Play 75 — Exam Generation Engine 📝
AI exam creation — question generation from learning materials, Bloom’s taxonomy distribution, MCQ distractor generation, IRT calibration.
Build an intelligent exam generation system. Extract content from learning materials (PDF/DOCX), generate questions across Bloom’s taxonomy levels, create plausible distractors grounded in common misconceptions, and calibrate difficulty via Item Response Theory.
Quick Start
cd solution-plays/75-exam-generation-engine
az deployment group create -g $RG -f infra/main.bicep -p infra/parameters.json
code .
# Use @builder to implement, @reviewer to audit, @tuner to optimizeArchitecture
Pre-Tuned Defaults
- Bloom’s: Remember 15% · Understand 25% · Apply 30% · Analyze 20% · Evaluate 7% · Create 3%
- Question Types: MCQ 50% · Short Answer 25% · Essay 15% · True/False 10%
- Difficulty: Easy 30% · Medium 45% · Hard 25% · Target mean 70%
- Distractors: 3 per MCQ · misconception-grounded · ±20% length tolerance
DevKit (AI-Assisted Development)
| Primitive | What It Does |
|---|---|
agent.md | Root orchestrator with builder→reviewer→tuner handoffs |
copilot-instructions.md | Exam domain (Bloom’s, distractors, IRT, question type constraints) |
| 3 agents | Builder (gpt-4o), Reviewer (gpt-4o-mini), Tuner (gpt-4o-mini) |
| 3 skills | Deploy (195+ lines), Evaluate (130+ lines), Tune (225+ lines) |
| 4 prompts | /deploy, /test, /review, /evaluate with agent routing |
Cost Estimate
| Service | Dev | Prod | Enterprise |
|---|---|---|---|
| Azure OpenAI | $25 | $200 | $600 |
| Cosmos DB | $3 | $50 | $180 |
| Blob Storage | $2 | $15 | $40 |
| Azure Functions | $0 | $30 | $150 |
| Azure AI Content Safety | $0 | $20 | $60 |
| Azure AI Search | $0 | $75 | $250 |
| Key Vault | $1 | $3 | $10 |
| Application Insights | $0 | $15 | $50 |
| Total | $31 | $408 | $1,340 |
vs. Play 74 (AI Tutoring Agent)
| Aspect | Play 74 | Play 75 |
|---|---|---|
| Focus | Real-time Socratic tutoring | Exam/assessment generation |
| Interaction | Multi-turn conversation | Batch generation + export |
| Output | Personalized dialogue | Exam PDF with answer key + rubrics |
| Calibration | Adaptive difficulty per student | IRT calibration across population |
📖 Full documentation · 🌐 frootai.dev/solution-plays/75-exam-generation-engine · 📦 FAI Protocol
FAI Manifest
| Field | Value |
|---|---|
| Play | 75-exam-generation-engine |
| Version | 1.0.0 |
| Knowledge | F1-GenAI-Foundations, R1-Prompt-Engineering, T2-Responsible-AI |
| WAF Pillars | responsible-ai, reliability, performance-efficiency |
Last updated on