I updated my Kindle clipping notebook to handle the Kindle Scribe (with its irregular line lengths).

A quick experiment with Intel's OpenVINO notebook example 213 proved that it was trivial to ingest the automatically created kindle clippings as the local source text for OpenVINO's local language model. The LLM is fast but requires carefully worded questions and is not production ready. But with the right LLM, who knows?

https://github.com/frasertajima/kindle_clippings/tree/main