5_mins_rag_no_gpu

RAG in 5 Minutes

This implementation is tied to the YouTube video on NVIDIA Developer.

This is a simple standalone implementation showing a minimal RAG pipeline that uses models available from NVIDIA API Catalog. The catalog enables you to experience state-of-the-art LLMs accelerated by NVIDIA. Developers get free credits for 10K requests to any of the models.

The example uses an integration package to LangChain to access the models. NVIDIA engineers develop, test, and maintain the open source integration. This example uses a simple Streamlit based user interface and has a one-file implementation. Because the example uses the models from the NVIDIA API Catalog, you do not need a GPU to run the example.

Steps

1. Create a python virtual environment and activate it:

   python3 -m virtualenv genai
   source genai/bin/activate
   

2. From the root of this repository, GenerativeAIExamples, install the requirements:

   pip install -r community/5_mins_rag_no_gpu/requirements.txt

3. Add your NVIDIA API key as an environment variable:

   export NVIDIA_API_KEY="nvapi-*"

   If you don't already have an API key, visit the NVIDIA API Catalog, select on any model, then click on Get API Key.

4. Run the example using Streamlit:

   streamlit run community/5_mins_rag_no_gpu/main.py

5. Test the deployed example by going to http://<host_ip>:8501 in a web browser.

   Click Browse Files and select your knowledge source. After selecting, click Upload! to complete the ingestion process.

You are all set now! Try out queries related to the knowledge base using text from the user interface.