RAG Ingestion Workflow Example

This repository contains a complete example of how to ingest documents to be used later for RAG. Upload a file to an S3 bucket (minio in this demo) and see it appear in a vector database.

This demo:

• Shows how to use restate a reliable and resilient recipient of webhooks
• How to use LangChain with restate
• Using restate's workflows in Python.

Live demo:

Setup ollama locally to download the model

mkdir ollama/

docker-compose pull
docker-compose build
docker-compose up

From a separate terminal, download the embedding model. We are using mxbai-embed-large

curl http://localhost:11434/api/pull -d '{ "name": "mxbai-embed-large" }'

All done, now you can go and upload files to this bucket!

• Minio console Minio - user/password: minioadmin
• Qdrant Qdrant

Demo script

• Upload any .txt / .pdf file into the docs bucket
• Watch new vectors appear at docs collection

To teardown

docker-compose down --remove-orphans

Browse the code / Local development

Using virtual environments:

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt