# Changelog of Alternative Social Media Trend Monitor & OSINT (`thescrapelab/alt-social-narrative-trend-intelligence`) Actor

- **URL**: https://apify.com/thescrapelab/alt-social-narrative-trend-intelligence/changelog.md
- **Full Actor documentation**: https://apify.com/thescrapelab/alt-social-narrative-trend-intelligence.md

## Changelog

### 0.1.8 — 2026-09-16

- Fixed short watchlist queries so terms such as `AI` no longer match unrelated substrings in words like `available`, `campaign`, or `air`.
- Preserved matches for punctuation variants such as `A.I.`, exact hashtags, phrases, and domains.

### 0.1.7 — 2026-09-16

- Cleaned greentext quote markers, standalone years, and stray punctuation from singleton titles.

### 0.1.6 — 2026-09-16

- Preserved short source subjects such as product or software version names in deterministic titles.

### 0.1.5 — 2026-09-16

- Made singleton narrative titles client-readable by using the source thread subject or opening sentence.

### 0.1.4 — 2026-09-16

- Filtered dates and URL noise from deterministic narrative titles and preserved meaningful phrase order.
- Added explicit analyzed evidence, delivered evidence-row, and dataset-composition counts to `OUTPUT`.
- Added regression coverage for readable non-AI titles and partial-delivery row accounting.

### 0.1.3 — 2026-09-16

- Grouped the Console form and hid expert-only failure and upstream-cost controls behind safe defaults.
- Added ranked, self-contained narratives with time ranges, source breakdowns, watchlist matches, and bounded evidence previews.
- Expanded `OUTPUT` with effective settings, delivered top narratives, source summaries, warnings, and consolidated costs.
- Reworked Store metadata and README copy for client-facing SEO and added the Actor logo.

### 0.1.2 — 2026-09-16

- Added PPE-aware `narrative-result` delivery with customer spending-limit enforcement.
- Kept evidence and source-status rows free of duplicate event charges.
- Added generated, delivered, and charged narrative counts to the run summary.
- Added OpenRouter token and reported-cost telemetry to `AI_USAGE`.
- Disclosed the bounded public evidence excerpts sent to OpenRouter when AI is enabled.

### 0.1.1 — 2026-09-16

- Corrected strict-mode summaries so a terminal failure is also saved as `failed`.
- Replaced misleading per-record-type output links with one honest complete-results link.
- Reduced the default operator upstream-charge ceiling from $0.50 to $0.10.
- Added parent-deadline headroom and bounded child timeouts for stalled upstream runs.
- Validated 256 MB as the production default through balanced and five-source deep cloud stress tests.

### 0.1.0 — 2026-09-16

- Added private beta orchestration for five stable portfolio source Actors.
- Added deterministic normalization, clustering, scoring, and evidence citations.
- Added optional evidence rows and bounded AI relabeling with deterministic fallback.
- Added hard upstream charge limits, a two-run concurrency limit, and an operator kill switch.
- Enforced fresh-run-only storage with no named or cross-run state.
- Kept Gab target-only after cloud contract testing confirmed its removed search fields are rejected.
- Aligned the saved run summary with the failed terminal status when fresh filtering yields zero usable evidence.