Manthan

Manthan is a platform for structured civic discussions - Opinions backed by facts and data, with Public sentiment clarity at a glance , implemented using real time polling system.

Comment 1

Inspiration

I wanted to know what general public actually think , their collective sentiment and opinions , backed by data, facts, and numbers , not just scattered opinions everywhere , drowned in noise.

Every time a major event happens, I find myself jumping between different platforms , and news comments trying to gauge public sentiment , Everything disorganized and scattered . But what do I find? Engagement farming. Rage bait.( Word of the year btw ) The loudest voices get amplified , follower counts determine whose opinion "matters." There is no metric or a one stop method through which , I can get the general sentiment , at a single look , which is also backed by numbers and the opinions are based on real facts.

I asked myself: Why isn't there a platform where I can see what people actually think , at a glance , without the chaos ?

That's why I built Manthan. A polling-based social platform where:

  • Every vote counts equally , no followers, no clout, no verification badges
  • Opinions are structured and backed by context and sources
  • You see real percentages in real time
  • The submitter's name and profile is hidden , the question matters, not who asked it

What It Does

Manthan is a civic polling platform with three main sections:

Feed Sections:

  • Featured: Curated questions that remain the same for everyone , capturing a true collective snapshot , and also providing a large data pool
  • Trending: Questions ranked by votes, comments, and recency
  • Community: User-submitted questions for all kinds of discussions

Question Types:

  • Opinion: Yes / No / Unsure polls with percentage breakdowns
  • Choice: Multiple-choice questions , also with the percentage breakdowns for each option
  • Discussion: Comment-driven conversations without voting

Unique Features:

  • Opinion Shift %: Shows what percentage of users changed their vote , proof that real discourse is happening
  • 7-Day Trend Charts: Visualize how opinions shift over time

-These two features can act as time capsule for future reference , imagine looking back in 5 years to see how public opinion on AI or climate change , evolved.

-Sharing and Saving: Functionality to Save or Share any question

-Mobile Friendly UI

AI-Powered Features - All Powered By Gemini 3 :

  • Pre-Submit Review: AI checks neutrality of the question, suggests sources by searching the web in real time , detects duplicate question already posted before on the platform , fact-checks the context and also can provide enhanced context
  • Global Smart Search: Natural language queries like "most controversial questions this week" , "most polarizing topics" or any random question related to anything like "questions related to crypto" , implemented using AI semantic search using the Gemini API , using custom algorithms in the background using API function calling
  • Translation: One-click translation for regional languages
  • Comment Summarization: TL;DR of the sentiment of people by analyzing the comment section , unlocks after a certain threshold of comments

How I Built It

  1. Prototyping: Started with Google AI Studio to build the prototype mainly frontend
  2. Full Development: Built using Antigravity for backend and database - Supabase integration—database, auth, storage, and Edge Functions
  3. For Security Checks: Used AI such as Claude

Tech Stack:

  • Frontend: React 19, TypeScript, Tailwind CSS
  • Backend: Supabase (PostgreSQL, Auth, Storage, Edge Functions) , Node.js
  • AI API: Google Gemini 3 Flash
  • Build: Vite
  • Deployment: Vercel

Challenges I Ran Into

  • Backend Setup: Configuring Supabase RLS policies, database functions, and Edge Functions took some effort
  • AI Semantic Search: Getting natural language search to work reliably took multiple iterations , but after tinkering and playing with AI model and using its different API functionalities like custom function calling, made the cut , its still not perfect but works relatively very good
  • Pre-Submit Review: Handling Gemini API rate limits and structuring prompts for consistent JSON outputs needed some iterations
  • Video Editing: Editing the demo video on limited hardware with constant crashes was honestly the most frustrating part

Accomplishments I am Proud Of

  • Never gave up: Every blocker eventually had a solution
  • Gained problem-solving skills: Many issues were solved using my own logic and referring to the documentation
  • End-to-end ownership: Built a complete full-stack application from scratch for the first time -frontend, backend, database, auth, storage, AI integration , all working together

What I Learned

  • How to architect a full-stack application with tight frontend-backend-database integration
  • PostgreSQL concepts: RLS policies, database functions, triggers
  • Supabase Edge Functions for secure serverless API calls
  • API integration: Managing rate limits, error handling, caching and function calling
  • Real-world product thinking: Building features that solve actual problems

What's Next for Manthan

  • News Integration: A dedicated section where users can react to breaking news in real-time.
  • Go Global: Currently India-focused, but would love to expand worldwide
  • Advanced Analytics: Demographic breakdowns and regional sentiment maps
  • Partnerships: Collaborate with news organizations and researchers who need genuine public sentiment data
  • Scalability: Have to improve the tech stack and use more advanced concepts , techniques to make the application scalable

Built With

Share this project:

Updates

posted an update —

My AI features are getting timed out , due to high demand for the Gemini 3 Flash Model , as i can see in the logs of my edge functions , if the AI features are not working or taking too long to respond , Please Try them again later.

Log in or sign up for Devpost to join the conversation.