters
male or female (any case); anything else is dropped.
Possible values: <= 16 characters
Possible values: <= 16 characters
Possible values: <= 200 characters
Possible values: <= 200 characters
Possible values: <= 200 characters
Possible values: <= 100 characters
reading object[]required
model, user (from a fix) or regex.
dropped object[]
model, user (from a fix) or regex.
questions object[]required
pair, tag or yesno.
The tag the question is about (NAME for the name law).
options object[]required
fix objectrequired
The user's answer to a parse question or a tap on a chip: {start, end, tag}; tag: "O" =
not a fact. start / end are code-point offsets into the NORMALIZED line (what
POST /people-search/parse read). The schema's own caps — at most 8 fixes, tag at most
16 code points, start / end at most 200 — are checked first and are a 400
(input_too_long / input_invalid). Within them, a fix is kept only when its tag is a
tagger tag ([A-Z_]{1,16}: O, NAME, LOC, …) and 0 ≤ start < end ≤ the normalized
line's length — any other is dropped, not refused (a stale fix of an edited line must not
fail the request).
Possible values: >= 0 and <= 200
Possible values: >= 0 and <= 200
Possible values: <= 16 characters
default object
The user's answer to a parse question or a tap on a chip: {start, end, tag}; tag: "O" =
not a fact. start / end are code-point offsets into the NORMALIZED line (what
POST /people-search/parse read). The schema's own caps — at most 8 fixes, tag at most
16 code points, start / end at most 200 — are checked first and are a 400
(input_too_long / input_invalid). Within them, a fix is kept only when its tag is a
tagger tag ([A-Z_]{1,16}: O, NAME, LOC, …) and 0 ≤ start < end ≤ the normalized
line's length — any other is dropped, not refused (a stale fix of an edited line must not
fail the request).
Possible values: >= 0 and <= 200
Possible values: >= 0 and <= 200
Possible values: <= 16 characters
fixes object[]
Possible values: >= 0 and <= 200
Possible values: >= 0 and <= 200
Possible values: <= 16 characters
LinkedIn profile links found in the line (with or without a scheme, any subdomain). They are
never read as an @handle or a place; a client opens the LinkedIn by-link flow
(POST /people-search/linkedin/{reportId}) for them.
{
"ok": true,
"took_ms": 0,
"fields": {
"name": "string",
"username": "string",
"phone": "string",
"location": "string",
"state": "string",
"country": "string",
"info": "string",
"gender": "string",
"language": "string",
"work": "string",
"interests": "string",
"life": "string",
"brand": "string"
},
"ready": true,
"reading": [
{
"tag": "string",
"text": "string",
"start": 0,
"end": 0,
"confidence": 0,
"source": "string"
}
],
"dropped": [
{
"tag": "string",
"text": "string",
"start": 0,
"end": 0,
"confidence": 0,
"source": "string"
}
],
"residue": "string",
"questions": [
{
"kind": "string",
"text": "string",
"start": 0,
"end": 0,
"read": "string",
"alt": "string",
"ask": "string",
"why": "string",
"p": 0,
"options": [
{
"answer": "string",
"tag": "string",
"label": "string",
"fix": {
"start": 0,
"end": 0,
"tag": "string"
}
}
],
"default": {
"start": 0,
"end": 0,
"tag": "string"
}
}
],
"fixes": [
{
"start": 0,
"end": 0,
"tag": "string"
}
],
"negative_topics": [
"string"
],
"linkedin": [
"string"
],
"truncated_tokens": 0
}