This was my first crack at building a self-hosted AI/ML system. I bit off more than I could chew with RAG pipelines and cognition cycles. Burned it down to come back with something cleaner.
Ollama (Qwen), PostgreSQL, FastAPI, n8n, Caddy proxy.
The live system is still up at cognitive.wadiem.cloudns.be but I'm rebuilding it with a simpler, more focused approach.