Skip to content

README_HF.md

Latest commit

 

History

History
50 lines (37 loc) · 1.84 KB

README_HF.md

File metadata and controls

50 lines (37 loc) · 1.84 KB
Raw
title Smart Study Agent
emoji 🎓
colorFrom blue
colorTo purple
sdk streamlit
sdk_version 1.56.0
app_file app.py
pinned true
license mit
short_description Adaptive AI study agent with OPEAA loop, RL & FSRS

🎓 Smart Study Agent

An adaptive AI study agent — a reinforcement-learning policy decides what you study next, an FSRS memory model decides when you review, and an LLM generates the quizzes in between.

Features

  • 5-phase OPEAA loop — Observe → Plan → Act → Evaluate → Adapt
  • RL decision layer — Q-learning + LinUCB bandit pick the action, not the LLM
  • POMDP belief state persisted across sessions
  • FSRS spaced repetition — per-topic recall probability, stability & due dates (same algorithm family as modern Anki)
  • 🃏 Anki export — download your generated question bank as a styled .apkg deck
  • Interactive concept graph — draggable prerequisite DAG, mastery color-coded
  • Multi-format input — PDF, TXT, MD, DOCX, PPTX, VTT, SRT
  • Multi-student support with peer comparison
  • MCP server & Claude Agent Skill in the GitHub repo — drive the agent from Claude

Usage

  1. Create a student in the sidebar
  2. Go to 📖 Study Session
  3. Click "Use Sample ML Lecture" or upload your own PDF
  4. Run all 5 OPEAA phases in order

Tech Stack

Streamlit · Claude API (claude-opus-4-6) · NetworkX + pyvis · Q-learning · FSRS (py-fsrs) · genanki

Author

Built by Haofei Sun · GitHub · HF Space

Note: This Space uses HF Inference Providers (Kimi-K2) for free LLM access. Local users with an ANTHROPIC_API_KEY can switch to Claude claude-opus-4-6 automatically.

License

MIT