Custom RAG App

General

An app that utilizes streamlit to offer a user-friendly interface for creating a rag pipeline. The app allows the user to upload his/her own PDF fies, which are then processed and stored in a new (or existing) vector database.

The user can then query the database, and receive the most relevant response generated from an llm model. The LLM that we chose to use is the gemma:2b model, and for model serving we use Ollama.

The app also offers the ability to delete an existing database, although currently this functionality is limited for Windows.

Setup

In order to get the app running, first you need to clone this repository. This can be done with the command:

git clone https://github.com/Jimlibo/custom-rag-app.git

After you have cloned the repository, you can navigate to the app's directory:

cd custom-rag-app

Run using Docker

If you have docker and docker-compose installed, you can run the app with the following commands:

docker-compose up

Run using Streamlit

If you do not have docker installed, you can first install the required packages from requirements.txt and then run the app through streamlit. This can be done with the following commands:

cd src
pip install -r requirements.txt
streamlit run app.py

In order to run inference on a LLM, you also need to have the Ollama service up and running. If it's the first time you run the app, you have to pull the gemma:2b model from ollama repository before serving it. To pull the model, open another terminal and execute the command:

ollama pull gemma:2b

To serve the model, run:

ollama serve

⚠️ If the app does not open automatically, you can navigate to http://localhost:8501/ in the browser of your choice.

License

Distributed under the MIT License. See LICENSE for more information.