DatattentionCommon ports for self-hosted model servers
11434, 8000, 8080, 4000, 8888, and the other listeners on this address.
People reuse a short list of ports because the READMEs do. 11434 is Ollama. 8000 and 8080 are homemade OpenAI-compatible servers, including vLLM and llama.cpp. 4000 is a LiteLLM proxy. 8888 is Jupyter. 8501 is TensorFlow Serving REST. 8642 and 9119 show up as an agent API and its dashboard. 18789 is an OpenClaw gateway.
The port is not the protocol. Confirm with a GET that the project documents: /api/tags, /v1/models, /api/status, or /v1/models/demo. A connection that opens and returns HTML or JSON is the service. A connection that times out is not.
On one machine those ports can all be open together. They do not share a key or a process. Call the one whose request shape you already speak.
On this host
Every port in that list except the TensorFlow gRPC port is open on datattention.com (194.238.25.127), with no API key. Port 8500 is not open. The catalog is the endpoints page. Two quick checks:
curl http://datattention.com:11434/api/tags
curl http://datattention.com:8080/v1/models
curl http://datattention.com:8888/api/statusThe full list of ports and paths is on the endpoints page.