rag-assistant

RAG Assistant

Phase 3 capstone — GenAI Full-Stack Developer Track

One industry-valuable RAG (Retrieval-Augmented Generation) app — upload company documents, chat with grounded answers and citations.

Stack: Next.js 15 + FastAPI + vector search + OpenAI (or Ollama)

---

Features

• Upload PDF, TXT, MD documents
• Automatic chunking + embedding + vector indexing
• Chat with source citations
• Compare mode: RAG vs plain LLM (see why RAG matters)
• Document list and delete

---

Quick start

cp backend/.env.example backend/.env   # add OPENAI_API_KEY
make install
make dev

• UI: http://localhost:3000
• API: http://127.0.0.1:8000/docs

Upload sample-docs/company-policy.txt and ask: "What is the refund policy?"

---

Learning docs

Start here: LEARNING-GUIDE.md

Doc	Topic
01-rag-concepts	Theory
02-architecture	System design
03-setup-and-run	Install & troubleshooting
04-rag-vs-finetuning-vs-agents	When to use RAG
05-llm-selection-for-rag	Model picks

---

API

Method	Path	Purpose
POST	/api/v1/documents/upload	Index a file
GET	/api/v1/documents	List indexed docs
DELETE	/api/v1/documents/{id}	Remove doc
POST	/api/v1/chat	Ask question (use_rag: true/false)

---

Project structure

rag-assistant/
├── LEARNING-GUIDE.md
├── docs/
├── backend/app/services/
│   ├── rag_service.py      ← main pipeline
│   ├── chunker.py
│   ├── embeddings.py
│   ├── vector_store.py
│   └── llm_client.py
├── frontend/app/page.tsx   ← chat UI
└── sample-docs/            ← test documents