WEBroots

Inspiration

Environmental awareness has a grassroots problem. The information exists: air quality indexes, carbon data, emissions breakdowns; but it lives scattered across government portals, research papers, and apps built for experts, not everyday people We asked ourselves: what if understanding your local environmental impact felt as simple as a Google search? WEBroots was born from that question — a grassroots platform, built from the ground up, for the people who live closest to the problem.

What it does

WEBroots lets users:

• Ask any question regarding local sustainability
• Instantly see a real-time Waste Emissions Breakdown
• Receive three personalized, hyper-local sustainability actions they can take today, generated by ASI1-Mini
• View a dynamic infographic of their environmental snapshot, generated by Cloudinary

How we built it

We built WEBroots from scratch using a multi-agent architecture powered by Fetch.ai:

• Three specialized uAgents handling distinct parts of the pipeline — Emissions Scout, Action Router, and Media Composer
• A Python FastAPI backend orchestrating the agent chain and serving data to the frontend
• A Next.js + Cloudinary React frontend for a clean, accessible user experience
• ASI1-Mini powering hyper-local, context-aware sustainability recommendations
• OmegaClaw routing user messages from Telegram directly into the agent pipeline

Challenges we ran into

• Getting ASI1-Mini to generate location-specific actions rather than generic advice required significant prompt engineering from UCLA
• The full agent chain introduces latency before a response reaches the user, keeping the pipeline fast enough to feel responsive required parallel API fetching and fallback logic
• The correct path from agent registration to OmegaClaw skill routing required significant trial and error to figure out
• Getting three independent uAgents to reliably pass structured messages in sequence required careful schema design upfront. Address mismatches between agents caused silent failures that were difficult to debug, making the end-to-end chain one of the most technically demanding parts of the build.

Accomplishments that we're proud of

• A fully functional end-to-end demo: from a simple sustainability question → real-time emissions data → AI-generated local actions → personalized Cloudinary infographic → delivered via Telegram
• Making sustainability information feel personal and local, not abstract and global
• Building a multi-agent system where each agent is specialized

What we learned

• Multi-agent architecture is only as strong as the schemas connecting agents — getting models.py right first saved hours of debugging later
• Prompt engineering for specificity is entirely in how you structure the context you give the model
• Real-world APIs have gaps, inconsistencies, and rate limits that documentation doesn't warn you about

What's next for WEBroots

• Community layer: Let users share their actions, compare neighborhood emissions, and see collective impact growing over time — making the grassroots feel real
• Personalized tracking: A dashboard where users can track their action history, emissions trends, and CO₂ savings over weeks and months
• Expanded data sources: Pulling in water quality, noise pollution, urban heat island data, and local green spaces for a fuller picture
• Smarter localization: Improving ASI1-Mini's ability to surface genuinely local resources — community gardens, local repair cafes, neighborhood sustainability events

Built With

• agentverse
• uagents