🚀 RAG From Scratch (FastAPI + Groq)

A lightweight Retrieval-Augmented Generation (RAG) system built using FastAPI and Groq LLM.

✨ Features

• 📄 Upload PDF & TXT documents
• 🔍 Semantic search (cosine similarity)
• 🤖 LLM-powered answers (LLaMA 3 via Groq)
• ⚡ FastAPI backend
• 🌐 Simple web UI

🛠️ Tech Stack

• Python
• FastAPI
• Groq API
• HTML/CSS

▶️ Run Locally

pip install -r requirements.txt
uvicorn main:app --reload

Open: http://localhost:8000

Folder Structure

rag_project/
│
├── main.py              ← FastAPI server (the "waiter")
├── rag_engine.py        ← All RAG logic (the "kitchen")
├── requirements.txt     ← Python packages to install
├── .env                 ← Your API key goes here
│
├── static/
│   └── index.html       ← Chat UI (opens in browser)
│
└── documents/
    └── sample_handbook.txt  ← Sample document to test with

---

Setup (do this once)

Step 1 — Open the folder in VS Code

File → Open Folder → select rag_project

Step 2 — Open the VS Code terminal

Terminal → New Terminal (or press Ctrl+`)

Step 3 — Create a virtual environment

python -m venv venv

Activate it:

• Windows: venv\Scripts\activate
• Mac/Linux: source venv/bin/activate

You'll see (venv) appear in the terminal. Good.

Step 4 — Install packages

pip install -r requirements.txt

Step 5 — Add your API key

Open .env and replace your-api-key-here with your real key:

ANTHROPIC_API_KEY=sk-ant-...

Get a key at: https://console.anthropic.com

---

Running the server

uvicorn main:app --reload

You'll see:

INFO:     Uvicorn running on http://127.0.0.1:8000

Open your browser at: http://localhost:8000

The --reload flag means the server restarts automatically whenever you save a file. Great for development.

---

Using the app

1. Click "Click to upload" in the sidebar
2. Select documents/sample_handbook.txt (or any .txt/.pdf)
3. Click "Upload & Index" — watch the terminal as it chunks and embeds
4. Type a question like: "What is the vacation policy?"
5. See the retrieved chunks + generated answer

---

API Endpoints

Method	URL	What it does
GET	/	Opens the chat UI
POST	/upload	Upload + index a document
POST	/ask	Ask a question, get an answer
GET	/status	See what's indexed
DELETE	/reset	Clear all indexed docs

You can also test the API directly at: http://localhost:8000/docs (FastAPI gives you a free interactive API explorer)

---

How the RAG pipeline works

Your document
     ↓
[Load]  →  Read the raw text
     ↓
[Chunk]  →  Split into ~400 char pieces with overlap
     ↓
[Embed]  →  Convert each chunk to a vector (list of numbers)
     ↓
[Store]  →  Keep vectors in memory

--- When you ask a question ---

Your question
     ↓
[Embed question]  →  Same embedding model
     ↓
[Similarity search]  →  Find chunks with closest vectors
     ↓
[Augment prompt]  →  question + top 3 chunks
     ↓
[Generate]  →  Claude reads context and answers
     ↓
Answer!