@Dmitry Shapiro From your session I understood that You are using both: Ingesting data into Gbrain (the issue probably could malicious gmails and prompt injection), and structuring the data with LLM aproach (by the way OKF could be even better). At the moment my setup is this: - Hermes on a VPS - Honcho local memory - QMD for .MD files - OKF/Obisidan mix of structuring the incoming data But how to inject data to Gbrain and take it to the Wiki - I have no idea, should i be next to the wiki etc. It's hard for me to visualize. Any advice on that ? Also: I'm using fireflies for meeting transcripts, sometimes to record them even offline. The issue is that sometimes on one meeting somone is rejoing the meeting with a phone numer, somone else is with few people on one mic etc. I made a simple gate. So when a new FF transcript is pulled, Hermes ask me with some quotes if they are correct (as sometimes ff fudges this up, specially when there are few people with the same first name in the org) and I give him info: Hey this and this person was on 1 mic etc. After the confirmation it makes 2 summaries one for the peopple from the meeting + additional personal notes for me. This a proach helps a little specially for the action plan and tasks to correct persons. But I can't imaging how would Gbrain do this on a autopilot.

@Dmitry Shapiro how do you scrape the data from linkedin & twitter and not get banned ? :)