# Changelog of Amazon Review Analysis Scraper (`automation-lab/amazon-review-demand-signals`) Actor

- **URL**: https://apify.com/automation-lab/amazon-review-demand-signals/changelog.md
- **Full Actor documentation**: https://apify.com/automation-lab/amazon-review-demand-signals.md

## Changelog

### 0.1.2 — 2026-09-15

- Documented the optional OpenAI label-enrichment payload, user-provided credential handling, 25-second fallback, provider retention boundary, and Apify run-storage deletion boundary.
- Clarified that review text and quotes are not transferred to OpenAI and that accepted model labels are stored only as part of normal dataset results.

### 0.1.1 — 2026-09-12

- Added opt-in per-product fault isolation for multi-ASIN batches. Successful summaries are preserved when another product cannot be acquired, and each failed product is recorded in the machine-readable `FAILED_PRODUCTS` key-value store record without a product-analysis charge.
- Runs still fail loudly when no product can be analyzed; the previous fail-fast behavior remains the default.
- Increased run memory to 512 MB so protected-source retries remain reliable across multi-product batches.

### 0.1.0

- Added Amazon.com ASIN and product/review URL analysis.
- Added deterministic complaint, feature-request, unmet-need, and praise clusters with source-attributed quotes.
- Added product and marketing action recommendations, SHADER-first sessions, residential fallback, and zero-review-shell rejection.
- Added optional user-key OpenAI cluster label enrichment; extraction and analysis remain fully functional without a model key.
- Added clause-level polarity checks and product-relevant request matching to prevent neutral anecdotes, brand substrings, and praise-only evidence from producing false demand actions.