Comunicar (ARGOS)

An advanced, autonomous, AI-powered intelligence engine designed to detect trends, eliminate noise, and transform raw data into high-value strategic narratives. This is not just a news aggregator; it is an Autonomous Media Engine.

The Core Value Proposition

Modern media is saturated with algorithmic noise. ARGOS was designed to cut through this clutter by applying a robust intelligence fusion approach rather than simple aggregation:

• Massive +50 Data Ingestion: Real-time extraction from governmental, financial, social (X/Twitter), and proprietary sources.
• Trend Genesis (The Birth of News): The AI fuses Raw Data + Public Sentiment + Expert Bioprofiles to detect and publish emerging social and economic narratives before they hit mass media.
• AI E-E-A-T Architecture: Content is authored by specialized AI-Agent Personas with expert biographies (LinkedIn-grade), ensuring maximum SEO authority and trust.
• Smart Monetizable RSS: A proprietary syndication engine that transforms intelligence into a B2B data product.

The 8-Step Autonomous Pipeline

The core of ARGOS is a highly sophisticated data pipeline that connects demand signals with raw information, feeding precise contexts to our AI agents:

1. Google Trends RSS Ingestion: Continuously fetches geo-targeted search volume data (e.g., Argentina trends) to capture what the audience is searching for right now.
2. Agent Keyword Matching: Matches trending terms against specific keyword arrays assigned to our specialized AI agents.
3. Parallel RSS Fetching: Simultaneously polls over 50 weighted RSS feeds via Promise.allSettled(), resolving in under 5 seconds.
4. Trend Filtering: Discards all incoming articles that do not contain the active trending term.
5. Algorithmic Deduplication: Analyzes token similarity across headlines. If multiple outlets report on the same event, the engine discards duplicates, favoring the source with the highest assigned "weight" (1-10 scale).
6. Multi-criteria Ranking: Calculates a composite score based on source authority, title match, keyword presence, and recency decay.
7. Dynamic Prompt Injection: Instead of generic instructions, the selected agent is injected with real-time signals. E.g. "BCRA reservas - 320,000+ searches. Analyze the implications based on these 5 actual headlines..."
8. Generative Publishing: The agent authors a 400-600 word strategic analysis. The system calculates a dynamic impact score (boosted by search traffic) and publishes it instantly.

Architecture & Performance

The platform is orchestrated via Docker Compose and structured in highly performant pillars:

• The Intelligence Engine (NestJS + BullMQ): The backend core handles article generation, scoring, and queuing. We recently migrated to this robust stack, backed by PostgreSQL (Prisma ORM) for strict, relational storage of articles and expert personas.
• Edge-First Frontend (Next.js): The UI is optimized for instant rendering and high-performance indexing, achieving an LCP of < 1.2s and a Google PageSpeed score of 95+.
• Zero-Downtime CI/CD: A GitHub Actions pipeline automatically pulls the latest code, executes Prisma migrations, and performs rolling restarts via PM2 without dropping incoming requests.

The Result

The ARGOS engine successfully demonstrates how applied AI can move beyond simple summarization. By marrying quantitative demand signals (Google Trends) with qualitative intelligence (specialized AI personas), it produces actionable, high-impact content at machine speed.