TrashTheTest
TrashTheTest: AI Study Platform for Academic PDFs
- Status
- Active
- Date
- October 2026
- Category
- Applications
Summary
An open-source study platform that turns textbooks, lecture notes and syllabi into grounded notes, flashcards, practice questions, an AI tutor and a publishable study website.
Description
TrashTheTest treats a document as a course: upload a textbook, notes, syllabus or previous-year papers and it builds a structured academic model of pages, topics, subtopics and concepts, each with page-level provenance. Everything it generates is grounded in retrieved chunks, validated before persistence and traceable to a citation, while unsupported material becomes an explicit gap. Notes, flashcards, practice questions and a project-scoped tutor all read from that same retrieval layer, ending in a shareable site. Ingestion runs as resumable PostgreSQL-backed jobs, with 777 backend tests.
Technology
- Python 3.13
- TypeScript
- FastAPI
- Pydantic v2
- SQLAlchemy 2 async
- Alembic
- PostgreSQL 16
- pgvector (HNSW)
- tsvector full-text search
- PostgreSQL SKIP LOCKED job queue
- Redis
- Celery
- React 19
- Vite
- Tailwind CSS
- TanStack Query
- wouter
- Radix UI
- Qwen3-Embedding-0.6B
- Reciprocal rank fusion
- bge-reranker-v2-m3
- Groq
- OpenRouter
- PyMuPDF
- pdfplumber
- PP-OCRv6 tiny
- ONNX Runtime
- ReportLab
- Docker Compose
- S3-compatible storage
Features
- Structured ingestion: page-level extraction, native text first, targeted OCR and semantic chunking.
- Grounded study notes: topic-by-topic notes with citation checks and a PDF with TOC and gaps report.
- Retrieval practice: source-anchored flashcards and multiple-choice questions with citations.
- Project-scoped AI tutor: academic questions answer with sources, and sessions persist.
- Publishable study sites: notes, cards and questions rendered into a sanitised site with public slugs.
Links
Images
Machine-readable: /projects.json · /llms.txt · /llms-full.txt