Llamaindex RAG Mindful

RAG chatbot that answers questions on Mindful Substack articles, grounded in the actual article text instead of the model's own knowledge.

Setup

1. $ git clone git@github.com:JansenMok/Llamaindex-RAG-Mindful.git
2. $ pip install llama-index llama-index-llms-google-genai llama-index-embeddings-google-genai streamlit
3. Set GOOGLE_API_KEY env var
4. Add article .txt files to articles/
5. $ streamlit run main.py

Pipeline

1. Chunking: articles loaded via SimpleDirectoryReader, split into ~500-token chunks
2. Embedding: chunks embedded using Google's text-embedding-004
3. Retrieval: vector store index (GPTVectorStoreIndex), default top-k similarity search per query
4. Generation: retrieved chunks + question passed to gemini-2.0-flash for a grounded answer

Model Choice

gemini-2.0-flash for generation and text-embedding-004 for embeddings, both via the google-genai SDK, chosen for fast inference and a generous free tier that lets you iterate on retrieval quality without API cost getting in the way.

Interface

Streamlit web app: type a question, get a grounded answer pulled from the Mindful Articles corpus.

Example

Query: "what does the author say about morning routines?" Retrieved: top relevant chunks from that article Response: grounded answer citing the article directly