Skip to content

lumeirne/nova_hackathon

Repository files navigation

🏙️ CityPulse: Smart Urban Issue Resolution

CityPulse is a next-generation civic technology platform designed to revolutionize how citizens report and municipalities manage non-emergency urban issues (e.g., potholes, broken streetlights, water leaks, and public hazards).

🎯 The Purpose

Traditional municipal reporting systems are often plagued by manual triage bottlenecks, duplicate reports, and delayed response times. CityPulse solves this by placing an intelligent, multi-agent AI pipeline between the citizen and the city departments.

By automating the intake and triaging process, CityPulse:

  • Empowers Citizens with a seamless, conversational reporting interface (complete with GPS mapping and multimodal image uploads).
  • Accelerates City Operations by instantly categorizing, deduplicating, and prioritizing incoming issues.
  • Ensures Accountability by auto-assigning reports to the correct departments and calculating dynamic SLA (Service Level Agreement) deadlines.

🚀 Powered by Amazon Nova Foundation Models

At the core of CityPulse is a sequential 4-Agent Pipeline powered by Amazon Bedrock's state-of-the-art Nova Foundation Models. The system orchestrates these models to handle complex semantic reasoning, multimodal analysis, and dense vector comparisons in real-time (under 10 seconds per report).

1. The Classifier Agent 🧠

  • Model Used: Amazon Nova Lite (Multimodal)
  • Role: Analyzes the citizen's natural language description and uploaded images to accurately classify the issue into strict municipal categories and sub-categories. Nova Lite's multimodal capabilities ensure that visual evidence (like a photo of a sinkhole) is cross-referenced with the text to establish high-confidence categorizations.

2. The Deduplicator Agent 🔍

  • Model Used: Amazon Nova Embeddings
  • Role: Prevents operational bloat by catching identical or related reports. It uses Nova Embeddings to convert incoming issue descriptions into dense semantic vectors. By combining a vector similarity search (via ChromaDB) with geographic proximity checks (Haversine distance bounding), the agent accurately links new reports to existing open tickets.

3. The Prioritizer Agent 🚨

  • Model Used: Amazon Nova Lite
  • Role: Acts as the essential triage engine. Nova Lite evaluates multiple dimensions—such as safety risk, infrastructure criticality, population impact, and time sensitivity—to assign a rigid 1-10 priority score. The model explains its reasoning, ensuring transparent and justifiable resource allocations.

4. The Router Agent 🎯

  • Model Used: Amazon Nova Lite
  • Role: The final decision-maker. It synthesizes the category, priority, and original context to route the issue to the exact city department equipped to handle it (e.g., "Roads & Transportation" vs. "Water & Sewage"). Additionally, it dynamically generates an SLA deadline based on the incident's severity.

CityPulse proves that with the speed, affordability, and deep reasoning capabilities of the Amazon Nova model family, smart cities can bridge the gap between citizen frustration and municipal action.


📱 Example: Reporting a Water Leak

CityPulse features an intuitive conversational interface for citizens to report issues without filling out complex forms. Here is an example flow:

  1. Citizen: "There's a massive water leak bursting from a fire hydrant on Main Street."
  2. CityPulse AI: "Thank you for the description! 📸 Would you like to attach any photos of the issue? You can drag-and-drop images or skip this step by typing 'skip'."
  3. Citizen: (Uploads a photo of the leaking hydrant)
  4. CityPulse AI: "Great! Now I need the location. 📍 Please provide the location of the issue."
  5. Citizen: "123 Main Street"
  6. CityPulse AI: "Almost done! 📋 Would you like to leave your contact information for follow-ups? Type 'skip' to submit anonymously."
  7. Citizen: "John Doe, john@example.com"
  8. CityPulse AI: "✅ Your report has been submitted successfully! 🎫 Ticket Number: CP-2026-0012"

Behind the scenes, the 4-Agent Pipeline instantly:

  • Interprets the image and text to classify it as Water & Sewage -> Leak.
  • Verifies if someone else on Main Street already reported the hydrant (Deduplication).
  • Assigns a Priority 9/10 due to water waste and infrastructure damage.
  • Routes it to the Water & Sewage Department with a strict 24-hour SLA deadline.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors