genericmaster/tla-advisor- ? reverse-engineered prompt

Build me a simple Python app for a feedback based RAG system for technical lab assistants.

I want it to take in notes or documents, store them in a local vector database, and answer questions using retrieved context. Use Ollama for both embeddings and response generation, and make it easy to swap out models later if needed. The app should have a clean startup flow that loads environment settings, creates the Ollama client, sets up the embedder, and connects to the vector store.

Please keep the code organized so the embedder, vector store, and generator are separate pieces, and make the prompt formatting for retrieved context a small standalone helper. I also want a basic way to add documents, search them, and generate a helpful answer from the matched chunks. If you need to check current docs for Ollama or Chroma setup details, go ahead and look them up online.