Here's a neat trick for libraries that are under-documented but have comprehensive tests: use the CLI combo of llm and files-to-prompt to generate your own private documentation directly from their test suite! https://til.simonwillison.net/llms/docs-from-testsFull transcript here: https://gist.github.com/simonw/351cffbd254af5cbf329377fb95fcc13

OpenAI reduced a slightly confusing feature today called "Predicted Outputs", which lets you spend more money on a prompt to have it return faster if the output is mostly predictable based on the input (e.g. a prompt that makes edits to some code) - detailed notes here: https://simonwillison.net/2024/Nov/4/predicted-outputs/

@xsc for personal stuff absolutely - and yes for most of the things I'm building on the APIs

@daniel yeah it's really good - Claude 3.5 Sonnet has been my daily driver since they release itBest model I've used for code too

Added a section helping explain what's going on with our dog Cleo: https://gist.github.com/simonw/d680d5f27b50de4efc26ed2a673292fd

@matt I tried it on two documents so far and it looked reasonable, but I've not done a remotely robust comparison of it yet

Anthropic added PDF support to their Claude API this morning, I have a new release of the llm-claude-3 plugin that supports that as a new attachment type: llm install llm-claude-3 --upgrade llm -m claude-3.5-sonnet 'extract text' -a mydoc.pdfDetails: https://simonwillison.net/2024/Nov/1/claude-api-pdf-support-beta/

@tuananh doesn't help much if the user can do anything that you don't want the LLM to also be able to do

Hard-hitting coverage of the Halloween Wars format changes here from the Miami Student newspaper https://www.miamistudent.net/article/2022/10/halloween-wars-review

Got a bit behind on my weeknotes, so here are my monthnotes for October instead https://simonwillison.net/2024/Oct/30/monthnotes/

@xsc yes, if you have the llm-gemini plugin installed and configured with an API keyYou could try using this script here (or using Google's AI Studio tool) ti check it's not an LLM bug: https://til.simonwillison.net/llms/prompt-gemini

Bonus tool: if you save the raw API JSON as a Gist you can add the Gist ID to this URL to serve up a page that lets other people play back your audioHere's my example with a system prompt specifying a "thick French accent":https://tools.simonwillison.net/gpt-4o-audio-player?gist=4a982d3fe7ba8cb4c01e89c69a4a5335

@ame let me know if it works, I'm not sure how best to test it!

@akaihola @mesirii using https://pypi.org/project/genson/ to summarize the JSON is an interesting ideaCurrently I avoid reading the whole stream into memory at once which means I can't easily do a two step process where it first reads the whole thing and then replays it later, but I'm sure that could be fixed

Here's a short audio clip https://static.simonwillison.net/static/2024/russian-pelican-in-spanish.m4a

Paul Calcraft extended this idea into an implementation of Pictionary where different vision LLMs generate SVGs and race to guess what the others are drawing and it is absolutely brilliant https://twitter.com/paul_cal/status/1850262678712856764

@freddy @jaffathecake I think I can even get away with not serving the images from a separate domain if I instead inline them as base64 SVG in the img sec attribute(Running off a separate domain is OK for me but makes things harder for my users if I release open source code for other people to self-host)

@andre a kind of dirty trick would be to timestamp the data when the submit fires and then, on a subsequent page load, if Django is displaying a green "your data has been saved" message assume that any timestamped persisted submit data within last 5 seconds can be safely discarded

@johnstonphilip @mahryekuh I guess you could encrypt it with a public key in the JavaScript and have a private key server-side that is only available when the user is signed into their account

@sdruskat looks like the circle size is based on the number of dependencies