lecture-image-generator

# Lecture Image Generator ## Overview Create lecture visuals as code, not as one-off notebook output. Keep generation scripts reproducible, lecture-local, and aligned with the existing `lectures/*/generate_*.py` patterns in this repository. Read [`references/repo-patterns.md`](references/repo-patterns.md) when you need concrete examples from existing lecture scripts. Copy or adapt [`assets/image_generation_template.py`](assets/image_generation_template.py) when starting a new generator from scratch. ## Workflow ### 1. Inspect the lecture context - Read the target `lectures/XX/lecture.md` before naming images or choosing layouts. - Inspect existing `generate_*.py` files in the same lecture before introducing a new script. - Reuse existing filenames when updating visuals so slide links stay stable. - Use a new script only when the lecture has no suitable generator or when a topic is clearly separate. ### 2. Start from the lecture template or an existing script - For a new generator, copy `assets/image_generation_template.py` into the lecture directory and rename it to a descriptive `generate_*.py` filename. - For an existing generator, preserve the current structure unless it is actively blocking the task. - Keep scripts runnable from the lecture directory with relative paths such as `images/...`. ### 3. Keep generation deterministic and lecture-local - Create the `images/` directory in code if needed. - Set a fixed random seed before any stochastic layout or sampling step. - Prefer a shared helper for repeated save-and-style logic. - Close figures after saving to avoid memory leaks in multi-image scripts. - Print progress messages when generation is non-trivial. ### 4. Organize code around output images - Use one function per distinct figure or figure family. - Give each function a clear docstring that explains the teaching purpose of the image. - Return reusable objects only when that materially simplifies related figures. - Keep `main()` as the place that orchestrates all outputs for the script. ### 5. Match repository conventions - Save into `images/<descriptive-name>.png` unless the lecture already uses another extension. - Prefer descriptive filenames that map cleanly to slide content. - Use NetworkX, matplotlib, and NumPy directly unless the lecture already depends on another library. - Keep figure sizes and DPI appropriate for slide use, not paper publication. ### 6. Verify outputs before finishing - Run the generator after editing it. - Confirm that expected files were written under `images/`. - If the task includes slide updates, verify that every referenced filename exists. - If the task is primarily about slide layout or Marp authoring, switch to or combine with `marp-lecture-author`. ## Patterns To Preserve - Prefer simple, readable code over abstract helper layers. - Keep each script understandable by an instructor editing it later. - Use inline comments sparingly and only where the visualization logic is not obvious. - Avoid notebook-only idioms or hidden state. ## Outputs - `lectures/XX/generate_*.py` - `lectures/XX/images/*.png` - Optional small adjustments to `lectures/XX/lecture.md` when image names or ordering change ## Avoid - Do not hardcode absolute paths. - Do not save figures outside the lecture directory. - Do not introduce a new generator file when a small edit to an existing one is enough. - Do not change image filenames casually when slides already reference them.