[_boffin_]

Thanks for posting this. Currently building out something that is consolidating events from numerous sources and then automatically processing of those.

This post helped to give me some additional ideas on the processing of those events.

[threecheese]

Interesting, and something I’ve been thinking about. What is your architecture? Using any available workflow products for “retrieve/receive event from X, figure out what do, then do that in Y”?

[_boffin_]

Right now, since it’s currently being built out for a single user, I’m just using NestJS for the backend, TypeScript and Vite for the browser extension, Python for the daemon, Postgres for the DB and ELK for logging and embeddings storage.

I have a few sources built out right now and some fun utilities thanks to the browser extension, but in the end, I’m expecting to have 15-20 different data sources that are sent by the browser extension, collected by the daemon, and polled via crons in NestJS.

I’m taking ideas i gleaned from the AI-Town paper about taking “memories” and then generating “reflections” and then doing some other stuff on top of that and also building out a knowledge graph of everything i do, read, listen to, etc..

I’m currently figuring out the standardization of the disparate events into memories to then transform into reflections and knowledge graphs.

I’d love to hear more about what you’ve been thinking of and the issues you’ve had.

Edit: One cool thing you can do with the browser extension for this stuff is to send your cookies and headers to the database and then use in the cron jobs to get the data—think pandora.