📖 Mini RAG (LangChain)

Ask-Your-Docs (Tiny RAG):
Load a handful of local files, build a vector index, and answer questions grounded in those files with short citations.

This repo is designed as a minimal, educational project so you can learn how Retrieval-Augmented Generation (RAG) works using LangChain.

---

🔍 What is RAG?

Retrieval-Augmented Generation (RAG) is a simple but powerful pattern:

1. Ingest your documents → split them into chunks → embed them into a vector database.
2. Retrieve the most relevant chunks for a user query.
3. Generate an answer with an LLM using only the retrieved context.

This makes LLMs factual and grounded in your own data.

---

🛠 What we use

• Chroma for the vector store (persists locally).
• HuggingFace embeddings (all-MiniLM-L6-v2) – works offline, no API key needed.
• OpenAI LLM for generation by default (configurable to Ollama if you want local models).
• Python scripts:
  • app/ingest.py → load, split, embed, and index your docs
  • app/rag.py → retrieve + generate pipeline
  • app/streamlit_app.py → simple UI to ask questions

---

⚡ Quickstart

Clone the repo and set up your environment:

git clone https://github.com/yourusername/mini-rag-langchain.git
cd mini-rag-langchain

# Create venv
python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Copy config
cp .env.example .env   # Add your OPENAI_API_KEY if you want to use OpenAI

 # Prepare your docs: Drop some .pdf, .md, or .txt files into data/raw/
mkdir -p data/raw

# Run data ingestion - Build the index
python app/ingest.py

# Run Streamlit UI to query dB
streamlit run app/streamlit_app.py

Streamlit UI