🧠 RAG Learning Guide

Interactive visual guide to Retrieval-Augmented Generation — built from whiteboard notes.

Live demo: https://shri-phnx.github.io/rag-learning-guide/

---

What's Inside

Tab	Content
🗺️ Pipeline	End-to-end RAG flow — Indexing (DL → TS → Embedding → Vector Store) + Query (Retriever → SP+R+UQ → LLM → G). Tap any stage for details.
📖 Context	What RAG is, why it exists, the core insight, when NOT to use it, common pitfalls, RAGAS evaluation metrics, and real-world use cases.
💻 Code	Syntax-highlighted Python/LangChain code for every single stage, plus a full end-to-end Streamlit pipeline. Copy button included.
⚡ Tech Stack	5 layers compared (UI, Orchestration, Embedding, Vector DB, LLM) with effort ratings, recommendations, and a RAG vs Fine-tuning decision guide.

---

Tech Stack Used in the Guide

• UI: Streamlit · Gradio · Flask · FastAPI
• Orchestration: LangChain (LC) · LlamaIndex · Haystack
• Embedding: OpenAI Ada-002 · HuggingFace BGE · Cohere
• Vector DB: FAISS · Chroma · PgVector · Pinecone
• LLM: GPT-4 · Claude 3.5 · Gemini 1.5 · LLaMA 3

---

Run Locally

npm install
npm run dev

Open http://localhost:5173/rag-learning-guide/

The Augmented Prompt Formula

SP  +  R  +  UQ  →  LLM  →  G
│       │       │              │
System  Retrieved  User        Grounded
Prompt  Chunks     Query       Answer

---

Built from whiteboard notes — my understanding of how RAG works and how to implement it.