Skip to main content
Retrieval Augmented Generation (RAG) adds targeted context to LLM calls beyond training data. This example demonstrates a simple RAG workflow using pgvector.
Instead of dedicated vector databases, we use the PostgreSQL database you’re already running with the pgvector extension.

​
Why PGVector?

  • Simple Setup: Uses your existing PostgreSQL database - no additional infrastructure
  • Production Ready: Battle-tested approach used across all Datalumina projects
  • Cost Effective: No separate vector database costs or complexity

​
Quick Setup

1

Checkout Example Branch

git checkout example/pgvector-rag
2

Install Dependencies

The pgvector extension is included in the Docker setup
3

Insert Your Data

Use the vector insertion example to add your documents:
python examples/insert_vectors.py
4

Test the Workflow

python playground/workflow_playground.py

​
Components

​
RAG Service Core

Core RAG implementation
  • Document chunking and embedding
  • Vector similarity search
  • Context retrieval and ranking

​
Vector Management

Insert vectors example
  • Document processing and embedding
  • Batch insertion into pgvector
  • Metadata and indexing setup

​
Workflow Integration

Workflow playground
  • Complete RAG workflow testing
  • Query processing and response generation
  • End-to-end validation

​
Modular Design

Flexible architecture
  • Swap embedding models easily — Customize retrieval strategies
  • Integrate any RAG approach

​
Key Features

​
Modular Integration

The Launchpad’s modular design lets you integrate virtually any RAG workflow you engineer into the system.

​
PostgreSQL Native

Leverages your existing database infrastructure with pgvector for vector operations.

​
Production Validated

Based on patterns used successfully across multiple production deployments.

​
Resources

  • PGVector RAG Service: Complete service with embedding, retrieval, and context management
  • Insert Vectors Example: Shows how to process documents and insert vector embeddings
  • Workflow Playground: Test the complete RAG workflow with your own queries
This example gives you the foundation to understand how RAG integrates with workflows. The Launchpad’s flexibility means you can adapt this approach to your specific requirements and data sources.
⌘I