AI Issue Assistant — Turning Past Issues into Useful Insights
This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend.
What I Built
I built AI Issue Assistant, an AI-powered issue tracking tool designed to help teams manage and understand recurring issues more effectively.
The idea came from a simple problem: when a similar issue happens again, we often have to manually search through old tickets to understand what happened before.
AI Issue Assistant helps by:
- Creating and managing issues through a simple interface
- Generating vector embeddings for issues
- Finding similar historical issues using semantic similarity
- Making it easier to identify recurring problems
- Providing a foundation for future AI-powered features such as issue summaries and Root Cause Analysis (RCA)
Instead of treating an issue tracker as just a list of tickets, I wanted to make it a tool that can learn from the team's historical issues and make that information easier to use.
Demo
Live Demo: https://ai-issue-assistant.vercel.app/
Backend API:
Code
GitHub Repository:
https://github.com/adarsh1114/ai-issue-assistant
https://github.com/adarsh1114/ai-issue-assistant-backend
How I Built It
The project is built with a modern full-stack JavaScript/TypeScript stack.
Frontend
- React
- TypeScript
- Tailwind CSS
Backend
- Node.js
- Express.js
- MongoDB
- Mongoose
AI
The core AI functionality uses Gemini Embeddings to convert issue descriptions into numerical vector representations.
I used the gemini-embedding-001 embedding model to generate embeddings, which are then stored with the issue data. These vectors can be used to find issues that are semantically similar even when they don't use exactly the same words.
For example:
"User cannot complete EPF authentication"
and
"EPF login is failing for a customer"
may describe a similar problem even though the wording is different.
This is where semantic search becomes more useful than a simple keyword search.
Why Does Open Innovation Matter?
Open innovation makes it possible to experiment, learn and build useful AI features without needing a large infrastructure or an expensive proprietary AI platform.
For this project, I could combine open-source technologies such as React, Node.js, Express and MongoDB with accessible AI tooling and build the complete application myself.
The most important part for me was not simply adding an AI API to an application. It was understanding how AI features such as embeddings and semantic similarity can solve an actual software-development problem.
Open innovation also makes it easier for developers to learn by building, modify existing tools, and create solutions tailored to their own problems.
My Agent Session
I used AI coding assistance during development to help with implementation, debugging and exploring the architecture of the project.
Prize Categories
- AI / Open Innovation
- Developer Tools
What I Learned
Building this project helped me understand that adding AI to an application is not just about generating text.
I learned how to work with:
- Embeddings
- Vector representations
- Semantic similarity
- AI-powered search
- Gemini APIs
- MongoDB data modelling
- Full-stack integration of AI features
The project started with a simple issue tracker idea and evolved into an experiment in making historical issue data more useful through AI.
Top comments (0)