Welcome to the PeP5 Wiki

Relevant source files

The following files were used as context for generating this wiki page:

• Models/ProgrammingExamModels.cs
• DEPLOY-GCLOUD.md
• Services/OpenAIService.cs
• PeP.ExamApp/Installer/README.md
• wwwroot/js/pep-ide.js
• Services/CodeExecutionService.cs
• PeP.ExamApp/Pages/AntiCheatPage.xaml.cs
• Controllers/ExamAppController.cs
• Data/ApplicationDbContext.cs
• Models/ExamModels.cs
• PeP.ExamApp/Pages/ExamCodePage.xaml.cs
• PeP.ExamApp/Pages/TeacherAuthPage.xaml.cs
• deploy-gcloud.ps1
• PeP.ExamApp/Installer/build-installer.ps1

PeP5 (Programming Examination Platform) is a comprehensive system for creating, administering, and evaluating programming exams. It consists of a web-based platform for instructors to design exams and a secure, locked-down desktop application for students to take them. The system supports both traditional multiple-choice questions and complex, project-based programming tasks.

Key features include a secure desktop client with anti-cheat measures, automated code execution against test cases, and AI-powered services for generating questions and evaluating student code submissions. The platform is designed for cloud deployment, specifically on Google Cloud Platform, and utilizes a range of modern technologies including .NET, Blazor, WPF, and external APIs for code execution and AI.

Architecture Overview

The PeP5 system is composed of several key components that interact to provide a complete examination experience. The main components are the PeP Web App, the PeP Exam App (desktop client), a SQL Server database, and external services for specialized tasks.

This diagram illustrates the high-level architecture of the PeP5 platform.

Database Schema

The core of PeP5 is its relational database, which stores all information about courses, exams, users, and attempts. The system uses Entity Framework Core for data access. There are two main types of exams: standard multiple-choice exams and programming exams, each with its own set of related tables.

Entity Relationship Diagram

The following diagram shows the key entities and their relationships for the programming exam module.

Key Data Models

Standard Exam Models

These models support traditional multiple-choice exams.

Class	Description	File
Exam	Represents a standard exam with properties like duration, scoring type, and a collection of questions.	Models/ExamModels.cs:33
Question	A single question within an exam, which can be multiple-choice or multiple-answer.	Models/ExamModels.cs:111
QuestionChoice	A possible answer for a Question.	Models/ExamModels.cs:141
ExamAttempt	A student's attempt at a standard exam, tracking score and status.	Models/ExamModels.cs:163
StudentAnswer	A student's submitted answer for a specific question in an attempt.	Models/ExamModels.cs:197

Programming Exam Models

These models are specifically designed for project-based programming exams.

Class	Description	File
ProgrammingExam	Defines a programming exam, including the language, project name, and starter files.	Models/ProgrammingExamModels.cs:8
ProgrammingTask	A specific task or problem within a programming exam, with instructions and points.	Models/ProgrammingExamModels.cs:90
TestCase	A test case for a ProgrammingTask used for automated evaluation, with input, expected output, and points.	Models/ProgrammingExamModels.cs:119
ProjectFile	A starter file (source code, data file, etc.) provided to the student at the beginning of the exam.	Models/ProgrammingExamModels.cs:152
ProgrammingExamAttempt	A student's attempt at a programming exam, tracking status, score, and containing all student-modified files.	Models/ProgrammingExamModels.cs:189
StudentProjectFile	A student's version of a project file, storing their code and modifications.	Models/ProgrammingExamModels.cs:258
TaskProgress	Tracks a student's progress on a single ProgrammingTask, including AI feedback and scores.	Models/ProgrammingExamModels.cs:282
TestCaseResult	The result of running a single TestCase against the student's code.	Models/ProgrammingExamModels.cs:348
CodeSubmission	A snapshot of the student's code, created on auto-save, manual save, or run.	Models/ProgrammingExamModels.cs:377

Sources: Models/ProgrammingExamModels.cs, Models/ExamModels.cs

AI Services Integration

PeP5 leverages the OpenAI API for two primary purposes: generating exam content and evaluating student code. The OpenAIService class encapsulates this functionality.

Sources: Services/OpenAIService.cs

AI-Powered Exam Generation

The system can automatically generate both multiple-choice questions and complete programming exam projects based on a given topic, difficulty, and language.

• Question Generation: The GenerateQuestionsAsync method creates a set of multiple-choice questions by providing a structured prompt to the AI. It requests a strict JSON output and performs multiple attempts and parsing to ensure valid data is returned.
• Programming Exam Generation: The GenerateProgrammingTasksAsync method generates a full-fledged programming exam, including a project name, description, starter files (with TODO comments), and multiple tasks with instructions and test cases. It is capable of creating data files (e.g., CSV, JSON) for tasks involving file I/O.

Sources: Services/OpenAIService.cs:91-247, Services/OpenAIService.cs:257-410

AI Code Evaluation

A core feature is the automated evaluation of student code submissions using an AI model. The EvaluateCodeAsync method sends the student's code, the task instructions, and a detailed system prompt to the AI for assessment.

This sequence diagram shows the code evaluation flow.

The evaluation returns a detailed breakdown of the student's performance based on several criteria:

Metric	Description
Score	The final score awarded for the task, out of maxPoints.
CodeQualityScore	A score from 0-100 for structure, readability, and naming.
CorrectnessScore	A score from 0-100 assessing if the logic implements the task requirements.
EfficiencyScore	A score from 0-100 for the algorithmic efficiency of the solution.
CompletionPercentage	A score from 0-100 indicating how much of the task was addressed.
Feedback	Detailed constructive feedback on the overall submission.
CodeSnippets	Specific snippets from the student's code with targeted comments.

Sources: Services/OpenAIService.cs:17-36, Services/OpenAIService.cs:644-825

Code Execution Service

To run student code against test cases, PeP5 uses the CodeExecutionService, which acts as a client for the external Piston API. This service is responsible for packaging the student's files, sending them for execution, and interpreting the results.

The following diagram shows the process of executing a single test case.

A key implementation detail is the InjectDataFiles method. Since the Piston API does not support a virtual file system, this method reads data files (like .csv or .txt) and embeds their content as strings within a helper source file (e.g., DataFiles.java or data_helper.py). This allows the student's code to "read" files in the sandboxed execution environment.

Sources: Services/CodeExecutionService.cs:194-281

Secure Exam App (PeP.ExamApp)

The PeP.ExamApp is a WPF desktop application that provides a secure environment for students to take exams.

Authentication and Authorization Flow

Before starting an exam, a student must go through a multi-step authentication and authorization process, which involves the teacher's approval.

This flowchart outlines the student's journey from login to starting the exam.

Anti-Cheat System

The application performs a series of security checks before allowing an exam to begin. These checks are implemented in the AntiCheatPage.

The following checks are performed:

• Administrator Privileges: Verifies the application is running with admin rights.
• Monitor Count: Ensures only one display is connected.
• Remote Session: Detects if the session is a remote desktop connection.
• Virtual Machine: Checks if the OS is running inside a VM.
• Debugger Attached: Verifies no debugger is attached to the process.
• Blacklisted Processes: Scans for and flags processes associated with screen recording, remote access, or cheating tools.
• Screen Recording: Checks for active screen recording sessions.

If any blocking issues are found, the "Launch Exam" button is disabled. For blacklisted processes, the student is given an option to automatically close them.

Sources: PeP.ExamApp/Pages/AntiCheatPage.xaml.cs:69-148

IDE and Editor

For programming exams, the exam runner page hosts a WebView2 control that loads the web-based IDE. The front-end interaction with the Monaco Editor is managed by pep-ide.js. This JavaScript interop layer provides functions to create, update, and interact with the editor from the Blazor application.

Key functions include:

• createEditor: Initializes a new Monaco editor instance.
• getContent/setContent: Gets or sets the editor's content.
• OnEditorContentChanged: A callback that invokes a .NET method when the content changes.
• OnEditorSaveRequested: A callback that invokes a .NET method when Ctrl+S is pressed.
• setMarkers: Displays error or warning markers in the editor, typically for compilation errors.

Sources: wwwroot/js/pep-ide.js:2-80

Installer and Auto-Update

The PeP.ExamApp installer is built using Inno Setup. The build process is automated with a PowerShell script (build-installer.ps1).

The application supports auto-updates by checking an API endpoint on startup.

• Endpoint: GET /api/exam-app/update?currentVersion=X.X.X
• Configuration: The update information is managed in the main web application's configuration.

{
  "ExamAppUpdate": {
    "Version": "1.1.0",
    "DownloadUrl": "https://yourserver.com/downloads/PeP.ExamApp_Setup_1.1.0.exe",
    "ReleaseNotes": "- Bug fixes\n- Performance improvements",
    "ReleaseDate": "2025-12-22",
    "IsMandatory": false,
    "MinimumVersion": "1.0.0",
    "Checksum": "sha256hashofinstallerfile"
  }
}

Sources: PeP.ExamApp/Installer/README.md:55-77, Controllers/ExamAppController.cs:29-50, PeP.ExamApp/Installer/build-installer.ps1

Cloud Deployment

The platform is designed for deployment on Google Cloud Platform. The DEPLOY-GCLOUD.md guide and deploy-gcloud.ps1 script outline the infrastructure setup.

Google Cloud Services

Estimated Monthly Costs

The following table provides a cost estimate based on the recommended specifications.

Service	Specification	Est. Cost
Cloud SQL for SQL Server	db-custom-1-3840 (1 vCPU, 3.75GB RAM)	~$50-80
Cloud Run	1 vCPU, 512MB RAM, auto-scaling	~$5-20
Cloud Storage	For ExamApp installer hosting	~$1
Total		~$56-100/month

Note: Costs are estimates and vary based on actual usage.

Sources: DEPLOY-GCLOUD.md:10-18