Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/calvin-xu/auto-incremental-reading
Easy incremental reading with auditable LLM-generated Anki decks
https://github.com/calvin-xu/auto-incremental-reading
anki education gpt incremental-reading llm spaced-repetition
Last synced: 2 days ago
JSON representation
Easy incremental reading with auditable LLM-generated Anki decks
- Host: GitHub
- URL: https://github.com/calvin-xu/auto-incremental-reading
- Owner: Calvin-Xu
- Created: 2025-01-17T12:17:32.000Z (4 days ago)
- Default Branch: main
- Last Pushed: 2025-01-18T06:19:15.000Z (3 days ago)
- Last Synced: 2025-01-18T07:19:50.320Z (3 days ago)
- Topics: anki, education, gpt, incremental-reading, llm, spaced-repetition
- Language: Python
- Homepage:
- Size: 925 KB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Auto-Incremental-Reading
Motivation:
- There are books about things I have a working knowledge of (["Do not learn if you do not understand"](https://www.supermemo.com/en/blog/twenty-rules-of-formulating-knowledge)!) and want to read, but are hundreds of pages and I do not have enough activation energy.
- Incremental reading & making flashcards take a lot of time
Solution:
- We can generate cards with LLMs cheaply. Just suspend or delete ones that don't work or you already know.
- Let's read the book incrementally while studying cards by putting book sections on cards.
## What this tool does:
Helps you generate cards like this for an entire book with no initial effort.
|  |  |
| ------------------------------------------------ | ---------------------------------------------- |
(credit: [Fluent Python, Second Edition](https://www.fluentpython.com/))
- Takes a pdf file w/ a valid Table of Contents. Finds all the leaf sections and their page ranges
- Generates a directory tree matching ToS structure; saves cropped bitmap images to directories of leaf sections
- For each leaf section, uses the OpenAI multimodal API to generate flashcards; saves to `flashcards.json` per-section
- Tangle CSV that can be imported to Anki
- Code fields can be rendered to HTML using Pygments to support syntax highlighting
Each note is also associated with
- `source`: ToS path to section
- `source_imgs`: image tags pointing to each page in the section
Images of pages are renamed & copied to `img_assets`; copy them to your Anki media folder to view them while studying / reviewing.
## Example Usage
```python
from flashcard_creator import (
create_flashcards_for_book,
create_anki_deck,
AnkiDeckConfig,
FlashcardPromptConfig,
)
from pdf_toc_reader import process_pdf_and_create_structure
book_title = "Fluent Python: Clear, Concise, and Effective Programming"
book_path = "fluent_python.pdf"
custom_prompt = FlashcardPromptConfig(
cards_per_page=3,
model="gpt-4o",
temperature=0.2,
template="""Your custom template here, or use default""",
)
# Create directory tree matching ToS
directory = process_pdf_and_create_structure(
book_path,
skip_keywords={
"copyright",
"about the author",
"further reading",
},
)
# Generate flashcards for each leaf ToS section
create_flashcards_for_book(
book_title=book_title,
mapping_file=f"{directory}/directory_mapping.json",
log_file=f"{directory}/flashcard_creation.log",
prompt_config=custom_prompt,
)
# Create the Anki deck
result = create_anki_deck(
directory=directory,
config=AnkiDeckConfig(
fields=["Front", "FrontCode", "Back", "BackCode", "Comments"],
code_fields={"FrontCode", "BackCode"},
render_markdown=True,
syntax_highlight_html=True,
separator=",",
),
)
print(f"\nCreated Anki csv at: {result.csv_path}")
print(f"Media assets directory: {result.img_assets_dir}")
```
### Notes
- Mapping and logging files are created in the ToS tree directory.
- Currently uses no batching or concurrent requests. It is not that slow for a whole book and I get rate-limited.
- Works for books with no selectable text. But needs a ToC where leaf sections have sane lengths.