A Gradio WebApp / Huggingface Spaces to perform retrieval augmented generation (RAG) on The Ranveer Show podcast using Langchain, OpenAI Ada Embeddings and FAISS.
Obtain an API key from OpenAI and set it as the value of the api_key variable in the script.
pip install -r requirements.txt
export OPENAI_API_KEY=<API_KEY>
python app.py
- Replace with an open-source alternative to OpenAI ada embeddings. See embeddings.
- Docker setup for this webapp.
- Trial with GPU version of FAISS (not compatible yet for Python 3.10)
- Should retrieve embeddings from a known cache (such as Pinecone DB) to avoid re-encoding already visited videos.
- Custom LlamaHub loader for Podcasts (not just a single video).
- Translation/Transcribe the hindi podcast using technique similar to BongoVaad