Embeddings just took 42 minutes to run?

[simonw]

https://github.com/simonw/til/actions/runs/15311283313

I'm going to disable that.

[simonw]

I'm tempted to switch to running https://huggingface.co/TaylorAI/gte-tiny instead, it's only about 80MB, should run well (and cache the model too) in GitHub Actions.

I could use llm embed-multi for that.