⚡ NexusAI — Full Stack Agentic AI Platform

A production-ready full stack platform with AI agents, RAG pipeline, real-time streaming, and enterprise-grade infrastructure — built from scratch.

---

👋 About This Project

I built NexusAI because I wanted a single platform that combines everything I've been learning — full stack development with TypeScript and Python, agentic AI with LangGraph, real-time WebSocket communication, RAG pipelines, and production DevOps with Docker and Kubernetes.

Most AI projects I found online were either just a Python script calling an API, or a simple chatbot with no real backend. I wanted to build something that feels like a real production system — with proper authentication, background workers, observability, and a clean architecture.

This is that project.

---

🧠 What It Does

NexusAI is a full stack AI agent platform where users can:

• Chat with AI agents that can use tools (web search, database queries, file reading)
• Upload documents and query them using RAG (Retrieval-Augmented Generation)
• Stream responses in real time via WebSocket — no waiting for the full response
• Manage conversations with full history persistence
• Monitor everything — agent runs, API latency, token usage, errors

---

🏗️ Architecture

┌─────────────────────────────────────────────────┐
│              Frontend (Next.js 15)               │
│   React 19 · TypeScript · Tailwind · Zustand    │
└───────────────────┬─────────────────────────────┘
                    │  REST + WebSocket
                    ▼
┌─────────────────────────────────────────────────┐
│               Backend (FastAPI)                  │
│                                                 │
│  ┌──────────────────────────────────────────┐   │
│  │           AI Agents (LangGraph)          │   │
│  │  Planner → Tool Use → Critic → Response  │   │
│  │  Tools: search · code · database · files │   │
│  └──────────────────────────────────────────┘   │
│                                                 │
│  ┌──────────────────────────────────────────┐   │
│  │            RAG Pipeline                  │   │
│  │  Ingest → Chunk → Embed → Store → Search │   │
│  └──────────────────────────────────────────┘   │
│                                                 │
│  Auth · Rate Limiting · Webhooks · Admin Panel  │
│  Celery Workers · Prometheus · Sentry           │
└────────┬──────────┬──────────┬──────────────────┘
         │          │          │
         ▼          ▼          ▼
    PostgreSQL    Redis    Vector DB
    MongoDB               (Qdrant/Chroma)

---

🛠️ Tech Stack

Frontend

Technology	Purpose
Next.js 15 + React 19	UI framework with App Router
TypeScript	Type safety across the entire frontend
Tailwind CSS v4	Styling
Zustand	State management
WebSocket client	Real-time streaming

Backend

Technology	Purpose
Python 3.11 + FastAPI	High-performance async API
LangGraph + LangChain	Multi-agent orchestration
Pydantic v2	Data validation
SQLAlchemy (async)	ORM for PostgreSQL
Celery + Redis	Background task queue
JWT + OAuth2	Authentication

Databases & Infrastructure

Technology	Purpose
PostgreSQL	Primary relational database
MongoDB	Document storage
Redis	Caching, sessions, task queue
Qdrant / ChromaDB	Vector database for RAG
Docker + Kubernetes	Containerization and orchestration
GitHub Actions	CI/CD pipeline
Prometheus + Sentry	Monitoring and error tracking

---

✨ Key Features

🤖 Multi-Agent AI System

Built with LangGraph — agents follow a Planner → Tool Use → Critic → Synthesizer pipeline. Each agent can call tools autonomously, maintain memory across sessions, and stream tokens back to the frontend in real time.

📄 RAG Pipeline

Upload PDFs, DOCX, or plain text files. The system parses, chunks, embeds, and stores them in a vector database. Agents automatically search the knowledge base when answering questions.

⚡ Real-time Streaming

WebSocket-based streaming means users see tokens as they're generated — not a loading spinner for 10 seconds. Includes tool call visualization so users can see what the agent is doing.

🔒 Production Auth

JWT access tokens + refresh tokens, API key support, and Google OAuth2. HTTP-only cookies on the frontend. Role-based access control.

📊 Observability

Full tracing via Logfire (for PydanticAI) and LangSmith (for LangChain). Prometheus metrics endpoint. Sentry for error tracking. Every agent run, tool call, and LLM request is traced.

🖥️ Admin Panel

SQLAdmin panel with authentication — manage users, view database records, monitor background tasks via Celery Flower.

---

🚀 Getting Started

Prerequisites

• Python 3.11+
• Node.js 18+ (or Bun)
• Docker + Docker Compose
• An OpenAI or Anthropic API key

1. Clone the repo

git clone https://github.com/arpitkasaudhan/ai-fullstack-platform.git
cd ai-fullstack-platform

2. Set up environment variables

cp .env.example .env
# Edit .env and add your API keys

3. Start with Docker (easiest)

docker-compose up -d

4. Or run locally

Backend:

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

Frontend:

cd frontend
bun install
bun dev

5. Access the app

Service	URL
Frontend	http://localhost:3000
API	http://localhost:8000
API Docs	http://localhost:8000/docs
Admin Panel	http://localhost:8000/admin

---

📁 Project Structure

ai-fullstack-platform/
├── backend/
│   ├── app/
│   │   ├── main.py              # FastAPI app entry point
│   │   ├── api/
│   │   │   └── routes/v1/       # API endpoints (auth, chat, rag, users)
│   │   ├── core/                # Config, security, middleware
│   │   ├── db/                  # Database models (SQLAlchemy + MongoDB)
│   │   ├── schemas/             # Pydantic request/response schemas
│   │   ├── repositories/        # Data access layer
│   │   ├── services/            # Business logic
│   │   ├── agents/              # LangGraph AI agents
│   │   ├── rag/                 # RAG pipeline (ingest, embed, search)
│   │   └── worker/              # Celery background tasks
│   ├── tests/                   # pytest test suite
│   └── alembic/                 # DB migrations
├── frontend/
│   ├── src/
│   │   ├── app/                 # Next.js App Router pages
│   │   ├── components/          # React components
│   │   ├── hooks/               # useChat, useWebSocket, useAuth
│   │   └── stores/              # Zustand state stores
│   └── e2e/                     # Playwright end-to-end tests
├── docker-compose.yml
├── Makefile
└── README.md

---

🔌 API Overview

Method	Endpoint	Description
POST	/api/v1/auth/login	Login and get JWT tokens
POST	/api/v1/auth/refresh	Refresh access token
GET	/api/v1/chat/conversations	List user conversations
POST	/api/v1/chat/message	Send message to AI agent
WS	/api/v1/chat/ws	WebSocket for streaming
POST	/api/v1/rag/upload	Upload document to knowledge base
GET	/api/v1/rag/search	Search knowledge base
GET	/api/v1/users/me	Get current user profile

Full interactive docs available at /docs (Swagger) and /redoc.

---

🧪 Running Tests

# Backend tests
cd backend
pytest tests/ -v

# Frontend tests
cd frontend
bun test

# E2E tests
bun playwright test

---

🐳 Docker Services

services:
  backend    # FastAPI app
  frontend   # Next.js app
  postgres   # Primary database
  mongodb    # Document store
  redis      # Cache + task queue
  celery     # Background workers
  qdrant     # Vector database
  flower     # Celery monitoring UI
  prometheus # Metrics collection

Start everything: docker-compose up -d

---

📈 What I Learned Building This

• LangGraph agent patterns — how to properly structure multi-step agents with tool use, memory, and human-in-the-loop checkpoints
• WebSocket streaming with FastAPI and how to handle backpressure and reconnection on the frontend
• RAG pipeline design — the difference between naive chunking and recursive character splitting, and why reranking matters
• Async Python — SQLAlchemy async sessions, async Celery tasks, and how to avoid common pitfalls
• Production TypeScript patterns — Zustand for state, React Query for server state, and proper error boundaries

---

🤝 Contributing

Contributions, issues and feature requests are welcome. Feel free to open an issue or submit a PR.

---

📄 License

MIT License — see LICENSE for details.

---

Built with ❤️ by Arpit Kasaudhan

GitHub • LinkedIn