Anki Card Generator

Generates Anki flashcard decks from PDF lecture notes using the Claude API. One command reads your PDF, calls Claude, and produces an .apkg file ready to import into Anki.

Prerequisites

• Python 3.13+
• uv — Python package manager
• Anki desktop app (2.1.50 or newer)
• An Anthropic API key (free tier works)

Setup (first time only)

1. Clone the repo

   git clone https://github.com/pclutton/anki-cards
   cd anki-cards
   

2. Install dependencies

   uv sync
   

3. Add your API key

   Copy .env.example to .env and paste in your key:

   ANTHROPIC_API_KEY=sk-ant-...
   

   Get a free key at console.anthropic.com.

The Source Material/, Topics Requested/ and Card Decks/ folders are created automatically the first time you run the script.

Daily use

Generate cards from a PDF:

uv run make_cards.py "Source Material/Week 3 - Genetics.pdf"

Or just the filename if it's already in Source Material/:

uv run make_cards.py "Week 3 - Genetics.pdf"

With a topic focus file:

uv run make_cards.py "Week 3 - Genetics.pdf" --topic "Topics Requested/krebs-cycle.md"

Topic file only (no PDF — uses Claude's general knowledge):

uv run make_cards.py --topic "Topics Requested/krebs-cycle.md"

Text only — skip images (faster, lower cost):

uv run make_cards.py "Week 3 - Genetics.pdf" --text-only

Strip copy-protection from a PDF before processing:

uv run make_cards.py --unprotect "locked-lecture.pdf"

This saves a clean copy to Source Material/.

Check what's waiting to be processed:

uv run make_cards.py --inbox

The finished .apkg file lands in Card Decks/. Double-click it — Anki imports it automatically.

Before calling the API the script prints an estimated cost, and afterwards the actual cost. A 40-page illustrated chapter runs to roughly $0.70–0.90; text-only is much less. Above $1 it asks you to confirm — pass -y to skip that prompt.

Customising card style

STYLE.md controls everything about how cards are written: how many, how they're split, which card type is used, and where diagrams go. It's an ordinary text file — open it in VS Code and edit it. Changes take effect on the next run.

It's worth reading once before your first real deck. The two things most worth tuning are the card count target (currently ~90–120 per chapter) and the worked examples at the bottom — the examples do more to shape the output than the rules above them, so adding one in your own subject is the fastest way to improve results.

Re-running a lecture

Re-running the same PDF and importing again updates existing cards rather than duplicating them — so you can tweak STYLE.md, re-run, and your review history is preserved.

The one exception: cards are matched on their question text. If Claude rewords a question between runs, that counts as a new card and the old one stays behind. If you re-run a lecture several times it's worth checking the deck for strays with Anki's Browse window.

Writing topic request files

Create a .md file in Topics Requested/. Example:

## Deck
Anatomy S2::Krebs Cycle

## Source File
Week 4 - Metabolism.pdf

## Special Emphasis
- The role of acetyl-CoA
- Products per turn of the cycle
- Regulation points

## Special Emphasis sections get more cards, exam-style phrasing, and an exam-priority tag. If ## Source File is (none), Claude generates from general knowledge.

Folder structure

anki-cards/
  make_cards.py          <- the main script
  STYLE.md               <- edit this to change card style
  Source Material/       <- drop your PDFs here
  Topics Requested/      <- drop topic .md files here
  Card Decks/            <- finished .apkg files (double-click to import)
  .env                   <- your API key (never committed)