Skip to content

Colbstang/studyos

Repository files navigation

Study OS

Your AnKing deck, but it tells you what to study — and tutors you on it. All inside Anki.

Study OS takes an AnKing/Step deck and turns it into something that knows what you're shaky on, maps how every topic connects, and answers questions about the card in front of you — without ever leaving Anki.

🛠️ Heads up — read this first

Fair warning: I vibe-coded this. It was built fast, mostly by throwing a lot of AI at it, for one med student's own Step studying. It's early and rough — some things won't work perfectly, and plenty of it is tuned to one person's workflow.

Your Anki data is safe, though. Study OS only ever reads your collection, from a copy — it never edits, deletes, or reschedules real cards. Worst case it shows something wrong or just breaks; it can't lose your deck. The one exception: it can unsuspend cards on request, and that's one Ctrl+Z away.

The concept graph around RBC — connected topics, color-coded by mastery Every topic in your deck becomes a node, wired to the ones it connects to. Click any node to open it.

Today — the Director's daily plan It tells you what to study today — weakest stuff first, paced so you don't drown.

Progress — retention by system Honest mastery, by system, on live FSRS — where you actually stand, not "100% (suspended)."

Overview map — the whole deck as one graph Your entire deck as one connected map.

What it does

  • Sees the whole deck as a map. It reads the AnKing tags and builds a graph of every topic ("objects") and how they connect — so RBCs visibly sit next to hemolysis sit next to complement, and you can walk between them.
  • Tells you what to study. Every topic is scored by how shaky it is (lapses, FSRS retention), and the result is a daily plan — weakest stuff first, paced so you don't drown. It greets you with today's plan when Anki opens.
  • Tutors you on the card in front of you. Ctrl+J on any card opens a tutor that answers using your own cards, running on your own Claude / ChatGPT / Gemini login — so there's no extra subscription to buy and nothing to set up if you already use one.
  • Puts resources one click away. AMBOSS, Boards & Beyond, Sketchy, and more show up right on the card and open in your logged-in account. It also figures out which AMBOSS questions cover the current topic, straight from your own tags.
  • Honest mastery, not fake 100%. Each topic shows real recall probability and the actual card images, instead of "100% (suspended)."

No copying — just links into what you already pay for

Study OS never copies AMBOSS / First Aid / UWorld / B&B content. It links into the subscriptions you already pay for (they open in your own browser, logged in as you) and uses your own card tags. None of their text, images, or questions get stored or shown anywhere — a guard in the code blocks it, and nothing copyrighted lives in this repo.

The AI tutor works the same way: it runs on a coding-agent CLI you already have, with the API key stripped out so it physically can't run up a bill. No CLI installed? Everything else still works — just without the tutor.

Bring your own

Study OS ships none of the copyrighted material it works with. You bring:

  • A deck — an AnkiHub/AnKing Step deck (or anything tagged the AnKing way: #AK_Step…, ^Systems, vendor tags).
  • PDFs (optional) — your own First Aid / Pathoma PDFs, pointed at with the STUDY_TEXTBOOK_DIR env var. Nothing's bundled; the page-preview just lights up if the files are there.
  • Subscriptions — AMBOSS, UWorld, B&B. Study OS links you in; you stay logged in as yourself.
  • Optionally, a claude / codex / gemini CLI — logged in, to power the tutor. Totally optional.

Setup

Setup is the rough part right now. A coding agent (Claude Code, Cursor, etc.) pointed at this repo and AGENTS.md can handle most of it. Otherwise:

Requirements:

  • Anki desktop + the AnkiConnect add-on, installed and running.
  • A Python 3.11+ with a few packages (Anki's built-in Python won't have them, so use a system/conda one):
    python3 -m pip install fastapi "uvicorn[standard]" httpx networkx
    # or single-sourced: python3 -m pip install -e ./backend
    # optional: pymupdf  (only for textbook page previews)
  • (Optional) a claude / codex / gemini CLI for the tutor.

Steps:

  1. Drop the add-on into Anki. Symlink (or copy) the anki_addon/ folder into Anki's add-ons folder as studyos, then restart Anki:

    # macOS:
    ln -s "$HOME/Desktop/studyos/anki_addon" \
      "$HOME/Library/Application Support/Anki2/addons21/studyos"

    (Linux: ~/.local/share/Anki2/addons21/ · Windows: %APPDATA%\Anki2\addons21\ — copy the folder there as studyos.)

  2. Point it at the right Python. Open Tools → Add-ons → Study OS → Config and set python_path to the interpreter the packages were installed into. A wrong path shows a warning; the log is at $TMPDIR/studyos-server.log.

  3. Open Anki and let the wizard run. A first-run wizard finds the deck and builds the map (~1–2 min). After that:

    • The Study OS button in Anki's top toolbar opens the sidebar.
    • Ctrl+J in the reviewer asks the tutor about the current card.
    • Tools → Study OS has the rest (windows, full-screen, today's briefing, server on/off).

Status

A personal project, tuned to one workflow (Step 1, AnKing v12, a specific set of vendors) and shared as-is — some defaults are still personal, and it's actively being worked on. For what's built versus what's coming (new-material onboarding, in-app review, mobile, and more), see docs/ROADMAP.md.

License

MIT © 2026 Colby Grames.

About

Your AnKing deck, but it tells you what to study and tutors you — inside Anki.

Topics

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors