Creative RAG Assistant

The idea of this project is to help a creative project team by storing it's documentation files in a vector database which is used in a RAG approach to power a useful team assistant chatbot based on an LLM. Creative teams often accumulate mdeia-rich content like images, videos, audio files, and text documents. This project aims to help these teams to store and retrieve their files in a multimodal way.

Pre-requisites

1. Create vector database

   a. Local setup: Qdrant, Weaviate, ...

   b. Cloud setup: using API keys from hosted vector database

2. Install requirements of a multimodal embedding model (e.g. ImageBind)

3. Setup LLM

   a. Local setup: Ollama -> ollama run llava (or any other local multimodal model, like Bakllava)

   b. Cloud setup: OpenAI API, Claude API, ...

How to use

1. Store files in the vector database

   a. Put files into folder and add that folder to .env file

   b. Run python load_data_into_qdrant.py

2. Run RAG application

   python rag_pipeline.py