Back to Projects
Full Stack
CopyGuard
A local-first platform for detecting suspicious document pairs in Turkish student assignments. It combines MinHash, TF-IDF, sentence embeddings, and section/block overlap with optional LLM evaluation, then presents results through run-based analysis history and exportable reports.
Tech Stack
PythonFastAPINext.jsSQLiteSentence TransformersSSE
Key Features
- Hybrid analysis using MinHash, TF-IDF, sentence embeddings, and section/block overlap
- Optional full-AI mode via OpenRouter, OpenAI & Anthropic-compatible APIs
- Side-by-side comparison of hybrid and full-AI results
- Live progress tracking via Server-Sent Events
- Supports .docx, .pdf, .pptx, .md and .txt documents
- Searchable HTML reports, CSV exports and visual heatmaps
- Per-course and per-assignment file organization