Gmail to Vector Embeddings with PGVector & Ollama

This workflow automates the extraction and vectorization of Gmail emails into a Postgres database enhanced with pgvector. It uses the open-source Ollama with nomic-embed-text model for generating embeddings locally, ensuring privacy and no external dependencies. Emails are stored in both structured tables (with metadata like ID, date, sender) and a vector embeddings table, linked by Gmail message ID for hybrid queries.

To use, specify your Gmail account creation date and manually run the workflow to bulk-import historical emails in monthly batches. Once populated, activate the Gmail Trigger node to continuously process new incoming emails. This setup supports semantic similarity searches, allowing queries like 'What hotel did I stay in last summer?' via vector search on email content.

Benefits include 100% local operation for data sovereignty, dual storage for structured and semantic access, and foundation for RAG applications. It saves hours of manual data organization and enables AI-powered insights from email archives without cloud costs.

Ideal use cases: Personal knowledge bases from emails, business email analytics for customer interactions, compliance archiving with search, or integrating with chatbots for querying history (pairs with Telegram RAG workflows). Perfect for users wanting offline AI vector stores.