SIH Tech Tidbits

Useful tips, libraries and tools from the Sydney Informatics Hub team

Local RAG with LLM


Don't want to share your private data? You can run your own local LLM to query your own documents in a couple steps with a RAG!

  1. Get a local LLM
  2. Get Ollama
  3. Follow the instructions to install.
  4. Open a Terminal and download a model (e.g. llama3.1).
  5. And serve the model locally.
ollama pull llama3.1
ollama serve
  1. Get the interface: We will use LangChain to handle the vector store and Steamlit to give you a fancy front-end to interact with the LLM and data upload.
git clone https://github.com/Sydney-Informatics-Hub/LLM-local-RAG/
cd LLM-local-RAG
conda create -n localrag python=3.11 pip
conda activate localrag
pip install langchain streamlit streamlit_chat chromadb fastembed pypdf langchain_community
  1. Start chatting!
streamlit run app.py

Upload your docs and start asking questions!