Curator AI: Knowledge Engine

A premium Retrieval-Augmented Generation (RAG) platform built on Endee, a blazingly fast open-source vector database. Curator AI transforms static documents into a dynamic, searchable, and intelligent private database with a state-of-the-art UI/UX.

• Live Demo*: Live

---

Key Features

1. AI Knowledge Assistant (Premium RAG)

Ask complex questions about your private documents. The system retrieves the most relevant context from Endee to provide accurate, fact-grounded answers.

• Visual OCR Support: Integrated Gemini Vision reads handwritten notes or scanned PDFs.
• Stateful Memory: Maintains context for fluid, multi-turn technical conversations.
• Sleek Glassmorphism UI: A modern, theme-aware interface that adapts to Light and Dark modes.

2. Intelligent Knowledge Analytics

Gain deep insights into your knowledge base performance and user curiosity.

• Real-time Velocity Tracking: Monitor query volume and retrieval speeds via dynamic area charts.
• Topic Clustering: Identify "Market Analysis" trends and other most-queried topics automatically.
• Document Health: Track ingestion statistics and growth metrics (+12% vs LW).

---

How It Works

Step	Functionality	Powered By
1. Text Extraction	Parses PDF, MD, and Text (including Vision OCR)	PyMuPDF + Gemini-Flash
2. Vectorization	Converts text into 384-dim semantic embeddings	S-Transformers
3. Vector Storage	Blazing-fast indexing and similarity search	Endee Vector Database
4. Retrieval	Finds the top context chunks for any query	Endee.query()
5. Generation	Generates professional, grounded answers	Google Gemini 2.0-Flash

System Architecture

graph TD
    A["Documents (PDF/MD/TXT)"] --> B["Python Extraction / Vision OCR"]
    B --> C["Chunking & Embedding"]
    C -->|384-dim vectors| E[("⚡ Endee Vector Store")]
    
    F[" User Question"] --> G["Semantic Embedding"]
    G -->|Query vector| E
    E -->|"Top Context Matches"| H["LLM Prompt Assembly"]
    F --> H
    
    H --> I[" Google Gemini LLM"]
    I --> J[" Fact-Grounded Answer"]

    style E fill:#1a1a2e,stroke:#e94560,stroke-width:2px,color:#fff
    style J fill:#0f3460,stroke:#16213e,stroke-width:2px,color:#fff

Loading

---

Quick Start (Local Setup)

1. Clone & Install

git clone https://github.com/Vikash9546/endee.git
cd endee/assignment
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

2. Launch Endee Database

You can run Endee via Docker or by building the source:

# Via Docker
docker run -p 8080:8080 endeeio/endee-server:latest

3. Run the Dashboard

Ensure your .env file contains your GEMINI_API_KEY:

streamlit run app.py

---

📁 Project Structure

• assignment/app.py: Main entry point (State management & Analytics logic).
• assignment/ui.py: Premium Styling Engine (Glassmorphism & Theme detection).
• assignment/logic.py: Core RAG Pipeline (Endee Indexing, Chunking, LLM).
• assignment/assets/: Branding and profile media.
• assignment/stats.json: Persistent analytics storage.

---

Cloud Deployment

Curator AI is optimized for Streamlit Cloud and Railway.

1. Database (Railway): Deploy endeeio/endee-server and set PORT to 8080.
2. Frontend (Streamlit Cloud): Connect your repo and set GEMINI_API_KEY and NDD_URL as secrets.