🦅 Panoptes: Proactive AI Command for San Francisco

Panoptes — Beyond Reactive Policing.

---

👨‍💻 Quickstart Guide

# Clone the repository
git clone https://github.com/MatthewTran22/SparkHacks.git

# Install Frontend dependencies
npm install
npm run dev

# Setup Backend (FastAPI + Nemotron)
cd backend
python -m venv venv
#For MacOS/Linux
source venv/bin/activate
#For Windows
source venv/Scripts/activate
pip install -r requirements.txt
python app/main.py

Local LLM Setup

Follow the Instructions linked here

• Nemotron setup: https://github.com/NVIDIA/dgx-spark-playbooks/tree/main/nvidia/nemotron
• VLM setup: https://github.com/NVIDIA/dgx-spark-playbooks/tree/main/nvidia/live-vlm-webui

---

🏗 Project Structure

• app/: Next.js frontend with Tactical Command UI.
• backend/: FastAPI agentic service with NVIDIA Nemotron integration.
• lib/: High-fidelity geographic data (SF Neighborhoods fixed with convex hulls).
• docs/: PRDs, Data Schemas, and the Simulation Blueprint.

---

Judges / VCs: Why DGX Spark?

Panoptes requires massive parallel processing of real-time stressors and LLM contexts. The DGX Spark allows us to unify our vector memory and crime databases, hitting < 2s end-to-end latency—critical for high-stakes law enforcement.