StreamSarthi

Inspiration

Urban streams are often polluted and ignored. We wanted to make it easier for people to monitor them and also make their observations more reliable. Our idea was to make people return to check streams while rewarding them for accuracy.

What it does

StreamSaathi Guardians is a mobile friendly app where users check stream health using photos and simple questions. AI gives a second opinion and other users can verify the observations. Difficult cases go to experts. Users earn points for accurate work and can track the health of streams over time.

How we built it

We built the frontend using Next.js TypeScript and Tailwind. The backend uses Python FastAPI and Supabase PostgreSQL for data. Gemini Vision is used for image analysis. We also built Python services for scoring verification levels and badges.

Challenges we ran into

Our main challenge was making the game engaging without encouraging users to submit low quality data. We also had to make crowd voting fair and prevent people from simply trusting AI. Privacy and deployment issues were also important challenges.

Accomplishments that we're proud of

We built a working system that connects gamification with data quality. We are especially proud of the blind verification game Stream Stations and the system that sends only difficult cases to experts.

What we learned

We learned that rewards work better when they are connected to accuracy instead of quantity. We also learned that clear feedback and showing AI confidence can make people trust and use AI more responsibly.

What's next for StreamSaathi

Our next goal is to pilot StreamSaathi with college eco clubs and local communities and test it on real streams.

We want to add offline support so users can collect data even when there is poor internet connectivity.

We plan to add Hindi and other regional language support with voice guidance so more people can use the platform easily.

We want to create more expert verified gold photos to improve the verification system and make user skill scores more accurate.

We also plan to add team based challenges where college clubs and local communities can work together to monitor streams.

In the future users could create local stream groups and organise cleanup drives or monitoring campaigns directly through the platform.

We want to improve the stream health map and show long term changes in water quality and environmental conditions.

We also plan to add better AI support for identifying litter vegetation invasive species and other visible stream problems.

Another goal is to add notifications for important changes in a user's adopted stream.

We want to connect verified data with OneAquaHealth so the collected information can be useful for larger environmental research.

In the long term we hope StreamSaathi can grow from a college project into a community based platform where citizens scientists students and environmental experts work together to protect local streams.

Built With

• fastapi
• gemini
• github-actions
• nextjs
• postgresql
• pydantic
• python
• react
• render
• supabase
• tailwindcss
• typescript
• vercel