Inspiration
We saw local concerns in social media threads vanish quickly, making it hard for residents and leaders to spot ongoing issues. Civic Insight was born to capture and amplify these real-time community voices without requiring new platforms.
What It Does
Civic Insight pulls recent social media posts for a city, uses AI to summarize and categorize them into areas like Housing, Events, and Safety. Users get a clean dashboard to browse summaries alongside original post snippets for context, helping residents stay informed and engaged.
How We Built It
Using FastAPI, we fetch posts from Reddit via PRAW and classify them with Anthropic’s Claude, returning structured JSON through an API. A Redis cache speeds up repeat queries. The frontend is a responsive React app built with Vite and Tailwind, all containerized with Docker for easy deployment.
Challenges We Ran Into
Filtering out irrelevant or offensive posts without losing signal, managing API rate limits, maintaining consistent category labels across complex posts, and delivering a polished end-to-end product within 24 hours were our main challenges.
Accomplishments We’re Proud Of
We built a fully functional pipeline with fast responses, a clean, mobile-friendly UI, and modular code that can easily incorporate new data sources. All core features worked reliably by demo time despite the tight timeline.
What We Learned
Effective prompt design dramatically improves AI accuracy, showing raw post snippets builds user trust, and caching is vital for speed. Clear task division and agreeing early on priorities helped us move fast.
What’s Next for Civic Insight
We plan to add more data sources like Nextdoor and 311 feeds, implement trend analytics, enable user feedback for better curation, offer dashboards for city officials, and strengthen moderation and privacy protections. What started as a hack can grow into a powerful tool for community engagement.
Log in or sign up for Devpost to join the conversation.