Скачать или смотреть Comparing Emacs region/speech or LLM queries to embeddings, LLM agents in work with Emacs Lisp

Video demonstrates how query, or text, question, command, or speech, could be compared to embeddings in the database to find those distinctive commands.

This is one way how LLM agents can be invoked in the background:

query is converted to embedding or multiple embeddings;
multiple commands are found
program may execute multiple commands
program could output more possible outcomes
possible outcomes could lead to more program execution

How could user prepare it?

User could say what he wants and his wishes could be stored in the database. User could say "I like sport, nature, and animals", and could describe more of his personal wishes. User could say that he follows football club Manchester United, or otherwise could speak to LLM. Queries of user could be split into various embeddings.

Use could say to computer what to do during or after certain times of the day, to maybe turn off the lights or turn on specific music.

Then on each query to LLM, the program behind, can check for specific embeddings, and invoke commands, and engage various agents in background to work for the user.

To maybe call business partner, and open up notes on screen, and inform user which is next command to be executed.

To maybe insert next command into the waiting list and inform user about it.

There are so many combinations.