Actionable civic insights through human–AI collaboration
Built at Polderr Hackathon · 9 November 2025
- PowerPoint slides: CivicLens Pitch Deck
- Demo video: YouTube Showcase
Municipalities and civic actors often struggle to:
- Keep a high-level overview of local issues
- Track short-term and long-term trends
- Detect misinformation early
- Extract actionable insights from fragmented online discussions
Relevant information exists but is scattered across forums, local news, and discussions, making synthesis slow and manual.
CivicLens aggregates public online sources and turns them into structured, actionable civic intelligence, combining AI analysis with human collaboration.
Key idea: Humans and AI work together on real civic problems, using shared context and verified sources.
- Local search engine for news and discussions (USP)
- AI-assisted summarization of events, topics, and discussions
- Human–AI collaboration via topic-based forums (USP)
- Trends overview (short & long term)
- Misinformation detection and flagging
- Ready-to-send reports generated from discussions
- Supports multiple input formats: forums, articles, posts
Main: Online forums, local news feeds
Considered: Social media, additional online forums
Assumption: Public websites can be legally parsed as long as no personal data (e.g., usernames) is stored.
Backend http://localhost:8000/docs
pip install -r requirements.txtbrew install ffmpeg
ollama pull qwen3-vl:4b-instruct
python -m api.mainFrontend http://localhost:3000
cd munincipalitator3000
npm run devpython main_generate.py
- locally hosted via ollama : qwen3-vl:4b-instruct for image to text for eval newspapers etc...
- azerion hosted
- elevenlabs-scribe for speech to text
- gemini-embedding-001 for embeddings
- mistral-large-2407-v1:0 for nlp decision making
Showcase via sample.png and sample.mp4
To activate, use:
Hey, PolderrAI

