Chat with Internal Documents: Ollama RAG + Supabase + Google Drive

This n8n workflow creates a conversational AI interface to query internal documents using Retrieval-Augmented Generation (RAG). It automatically ingests files from a Google Drive folder, supporting PDFs, Excel, Google Docs, and text files. Content is extracted, preprocessed, split into chunks, embedded via Ollama, and stored in Supabase Vector DB for efficient semantic retrieval. Old embeddings are purged on updates for accuracy.

The chat agent, triggered by webhook or interface, uses Ollama LLM to process user queries. It calls a custom tool to search the vector store, fetching relevant document chunks. Responses are grounded in retrieved context, ensuring accuracy within provided texts. PostgreSQL stores conversation memory for context-aware continuity, enabling multi-turn chats.

Benefits include hands-free document search, real-time updates, and reduced manual effort. Ideal for marketing teams handling business directories, sales docs, or research files—quickly answer queries like 'What are our key partners?' without sifting through folders. Scalable for knowledge bases, support bots, or internal Q&A, saving hours weekly.