DatattentionDefaults that leave a model API open
The bind address and the missing key are the two settings that make a model port public.
Most local model servers default to a loopback address and no authentication, because they were written for a laptop. The open case is the laptop default plus one changed line: the listen address becomes all interfaces, and the key check stays absent.
For an Ollama-style server that line is OLLAMA_HOST=0.0.0.0:11434. For a vLLM or llama.cpp style server it is --host 0.0.0.0 --port 8000 with no --api-key. A LiteLLM proxy does the same on port 4000 when its config lists no master key. Jupyter does it on 8888 when the token is empty.
The practical test is from a second network. GET /api/tags or GET /v1/models either answers or it does not. A timeout means the port is closed. A JSON body means the default stuck.
On this host
On datattention.com those defaults are the published setup. Port 11434 answers tags, and port 8000 answers models. Neither asks for a key.
curl http://datattention.com:11434/api/tags
curl http://datattention.com:8000/v1/modelsThe full list of ports and paths is on the endpoints page.