scripture_rag

A repo for me to play around with RAG (retrieval augmented generation) using LDS scriptures

Usage:

To set up your environment and download data, run startup.sh. This will download data into the data directory and create an environment ollama_env

In order to run the python files, run source ollama_env/bin/activate. You will also need to have an ollama server running locally, and you need to have pulled the models for 'llama3' and 'nomic-embed-text'

To create the feature store which will be used in RAG, run create_embeddings.py. This will need to be done before running answer_scripture_questions.py

Finally, once the above steps are done, to ask a question about the scriptures run answer_scripture_questions.py.

Using a standalone Chroma FastAPI server

You can run ChromaDB as its own FastAPI server and point the code at it:

• Start the Chroma server (ensure chromadb is installed in your venv):

  • ./run_chroma_server.sh
  • or manually: chroma run --path chroma_dir --host 0.0.0.0 --port 8000

• Configure clients to use the server by editing config.py:

  • CHROMA_USE_HTTP = True
  • CHROMA_SERVER_HOST = "localhost" (or your host/IP)
  • CHROMA_SERVER_HTTP_PORT = 8000

With this configuration, create_vector_store.py and answer_scripture_questions.py connect to the Chroma server via REST and store/query vectors on that server.

Retrieval FastAPI Service

This repo now includes a standalone FastAPI microservice that wraps the scripture retrieval logic from llm_retrieval.py.

• Start the service (ensure dependencies are installed):

  • uvicorn retrieval_service:app --host 0.0.0.0 --port 8001

• Endpoint:

  • POST /retrieve
    • Request body: { "question": "<your question>", "generative_model": "gpt-4o-mini" }
    • Response: { "scriptures_string": "<tab-separated scripture refs and text>" }

• Health check:

  • GET /health -> { "status": "ok" }

The service initializes the scriptures index and a Chroma collection on startup. It supports using a standalone Chroma server when configured via config.py (CHROMA_USE_HTTP, CHROMA_SERVER_HOST, CHROMA_SERVER_HTTP_PORT).