Troubleshooting

What to do when a client can't reach a model you share (Share a Model).

An OpenAI SDK gets 401 through --auth

OpenAI SDKs send their own Authorization: Bearer, so through a tunnel with --auth pass the password as a Basic header instead.

from base64 import b64encode
from openai import OpenAI

client = OpenAI(
    base_url="https://qwen.<user>.<domain>/v1",
    api_key="unused",
    default_headers={"Authorization": "Basic " + b64encode(b"friend:secret").decode()},
)

/api/... returns 404

Only /v1's inference paths are served, on purpose: call the model at /v1/chat/completions, or another of them, such as /v1/embeddings.