llm

A collection of fun things to do with llms. Mostly on the command line. Mostly using https://github.com/simonw/llm.

Describe a web page:

$ lynx -dump https://archive.is/EcQfE | llm -m 4o -s "What's the key point? ELI5. Foramt as a note, be lanconic."
wget https://www.quantamagazine.org/how-ai-revolutionized-protein-science-but-didnt-end-it-20240626/ -O - | html2text --ignore-links --ignore-images --mark-code | llm -m 4o -s "Summarize this article. For each part return two or three sentences, except part three where you can return up to six sentences.  Focus particularly on the impaact to scientist's experimental work" | mdformat --wrap 110 - 

Write a git commit message:

$ git commit -aeF <(git diff  | llm -m 4o -s "Write a git-style commit message for this change. Don't yammer. You can be profane, but don't be offensive")
Some real output from mr_roboto:
Add collision detection and event-driven stopping

- Integrated distance sensor check to halt operations if obstacle is detected within defined minimum distance.
- Refactored movement functions (`move_forward`, `move_backward`, `turn_left`, `turn_right`) to include event-driven stopping mechanism.
- Fixed headlight pin assignment in `pins.py` from 9 to 18 to match new hardware configuration.

Prompting doesn't have to be complicated!

notes/attachments/simple_prompt_screenshot.png