These pages are notes on how I set this up. They are not delivery case studies.
If you want me to leave it running, that is on RAG and agents. Madrid.
- Set up an internal RAG A bounded set of docs, search with a source, and a chat or API the team can use.
- Data and access for a private GPT What goes in, who may ask, and what stays out of a public model.
- Agent, chatbot or n8n How to pick the deliverable. Logs, a perimeter and an owner for the deploy.
- Why not to paste internal PDFs into ChatGPT The habit of pasting documents into a public chat, and the system that replaces it.
- From the wiki or Drive to answers with a source Ingestion, index and citations so each answer can be checked in the original document.
- n8n on email or CRM What the flow may read or write, and where a human confirms before sending or updating.