AI Chatbot with RAG: Smart Workflow Using Google Gemini

Transform your documents into an intelligent, conversational knowledge base with this powerful n8n workflow that combines Google Gemini AI with Supabase vector storage. This complete RAG (Retrieval-Augmented Generation) system processes your company documents and creates a smart chatbot that provides accurate, context-aware answers based strictly on your uploaded content.

What this workflow does

This comprehensive two-in-one architecture delivers both document processing and conversational AI in a single workflow template. The ingestion pipeline processes and stores your uploaded documents using Google Gemini's embedding model (models/gemini-embedding-001), while the chat interface leverages Gemini 2.5 Flash to generate intelligent responses. The system maintains persistent conversational memory through PostgreSQL, tracking chat histories per sessionId to ensure contextual continuity across conversations. Each browser window gets a unique session ID, keeping user conversations completely separate.

Key technical capabilities

• Vector-powered accuracy through Supabase pgvector integration that retrieves the top 5 most relevant document chunks
• Global error handling with built-in error triggers for API rate limits, parsing failures, and bad requests
• State-of-the-art Google Gemini AI models for embeddings and chat generation
• Automatic session management preventing conversation cross-contamination

Use cases

Perfect for SaaS operators building customer support knowledge bases from documentation, automation engineers creating internal FAQ systems from company policies, or any organization needing to make large document collections instantly searchable and conversational. Ideal for HR departments, technical documentation teams, and customer success teams who need AI-powered document querying without hallucination risks.

Technical details

Built with essential n8n nodes including form trigger for document upload, code nodes for processing logic, set nodes for data manipulation, conditional if nodes for flow control, error triggers for robust handling, and sticky notes for workflow documentation. Integrates seamlessly with Google Gemini API and Supabase PostgreSQL with vector extensions.