Collects the latest G2 product reviews from public RSS feeds. No login, no browser, no residential proxy. Built for incremental monitoring, not full review history.
All notable changes to this actor are recorded here. This project follows
semantic versioning .
0.1.0 — unreleased
First working version. Not yet published to Apify Store.
Added
Collects the latest G2 reviews from the public product RSS feed
(/products/{slug}/reviews.rss), with no login, browser or proxy.
Accepts product slugs or G2 product URLs, up to 100 products per run. A
product listed twice is collected once; only the validated slug is kept.
Flat output of 15 fields per review, suitable for JSON, CSV and Excel export.
Star rating filter on the rating each review actually displays, in steps of
0.5. G2's coarser feed bands are used only to fetch candidates; every review is
re-checked against its displayed rating.
Incremental collection through publishedAfter (strictly after the given
instant, compared in UTC) and knownReviewIds (up to 10,000 IDs).
Deduplication across feeds by review ID and by normalised review URL. Where
copies of one review name exactly one ID, a copy that lacks it adopts that
verified ID; conflicting identities are counted and never merged.
Per-product limit, a run-wide review ceiling and a maximum price worked out
and logged before the first request.
Pay-per-event billing: actor-start and review-saved. A review is charged
only after the dataset has accepted it. Duplicates, filtered-out reviews, empty
results, unknown products and failed stores are not charged.
Run summary written to the OUTPUT key of the key-value store, with
per-product status and the counts from every filtering stage. The dataset holds
review rows only.
Per-product error classification: succeeded, partial, no_reviews,
invalid_product, blocked, rate_limited, temporary_error, parse_error.
One failing product does not stop the rest of the run.
Retries with exponential backoff for temporary failures, honouring
Retry-After on rate limits, capped at 60 seconds.
Run-wide request concurrency limit (1–5, default 3) shared by all products.
Reviewer display names excluded by default.
Known limitations
The public feed is a snapshot with no paging. The default feed carries roughly
the 25 most recent reviews; merging the per-rating feeds adds more recent
candidates, but they overlap, so the number of distinct reviews varies by
product. maxReviewsPerProduct accepts up to 125 as a ceiling, not a promise.
Full review history is out of scope.
Products are processed sequentially, so run time grows with the number of
products.
Duplicate storage and duplicate charging are prevented within a single
process, but not across a platform restart or a request whose outcome is
unknown. Use knownReviewIds or publishedAfter to control this.
knownReviewIds is capped at 10,000 entries; automatic state between runs is a
candidate for a later version.
Parsing depends on the structure of G2's public feed and may need updating if
that structure changes.
