# Changelog of Brave Discussions Scraper (`searchapi/brave-discussions-scraper`) Actor

- **URL**: https://apify.com/searchapi/brave-discussions-scraper/changelog.md
- **Full Actor documentation**: https://apify.com/searchapi/brave-discussions-scraper.md

## Changelog

### Unreleased - 2026-10-02

- Corrected date-only publication labels to `YYYY-MM-DD` plus `publishedAtPrecision: day`, preserving source-zoned timestamps without adding time, timezone, or fractional-second precision. Invalid and partial dates remain raw-only.
- Corrected the dataset view to show panel upvotes, comments, and discussion excerpts without calling them top-answer votes or questions.
- Added publication, calendar, deduplication, limit, and schema-label regressions and tightened the dataset audit to reject the actual pre-fix run's invented timestamps. The four-row successful run on build 2.1.2 remains immutable evidence, not a validation of this offline correction.

### 2.1.1 - 2026-10-02

- Replaced the mismatched Chromium crawler with direct bundled stock Firefox, explicit TLS verification, matching pinned runtime dependencies and an executable build check.
- Reduced default/prefill traffic to five rows, one page, zero retries, direct access; unavailable requested proxies fail closed.
- Added exact search URL validation and observed, query-preserving pagination; source restrictions and unknown empty pages remain terminal.
- Corrected field provenance: comments are not answers, individual answer votes are not thread votes, and a metadata date is not a community name. Narrowed forum recognition and validated outgoing URL schemes.
- Added offline DOM, runtime, pagination, proxy, and failure tests. No live recovery is claimed; retained sample remains historical, with the incorrect answer-count alias removed.

### 2.1.0 - 2026-09-22

- Updated the browser runner to Chrome/Chromium and added current Brave `#discussions` panel extraction.
- Added community, thread, source, comment, upvote, date, discussion-type, and forum metadata while preserving inline Q\&A support.
- Added a 36-field dataset contract, README, sample audit, and generated-by metadata.