is anyone out there building / have built ai text agents? if so I'd love to connect and share ideas / best practices. here's what I've done so far: n8n workflow that's triggered by an inbound text message webhook (via twilio API). that then triggers a node to grab the last 50 text messages to AND from that number combine those into one, chronologically ordered variable pass that to an LLM with a sales prompt to generate a response send that response to lead I have had issues with the LLM coming up with a "good" response tho. I was thinking it might be better to create a "sales training" type KB, and have that supplement the prompt, or possibly even break the LLM into two nodes (although less cost effective) to first determine strategically what kind of response it should have, based on the sales training KB, and then pass that to another LLM to actually generate the response. I'm thinking that would increase the quality of the responses, but again might be cost prohibitive since there would be two LLM calls per inbound text anywho, would love to hear the tech stack others are using / have used successfully, and how they overcome the QC issues when relying on their LLM to generate a response.