One thing I find particularly interesting here: The general technique they describe for automatically generating the memory stream and derived embeddings (as we

One thing I find particularly interesting here: The general technique they describe for automatically generating the memory stream and derived embeddings (as well as higher-level inferences about that they call "reflections"), then querying against that in a way that's not dependent on the LLM's limited context window, looks like it would be pretty easily generalizable to almost anything using LLMs. Even SQLite has an extension for vector embedding search now [1], so it should be possible to implement this technique in an entirely client-side manner that doesn't actually depend on the service (or local LLM) you're using. [1]: https://observablehq.com/@asg017/introducing-sqlite-vss https://observablehq.com/@asg017/introducing-sqlite-vss