my-professor

The learn-loop's practice-and-assess step. A local cram tool: quiz yourself on a topic drawn from a document corpus, grade your answers against the cited source, and track what you have and haven't mastered — so the next session hits your weakest topics first.

It sits on two MyThingsLab core seams: mythings.corpus (shortlist-and-cite the relevant excerpts) for grounding, and mythings.mastery (an append-only local ledger of graded attempts) for the feedback that closes the loop my-glossary only opened.

Usage

# Quiz yourself on a topic from a PDF / notes corpus
myprofessor quiz "EM algorithm" --corpus ~/Desktop/unsupervised_learning.pdf --engine claude

# Answer, and record the graded attempt to the mastery ledger
myprofessor grade "EM algorithm" \
  --answer "EM alternates an E-step and an M-step to fit latent-variable models" \
  --corpus ~/Desktop/unsupervised_learning.pdf --engine claude \
  --ledger .mythings/mastery.jsonl

# What should I study now? (weakest / most overdue first; --all for full standing)
myprofessor due --ledger .mythings/mastery.jsonl

--engine noop (the default) makes zero Engine calls: quiz prints the source excerpts with no questions, grade returns a fixed partial stub. Use --engine claude for real questions and grading. --corpus is repeatable and accepts files or directories (.pdf, .md, .txt, .rst, .tex); --cache memoises PDF text extraction across runs.

How it works

• quiz shortlists the corpus for the topic, makes one Engine call to write N questions with their expected key points (cite-only — questions may rest only on the shown excerpts), and prints them with their sources.
• grade shortlists the same corpus, makes one Engine call to score the answer (verdict + 0–1 score + the gaps it missed), and appends an Attempt to the local mastery ledger. Never a PR — a graded answer is local state.
• due rolls the ledger up into a recency-decayed score per topic and orders them weakest / most overdue first, the signal the study loop re-ranks on.

Exactly one Engine call per run; retrieval is deterministic and citations are validated after the call, never inside it. No Workspace, no PR, no GitHub.

Install (development)

python -m venv .venv && source .venv/bin/activate
pip install -e ../my-things-core -e ".[dev]"
pytest

License

MIT — see LICENSE.