AI Rubber Duck For Gitlab's Issues - Smart Debugging Assistant

🏆 Google AI in Action Hackathon 2025 - GitLab Challenge

An AI-powered debugging assistant that helps developers solve problems through interactive questioning directly within GitLab issues. Built with Google Cloud AI and GitLab webhooks to accelerate software development.

Table of Contents

1. About
2. Getting Started
3. Setup
4. How It Works in the Backend
5. Examples
6. Future Work

---

About

What is AI Rubber Duck?

AI Rubber Duck is an intelligent debugging companion that integrates directly into your GitLab workflow. When developers get stuck, they create an issue with "Rubber Duck Help Me" in the title, and our AI assistant provides contextual guidance through Socratic questioning.

Key Features

• 🤖 AI-Powered Assistance: Uses Google Gemini AI for intelligent responses
• 🔗 GitLab Integration: Seamless webhook integration with GitLab issues
• 📚 Context-Aware: Understands your entire project structure
• 💬 Interactive Conversations: Supports multi-turn debugging sessions
• ⚡ Real-Time: Instant responses to issue creation and comments

Technologies Used

• Google Cloud: Firestore (database), Gemini AI (language model)
• GitLab: Webhooks API, Issues API, Repository API
• Backend: Python Flask, python-gitlab library
• Platform: Web-based service

Demo & Links

• 🎬 Demo Video:

  

  Click the image above to watch the demo video

• 🌐 Website: [Ai Rubber Duck For Gitlab's Issues]

• 🎮 Live Demo: [Demo Gitlab Repo With Issues and Webhook Configured]

---

Getting Started

Configure GitLab Webhook

To enable the AI Rubber Duck functionality in your GitLab project:

1. Access Webhook Settings

   • Go to your GitLab project → Settings → Webhooks

2. Configure Webhook Details

   • URL: https://rubber-duck-ai-957711343120.us-central1.run.app/webhook (or ngrok URL for testing)
   • Secret Token: Use same value as GITLAB_WEBHOOK_SECRET should be nkcuxx7uvUsywxT
   • Trigger Events: Enable the following:
     • ✅ Issues events
     • ✅ Comments (Note events)
     • ✅ Merge request events

3. Save It

   • Click Add webhook to save

Using the AI Rubber Duck

Once configured, simply create an issue with "Rubber Duck Help Me" in the title or description, and the AI will analyze your project context and provide helpful assistance!

---

Setup

Prerequisites

• Python 3.8+
• GitLab account with project
• Google Cloud account
• Public URL (ngrok for testing)

Installation and Configuration

1. Clone Repository

   git clone <repository-url>
   cd version_2

2. Install Dependencies

   pip install -r requirements.txt

3. Google Cloud Setup

   • Enable Firestore API and Google AI API
   • Create service account with Firestore Admin role
   • Download key as account-key.json

4. Environment Configuration

   # Create .env file
   FLASK_APP=app/app.py
   APP_GITLAB_URL=https://gitlab.com
   APP_TARGET_GITLAB_TOKEN=your_gitlab_token
   APP_GOOGLE_AI_API_KEY=your_google_ai_key
   GITLAB_WEBHOOK_SECRET=your_webhook_secret
   GOOGLE_CLOUD_PROJECT=your_project_id
   GOOGLE_SERVICE_ACCOUNT_PATH=account-key.json

5. Run Application

   flask run -p 5000

---

How It Works in the Backend

Architecture Overview

┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│   GitLab        │    │   Flask Web      │    │  Google Cloud   │
│   Platform      │    │   Application    │    │   Services      │
│                 │    │                  │    │                 │
│ ┌─────────────┐ │    │ ┌──────────────┐ │    │ ┌─────────────┐ │
│ │   Issues    │◄├────┤►│   Webhook    │ │    │ │  Firestore  │ │
│ │   Comments  │ │    │ │   Handler    │◄├────┤►│  Database   │ │
│ │   Merges    │ │    │ │              │ │    │ │             │ │
│ └─────────────┘ │    │ └──────────────┘ │    │ └─────────────┘ │
│                 │    │         │        │    │                 │
│ ┌─────────────┐ │    │ ┌──────────────┐ │    │ ┌─────────────┐ │
│ │ Repository  │◄├────┤►│   Business   │ │    │ │   Gemini    │ │
│ │   Content   │ │    │ │    Logic     │◄├────┤►│     AI      │ │
│ │     API     │ │    │ │   Handler    │ │    │ │   Models    │ │
│ └─────────────┘ │    │ └──────────────┘ │    │ └─────────────┘ │
└─────────────────┘    └──────────────────┘    └─────────────────┘

Data Flow:
1. Developer creates issue with "Rubber Duck Help Me"
2. GitLab sends webhook to Flask app
3. Handler fetches project context from Firestore
4. AI processes issue + context → generates response
5. Response posted back to GitLab issue

Core Components

1. Webhook Handler (app/app.py)

   • Receives GitLab webhooks
   • Validates secret tokens
   • Routes events to appropriate handlers

2. Business Logic (app/handler.py)

   • Processes issue events
   • Manages project registration
   • Orchestrates AI responses

3. GitLab Integration (src/gitlab_integration.py)

   • Fetches issue details and comments
   • Posts AI responses back to issues
   • Retrieves repository content

4. AI Integration (src/google_ai_integration.py)

   • Configures Google Gemini AI
   • Generates contextual responses
   • Supports Socratic questioning mode

5. Data Storage (src/firestore_integration.py)

   • Stores project metadata
   • Caches repository content
   • Tracks conversation history

Data Flow

1. Project Registration: First webhook triggers repository content fetch and Firestore storage
2. Issue Processing: AI receives issue + full project context to generate relevant questions
3. Response Generation: Gemini AI creates Socratic questions based on problem and codebase
4. Content Updates: Merge events automatically refresh repository context

AI Context Generation

The AI receives comprehensive context including:

• Project metadata and structure
• README and configuration files
• Relevant source code files
• Issue history and previous responses

---

Examples

Example 1: JavaScript Error

Issue Title: Rubber Duck Help Me - ReferenceError in Node.js server

User Problem:

Getting error: ReferenceError: ten is not defined
Line: ten(port, () => {

AI Response:

Looking at your error "ReferenceError: ten is not defined", I can see this is on the line where 
you're trying to start a server.

Can you tell me what you intended to write instead of "ten"? 
What function do you typically use to start a server listening on a port?

Example 2: Authentication Issue

Issue Title: Rubber Duck Help Me - User login keeps failing

User Problem:

My authentication endpoint returns 401 even with correct credentials

AI Response (with project context):

I can see you're using bcrypt for password hashing in your auth.js file. 

Let's think through this step by step:
1. Are you comparing the plain text password with the hashed password correctly?
2. Can you show me how you're calling bcrypt.compare()?
3. What does your user lookup query return - does it find the user?

Example 3: Database Connection

Issue Title: Rubber Duck Help Me - Database connection timeout

AI Response (after analyzing config files):

I notice you're using MongoDB connection pooling in your config. 

When you say "timeout", let's narrow this down:
1. Is this happening on the initial connection or during queries?
2. Have you checked if your connection string includes the correct timeout parameters?
3. Looking at your .env.example, are all required environment variables set?

Conversation Flow

Developer creates issue → AI asks clarifying questions → Developer responds → 
AI provides deeper analysis → Solution discovered → AI congratulates and summarizes

---

Future Work

Immediate Enhancements

• Multi-Language Support: Expand beyond Python/JavaScript to Java, C++, Go
• IDE Integration: VS Code extension for direct AI assistance
• Enhanced Context: Include test files and documentation in AI analysis
• Performance Optimization: Faster response times and better caching

Advanced Features

• Custom AI Training: Project-specific models trained on team's coding patterns
• Team Analytics: Insights into common debugging patterns and knowledge gaps
• Smart Suggestions: Proactive issue detection based on code changes
• Integration Expansion: GitHub, Bitbucket, Azure DevOps support

Enterprise Features

• Multi-Tenant Support: Organization-level management and analytics
• Advanced Security: SSO integration and audit logging
• API Rate Limiting: Usage controls and quota management
• Custom Deployment: On-premises and private cloud options

Research Directions

• Code Understanding: Better static analysis integration
• Learning from Interactions: AI improvement from successful debugging sessions
• Predictive Assistance: Anticipate issues before they occur
• Knowledge Graph: Build connections between common problems and solutions

---

Submission Links

📂 Source Code: [GitHub Repository URL]
📋 Devpost: [Hackathon Submission URL]

---

Built for Google AI in Action Hackathon 2025 - Accelerating Software Development with AI