This repository contains tools and scripts for using GraphRAG on your own custom CSV files, specifically designed to run on a free T4 instance offered by Google Colab.
To get started, clone the repository using the following command:
!git clone https://github.com/hcss-utils/rls-graphrag.git
Install and run ollama:
curl -fsSL https://ollama.com/install.sh | sh
ollama serve
Next, pull the base model required for your tasks (define it in ./rls/settings.yaml):
ollama pull llama3.1:8b
Then, create a new model with a higher num_ctx parameter (by default ollama uses 2048, but we need more than that-follow this comment or use this Modelfile).
ollama create -f Modelfile custom-model
ollama rm llama3.1:8b
ollama run custom-model
Embedding model:
ollama pull nomic-embed-text
Indexing:
python -m graphrag.index --root ./rls