AI RAG System — AI Operating System

A full-stack AI platform built as a browser-based desktop OS. Upload documents, query them with AI, run automations, and manage everything from a drag-and-drop desktop interface.

Architecture

React Desktop UI (ai-rag-chat-ui)
        ↓
FastAPI Backend (app.py)
        ↓
AI Router (router.py)
    ├── RAG — Pinecone + LangChain
    ├── Agent — Tool-based reasoning
    ├── Chat History — SQLite
    └── n8n — Workflow automation

Stack

• FastAPI — backend API
• LangChain — document loading, embeddings, agent reasoning
• Pinecone — vector database
• OpenAI — embeddings (text-embedding-3-small) and chat
• SQLite — persistence (chat history, documents, workflow logs)
• n8n — workflow automation
• React — desktop OS UI with draggable windows
• Docker — containerized deployment

Project Structure

ai-rag-system/
├── app.py                  # FastAPI app — all endpoints
├── router.py               # AI Router — intent detection and routing
├── agent.py                # AI Agent — tool-based reasoning
├── ingest.py               # Document ingestion pipeline
├── query.py                # RAG query against Pinecone
├── database.py             # SQLite persistence layer
├── requirements.txt        # Python dependencies
├── Dockerfile              # FastAPI container
├── docker-compose.yml      # All services (FastAPI + React + n8n)
├── n8n-workflows/          # Exported n8n workflow JSONs
├── data.db                 # SQLite database (auto-created, gitignored)
├── uploads/                # Uploaded documents (gitignored)
└── ai-rag-chat-ui/         # React desktop OS UI
    ├── Dockerfile           # React + nginx container
    ├── nginx.conf           # nginx config
    └── src/
        ├── App.js
        ├── apps/
        │   ├── ChatApp.js        # AI chat with history
        │   ├── FileManagerApp.js # Upload + document list
        │   ├── WorkflowApp.js    # Run RAG/agent workflows
        │   └── HistoryApp.js     # Browse chat/doc/workflow logs
        └── components/
            ├── Desktop.js        # Desktop with app icons
            ├── Window.js         # Draggable windows
            ├── Taskbar.js        # Bottom taskbar
            └── ToastContext.js   # Toast notifications

Quick Start (Docker)

docker-compose up --build

Service	URL
React UI	http://localhost:3000
FastAPI	http://localhost:8000
n8n	http://localhost:5678

Manual Setup

1. Install Python dependencies

pip install -r requirements.txt

2. Configure environment variables

Create a .env file:

OPENAI_API_KEY=your_openai_api_key
PINECONE_API_KEY=your_pinecone_api_key

3. Start FastAPI

uvicorn app:app --reload

4. Start n8n

n8n start

5. Start the React desktop UI

cd ai-rag-chat-ui
npm install
npm start

API Endpoints

Method	Endpoint	Description
POST	/upload	Upload and ingest a document
POST	/ingest-path	Ingest a document already on disk
GET	/chat	AI Router — smart intent-based response
GET	/ask	Direct RAG query
GET	/agent	AI agent with tool reasoning
GET	/history/chat	Chat history
GET	/history/documents	Uploaded documents
GET	/history/workflows	Workflow logs

AI Router Intents

Intent	Example	Action
KNOWLEDGE_BASE	"What are Michael's skills?"	RAG query
SUMMARIZE	"Summarize the document"	RAG + summarization
CHAT_HISTORY	"What did I ask before?"	SQLite history lookup
LIST_DOCUMENTS	"List my uploaded files"	SQLite documents query
N8N_WORKFLOW	"Run the ask workflow"	n8n webhook trigger
GENERAL	Anything else	Direct LLM response

Desktop Apps

App	Description
💬 AI Chat	Chat with the AI router, loads previous history
📂 File Manager	Upload documents, view ingested files
⚙️ Workflows	Run RAG or agent queries directly
🕓 History	Browse chat, document, and workflow logs

n8n Workflows

Import the JSONs from n8n-workflows/ into n8n via Workflows → Import from file.

For Docker, update HTTP Request node URLs from http://127.0.0.1:8000 to http://fastapi:8000.

Supported File Types

.pdf .txt .docx