Nexus — Multimodal Distributed AI Orchestrator

A distributed system to process multimodal inputs (Images/PDFs) with asynchronous task queuing, ML classification, and vector search.

Tech Stack

• FastAPI — REST API server
• Redis — Async task queue
• Scikit-learn — Document classification
• Pinecone — Vector search database
• Matplotlib — Real-time dashboard
• PyMuPDF + Pillow — PDF & Image processing

Features

• Upload PDF or Image documents via API
• Automatic ML-based categorization (technology, medical, legal, finance)
• Vector embeddings stored in Pinecone for similarity search
• Live performance dashboard with real-time graphs
• Fully async pipeline — never blocks on processing

How to Run

1. Install dependencies

pip install -r requirements.txt

2. Setup .env file

PINECONE_API_KEY=your_key_here PINECONE_INDEX=nexus-index

3. Start Redis

brew services start redis

4. Start API Server

uvicorn main:app --reload

5. Start Worker

python worker.py

6. Start Dashboard

python dashboard.py

API Endpoints

• GET / — Health check
• POST /upload — Upload a document
• GET /results — View all processed results
• GET /status — Check queue status

Project Structure

main.py — FastAPI server worker.py — Document processor classifier.py — ML classifier vector_store.py — Pinecone integration dashboard.py — Matplotlib dashboard