cardsmith: offline flashcards from your own PDFs, checked against the source

Description and use

cardsmith automatically generates flashcards from PDFs, PPTX, and text files, verifies quotes against the source, stores them locally with SM‑2 scheduling, and can export decks to Anki.

Problem

Creating flashcards from long PDFs or slide decks manually is time‑consuming, leading users to skip active recall and rely on weaker rereading.

Outcome

On a 3,248‑word biology chapter, cardsmith produced 152 cards in ~13 minutes, with 92.5% factual accuracy after manual review.

Workflow

• Point cardsmith at a PDF, PPTX, or text file
• cardsmith splits the document into 120‑900 word chunks
• each chunk is sent to a local Ollama model (default qwen3:4b)
• model generates flashcards with a verbatim source_quote
• cardsmith checks if source_quote is an exact substring of the chunk
• flagged cards are shown in preview for editing or deletion
• cards are stored in a local SQLite database with SM‑2 scheduler
• grade cards with four buttons to record recall quality
• export deck to .apkg via genanki for Anki use

Evidence

The mechanical quote check matched 142 of 152 cards to an exact substring of the source text. I then hand rated a 40 card sample against the source and found 37 factually accurate, 1 inaccurate, and 2 unusable, for 92.5 percent accuracy.

Source