Turn any folder of multimedia materials into an importable .apkg file through a multistage AI workflow.
Recommended (pipx):
git clone https://github.com/august-andersen/ankibot.git
cd ankibot
pipx install .Alternative (pip + venv):
git clone https://github.com/august-andersen/ankibot.git
cd ankibot
python3 -m venv .venv
source .venv/bin/activate
pip install .ankibot uses Anthropic API. Set your key one of two ways:
- Environment variable (takes priority):
export ANTHROPIC_API_KEY=sk-ant-... - First-run prompt: ankibot will ask for your key and save it to
~/.ankibot/config.json
cd ~/my-multimedia-materials
ankibotankibot scans the current directory, asks for a deck name and detail level, executes the workflow, and saves an .apkg file ready to import into Anki. Workflow:
- Classifies and extracts file contents with format-specific parsers and Claude vision for OCR/visual interpretation
- Content payloads are chunked into context-window-optimized batches
- Batches are sent to Claude via API with a system prompt and deck-specific context. Output is parsed, validated, and deduplicated.
- Cards are compiled into a styled Anki deck with formatting, tags, and a consistent note model, then packaged as an importable
.apkgfile.
| Type | Extensions |
|---|---|
.pdf |
|
| Word | .docx |
| PowerPoint | .pptx |
| Markdown | .md |
| CSV | .csv |
| Images | .jpg, .jpeg, .png |
| Level | Description |
|---|---|
| 1 | Basics only — key definitions, core facts, surface-level recall |
| 2 | Fundamental understanding — concepts, relationships, why things work |
| 3 | Highly detailed — nuance, edge cases, deep connections, advanced details |