RAG POC Mixtral

Proof of Concept for Local RAG Extraction. We are using, mainly:

• Mistral
• ollama
• llamaindex

Getting Started

Start a mongo server instance in Docker. Acts as unstructured data store:

docker run -p 27017:27017 --name rag-mongo -d mongo:latest

Start ChromaDB Server in Docker Container:

docker run -p 8000:8000 --name rag-chromadb -d chromadb/chroma:latest

Install ollama. On Linux:

curl https://ollama.ai/install.sh | sh

We will be using ollama to serve our local LLM as a POC. Keep this terminal open.

ollama serve

To interact with the model (and download it implicitly), you can also use a run command for a specific model:

ollama run dolphin-mixtral:latest
ollama run mixtral:latest

Note that some of these models are $> 26Gb$ in size.