TextNLPClassifierApp Runtime Architecture

TextNLPClassifierApp Runtime Architecture A architecture diagram generated by Archify. Google News RSS · Public News Feeds · Untrusted Public Web Google News RSS Public News Feeds Foxcape Engine · Camoufox Headless · Ingestion & Anti-Bot Sandbox Foxcape Engine Camoufox Headless Multi-Engine Harvester · Trafilatura / N4k / Readability · Ingestion & Anti-Bot Sandbox Multi-Engine Harvester Trafilatura / N4k / Readability F1 Consensus Selector · 5-Token Shingles · Trusted Core Processing Domain F1 Consensus Selector 5-Token Shingles Markdown Sanitizer · Editorial Normalizer · Trusted Core Processing Domain Markdown Sanitizer Editorial Normalizer State Machine · Lifecycle Transition Gate · Trusted Core Processing Domain State Machine Lifecycle Transition Gate ECP Snapshot · Anchors & Graph · Trusted Core Processing Domain ECP Snapshot Anchors & Graph Inherence Classifier · Tier 1 Multilingual NLP · Trusted Core Processing Domain Inherence Classifier Tier 1 Multilingual NLP AI Fallback · Tier 2/3 Vector & LLM · External AI Provider Gateway · Optional AI Fallback Tier 2/3 Vector & LLM Optional Storage Ledger · SQLite Store & FileStore · Trusted Core Processing Domain Storage Ledger SQLite Store & FileStore Observability · Telemetry & Quality Gate · Trusted Core Processing Domain Observability Telemetry & Quality Gate RSS links stealth DOM triplet JSON winner JSON clean MD eval inherence anchors & aliases LLM fallback atomic commit audit telemetry Untrusted Public Web Ingestion & Anti-Bot Sandbox Trusted Core Processing Domain External AI Provider Gateway Legend Backend Database Cloud External

Ingestion & Anti-Bot

  • • Camoufox stealth headless browser with automated anti-bot evasion
  • • Tri-engine simultaneous extraction via Trafilatura, N4k & Readability

Consensus & NLP

  • • Zero-dependency 5-token shingle consensus with F1 score selection
  • • Multilingual NLP classifier (6 languages) weighted by ECP anchors

AI & Fallback Resilience

  • • Tier 2 vector embeddings and Tier 3 LLM fallback for ambiguity (<0.60)
  • • Graceful degradation to deterministic Tier 1 heuristics on API failure

State & Persistence

  • • Finite state machine enforcing strict validated lifecycle transitions
  • • SHA-256 fingerprint deduplication with atomic SQLite & file persistence