Rails AI

A simple proof of concept demonstrating how to use Rails with Ollama LLM (locally) to generate related articles.

How it works

1. Generate Vector Embeddings: An article's title, tags, and body are used to create a vector embedding.

2. Find Related Articles: The vector embedding is compared to other article embeddings, and the closest vectors determine related articles.

Requirements

• pgvector: For storing and querying vector embeddings in PostgreSQL
• Ollama: A locally hosted large language model (LLM)

Setup

Install and run Ollama Server

brew install ollama
ollama serve
ollama pull llama3:latest

Seed data

Run the following command to populate your database with example data:

rails db:seed

Generate Embeddings

Generate vector embeddings for all articles:

Article.all.each {|article| article.generate_embedding! }

Notes

• Ensure pgvector is installed and configured in your PostgreSQL instance before running the application.
• For more information on how Ollama works or to troubleshoot installation issues, refer to the Ollama documentation.