Twine is a premium, retrieval-augmented intelligence system designed to synthesize complex knowledge into a high-fidelity digital persona. Built with a "Cyber-Industrial" aesthetic, it combines advanced RAG (Retrieval-Augmented Generation) with deep administrative oversight to deliver a state-of-the-art conversational experience.
Frontend: https://twine-rag.vercel.app/
Backend: twine-backend.onrender.com
A sophisticated, high-contrast entry point designed for secure administrative and demo access.
The core conversational engine featuring clean, industrial typography and smooth, rounded message geometry.
Real-time administrative oversight with glassmorphism panels and high-density information layout.
Deep operational telemetry featuring monotone data curves and glowing progress indicators for RAG accuracy and engagement.
- Intelligence: OpenAI GPT-4o +
text-embedding-3-small - Memory: ChromaDB (Vector) + PostgreSQL (Relational)
- Engine: FastAPI (Python 3.10+)
- Interface: React 18 + Tailwind CSS + Lucide
- Infrastructure: Vercel (Frontend), Render (Backend), Neon (DB)
- Hybrid RAG Architecture: Seamlessly routes queries between general knowledge bases and high-priority "Gold" facts taught directly by admins.
- Drive-Native Ingestion: Automatically synchronizes and indexes PDF/DOCX content from Google Drive.
- Cyber-Industrial UI: A unique design system utilizing deep glassmorphism (
backdrop-blur-3xl) and a curated Sky-Indigo palette. - Real-Time Teaching: Admins can use the
!learncommand to instantly update the engine's memory during live sessions. - Advanced Telemetry: Comprehensive analytics dashboard for monitoring engine velocity, accuracy, and user sentiment.
cd backend
pip install -r requirements.txt
uvicorn backend.main:app --reloadcd frontend
npm install
npm run devpython scripts/setup_drive.py© 2026 Twine Intelligence Engine | Built for High-Performance Persona Synthesis.