Hybrid Adaptive Agentic Architecture for Autonomous Content Curation and Delivery: A Serverless Implementation for Real-Time Tech News Syndication

Abstract: The exponential growth of unstructured online content has rendered manual curation and rule-based automation ineffective for consistent, high-quality social media posting. This paper presents a fully implemented Hybrid Adaptive Agentic Architecture that autonomously searches, summarizes, and posts the latest tech news to Twitter (X). The system comprises three specialized agents—News Search Agent, Summarization Agent, and Twitter Posting Agent—orchestrated by a single main Python script. A focused Hugging Face T5 summarization pipeline generates tweet-ready content with an Internal Quality Score (IQS), while a dedicated URLs tracking page ensures 100% deduplication. The entire pipeline is deployed using serverless computing on Google Cloud Run with two daily cron triggers (8:00 AM and 8:00 PM IST), enabling true 24/7 operation without any local hardware. A Reflection Agent provides continuous self-correction by tuning the IQS threshold based on performance telemetry stored in a dedicated persistent storage layer. A web front-end further allows custom genre prompting for on-demand posting. Experimental evaluation on live tech feeds demonstrates 98.7% posting success rate, zero duplication, and measurable quality improvement through adaptive tuning. The architecture overcomes the brittleness of traditional RPA while maintaining deterministic execution and transparent self-optimization.

Keywords: Agentic Architecture, Machine Learning Summarization, T5 Pipeline, Self-Correcting Systems, Serverless Computing, Content Curation, Persistent Storage Telemetry, Twitter API Integration