AI SOLUTIONS

Health RAG Pipeline – AI-Powered Medical Content System

Health RAG Pipeline is an end-to-end AI content automation system built for a health and pharmaceutical client, designed to generate expert-level Bulgarian medical articles grounded exclusively in verified source documents. Built in n8n and connected to Voyage AI, Supabase and Claude Sonnet through a custom RAG architecture, the system combines document ingestion, semantic search and AI writing into one fully automated workflow. The Document Intelligence Layer: the system monitors a shared Google Drive folder, automatically extracts text from incoming PDF and text files, splits it into semantic chunks and indexes each chunk as a 1024-dimensional vector embedding using Voyage AI’s specialized retrieval model. The Knowledge Retrieval Layer: when a topic is submitted, the system embeds the query and performs cosine similarity search across the entire document database using pgvector in Supabase, surfacing the most contextually relevant passages. The Generation Layer: the retrieved context is passed to Claude Sonnet with a strict prompt engineered to prevent hallucination, enforce grammatical Bulgarian and maintain the tone of a senior medical copywriter with 15 years of experience. The result is a structured health articles built entirely from the client’s own verified content, ready for editorial review.