fix(helpers): keep a pagination zero in as_page

[CaptainAni187]

as_page picks the pagination fields out of whichever key the endpoint used, and the candidates are chained with or:

total_count=(
    _get(data, "total_count")
    or _get(data, "total")
    ...
),
total_pages=_get(data, "total_pages") or _get(pagination, "total_pages"),

or skips a falsy value, and 0 is a perfectly real answer here. A page the server honestly reports as empty loses its zeros:

>>> from smallestai.atoms.helpers._envelope import as_page
>>> as_page({"data": {"agents": [], "total_count": 0, "total_pages": 0, "has_more": False}})
Page(items=[], total_count=None, total_pages=None, has_more=False)

total_count=None means "the server did not tell us", so a caller that pages until it has seen total_count items, or that renders "0 of N", cannot tell an empty result from a missing field.

It is also inconsistent depending on where the zero sits in the chain, because the last term of an or is returned whether or not it is truthy:

>>> as_page({"data": {"agents": [], "pagination": {"total": 0, "total_pages": 0, "has_more": False}}})
Page(items=[], total_count=None, total_pages=0, has_more=False)

Same payload, same kind of zero, one survives and one does not.

has_more is already correct, and it is the line directly underneath:

has_more=(
    _get(data, "has_more") if _get(data, "has_more") is not None else _get(pagination, "has_more")
),

That explicit is not None is there because False would hit the same trap. The numeric fields just never got the same treatment.

The change

A small _first_present helper that returns the first candidate which is not None, used for all three fields. Precedence is unchanged, so a flat field still wins over the nested pagination object whether or not it happens to be truthy. has_more moves onto the helper too, which is the same behaviour it had, just no longer spelled out inline.

Tests

Three added to tests/custom/test_as_page_envelope.py: the flat zeros, the nested zeros, and one pinning that a flat 0 still beats a nested 7 so the fix cannot quietly reorder precedence. All three fail on main. The existing seven are untouched and still pass.

_envelope.py is .fernignored, so a regen keeps this.

[crim-app]

crim doesn't review pull requests automatically here.

Comment crim review on this pull request whenever you want a review.

[CaptainAni187]

crim review

[crim-app]

crim is reviewing this pull request. Findings will be posted shortly.

[crim-app]

Good to merge

The bug where an honestly empty page pretended it had no idea how empty it was is now fixed, and _first_present finally teaches this code the difference between "zero" and "nothing." Ship it.

[crim-app]

LGTM

One-sentence assessment: A correct and well-scoped fix that stops as_page from discarding legitimate zero/false pagination values.

What this PR does

Replaces the a or b or ... truthiness chains in as_page with a _first_present helper that selects the first non-None value, so a server that honestly reports total_count: 0, total_pages: 0, or has_more: false no longer has those collapsed to None. The has_more rewrite is behaviorally equivalent to the prior is not None ternary, and precedence (flat fields before nested pagination) is preserved. Tests cover empty-page zeros, nested-pagination zeros, and flat-over-nested precedence.

Findings

No issues found.

[abhishekmishragithub]

Superseded by #120 (re-homed onto an upstream branch so CI could run; your commit is included with authorship preserved, and the test type-check gaps for #112/#118 were fixed there). Shipping in 5.12.1.