PlateWise is a production-ready, AI-native SaaS application designed to transform raw documents (PDFs, DOCX, TXT) into actionable intelligence. It uses advanced Hybrid RAG (Retrieval-Augmented Generation), Supabase for multi-tenant isolation, and a dynamic LLM cascade supporting both proprietary (Google Gemini, OpenAI, Claude) and open-source models (Llama 3, Mixtral via Groq).
Built originally for food delivery operations and restaurant compliance, the engine is fully generalized for any document-heavy enterprise workflow.
- Agentic RAG Upgrade: Query decomposition for compound questions, self-correcting retrieval on low similarity, confidence scoring with honest fallback, and instant semantic caching (< 15ms response time).
- Hybrid RAG Retrieval Engine: Combines
pgvectorsemantic embeddings with BM25 keyword search using Reciprocal Rank Fusion (RRF) for 94.2% hit rate. - Dynamic LLM Cascade: Intelligently routes queries with automatic fallback (Gemini 3.1 Flash Lite → Groq Llama 3.3 70B → Groq Mixtral 8x7B).
- Premium UI/UX: Built with React 19 & Tailwind CSS v4. Supports Light and Dark mode theming inspired by top-tier SaaS (Linear, Vercel).
- Secure Authentication & Billing: Integrated with Supabase Auth for Email/Password & Google OAuth, plus Stripe/Razorpay billing integration.
- Frontend: React 19, Vite, Tailwind CSS v4, Lucide Icons, React Router
- Backend: Python 3.11+, FastAPI, Uvicorn, SentenceTransformers (
all-MiniLM-L6-v2) - Database / Auth: Supabase (PostgreSQL + pgvector + GoTrue Auth)
- AI Inference: Google GenAI SDK, Groq SDK
- Document Parsers: PyMuPDF, python-docx, python-pptx, pandas
Follow these instructions to get PlateWise running on your local machine.
- Node.js (v18+)
- Python (3.10+)
- A Supabase account (Free tier is fine)
- A Groq API Key (Free)
- A Google Gemini API Key (Free)
Clone the repository and set up your .env file in the root backend directory:
cp .env.example .envOpen .env and fill in your Supabase, Gemini, and Groq keys.
Create a virtual environment and install the dependencies:
# Windows
python -m venv venv
.\venv\Scripts\activate
pip install -r requirements.txt
# Mac/Linux
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txtStart the FastAPI server:
uvicorn app.main:app --reloadThe API will run on http://localhost:8000.
Open a new terminal window, navigate to the frontend directory, and install npm packages:
cd frontend
npm installStart the Vite development server:
npm run devThe web app will run on http://localhost:5173.
- Frontend: Deploys seamlessly to Vercel using the included
vercel.jsonconfiguration. - Backend: Designed to be containerized and deployed to services like Render, Railway, or AWS App Runner.
- Database: Supabase handles cloud persistence automatically.
This project is licensed under the MIT License - see the LICENSE file for details.
Built with ❤️ for modern operations.