How We Built FeedSynth: Turning Community Feedback into AI-Powered Insights

shah prince — Sun, 07 Jun 2026 13:31:13 +0000

Building FeedSynth: An AI-Powered Community Feedback Memory System

Introduction

During hackathons, community events, and product launches, organizers receive a huge amount of feedback. Some people report issues, some share suggestions, while others talk about their overall experience. The problem is that most feedback gets buried in spreadsheets, forms, or chat messages, making it difficult to identify long-term patterns and recurring problems.

This challenge inspired us to build FeedSynth, an AI-powered Community Feedback Synthesizer that does more than just collect feedback. It remembers past feedback, tracks how sentiment changes over time, identifies recurring issues, and helps organizers understand what is happening in their community.

The main idea behind FeedSynth is simple: feedback should not be treated as isolated comments. Instead, it should become part of a memory system that continuously learns and improves.

The Problem

Most feedback tools focus on collecting data. They can show charts and ratings, but they often fail to answer important questions such as:

What issue keeps appearing every day?
Is community sentiment improving or getting worse?
Which event caused negative reactions?
What are users complaining about repeatedly?

For example, if ten different users report WiFi issues using different words such as "internet slow," "network lagging," or "WiFi disconnecting," most systems store them as separate comments.

As organizers, we wanted a smarter solution that could connect these pieces together and generate meaningful insights.

Our Solution

FeedSynth acts like an AI-powered memory system for community feedback.

Whenever new feedback is submitted, the system analyzes it and updates a structured memory state. Instead of only storing comments, it remembers trends, recurring complaints, important events, and sentiment changes.

This allows the platform to answer questions like:

What is the biggest recurring issue?
How did sentiment change after lunch delays?
What are the most requested improvements?

In short, FeedSynth transforms raw feedback into actionable knowledge.

How FeedSynth Works

The workflow is straightforward:

Users submit feedback through the platform.
The system analyzes ratings and comments.
Similar issues are grouped together.
Historical memory is updated.
Sentiment trends are calculated.
Insights are generated automatically.
Users can query the memory through an AI chat interface.

One of our favorite features is the Chronic Issues Registry. If multiple users repeatedly mention WiFi problems, the system recognizes that these reports are related and groups them under a single issue instead of treating them as separate complaints.

Technologies Used

We wanted the project to be lightweight, fast, and easy to deploy.

Frontend

React 18
Tailwind CSS
Babel Standalone
Google Fonts
SVG Icons

AI Integration

Google Gemini 1.5 Flash API
Custom rule-based AI fallback system

Storage

LocalStorage for persistent browser-based storage
JSON-based AI Memory Schema

Hosting

Lightweight Python server
Node.js static server

Architecture

The architecture follows a simple but effective design.

The frontend is built using React and Tailwind CSS. Users interact with dashboards, forms, timelines, and AI chat components.

Feedback data is stored locally using LocalStorage. This allows the application to remember data even after refreshing the browser.

The AI layer processes new feedback and updates the memory state. When a Gemini API key is available, Google Gemini performs the analysis. If no API key is provided, a local rule-based engine generates insights and tracks recurring issues.

This approach makes the application both powerful and easy to demonstrate during a hackathon.

Challenges We Faced

One of the biggest challenges was designing the AI Memory system.

Most feedback tools simply store comments in a database. We wanted our system to remember historical context and connect related feedback across multiple days.

Another challenge was grouping similar complaints together. Different users often describe the same issue in different ways, so we needed a strategy to identify recurring themes accurately.

We also spent time designing a dashboard that could present insights clearly without overwhelming users with too much information.

What Makes FeedSynth Different?

Many feedback platforms focus on collecting data.

FeedSynth focuses on understanding data.

Instead of showing hundreds of comments, it answers questions such as:

What is our biggest problem?
Which issue appears most often?
How has community sentiment evolved?
Which event had the largest impact?

The AI Memory feature is what makes FeedSynth unique. It continuously learns from previous feedback and builds a long-term understanding of the community.

Future Scope

There are many ways to expand FeedSynth in the future.

Some ideas include:

Integration with Discord, Reddit, and social media platforms
Cloud database support using Firebase
Real-time collaboration for organizers
Predictive sentiment analysis
Automated weekly reports
Community health scoring
Advanced analytics and trend forecasting

These additions could transform FeedSynth from a hackathon project into a production-ready platform.

Conclusion

Building FeedSynth was an exciting experience because it allowed us to explore the idea of AI Memory in a practical way.

Rather than simply collecting feedback, our platform remembers it, learns from it, and turns it into useful insights. By combining React, Tailwind CSS, LocalStorage, and Gemini AI, we created a system that helps organizers better understand their communities and make more informed decisions.

FeedSynth demonstrates that feedback becomes far more valuable when an AI can remember, connect, and learn from it over time.

DEV Community: shah prince