RAG Chat

A Retrieval-Augmented Generation (RAG) system that allows users to upload documents and ask questions about them through a conversational interface with FastAPI and SvelteKit.

Architecture

┌─────────────┐     ┌─────────────┐     ┌─────────────────┐
│   Svelte    │────▶│   FastAPI   │────▶│   PostgreSQL    │
│  Frontend   │◀────│   Backend   │◀────│   + pgvector    │
└─────────────┘     └──────┬──────┘     └─────────────────┘
                           │
                           ▼
                    ┌─────────────┐
                    │  Gemini API │
                    │ (Embeddings │
                    │   + LLM)    │
                    └─────────────┘

Key Design Decisions

Vector Database: PostgreSQL with pgvector in Docker:

• Persistence without extra infrastructure
• Transactional integrity with document metadata
• Simpler deployment (single database for all data)

Chunking Strategy: RecursiveCharacterTextSplitter with 1000 char chunks and 100 char overlap ensures:

• Chunks are large enough to preserve context
• Overlap prevents information loss at boundaries
• Recursive splitting respects natural text boundaries (paragraphs, sentences)

Embedding Model: Gemini's gemini-embedding-001 with 768 dimensions provides good semantic representation while keeping storage reasonable.

Retrieval: Top 5 chunks via L2 distance is a balanced default—enough context without overwhelming the LLM's context window.

Tech Stack

• Backend: Python, FastAPI
• Frontend: SvelteKit, TailwindCSS
• Database: PostgreSQL with pgvector extension
• LLM: Google Gemini (free tier)

Prerequisites

• Docker & Docker Compose
• Gemini API key

Setup

1. Clone and configure

git clone <repository-url>
cd RAGChat

cp .env.example .env
# Edit .env and add your GEMINI_API_KEY

Environment Variables

Variable	Description	Default
GEMINI_API_KEY	Your Google Gemini API key	(required)
FRONTEND_URL	URL where the frontend runs	http://localhost:5173
VITE_API_URL	Backend API URL for the frontend	http://localhost:8000

2. Start the database

docker compose up -d

3. Start the backend

cd backend
python -m venv env
source env/bin/activate  # On Windows: env\Scripts\activate
pip install -r requirements.txt
fastapi dev main.py

The API will be available at http://localhost:8000

4. Start the frontend

cd frontend
npm install
npm run dev

The app will be available at http://localhost:5173

Usage

1. Open http://localhost:5173 in your browser
2. Upload a PDF, TXT, or Markdown file
3. Ask questions about the document in the chat interface

API Endpoints

Method	Endpoint	Description
POST	/upload	Upload and process a document
POST	/chat	Send a message and get a response
GET	/documents/status	Check if documents are uploaded
GET	/conversations	List all conversation sessions
GET	/chat/history/{id}	Get messages for a conversation
DELETE	/clear-documents	Remove all documents and history

Project Structure

RAGChat/
├── backend/
│   ├── main.py          # FastAPI app, endpoints, RAG logic
│   ├── database.py      # SQLAlchemy models, DB setup
│   ├── models.py        # Pydantic request/response schemas
│   └── requirements.txt
├── frontend/
│   └── src/
│       ├── routes/      # SvelteKit pages
│       └── lib/
│           ├── api.js        # Backend API client
│           └── components/   # UI components
├── docker-compose.yaml  # PostgreSQL + pgvector
└── README.md