j6 S r SSKJr SSKJr SSKJrJrJrJ r SSK
J r J r J
r
SSKJr " S S\ 5 r " S S
\ 5 r " S S \ 5 r " S
S\ 5 r " S S\ 5 r " S S\ 5 r\ \\\\\4 r " S S\ 5 rg)u The defined return values an answer can take.
A question that names a figure should come back as that figure, not as a
sentence containing it. These models are the shapes an answer is allowed to
have. The caller's `want` picks the shape in `expectation`. Under `auto`
the `fixed` agent picks the payload from the `Payload` union in the same call
that answers.
Every shape carries the ids it was read from. Table cells arrive tagged
(`| $2.97 | `), so a scalar can name the exact cell its
value came from, and the value it reports can be checked against the text
printed in that cell — a grounding test no prose answer admits.
) annotations)Decimal)ListLiteralOptionalUnion) BaseModelFieldcomputed_field) as_numberc \ rS rSr% SrSrS\S' \" SS9rS\S ' \" S
S S 9r S
\S' \" S
SS 9r
S
\S' \" S
SS 9r S
\S' \" S
SS 9r S
\S' \
\SS j5 5 rSrg
)Scalar z$One figure, as the source prints it.scalarzLiteral['scalar']kindzThe figure exactly as printed in the source, including its currency sign, separators, and any trailing sign: '$14.47', '4.6x', '92.6%', '(0.02)'. No words, no sentence, no explanation. descriptionstrvalueNzIWhat the number counts: 'USD per share', 'USD thousands', 'percent', 'x'.)defaultr
Optional[str]unitzMThe period or as-of date the figure belongs to, as printed: 'March 31, 2026'.periodzOThe line item as printed in the source: 'Book Value Per Share of Common Stock'.labelzbThe id of the single cell or chunk this figure was read from. Must appear verbatim in the context. source_idc , [ U R 5 $ )a The figure as a plain number, sign applied and separators removed.
Derived from `value` rather than asked of the model. The conversion is
deterministic, and a model asked for it alongside the figure returns the
figure every time and the number only most of the time: one question put
four times came back with the figure four times and the number three.
A numeric column with unpredictable holes is worse than one computed in
a single place and tested there.
r r selfs =/home/mande/repo/quber/src/quber/playground/answers/shapes.pynumber
Scalar.number9 s $$ returnzOptional[Decimal])__name__
__module__ __qualname____firstlineno____doc__r __annotations__r
r r r r r r propertyr! __static_attributes__r$ r# r r r s . &D
&
?E3 _D- "c FM !e E= %
. I}
%
%r# r c r \ rS rSr% Sr\" SS9rS\S' \" SS9rS\S' S r S
\S ' \
\ SS j5 5 r S
r
g )PointH z$One labelled figure inside a series.z4What this point is: a period, a segment, a category.r r r zThe figure as printed.r Nr r c , [ U R 5 $ )zJThe figure as a number, read from `value` on the same terms a scalar's is.r r s r r! Point.numberO s $$r# r$ r% )r' r( r) r* r+ r
r r, r r r r- r! r. r$ r# r r0 r0 H sI .#YZE3Z#; \ rS rSr% SrSrS\S' \" SS9rS\S ' S
r g ) Unanswerableo zThe context does not contain the answer.
A distinct shape rather than an empty scalar: a caller in code must be able
to tell 'not present' from 'zero' without reading English.
unanswerablezLiteral['unanswerable']r z0One sentence: what was missing from the context.r r reasonr$ N)
r' r( r) r* r+ r r, r
rL r. r$ r# r rI rI o s$ %3D
!2$VWFCWr# rI c H \ rS rSr% Sr\" SS9rS\S' \" \SS9r S \S
' S r
g )
Answer} z>The single return value. Always this type; the payload varies.r )
discriminatorPayloadpayloadz7Chunk ids and/or table cell ids supporting the payload.)default_factoryr r@ cited_idsr$ N) r' r( r) r* r+ r
rR r, listrT r. r$ r# r rN rN } s+ H 62GW2 M Iy r# rN N)r+
__future__r decimalr typingr r r r pydanticr r
r quber.playground.answers.figuresr r r0 r5 r= rD rI rQ rN r$ r# r r[ s # 1 1 5 5 6,%Y ,%^ %I % .Y . 9 I X9 X e\9
:Y r#