A run started with the prefilled input alone now returns data: when the
target-apps field is empty, the apps are looked up from the search terms
first and then processed. Applies to the reviews, app-records and
similar-apps modes. The run only stops early when both the target-apps and
the search-terms fields are empty, and the message now names both.
Bulk reviews: every app in the target list is collected in one run.
Review filter suite: star levels, date window (absolute or relative), keyword
include/exclude, app version, minimum helpful votes, developer-reply-only,
de-duplication.
Review ordering (most relevant / newest / by rating) and a configurable
number of reviews per round trip.
Per-app ratings analytics rows: star mix in percent, ratings per 1,000
installs, average collected score, collected star counts, developer reply rate.
Optional AI classification of each review with the buyer's own provider key:
sentiment, sentiment score, emotion, topics, review category, spam flag,
one-line summary, detected language. Off by default; empty without a key.
Reviews are labelled as child rows and mirrored to a per-run child dataset.
Every row carries scrapedAt and the storefront it came from.
Original input keys and the three app modes are preserved unchanged.
