At CoalDev I built n8n agent workflows that stayed up. A business trigger starts the run, an LLM step classifies or drafts, embeddings go into Postgres with PGVector, and a failed run hits an error branch instead of disappearing. This is the automation I still take on as a one-week sprint.
0 likes · 4 views · Sep 24, 2026
Production n8n workflows with an LLM step, embeddings, and Postgres plus PGVector.
