An AI agent that automatically detects and fixes quality issues in assessment questions using Claude AI, with a Gradio UI for human review.
- 9-Module Pipeline — Processes reported questions through Detect → Fix → Report phases
- AI-Powered Analysis — Uses Claude AI for question analysis, automated fixes, and quality control
- Smart Fix Routing — Routes issues by problem type (spelling, grammar, comprehension, relevance, etc.)
- Severity Scoring — Weighted 4-dimension scoring to prioritize critical issues
- Analytics Dashboard — Plotly charts and executive summaries for actionable insights
- Human Review Queue — Priority-ranked review queue with accept/reject/edit workflow
- Backend: Python, SQLAlchemy, MySQL
- AI: Claude AI (Anthropic)
- UI: Gradio
- Analytics: Plotly
# Install dependencies
pip install -r requirements.txt
# Configure environment
cp .env.example .env
# Edit .env: set ANTHROPIC_API_KEY, MYSQL_HOST/USER/PASSWORD/DATABASE
# Create DB schema
python setup_db.py
# Launch UI (http://localhost:7860)
python app.py| Phase | Modules | Description |
|---|---|---|
| Detect | InputValidator → NLPScanner → Classifier → PatternDetector | Load Excel, analyze questions via Claude, classify issues, detect patterns |
| Fix | SeverityScorer → FixEngine | Compute severity scores, generate targeted fixes per problem type |
| Report | FinalQC → ReviewQueueBuilder → ReportGenerator + AnalyticsAgent | QC checks, build review queue, generate reports and charts |