Weaviate Ollama Simple RAG

🎯 Overview

This is a simple demo of how one can run Weaviate with Ollama modules and create simple RAG.

📦 Requirements

In order to be able to create Weaviate one needs at least.

1. Docker
2. Python3

💡 Running

In order to run the setup one needs to issue:

docker compose up -d

Please note that below operations take some time to succeed.

Pull Aya model. It's an open source multi linugal large language model from Cohere. Please note that you need at least 5GB of free space on your disk.

docker exec -i generative-ollama ollama pull aya

In order to run the python scripts, it's advised to setup a venv for a project.

python3 -m venv .venv
source .venv/bin/activate
pip3 install -r requirements.txt

In order to import data issue:

python3 import.py

🔍 Search

Data in this example has been vectorized using 2 embedding models (english only and multi lingual). Their target vector names are respectively english and multi_lang. Those 2 examples below shows a difference in querying english data using polish language and using 2 different target vectors:

• query_in_pl_using_english_model.gql
• query_in_pl_using_multi_lang_model.gql

📖 Prompts

Create a tweet prompt (generate_tweet.gql):

Use lot's of emojis to create a short (max 15 words) and passionate tweet in English language recommending a book from {author}, add hashtags at the end of the tweet, use this description: {description}

🤖 Code samples

All of those examples can be also executed using python scripts. If you haven't imported data issue:

python3 import.py

in order to perfom search queries and run RAG samples issue:

python3 query.py