Codefessor

Code Buddy - AI Detection Interview System

A web application that conducts AI-powered interviews to assess whether submitted code was written by a human or AI. Uses the Ribbon API to create personalized code review interviews.

Features

• Code Submission: Students can paste their code and provide basic information
• Automated Interview Generation: Creates personalized interview questions based on the submitted code
• AI-Powered Interviews: Uses Ribbon’s voice agent to conduct the interview
• AI Detection Analysis: Analyzes interview responses to determine likelihood of AI-generated code
• Real-time Status Updates: Tracks interview progress and provides real-time feedback
• Detailed Results: Provides scoring, confidence levels, and reasoning for the assessment

Setup Instructions

1. Prerequisites

• Node.js (version 14 or higher)
• A Ribbon API account and API key

2. Get Ribbon API Key

1. Sign up at Ribbon AI
2. Navigate to your API settings
3. Generate an API key
4. Copy the API key for use in step 4

3. Install Dependencies

cd "/Users/ciqbian/Desktop/VS/HT6 25/Code Buddy 2"
npm install

4. Configure Environment

1. Open the .env file
2. Replace your_ribbon_api_key_here with your actual Ribbon API key:

RIBBON_API_KEY=your_actual_api_key_here
PORT=3000

5. Start the Application

For development (with auto-restart):

npm run dev

For production:

npm start

6. Access the Application

Open your browser and navigate to:

http://localhost:3000

How It Works

1. Code Submission

• Students enter their name, email, programming language, and paste their code
• The system generates appropriate interview questions based on the code and language

2. Interview Creation

• Creates a Ribbon interview flow with customized questions
• Generates a unique interview link for the student
• Questions focus on code understanding, implementation details, and problem-solving process

3. Interview Process

• Students click the provided link to start the AI-powered voice interview
• The Ribbon AI agent asks questions about the code
• Students explain their code, discuss their approach, and demonstrate understanding

4. Analysis & Scoring

• Once completed, the system analyzes the interview transcript
• Uses multiple factors to determine AI likelihood:
  • Positive indicators (human-written): Personal authorship claims, mentions of debugging, technical terminology usage
  • Negative indicators (AI-generated): Mentions of AI tools, uncertainty about code details, very brief responses
• Provides a score from 0-100 and confidence level

5. Results

• Displays detailed analysis including:
  • AI detection score and confidence level
  • Assessment of whether code is likely human or AI-generated
  • Reasoning behind the score
  • Full interview transcript

API Endpoints

POST /api/submit-code

Submit code for interview creation.

Request Body:

{
  "code": "string",
  "language": "string",
  "studentName": "string",
  "studentEmail": "string"
}

Response:

{
  "success": true,
  "sessionId": "string",
  "interviewLink": "string",
  "interviewId": "string"
}

GET /api/interview-status/:interviewId

Check interview completion status and get results.

Response:

{
  "status": "completed",
  "analysis": {
    "score": 75,
    "confidence": "high",
    "aiLikelihood": "likely human-written",
    "reasoning": "Claims personal authorship (+15); Uses technical terminology (4 matches) (+10)"
  },
  "transcript": "string",
  "studentInfo": {
    "name": "string",
    "language": "string"
  }
}

Scoring Algorithm

The AI detection algorithm considers multiple factors:

Positive Indicators (Human-written)

• Personal Authorship (+15): Uses phrases like “I wrote”, “I coded”, “I implemented”
• Challenges Mentioned (+10): Discusses struggles, difficulties, or challenges
• Debugging Process (+12): Mentions debugging, fixing bugs, or errors
• Technical Terminology (+10): Uses relevant programming concepts and terminology

Negative Indicators (AI-generated)

• AI Tool Mentions (-25): References ChatGPT, AI, or generation tools
• Copy/Paste References (-15): Mentions copying or pasting code
• Uncertainty (-12): Shows lack of understanding with “don’t know” or “can’t explain”
• Brief Responses (-10): Very short or generic responses

Score Interpretation

• 70-100: Likely human-written (high confidence if 80+)
• 50-69: Possibly human-written (medium confidence)
• 30-49: Possibly AI-generated (medium confidence)
• 0-29: Likely AI-generated (high confidence if <20)

Supported Programming Languages

• JavaScript
• Python
• Java
• C++
• C
• C#
• Go
• Rust
• PHP
• Ruby
• Swift
• Kotlin
• Other (custom questions)

Customization

Adding New Languages

Edit the generateCodeQuestions function in server.js to add language-specific questions.

Modifying Scoring Algorithm

Update the analyzeForAIDetection function in server.js to adjust scoring factors and weights.

Interview Questions

Customize the base questions in the generateCodeQuestions function to focus on different aspects of code understanding.

Troubleshooting

Common Issues

1. “Missing required fields” error: Ensure all form fields are filled out
2. “Failed to create interview” error: Check your Ribbon API key is correct
3. Interview not starting: Verify the Ribbon API key has proper permissions
4. Results not loading: Check browser console for errors and verify API connectivity

API Rate Limits

Be aware of Ribbon API rate limits. The application includes basic error handling, but you may need to implement additional retry logic for high-volume usage.

Security Considerations

• Store API keys securely (never commit to version control)
• Implement input validation and sanitization
• Consider adding authentication for production use
• Monitor API usage and costs

License

This project is for educational purposes. Please ensure compliance with Ribbon AI’s terms of service.