Democracy starts with knowing where to go.
Poll-Rover is an AI-powered civic tech platform designed for the 2026 Indian Elections. It empowers every voter by providing a high-trust, accessible, and intuitive interface to discover their assigned polling stations, evaluate accessibility features, and navigate directly to the ballot box.
The Poll-Rover ecosystem is monitored via a suite of Key Performance Indicators (KPIs) to ensure the highest reliability for voters.
| Agent | Reliability (Success Rate) | Efficiency (Latency) | Quality Score |
|---|---|---|---|
| 🛰️ Harvester |  |
 |
 |
| 🛡️ Quality |  |
 |
 |
| 🏗️ Orchestrator |  |
 |
 |
| 🧑🤝🧑 Assist |  |
 |
 |
| 🩺 SRE Ops |  |
 |
 |
Note
Metrics are aggregated every Sunday at 23:59 UTC and published to the Weekly Performance Report.
Poll-Rover operates on a sophisticated 5-agent autonomous pipeline to ensure data reliability and voter support:
- 🛰️ Harvester Agent: Daily autonomous extraction of polling station data from ECI and State CEO portals (TN, KL, PY, AS, WB).
- 🛡️ Quality Agent: Runs post-harvest verification, cross-referencing coordinates with OpenStreetMap for 100% locational accuracy.
- 🏥 SRE Ops Agent: Automated health audits of the entire data mesh, ensuring zero stale entries and platform stability.
- 🧑🤝🧑 Citizen Assist AI: A multilingual (English, Tamil, Malayalam) support agent powered by localized LLMs (Ollama/Gemini) to answer voter queries instantly.
- 🏗️ Orchestrator: Manages the
YAML → Python → Zolabuild pipeline, converting raw civic data into a high-performance static visualizer.
Integrated header search bar for instant discovery. Type a Station ID, Name, or Landmark to fly directly to the booth location on the interactive map.
Found your station? Click "Get Directions" in the popup to launch turn-by-turn navigation via Google Maps deep-links.
Every station is evaluated and color-coded based on its accessibility rating:
- Green (High): Fully accessible with ramps, audio booths, and braille.
- Yellow (Moderate): Partial accessibility features present.
- Red (Limited): Basic facilities only.
A premium, institutional design system optimized for high-contrast legibility. Features a sleek Dark/Light theme toggle with persistence and cache-busting for a seamless experience.
Poll-Rover now uses a next-gen discovery engine that reduces the initial map payload from 27MB to 3KB (99.9% reduction). Data is partitioned into district-level chunks and fetched on-demand based on proximity and user interaction.
- Discover: Open the Live Map and explore your region.
- Filter: Use the Accessibility Filters (♿, 🔊, ⠿) to find stations meeting your specific needs.
- Search: Use the global search bar to jump to your booth.
- Navigate: Click a marker and select "Get Directions" to find your route.
- Assist: Click the Citizen Assist icon to ask the AI about voting dates, procedures, or station details.
- Core: HTML5, Vanilla CSS (Royal Civic Indigo), JavaScript
- Static Site Engine: Zola (High-performance SSG)
- Map Engine: Leaflet.js with Custom Data Layers
- Automation: GitHub Actions (Daily Deployment Pipeline)
- Data Engineering: Python 3.11 with YAML/GeoJSON processing
- AI Models: Llama 3.2 (via Ollama) & Gemini 2.0 Flash
Poll-Rover is configured for Daily Automated Deployments. The deploy.yml workflow:
- Harvests regional data.
- Audits data health.
- Builds the static site.
- Pushes live to GitHub Pages.
Democracy is a journey—Poll-Rover is your map. 🏛️🗳️✨