# Changelog of Global News Archive, Daily Intelligence & AI Briefs API (`thescrapelab/apify-the-rise-of-the-phoenix`) Actor

- **URL**: https://apify.com/thescrapelab/apify-the-rise-of-the-phoenix/changelog.md
- **Full Actor documentation**: https://apify.com/thescrapelab/apify-the-rise-of-the-phoenix.md

## Changelog

### 3.2

- Added per-record collection and freshness timestamps to article output, plus returned-result coverage metadata in the `OUTPUT` record without running an archive-wide aggregate query.
- Added reviewed, publishable starter tasks for a daily country monitor, AI business/risk briefs, and daily country intelligence briefs.
- Made daily briefs default to the previous completed UTC day and expose generation readiness, missing countries, evidence counts, quality state, and requested/resolved model provenance.
- Added strict grounded-output validation so unknown article references and malformed model responses are never published.
- Added summary run, country outcome, token, and OpenRouter cost ledgers with a fail-closed $0.50 daily budget contract.

### 3.1.1

- Replaced unavailable or inconsistent OpenRouter defaults with Gemini 2.5 Flash Lite for Basic and Gemini 2.5 Flash for Pro, both on Zero Data Retention-capable routes.
- Raised the bounded completion allowance to 1,024 tokens so complete strict Signal cards are not truncated; output-limit failures now have an explicit safe error code.
- Kept the $0.006 Basic and $0.012 Pro add-ons, with updated Zero Data Retention price guards and margin calculations.

### 3.1

- Added an explicit Basic/Pro choice for `ai_briefs` at a flat $0.006 or $0.012 add-on.
- Pinned each customer-visible tier to its model, OpenRouter Zero Data Retention price guard, and separate Apify custom event.
- Added tier provenance to every generated brief and to `OUTPUT.aiBriefs` so model, charge, and cost telemetry are auditable.
- Made reuse tier-aware: a Basic stored brief is never represented as a reused Pro brief, and vice versa.
- Added tests for Pro model routing, its upstream price ceiling, tiered event selection, pricing validation, and output contract.

### 3.0

- Added a simple `ai_briefs` mode that turns selected archive articles into structured Phoenix Signal cards.
- Integrated OpenRouter with strict JSON-schema output, bounded retries, Zero Data Retention routing, and provider price guards.
- Reused complete stored article briefs before calling OpenRouter so customers do not pay twice for existing AI work.
- Added business, risk, policy, technology, and general briefing focuses plus English or source-language output.
- Added a pre-AI phrase filter for headlines, SEO descriptions, and source tags.
- Added per-row model, token, OpenRouter request-cost, provenance, status, and safe failure telemetry.
- Added the then-current fail-closed flat $0.006 AI event only for successfully generated and validated OpenRouter summaries; v3.1 supersedes it with Basic and Pro events.
- Added charge-limit preflight, conservative Result-price headroom, and exact continuation after partially delivered pages.
- Capped AI mode at 25 results per run and kept estimate-only mode free of LLM calls.
- Added an AI briefs dataset view, updated schemas, container checks, tests, and Store-facing pricing guidance.

### 2.3

- Locked the production Actor to the tested Apify Python SDK 4 dependency set.
- Reduced the deployment payload to the archive API modules required at runtime.
- Added container-build and latest-CLI schema checks to GitHub CI.
- Authenticated runtime storage diagnostics with the Actor run token and normalized typed Apify Client 3 responses.
- Added structured deployment-contract checks for dependencies, Docker packaging, schemas, and public documentation.

### 2.2

- Added `daily_summaries` output mode for persisted country/topic LLM briefs.
- Added bounded country/date filtering, estimates, and cursor pagination for daily summaries.
- Added separate dataset views for archived articles and daily LLM summaries.
- Kept article and daily-topic importance scores on their documented 1-10 and 0-100 scales.