Spring AI RAG with PGVector and Ollama Integration

[smadil997]

This PR introduces a complete example of a Retrieval-Augmented Generation (RAG) pipeline using:

Spring AI

PGVector (running via Docker)

Ollama for embedding and chat models

Instead of installing a local PGVector DB manually, this setup runs it using Docker Compose (compose.yaml). Please ensure the database is running before starting the Spring Boot application.

⚙️ Key Features

Uses Ollama locally for LLM (gemma:3b) and embedding (nomic-embed-text:latest) models

PDF file upload endpoint at POST /upload for document ingestion

Validates PDFs for table of contents (TOC) – required by default reader

Embeds paragraphs from the PDF and stores them in a vector store using PGVector

Auto-creates vector_store table if not specified

Uses initialize-schema: true in application.yaml to recreate the schema on every app run (can be customized)

[smadil997]

Signed-off-by: Mohdadil [email protected]

[smadil997]

Hi @tzolov please have look on this?