DocuMind - Secure AI Document RAG Platform

DocuMind is a multi-user Retrieval-Augmented Generation (RAG) platform. It secures user documents through complete Google OAuth authentication, local/cloud PostgreSQL storage, and isolated ChromaDB vector indexing (metadata filtering).

---

Key Features

1. Google OAuth 2.0 Integration: Authenticate securely using Google Login, automatically provisioning user accounts.
2. Session Security (HttpOnly Cookies): Tokens are signed on the backend (JWT) and stored securely in HttpOnly, SameSite cookies to mitigate XSS risks.
3. Strict Document Isolation: PostgreSQL collections utilize metadata filters matching the authenticated database user_id. Users can never query, list, or delete another user's documents.
4. Relational Data Mapping: SQLAlchemy models track Users, Documents, and Vectors in PostgreSQL (Neon serverless setup with pgvector).
5. Real-time Status Polling: The frontend vault tracks document parsing and embedding status dynamically.
6. Animated Dark UI: Designed with TailwindCSS v4 and Framer Motion for a modern, glassmorphic dark-theme console experience.

---

Architecture Diagram

flowchart TD
    User([User]) --> |Uploads PDF| API_Upload[FastAPI /api/upload]
    
    subgraph Ingestion Pipeline
        API_Upload --> Extractor[Extract Text]
        Extractor --> Chunker[Chunking]
        Chunker --> Embedder[Embedding Model]
        Extractor --> Summarizer[LLM Summary Generation]
    end
    
    Embedder --> |Insert Chunks & Embeddings| DB[(Neon PostgreSQL\npgvector)]
    Summarizer --> |Insert Summary & Topics| DB
    
    User --> |Chat Query| API_Chat[FastAPI /api/chat]
    API_Chat --> Router{Intent Router}
    
    Router -->|GREETING / SMALL_TALK| LLM_Greet[LLM Greeting Prompt]
    Router -->|DOC_SUMMARY / DOC_OVERVIEW| DB_Sum[Fetch Stored Summaries\nfrom PostgreSQL]
    Router -->|DOC_QUERY| DB_Vec[pgvector Similarity Search\n+ BM25 Reranking]
    
    DB_Sum --> LLM_RAG[LLM RAG Prompt]
    DB_Vec --> LLM_RAG
    
    LLM_Greet --> Response[Streaming Response]
    LLM_RAG --> Response
    Response --> User

Loading

---

Folder Structure

DocuMind/
├── backend/
│   ├── alembic/                # Database migrations
│   ├── chroma_db/              # Persistent Chroma database storage
│   ├── src/
│   │   ├── auth/               # Auth package (Google verify, JWT, dependencies)
│   │   ├── database/           # DB package (SQLAlchemy connections and models)
│   │   ├── data_loader.py      # Document parser (PDF, TXT, CSV, DOCX, JSON, Excel)
│   │   ├── embedding.py        # Text splitter & Embedding pipe
│   │   ├── search.py           # RAG retrieval & Groq LLM logic
│   │   └── vectorstore.py      # Chroma Store client with user isolation
│   ├── Dockerfile
│   ├── main.py                 # FastAPI application router
│   └── requirements.txt        # Backend python dependencies
├── frontend/
│   ├── src/
│   │   ├── app/
│   │   │   ├── dashboard/      # Protected dashboard workspaces
│   │   │   ├── login/          # Google sign-in landing card
│   │   │   ├── globals.css     # Styling sheets with TailwindCSS import
│   │   │   ├── layout.tsx      # Root html layout shell
│   │   │   └── page.tsx        # Auto-routing landing page
│   │   └── middleware.ts       # Server-side auth route guard middleware
│   ├── package.json
│   └── tsconfig.json
├── docker-compose.yml          # FastAPI + local PostgreSQL config
└── README.md

---

Setup & Configuration

Prerequisites

• Node.js
• Python (3.10+)
• Google Cloud Console client credentials (configured origin )
• Groq API Key (for LLM RAG inference)

1. Environment Variables

Create .env in the backend/ directory:

GROQ_API_KEY="your-groq-api-key"
DATABASE_URL="postgresql://neondb_owner:xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx.tech/neondb?sslmode=require"
GOOGLE_CLIENT_ID="xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx.apps.googleusercontent.com"
JWT_SECRET="rag-for-docs-super-secret-key-change-this-in-production"

Create .env in the frontend/ directory:

NEXT_PUBLIC_BACKEND_URL="http://localhost:8000"

---

How to Run

Development Mode

Step A: Run the Backend

1. Open terminal in backend/ folder:

   cd backend
   # Ensure virtual environment is ready
   uv venv
   # Activate it (Windows)
   .venv\Scripts\activate
   # Install dependencies
   uv pip install -r requirements.txt

2. Apply database schemas to PostgreSQL:

   python -m alembic upgrade head

3. Run the FastAPI development server:

   uvicorn main:app --reload

   The backend API will run on http://localhost:8000.

Step B: Run the Frontend

1. Open another terminal in the frontend/ folder:

   cd frontend
   npm install
   npm run dev

   The client application will run on http://localhost:3000.

---

Production Deployment via Docker Compose

To deploy the stack locally with a local PostgreSQL server, run the following from the root workspace directory:

docker-compose up --build

This launches a PostgreSQL container mapped to port 5432 and builds/starts the FastAPI backend container listening on port 8000.