A simple ui tool to get a baseline model and evaluation set for a RAG application.
PS: Currently only supports webpages to be used as a knowledgebase and UI is barebones.
It's very easy to build an impressive looking demo but hard to build a useful application that gets used in production. In my experience, the domain expert and engineer are always different people so I wanted something that could harmonize their efforts. So RAG Alligator's goal is to solve that disconnect by allowing the domain expert to build out an evaluation set and a baseline model very quickly with no code and identify a target value for the for final model to achieve on the created evaluation set.
https://github.com/ST2-EV/ragalligator/assets/40699517/313ff13f-f022-4578-9d09-89bc77c58196
git clone [email protected]:ST2-EV/ragalligator.git
cd ragalligator
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
export COHERE_API_KEY=...
export OPENAI_API_KEY=sk-...
It uses cohere for the rag model and openai to run the evals.
mesop app.py