RAG with txtai
A Streamlit application, backed by txtai, that answers questions over a person's own data through vector-search RAG and graph-path-traversal RAG, run as a Docker container or a Python virtual environment.
Vector RAG runs a vector search over an Embeddings index to find the documents most relevant to a query, then places those matches into an LLM prompt to generate an answer. Graph RAG instead runs a graph path-traversal query: a gq: prefix expands an initial vector search using a graph network stored alongside the vector database, or a graph path query takes a list of concepts, finds the closest matching nodes, and traverses the graph to build context from related nodes.
It runs as a Docker container available on Docker Hub with GPU support (docker run --gpus=all ... neuml/rag), or installed directly with pip install -r requirements.txt and started with streamlit run rag.py inside a Python virtual environment.