rskbansal/hack-ai-thon

Hack-AI-thon

This is a Retrieval Augmented Generation based chatbot

The knowledge document was webscraped from this website

• However a simple script can extend this to any website.

• There are plans to include a UI element in the frontend to include whatever website a user wishes to have go through.

• Simple changes to the backend can make this very versatile chatbot to even accept knowledge documents from any text source.

• Further plans are to make this chatbot multi-modal.

The LLMs supported for generation are:

• Mistral

  • Link to download the model

• GPT

• LLama

  • Link to download model

• The LM used for generating embeddings is MiniLM-L6

INSTALLATION

• Clone this repository

git clone https://github.com/rskbansal/Hack-AI-thon.git
cd Hack-AI-thon

• Build the docker image

docker compose up -d

• Create a virtual environment

python -m venv venv
./venv/Scripts/activate

• Install the requirements.py file

pip install -r requirements.txt

• Install haystack

git clone https://github.com/deepset-ai/haystack.git
cd haystack
pip install --editable .
cd ..

• Run the required files

python ingest.py
python app.py

• Open the instance in your browser