Urban Assist

Inspiration

Our cities are alive, breathing ecosystems of data, energy, and human ambition. Yet, we interact with them through sterile apps and disconnected datasets. We get traffic alerts and pollution warnings, but we never truly understand the city's holistic pulse.

We were inspired by the Solarpunk vision of a harmonious future and the neurological concept of the subconscious mind. What if we could tap into our city's collective subconscious? What if we could understand its hidden stresses, its latent desires, its emergent patterns? We envisioned Urban Assist not as a simple assistant, but as a digital interpreter for the city's dreams, translating its complex, chaotic data streams into a language humans can understand, feel, and act upon. Our goal was to build a bridge between the citizen and the soul of their city, creating a truly symbiotic relationship.

What it does

Urban Assist is a hyperlocal, generative AI that transforms the city into a living entity you can communicate with.

• The 'City Dream' Interface: At its core, Urban Assist presents a daily "City Dream," a generative art piece (video and audio) created using Vision & Media models like SDXL and RunwayML. This isn't random art; it's a multi-modal translation of the city's data from the last 24 hours. A dream filled with fluid, green patterns and harmonious sounds indicates efficient transit and low pollution. A chaotic, glitchy, red-hued dream might signify traffic gridlock and high social stress detected from public data.

• Dream Interpretation with Agentic AI: You can ask the app, "What did our city dream about last night?" An AI agent, built using CrewAI, analyzes the "dream" and its underlying data to give you a narrative. For example: "The city had a restless dream. The downtown core felt 'constricted' (heavy traffic from 7-9 PM), while the northern parks felt 'calm and breathing' (high air quality and foot traffic)."

• Symbiotic Quests: Based on the dream's interpretation, Urban Assist generates "Symbiotic Quests"—hyperlocal, gamified tasks to help "heal" the city's stress points.

  • If the city dreams of "thirst" (low water levels in reservoirs, high usage), it might generate a quest to "Find and report 3 leaking public taps on your way to work."
  • If it dreams of "loneliness" (low footfall in a specific public square), it could create a quest for people to "Visit the 'forgotten square' and share a photo," partnering with a local café for a discount.

• Augmented Reality Overlays (The 'Wow' Factor): Using the IoT + AI track, you can point your phone at a building or street and see its contribution to the city's current state as an ethereal aura. A building with high energy consumption might have a faint, pulsating red glow, while a community garden contributes a vibrant, green shimmer.

How we built it

• The 'Unconscious Mind' - Backend: We used a multi-agent system (AgentOps Track) with CrewAI and LangChain. A 'Data-Gatherer' agent ingested real-time public data streams (simulated for the hackathon). A 'Dream-Weaver' agent then compressed this multi-dimensional data into a series of complex prompts for our generative models.
• The 'Dream' - Generative AI: We fed these prompts into Stable Diffusion XL and RunwayML Gen-2 (Vision & Media Track) to generate the daily video dreamscape. The audio landscape was created using generative models that translate data patterns into sound.
• The 'Interpreter' - Language Models: The conversational interface and dream interpretations were powered by the OpenAI API. We fine-tuned a model to speak in a more narrative, metaphorical style befitting an oracle.
• The 'Body' - Frontend: Our UI was mocked up in Figma with a focus on a clean, mystical aesthetic (Civic UX Track). The mobile app was built using React Native, allowing us to deploy the AR overlays.
• The 'Nervous System' - IoT & Data: We used open datasets like Air Quality Data and simulated a real-time IoT network for metrics like public transit capacity and energy grid load to make our model dynamic and responsive.

Challenges we ran into

1. Translating Data to Art: The biggest challenge was creating a meaningful and consistent mapping between quantitative data (e.g., 200 tons of CO2) and qualitative artistic expression (the color red, a glitchy effect). It required immense prompt engineering and creating a custom "aesthetic lexicon" for the AI.
2. Avoiding Dystopia: We had to be careful that our "dreams" and "auras" didn't feel like a dystopian surveillance tool. We focused the language and UI on collective well-being and empowerment, not on highlighting individual negative behavior.
3. Agent Coordination: Getting our CrewAI agents to work in concert—passing complex data structures from the data-gatherer to the dream-weaver without losing context—was a significant technical hurdle.

Accomplishments that we're proud of

We are incredibly proud of creating an AI that feels less like a machine and more like a mythical creature. The first time we saw a "City Dream" that perfectly captured a real-world event (a local festival creating a vibrant, pulsing dream), we knew we had created something special. We successfully built a multi-agent, multi-modal system that goes beyond simple utility to evoke emotion and inspire collective action.

What we learned

We learned that the future of AI in cities isn't just about optimization; it's about connection. Generative AI has a profound, untapped potential to act as an artistic and emotional bridge between humans and the complex systems we inhabit. By personifying the city, we can foster a sense of empathy and stewardship that raw data dashboards never could.

What's next for Urban Assist

• Emotional Nuance: We want to integrate more nuanced data, like anonymized public sentiment from social media, to allow the city to dream of "joy," "anticipation," or "frustration," making the experience even more profound.
• Citizen as Dreamer: Our ultimate vision is to allow citizens to contribute to the dream. Users could submit photos, sounds, or feelings about their neighborhood, which the AI would weave into the next day's collective dream, creating a closed-loop system of communication between the city and its inhabitants.
• Real-World Integration: Partner with our local municipality to integrate with real-time IoT sensors, turning Urban Assist from a proof-of-concept into a living, breathing digital soul for our city.

Built With

• crewai
• fastapi
• langchain
• python