Inspiration

Local governance is currently paralyzed by a dual bottleneck: inaccessible citizen reporting and fragmented civic planning[cite: 5]. We noticed that everyday citizens struggle to report infrastructure issues like potholes, broken streetlights, or sanitation failures because the bureaucratic channels are confusing and multi-layered. Meanwhile, municipal planners receive scattered, unstructured complaints through phone calls and disparate emails, making it impossible to conduct data-driven urban planning. CivicPulse was inspired by the need to bridge this gap, transforming civic frustration into structured, actionable data for community innovation.

What it does

CivicPulse is an intelligent civic triage platform[cite: 5]. For the citizen, it acts as a frictionless reporting tool where they can submit a simple text description and location of a community issue. Behind the scenes, the platform's AI engine instantly categorizes the issue, determines a preliminary severity level, and routes it to the specific municipal department's dashboard. It effectively translates unstructured human complaints into prioritized, geospatial data, allowing local authorities to allocate resources efficiently and citizens to track their report's resolution in real-time.

How we built it

Note: In accordance with the Graphiques Innovation Challenge guidelines, this submission represents a social-impact proposal supported by a pre-existing technical prototype.[cite: 5]

Our foundational prototype was built using Python to orchestrate a multi-agent AI architecture. We integrated the Gemini API to act as the cognitive engine, reading the unstructured text from simulated citizen reports and classifying them into distinct civic categories (e.g., Public Works, Sanitation, Utilities). We utilized scalable frameworks like FastAPI to build lightweight routing pipelines that simulate how these categorized tickets would be pushed to a localized government dashboard.

Challenges we ran into

One of the core challenges was designing an AI prompt structure that could accurately categorize vague or colloquial issue descriptions without misrouting them[cite: 5]. Additionally, conceptualizing a data model that could scale from a single local municipality up to a national infrastructure health index required significant architectural planning to ensure data privacy and integration compatibility with legacy government systems.

Accomplishments that we're proud of

We successfully developed a working AI triage backend that proves civic reporting does not require lengthy, complex forms[cite: 5]. By demonstrating that large language models can accurately parse and route everyday complaints, we've validated the core engine of our proposal: making civic engagement radically accessible for non-digital-native citizens.

What we learned

Building the prototype underscored just how fragmented local government data truly is[cite: 5]. We learned that the primary barrier to smart city innovation isn't a lack of government desire, but the overwhelming volume of unstructured data they are forced to process manually. We also discovered the immense potential of LLMs in serving as universal translators between casual citizen input and rigid bureaucratic requirements.

What's next for CivicPulse

Moving forward from this proposal, the next phase is building out the production-ready frontend interface optimized for mobile accessibility[cite: 5]. We plan to develop the macro-analytics tier—the country-level data model that aggregates local district data to identify systemic infrastructure failures. Ultimately, we aim to pilot CivicPulse with a forward-thinking local municipality, establishing API pipelines that feed our structured data directly into their existing work-order software.