The widget shows cost as "(partial)" when requests were unpriced or
unmetered and as "(approximate)" for an unknown source, but the AI
Insights snapshot flagged spend only for list-price or mixed sources and
the "estimated" value mode. A partial amount was therefore sent as exact,
and a partial amount with an estimated share lost its estimate flag too,
so the model presented a lower bound as the real weekly spend.

Spend now carries "incomplete" for the partial value mode and
"estimated" for approximate or unknown sources and any estimated share.

Docs: ai-insights.md snapshot contract. No CHANGELOG entry: AI Insights
is unreleased and its Added entry still describes the behavior.


Claude-Session: https://claude.ai/code/session_0161xP1fvVJ3SKr6kdENYb4f

Co-authored-by: Claude <noreply@anthropic.com>