CivicLens AI

Inspiration

CivicLens AI was inspired by a simple but powerful gap: cities generate massive amounts of public data, yet everyday people still struggle to answer basic questions like “Is it safe to go out right now?” or “What should I bring?”. Living in New York City, I noticed that critical information—weather alerts, 311 complaints, air quality, housing issues, disaster warnings—exists across dozens of platforms, but none of it is unified or translated into clear, actionable guidance for a layperson. CivicLens AI was created to act as a civic companion, not just a search tool.

---

What it does

CivicLens AI is an agentic AI platform that turns real-time civic, environmental, and public safety data into personalized, actionable recommendations.
Users can ask natural language questions like:

“Is it safe to attend an event in Brooklyn this afternoon? I forgot my jacket.”

The agent automatically:

• Understands context (location, time, activity)
• Fetches live data from trusted public APIs
• Evaluates safety, comfort, and risk
• Delivers clear decisions and next steps, not conditional explanations

It can also recommend nearby resources (e.g., stores), assess travel exposure, and ask clarifying questions only when necessary.

---

How we built it

We built CivicLens AI using DigitalOcean Gradient™ AI Platform with an agent-first architecture:

• Core AI Agent: A Gradient AI agent with carefully designed instructions focused on decision-making
• Function Routing: DigitalOcean Functions used to fetch live data (NOAA weather, NYC Open Data, CDC, FEMA)
• Knowledge Bases: Public health and civic datasets indexed using RAG for contextual grounding
• Real-Time APIs:
  • NYC 311, housing, and traffic data
  • NOAA weather and heat risk
  • CDC air quality and health indicators
  • FEMA disaster declarations and alerts
• Frontend Integration: Public agent endpoint + chatbot embed for a live demo experience

The system is modular, privacy-aware, and optimized for low-latency responses.

---

Challenges we ran into

• Turning data into decisions: Early versions over-explained. Refining the agent to provide direct recommendations required careful prompt and function design.
• Latency and API reliability: Some public APIs are slow or inconsistent, requiring optimization and fallback logic.
• Location handling: Balancing browser-based geolocation with privacy and graceful degradation was non-trivial.
• Scope control: Preventing the agent from drifting into speculation while still being helpful required strong guardrails in instructions.

---

Accomplishments that we're proud of

• Built a fully working agentic AI system, not just a chatbot
• Successfully integrated multiple real-time civic APIs into a single reasoning flow
• Designed an agent that asks minimal but intelligent clarifying questions
• Delivered actionable outputs instead of “if/then” explanations
• Created a public, live demo using DigitalOcean Gradient AI

---

What we learned

• Public APIs alone are not useful—reasoning and synthesis are everything
• Agent instructions matter as much as the model itself
• Function routing dramatically improves accuracy and trust
• Users value decisions first, context second
• Agentic AI is especially powerful for civic and public-good use cases

---

What's next for CivicLens AI

Next, we plan to:

• Add navigation-aware journey analysis (route safety, weather exposure)
• Integrate budget-aware recommendations (nearby stores, alternatives)
• Expand to more cities beyond NYC
• Add accessibility features (voice, multilingual support)
• Partner with civic organizations to deploy CivicLens AI at scale

CivicLens AI is a step toward making cities more understandable, safer, and more human-centered—powered by agentic AI.

Built With

• api
• cdc
• data
• digitalocean
• digitalocean-digitalocean-gradient-ai-platform-digitalocean-functions-javascript-node.js-html-css-opensearch-rest-apis-retrieval-augmented-generation-(rag)-noaa-weather-api-nyc-open-data-api-(311
• gps
• housing
• open
• openfema