basicRAGwithPython

A small RAG demo that chunks a Markdown document, embeds each chunk, stores the vectors in ChromaDB, reranks the best chunks, and can use Gemini 2.5 Pro to generate a final grounded answer.

Files

• main.py: runs the full demo pipeline.
• split_into_chunks.py: loads doc.md and splits it into text chunks.
• embeddings.py: loads the SentenceTransformer model and creates embeddings.
• vector_store.py: creates the ChromaDB collection, stores vectors, and queries it.
• similarity.py: shared vector math helpers, starting with cosine similarity.
• gemini_llm.py: calls Gemini to turn retrieved chunks into a final answer.

Gemini Setup

Create a .env file in the project root:

GEMINI_API_KEY=your_api_key_here

If GEMINI_API_KEY is not set, the app will fall back to the top reranked chunk.

Run

uv run python main.py

FastAPI

uv run uvicorn api:app --reload

Then open http://127.0.0.1:8000/docs to test the API.

Streamlit

uv run streamlit run streamlit_app.py

Run Both

PowerShell:

.\run_local.ps1

Shell:

./run_local.sh

This starts FastAPI and Streamlit together and stops both when you exit.

Python:

python run_local.py

If you are using the virtual environment directly:

.\.venv\Scripts\python.exe run_local.py